The Interpretation of National Income Estimates. B. D. Haig and S. S. McBurney

Transcription

1 The Interpretation of National Income Estimates B. D. Haig and S. S. McBurney

2 The Interpretation of National Income Estimates Summaries of the flows of total expenditure and income in the economy and in the principal sectors comprising it have been used in many kinds of economic inquiry. In particular, information of this type helps in the examination of current trends in the level of economic activity. This has led to the preparation of social accounts by official statisticians in many countries. In this book an explanation of the nature and purpose of such statistics is given. Topics considered include the economic design of the social accounts, estimation and reliability of the figures, and the types of question which these data help to answer. Principles are explained, but by liberal use of examples and by constant attention to applications for which the data are intended, their discussion is kept on a practical plane. The book should be of value to anyone using national income and social accounting figures. It should be of particular interest to Australian readers, as the examples are drawn mostly from Australian data. However, the principles described have a general application, and some references to overseas practice are made where they are needed to keep the discussion in perspective.

3 This book was published by ANU Press between This republication is part of the digitisation project being carried out by Scholarly Information Services/Library and ANU Press. This project aims to make past scholarly works published by The Australian National University available to a global audience under its open-access policy.

4 The Interpretation of National Income Estimates B. D. Haig and S. S. McBurney 1968 Australian National University Press Canberra

5 F ir s t published 1968 Copyright in all co u n tries su b scrib in g to the B erne Convention; reproduction in whole o r in p a rt, w ithout w ritten p e rm issio n of the p u b lish e rs, is forbidden Typed on IBM Executive Bold F ace No. 2 ty p e w riter and printed by G illingham P rin te rs P ty. Ltd., A delaide P rin te d and m anufactured in A u stralia R eg istered in A u stra lia fo r tra n sm issio n by post as a book N ational L ib ra ry of A u stra lia re g no AUS

6 Preface The chapters of this book w ere originally prepared as papers for a serie s of sem inars in the D epartm ent of Economic H istory, The A ustralian National U niversity, on the meaning and interpretation of national income estim ates. Although the papers have been revised in the light of discussion at the sem in ars, and some new m aterial - including a discussion of quarterly estim ates - has been incorporated, they retain th eir original function: to discuss some of the questions which a rise in using national income data in both sh o rt-term economic analysis.and long-term appraisal of economic trends. This work is therefore not an attem pt to provide a general introduction to methods of national income accounting or a full account of the sources and methods of estim ation. The fo rm er task is adequately covered for A ustralian re a d e rs in a num ber of other publications, including P ro fesso r R. I. Downing's National Income and Social Accounts (1966), and chapters in P ro fe sso r P.H. K arm el's Applied Statistics for Econom ists (1963). The la tte r task would involve a study much beyond the scope of a volume of the p resent size. It is hoped that in the absence of a detailed official volume of sources and methods the presen t work may a ssist u sers of national accounting figures by setting out in m ore detail than is at present available the chief considerations which affect the reliability and meaning of A u stralian national accounting s ta tis tic s. The first four chapters deal m ainly with the interpretation of the official estim ates of national income and expenditure contained in the annual Commonwealth Bureau of Census and S tatistics publication, A ustralian National Accounts, National Income and Expenditure. They include a discussion of the treatm en t of governm ents and financial e n terp rises in the social accounts, and the reliability of the estim ates. The next chapter outlines features of the quarterly estim ates, contained in official Bureau of Census and Statistics publications of Q uarterly E stim ates of National Income and Expenditure. The following three

7 vi P reface chapters cover some of the problem s which a rise in m easuring and using se rie s of national product estim ates for purposes of com parisons over tim e and between countries. This discussion is relevant, for exam ple, to the use of estim ates set out in P ro fesso r N.G. ButlinTs A ustralian Domestic Product, Investm ent and Foreign Borrowing (1962). The first four and the la st four chapters w ere prepared by different authors and there is unavoidably a difference in approach and in style between these chapters. No attem pt has been made to incorporate the chapters into a uniform presentation. Some topics are discussed in m ore than one chapter, but these discussions are approached from a slightly different point of view and it is thought that this may help to clarify different aspects of difficult theoretical problem s in the interpretation of national income estim ates. It also m eans that individual chapters cover the main points about the subject dealt with by the chapters, and are thus largely self-contained. Duplication in the original sem inar papers has, however, been deleted where it sim ply re ite ra ted m atters of fact. The first four chapters w ere w ritten by S. S. McBurney and the last four by B. D. Haig. We are grateful to a num ber of people for help during the preparation of this publication. Comments at the sem inars in the D epartm ent of Economic H istory w ere useful in revising the papers, M rs. A. Guenot typed the d rafts, and the staff of the A ustralian National University P re s s w ere helpful in preparing the m anuscript for publication. Finally, the sem in ars for which the papers w ere originally prepared w ere organised by P ro fesso r N.G. Butlin, and it is m ainly from his suggestion that the papers may be of in terest to a w ider audience that the p resen t publication originated. C anberra 1967 B. D. H. S. S. McB.

8 Contents Page P re fa c e v Introduction 1 1 The Concept of P roduction in Social A ccounting 5 The n atu re of social accounting 5 P roduction - g en eral asp e c ts 6 P roduction by governm ents 8 P roduction by financial e n te rp ris e s 11 G ro ss and net concepts 16 2 C lassificatio n of T ra n sactio n s by S ecto rs and by A ccounts 18 A ccounts of se p a ra te s e c to rs 18 Sub-division of accounts w ithin s e c to rs 19 O ther fa c to rs affecting the accounts 22 F u rth e r developm ents 26 3 E stim ation and R eliab ility of Social A ccounting S tatistic s - C oncepts 28 P rin c ip a l step s in estim atio n 28 A ccounting re la tio n s and co n sistency 29 B asic s ta tistic a l data - concepts 33 4 E stim ation and R eliab ility of Social A ccounting S tatistic s - C overage and A ccuracy 44 B asic sta tis tic a l data - coverage 44 B asic sta tistic a l data - re lia b ility 49 R eliability of social accounting e stim a te s 51 5 Q u arterly E stim a te s 59 The form and purpose of the q u a rte rly e stim a te s 59 Tim ing d iscre p a n c ie s 62 Seasonal adjustm ent 67 R eliab ility of q u a rte rly e stim a te s 72 6 N ational P ro d u ct as a M easure of Total Output 75 C oncepts of national product 77 S election 78 vii

9 viii Contents Page Duplication 81 The valuation of national product 86 7 Deflation of National Product 92 M ethods of deflation 92 Investm ent in stocks and d ep reciatio n 96 The b ase period 99 Quality change Com parison of the National Product and Income of C ountries 105 A lternative approaches 106 Interpretation of international com parisons 109 Sources of inform ation 113 F u rth er Reading 116 L ist of R eferences Cited 118

10 Introduction The f ir s t official estim ates of the national income and expenditure of A ustralia w ere published in a paper accompanying the Commonwealth Budget in These gave a detailed description of the activities of the governm ent secto r as well as the figures of total national income and expenditure. Publication of these estim ates has continued annually and in subsequent issues they w ere expanded to provide a social accounting presentation of the economic activities of all secto rs of the economy. In 1963 the Commonwealth Statistician issued the firs t com prehensive publication of national income estim ates, which provided a much g reater a rra y of statistical inform ation about the workings of the economy and. for the first tim e, gave estim ates adjusted for price changes, which sum m arised the A ustralian rate of econom ic growth since 1948/49. At about the sam e tim e as the detailed official publication was being prepared, other se rie s of national income figures w ere being compiled and issued. In 1960 the Commonwealth Statistician issued the first of a now reg u lar serie s of quarterly estim ates of national income and expenditure, and h isto rical serie s dating from 1860 w ere provided in 1962 in P ro fesso r N.G. B utlin s A ustralian Domestic Product, Investm ent and Foreign Borrowing (1962). While many attem pts had been made by A ustralian econom ists to m easure A ustralian national product since the la st century, P ro fesso r B utlin's publication was the first attem pt to provide a com prehensive long-run se rie s of total national product, using modern methods of com pilation and giving the supporting details needed for the application of the estim ates in h isto rical interpretation. The publication of these estim ates - both official and unofficial - has led in recent y e ars to a g reat in crease in the use of national income figures in the study of economic activities. The R eport of the Committee of Economic Enquiry (The Vernon Report) (Commonwealth of A ustralia 1965) used official annual figures to describe the course 1

11 2 National Income E stim a te s of A ustralian economic growth since World War II and as a b asis for economic projections up to 1974/75. Q uarterly and annual estim ates are used in cu rren t discussions of economic conditions, such as in the Commonwealth T reasu ry s annual economic surveys and in regular quarterly a rtic le s in the Economic R ecord on the A ustralian economy. In line with developm ents in overseas countries the national income estim ates have been used in the study of economic history and have, in this use, provided a convenient sum m ary of the main factors in A ustralian economic growth. It seem s true to say, however, that in spite of the im provem ents in basic statistics the use of national income estim ates in the interpretation of economic changes could be much further developed. This is for two main reasons. F irst, th ere are notable gaps in both the official and unofficial se rie s which lim it th eir applications. Thus, while h isto rical serie s exist up to 1938/39, they are not related to the official post-w orld War II estim ates either in concept or in statistical detail; the official quarterly estim ates do not elim inate the influence of price changes; annual estim ates do not provide details of capital depreciation in curren t replacem ent p rices or the stock of capital assets. Additional estim ates are needed to improve the adequacy of available figures for the appraisal of either long-term or sh o rt-term economic activities. The second reason is that up to the p resen t tim e there has been a lack of documentation of sources and methods and of the effect of these on the meaning and reliability of the estim ates. The national accounting figures are an attem pt to provide statistical m agnitudes for a num ber of economic concepts. There a re, however, many difficulties in providing the appropriate se rie s required for different types of economic analysis. This is partly because many economic concepts, p articu larly those which have originated in ab stract th eo ries, are difficult to define in a way which leads to an unambiguous m easure. The relationship of any statistical estim ate to the ideal theoretical c o rrelate is, accordingly, often a tenuous one; the concept of production, for exam ple, which is basic in the m easurem ent of national product is defined as much in term s of statistical conventions as of ab stract concepts of economic production; sim ilarly the meaning of consumption and investm ent expenditure, as used in national income accounting, can only be properly assessed by consideration of conventions used in th eir

12 Introduction 3 m easurem ent. In addition, national accounting estim ates are synthetic sta tistic s, obtained by sum m arising a great m ass of detailed inform ation into a few p articu lar estim ates. The basic data used in the estim ates vary considerably in reliability and thus the resulting figures differ in th e ir reliability, and som etim es, for purely statistical reaso n s, they may have features which are not obvious in the sim ple definitions. Thus the interpretation of the estim ates has to be made in the light of the adequacy and type of basic data used in th e ir estim ation. This study is largely an ex ercise in setting out some of the factors in the definition and estim ation of national income figures which are likely to affect the interpretation of available A ustralian estim ates. It does not attem pt to add to existing se rie s or to in terp ret the figures for any p articu lar purpose. It is, however, hoped that the discussion in different chapters may help in providing inform ation needed in one aspect of the interpretation of the available figures of A ustralian national product and the social accounting tables in which they are presented.

13 1 The Concept of Production in Social Accounting The nature of social accounting The purpose of social accounting is to am algam ate the accounts of individual tra n sacto rs (e.g. p articu lar e n terp rises, persons, governm ent authorities) into a set of accounts sum m arising the nation's business in a form suitable for economic analysis. It is concerned with the preparation of p articu lar types of economic statistic s. Many other types of economic statistics (e.g. life insurance statistic s, production statistics) are also am algam ations of figures relating to individual tra n sacto rs. Their purpose differs from that of social accounting data in that they are intended to provide details about p articu lar activities as such, w hereas social accounting data are concerned with the perform ance of the entire economy and the relationship of p articu lar activities to it. The difference of purpose is reflected in the classifications derived from the two types of am algam ation of accounts. For exam ple, life insurance statistics contain details about item s such as types of policy, different causes of discontinuance of policy, and classifications of prem ium s and claim s. These are im portant aspects of life insurance business, but they have no counterparts in other types of business and therefore cannot be added to details for other types of business to obtain sim ilar figures for the entire economy. In social accounting data som e of the m ore im portant classifications are production, incom e, expenditure, saving, borrowing, lending. F igures of this type, even if they relate to different kinds of tra n sacto r, can be totalled to yield inform ation of the sam e type for im portant secto rs of the economy or for the entire econom y. Most social accounting work has been built around the concept of production. As a larg e part of economic doctrine is concerned with the production, distribution and exchange of goods and serv ices, statistical data of this type have had a num ber of applications in applied 5

14 6 National Income E stim ates econom ics. Some applications to w elfare econom ics may only require estim ates of total production; others may require classifications of the total, e.g. by industry, by income sh are, o r by type of expenditure to which it is applied. Since the publication of Keynes s G eneral Theory (1936), and the growth of attention to analyses of income and expenditure which it prom pted, a need has arisen for m ore detail of the transactions of p articu lar secto rs and th eir relationship to the national totals. This has led to the publication by government statisticians of sets of social accounts showing transactions between m ajor secto rs of the economy. In the rem ainder of this chapter and in Chapter 2 som e of the basic concepts of social accounting are review ed briefly, to provide a basis for the discussion in la te r chapters. Chapter 1 considers the concept of production, which is fundamental in social accounting work, w hether we are concerned only with the national aggregates and their components or whether we are concerned with integrated sets of accounts. Chapter 2 considers the system s of accounts in which the resu lts a re presented. Production - general aspects The definition of production requires decisions on two m atters. F irstly, one m ust decide what item s a re included in production, that is one m ust define the flow of goods and serv ices in real term s. Secondly, one m ust decide the value to be applied to each item of production; this is essentially a m atter of choosing a set of value weights to be applied to the real quantities. If we a re given an estim ate of total production, however defined, we may inquire to what uses the goods and serv ices represented by it have been put. The principal uses which are commonly distinguished are consumption by residents or governm ent authorities, additions to capital a ssets (whether fixed capital equipment or stocks), and net ex p o rts.1 When expressed in money values the uses are called final expenditures. Given the estim ate of production, we may also inquire to whom it accru es. Division of the product among those who derive the benefit is indicated by a classification of income sh ares. Although production may be defined in various ways, the relations noted above a re independent of the definition chosen; total production, the to tal of The sam e id en tity is frequently rew ritten to show production plus im ports allo c a te d betw een the follow ing uses: consum ption by residents and g o vernm ent authorities, additions to c a p ita l assets, and to ta l exports.

15 Production in Social Accounting 7 the corresponding final expenditures, and the total of the corresponding income shares are n ecessarily identical. One of the conventions adopted in defining production for social accounting purposes is to include goods and serv ices bought and sold in the m arket, and to value them at th eir m arket p rices. In o rder to avoid m ultiple counting of goods which pass through sev eral stages of production, a distinction is drawn between final and interm ediate goods. Thus if a loaf of bread sells for twenty cents, it would not be c o rre c t to include in the estim ate of production the price of the flour used in producing the bread and the price of the wheat used in producing the flour in addition to the twenty cents for which the loaf of bread itself sells. To obtain the co rre ct figure, one may add up the values of the final' output (i.e. amounts corresponding to the value of the bread, but not to the wheat or flour which have gone into its production), or one may add up each p ro d u cer's contribution to the final product (i.e. the value added by his business, which is the value of his total sales less the value of interm ediate products purchased by him ), or one may add up the income shares obtained when the value added by each producer is distributed. Another im portant concept, 'factor co st, is derived from an analysis of the income sh ares. R eferring to the example of the preceding paragraph, every loaf of bread produced gives ris e to incomes of twenty cents (other things, such as im ports and stocks of wheat and flour, being equal). Some of the income is paid to the governm ent as indirect taxes (e. g. payroll tax) and the re s t is paid to the owners of productive re so u rces, either in the form of wages and sa la rie s, or as the various income sh ares making up the gross operating surplus of the en terp rises the activities of which contribute to the production of the bread. The paym ents to the owners of productive reso u rces are the factor cost corresponding to the output of bread. The im portance of the concept a rise s from the fact that it is the m arket price paid for the serv ices of the productive reso u rces. It is therefore a useful concept in applications involving examination of the demand for re so u rc e s, e.g. Keynesian a n a ly sis.2 In the case of trading2 ^ This term is used here to describe analyses of aggregate incom e and expenditure and their determinants which have developed since the publication of Keynes's General Theory (1936); it is not restricted to the m odel used in the General Theory,, 3 'Trading' enterprise is here used in its wide social accounting sense: any enterprise other than a financial enterprise.

16 8 National Incom e E stim ates e n te rp ris e s, fa c to r co st b e a rs a sim ple rela tio n sh ip to the m a rk e t p ric e of production (i.e. sale s of output le ss p u rch ases of goods and se rv ic e s from o th er p ro d u c e rs), as the two differ only by the am ounts of in d ire c t ta x es (net of su b sid ies). H ow ever, as la te r p a rts of this ch ap ter w ill show, m o re fundam ental d ifferences a ris e in connection with the output of governm ents and financial e n te rp ris e s (and th e re fo re in connection w ith the output of the econom y as a whole): the definition of th e ir output m u st be chosen with c a re if it is to be closely re la te d to the facto r co st of re s o u rc e s used. N on-m arket item s of c e rta in types a re included in production at im puted values. E xam ples a re the re n ta l value of ow ner-occupied houses and consum ption of farm produce by the fa rm e rs. C orresponding to the im puted production, th e re a re im puted incom es d istrib u ted to the o w n er-o ccu p ier and the fa rm e r re sp e c tiv e ly, and th e re a re also final expenditures re p re sen tin g th e ir im puted p u rch ases from th em selv es. T hese item s a re d iscu ssed in C hapter 6. P roduction by governm ents The tre a tm e n t of governm ent e n te rp ris e s, which sell goods or se rv ic e s at a p ric e, is co v ered by e a r lie r com m ents. Leaving th ese asid e, th e re a re d ifferences of opinion as to the ideal tre a tm e n t of g en eral governm ent activ ity. The p ra c tic e s followed differ le s s than opinions about th e ir adequacy, p a rtly because s ta tistic a l co n sid eratio n s lim it the freedom of p ractice if e stim a te s of definable concepts a re to be obtained. The debate about the adequacy of the conventional tre a tm e n t has not suggested an altern ativ e which is both c le a rly defined and m o re ap p ro p ria te fo r all of the m ain applications. The m ark et p ric e of m o st g en eral governm ent activity is n il, as it is not sold on the m a rk e t (there a re exceptions, such as fees of h o sp ita ls, schools, com pany r e g is tra r s, and other m iscellaneous ch arg e s of governm ents). C onsequently, if m a rk e t p ric e valuation w ere used for g en eral governm ent activity in the sam e way as for b u sin ess activity, the product (i.e. value added) in the g en eral governm ent se c to r would be a su b stan tial negative am ount because its p u rch ases of in term ed iate goods and se rv ic e s from o th er s e c to rs would exceed its s a le s to o th er s e c to rs (the fees and m iscellaneous ch arg e s noted above). The incom e s h a re s co rresp o n d in g to that

17 Production in Social Accounting 9 negative product would consist of: (i) indirect taxes paid by the general governm ent sector,4 (ii) the wages and sala rie s paid to employees engaged in general governm ent, and (iii) a much la rg e r negative item representing the operating loss on general governm ent activity (this loss is paid for, of course, from taxation and other revenue tra n sfe rre d to the governm ent). This is illu strated in the following example. Suppose that the governm ent purchases $100 m illion of goods and serv ices from other secto rs, pays wages of $50 m illion and indirect taxes of $5 m illion, and charges other secto rs $10 m illion by way of fees and other m iscellaneous charges. Then m arket price valuation of output would yield the following production account for the general governm ent sec to r: Production Account (m arket price basis) E xpenditure $m R eceipts $m Interm ediate purchases 100 Sales (fees, etc.) 10 Value added: wages 50 operating su rp lu s -145 indirect taxes 5-90 Output (m ark et price) 10 Output (m arket price) 10 In p ractice the m arket p rice valuation is not retained for the general governm ent secto r. The reason for government functions having no selling price (or only a nominal one) is in many cases quite unrelated to the valuation of the product. Several governm ent serv ices are provided as communal rath er than individual serv ices (e.g. roads, parks) and this ren d ers the collection of a price im practicable. Also som e serv ices (e.g. education) are deliberately made available to individuals free of charge (or for nominal fees) for social or political reaso n s, and not because it is considered the service is worth no m ore. If the figures are being used for w elfare analysis a m ore appropriate value weight should be applied to such serv ices. If the figures are being used for Keynesian analysis of the relation between 4 In d ire c t taxes m ay be paid from one authority to another w ithin the g en eral govern m en t sector, e.g. state governm ents pay p ay ro ll ta x to the C om m onw ealth govern m en t in A u stralia.

18 10 National Income E stim ates total employm ent and total expenditure, the concepts should be defined in such a way that an increase in the employment of, say, teach ers will be m atched by an in crease in final expenditure on education. E ither type of application calls for a variation of the m arket p rice basis for valuing output of the general governm ent secto r. The convention usually adopted in practice is to value the output of the general government secto r at cost, that is to value it at such a sum that the operating surplus on governm ent activity is zero. The data used above would yield the following production account of the general governm ent sector if the output w ere valued at cost: Production Account (cost basis) E xpenditure $m R eceip ts $m Interm ediate purchases Value added: wages 50 indirect taxes Sales: to other secto rs 10 to governm ent Output (cost) 155 Output (cost) 155 In these exam ples $145 m illion of expenditure has to be m et from taxation and other general governm ent funds, however it is classified. When the cost basis is used, the $145 m illion is regarded as an imputed purchase of its own output by the governm ent on behalf of the com m unity, but when the m arket price basis is used th ere is no output corresponding to the $145 m illion. The cost basis yields higher figures of governm ent product and national product than the m arket p rice basis - the difference is equal to the net adm inistrative expenditure by governm ents (i.e. total expenditure le ss fees, charges, e tc.). Corresponding to this it also yields higher figures of incom es (operating surplus is zero instead of a negative amount) and of final expenditure (sales of governm ent output to the governm ent itse lf). The p ractice of valuing the whole of general governm ent output at cost is often criticised on the grounds that some serv ices provided by governm ents are interm ediate output ra th e r than final output (e.g. defence, police fo rces, general adm inistration) and that the

19 P roduction in Social A ccounting 11 in term ed iate se rv ic e s a re duplicated in the national product e stim ates if the c o st of all governm ent functions is included. The extent of this duplication p ro b lem, its co n sequences, and a p a rtia l solution a re investigated in C hapter 6. H ere it is sufficient to note th a t, w hatever the m e rits o r d e m e rits of the conventional tre a tm e n t for w elfare a n a ly sis, it p rovides a satisfa c to ry concept of production for use in K eynesian a n a ly sis, since it m atch es an in c re a se in em ploym ent in governm ent activity with an in c re a se in final expenditure on the production' to which the em ploym ent gives r is e. A fter deducting in d ire c t taxes paid, this valuation of governm ent output, like m a rk e t p ric e valuation of trad in g e n te rp ris e s ' output, y ie ld s the facto r co st of the output. As noted above, fa c to r co st is a m e a su re of the dem and fo r produotive re s o u rc e s valued at th e ir own m a rk e t p ric e. A lternative valuations of governm ent output (e.g. m a rk e t p ric e s, or the co st of se rv ic e s o th er than those re g a rd e d as in term ed iate output) would not yield the facto r co st a fte r deducting in d irect taxes (net of su b sid ie s).^ P roduction by financial e n te rp ris e s The p rin cip al activity of th is group of e n te rp ris e s is borrow ing and lending of m oney, and th e ir m ain so u rce of incom e is the m argin betw een re c e ip ts of in te re s t and paym ents of in te re st and a d m in istra tiv e expen ses. The e n te rp ris e s include banks, money m a rk e t d e a le rs, life in su ran ce co m p an ies, superannuation funds, in stalm ent c re d it co m panies, building so c ie tie s, and the m ortgage lending d ep artm en ts of housing com m issio n s and of o th er public a u th o ritie s such as the W ar S ervice H om es D ivision of the D epartm ent of H ousing. T hese e n te rp ris e s call for sp ecial m ention in this d iscu ssio n of production, not p rim a rily because of the size of th e ir borrow ing and lending and th e ir re c e ip ts and paym ents of in te re s t, but because the valuation of the product of som e of them poses a problem s im ila r to th at posed by governm ents. They have su b stan tial running c o sts to pay and, like governm ents, they m eet them la rg e ly out of tra n s fe r incom e ra th e r than out of selling p ric e s ch arg ed for th e ir s e rv ic e s. The conventional m ethods of tre a tin g It is, of course, possible to derive a gross operating surplus consistent with any valuation of output m erely by balancing the production account. The resulting surplus, however, need not be a com ponent of factor cost even though its derivation is form ally sim ilar to that of the gross operating surplus of a trading enterprise (w hich is a com ponent of factor cost). For instance our first exam ple production account illustrates market price valuation of governm ent output; the negative operating surplus of -$145 m illion is not a factor cost as it is not in any sense a remuneration to attract productive resources into service.

20 12 N ational Incom e E stim ates such e n te rp ris e s in the social accounts a re le ss stan d ard ised than the m ethods of tre a tin g governm ents; in p a rtic u la r, the p ra c tic e followed in the A u stralian official e stim a te s d iffers from the p ra c tic e s m ost com m only followed by o th er co u n tries. The a d m in istrativ e expenses a ttrib u tab le to som e e n te rp ris e s within th is se c to r m ay not be ch arged against th e ir earn in g s, but m ay be tre a te d as expenditure by another entity. Thus the expenses of the W ar Service H om es D ivision and of som e governm ent superannuation funds a re included in the output of and final expenditure by public a u th o ritie s. The expenses of p riv ate superannuation funds m ay also be charged as expenses of the b u sin ess which o p erates them, and may thereby be included in the value of its output. No special p ro b lem s a re caused by th is group of financial e n te rp ris e s. The prem iu m s paid to life in su ran ce com panies and the contributions paid to superannuation funds (other than those m entioned in the la st p arag rap h ) have to cover not only the a c tu a ria l co st of the benefits provided, but also the a d m in istrativ e ex penses of the life offices or the superannuation funds. In the social accounts the p rem iu m s and contributions a re broken into an expense com ponent (equal to the actual expenses) and the balance (which is tre a te d as p erso n al saving). The expense com ponent is tre a te d as a sale of output by financial e n te rp ris e s to p e rso n s, and is included in p erso n al consum ption expenditure. As the value of the sale is equal to the actual expenses th e re is no operating su rp lu s, and the value added (at facto r cost) by th is group of financial e n te rp ris e s is equal to th e ir wage and sala ry b ill. The rem aining e n te rp ris e s ch arg e le ss for th e ir se rv ic e s than they pay for a d m in istrativ e ex p en ses, and th e re is th e re fo re an operating lo ss to be m et from th e ir net in te re s t earn in g s. In the re m a in d e r of th is ch ap ter the tre a tm e n t of th e se e n te rp ris e s is illu stra te d by specific re fe re n c e to th e ir m o st im portant type (the banks), but the p rin c ip le s also apply to other e n te rp ris e s in this group. If banks ch arg e c u sto m e rs $2 m illion in fees and com m ission for keeping c u rre n t accounts and fo r c e rta in o th er s e rv ic e s, pay $5 m illion for goods and serv ic e s purch ased from other b u sin e sses, $5 m illion in wages and s a la rie s and $1 m illion in in d irect ta x e s, th e ir production account would appear as follow s, if th e ir output w ere valued at m a rk e t p ric e s:

21 Production in Social Accounting 13 Production Account of Banks (m arket price basis) Expenditure $m R eceip ts $m Interm ediate p u rch ases 5 Sales (fees) 2 Value added: wages 5 operating su rp lu s -9 in d irect taxes JL -3 Output (m arket price) 2 Output (m arket p rice) 2 This b asis is generally considered to have some unsatisfactory features which have prom pted a search for alternative bases of valuation. It is objected that in o rd er to a ttra ct funds from which their in terest earnings are derived, banks provide serv ices to custom ers at less than cost (and at less than th eir true value). It is also objected that it is unreal to rep resen t this industry as paying expenses g re a ter than the value of its output, even when it is prospering. For these reasons it is held that the m arket p rice undervalues the output of these enterp rises just as it undervalues the output of governm ents - the banks, like the governm ents, can afford to ignore the m arket m echanism in pricing th eir output or serv ices because they have substantial tra n sfe r income to cover deficiencies in tjie service charges. An alternative to m arket price valuation is therefore sought for the output of banks, just as it was sought for the output of governm ents. In the B ritish official estim ates the m arket price basis is retained, as the alternatives a re not entirely satisfactory in principle, and they pose some difficult statistical problem s. The figures of value added in the official estim ates are presented, how ever, not as the sum of wages and the negative operating surplus (as in the account above) but as the sum of w ages, bank profits (as ordinarily understood), and a negative adjustm ent equal to net in te re st received by the banks. The sam e bank product and national product figures are obtained from either presentation, but this procedure has the effect of setting out details of the total in a form which can be m ore readily related to other data on incom es of e n terp rises. In the A ustralian official e stim a te s the convention adopted is to value

22 14 N ational Income E stim ates the output of banks at cost (i. e. to value it at such a sum that the operating surplus of banks is zero), and to impute a sale of p art of the output to the banks them selves. The data for the exam ple given in the previous table then yield the following production account: Production Account of Banks (cost basis) Expenditure $m R eceip ts $m Interm ediate purchases 5 Value added: wages indirect taxes Sales: to other secto rs to banks 2 9 Output (cost) 11 Output (cost) 11 The treatm ent is the sam e as that adopted for valuing the output of governm ents. That basis is adopted because the serv ice provided by banks (keeping custo m ers' accounts, exchanging cheques, e tc.), like some adm inistrative functions of governm ents, is a n ecessary expense of keeping the economy running. Because of the sim ilarity to those functions of governm ent, the sam e basis of valuation seem s appropriate. The objections to, and defence of, the conventional treatm ent of governm ent also apply to this treatm en t of banks. The replacem ent of m arket price valuation by valuation at cost has the effect of raising the estim ates of bank product and national product by the net expenses of banks (i.e. th eir adm inistrative expenses less service charges). Corresponding to th is, there are higher incomes (operating surplus is zero instead of a negative amount) and higher final expenditures (sales of banking output to the banks them selves on behalf of th eir custom ers). Value added (at factor cost) by banks, according to this treatm en t, is equal to th eir wage and sala ry bill. The practice followed in the official estim ates of many countries is to add to the actual service charge by the banks an additional imputed service charge which brings the value of banking output up to what is considered a m ore appropriate figure, and to break up the imputed charge among p articu lar classes of custom ers and derive the

23 P roduction in Social A ccounting 15 consequential effect on the national product fig u re s. One effect of im puting such a ch arg e is to r a is e the product of banks above the m ark et p ric e valuation by the am ount of the im puted ch arg e. In so fa r as the im puted charge is tre a te d as a paym ent by p e rso n s o r by governm ents it is added to final expenditure (personal consum ption, or expenditure on goods and se rv ic e s by public au th o rities). In so fa r as it is tre a te d as a paym ent by e n te rp ris e s it is im puted in term ed iate expenditure, and the product of o th er in d u strie s is to th at extent reduced. The effects of su bstituting th is b a sis of valuation for the m ark et p ric e b a sis a re th e re fo re as follows: (i) national product is in c re a se d by the im puted charge to p e rso n s or governm ents; (ii) final expenditure is also in c re a sed by th at am ount; (iii) the product of banks is in c re a sed by the to tal am ount of the serv ice ch arg e to all c u sto m e rs, and the product of o th er in d u strie s is reduced by the am ount of the im puted ch arg e to o th e r e n te rp ris e s. T here is thus som e re d istrib u tio n of product to banks from other in d u stries. The im putation could take any one of sev e ra l fo rm s. C onsider the following exam ple: banks re c e iv e in te re s t of $50 m illion and actual serv ice ch arg e s of $3 m illion; they pay in te re st of $22 m illion, in d irect taxes of $1 m illion, and other expenses (including w ages and sala rie s) of $10 m illion, and m ake a p ro fit of $20 m illion. If the output of banks w ere valued at co st the im puted ch arg e would be $8 m illion (expenses and taxes $11 m illion, le ss actual ch arg e s of $3 m illion); if it w ere valued at c o st plus p ro fit, the im puted ch arg e would be $28 m illion (expenses and ta x es of $11 m illion, plus p ro fit of $20 m illion, le ss actual ch arg e s of $3 m illion). The ch arg e s could be im puted en tire ly to b o rro w e rs, en tire ly to d e p o sito rs, o r p a rtly to each. If an im puted ch arg e of $8 m illion w ere co n sid ered as paid en tirely by b o rro w e rs, the actu al in te re s t of $50 m illion paid by them would be reg ard ed as equivalent to paym ents of $8 m illion of im puted serv ice ch arg e s and $42 m illion of in te re st p ro p e r. If an im puted charge of $8 m illion w ere co n sid ered as paid en tirely by d ep o sito rs, the actual in te re st of $22 m illion receiv ed by them would be reg ard ed as equivalent to in te re s t p ro p e r of $30 m illion (actual $22 m illion plus im puted $8 m illion), le s s im puted serv ic e ch arg e s of $8 m illion. The ^ The A ustralian p ractice (described in the preceding paragraph) im putes a sale by banks to them selves rath er than to custom ers.

24 16 National Income E stim ates practice recom m ended by the United Nations Statistical Office in A System of National Accounts and Supporting Tables (U.N. 1964b), and followed by a num ber of countries, is to value the services of banks, like the output of other e n terp rises, at cost plus profit, and to reg ard the imputed charge as paid entirely by depositors. In term s of the example above, this would involve regarding the actual in terest of $22 m illion received by depositors as equivalent to in terest proper of $50 m illion, less imputed service charges of $28 m illion. This practice leads to a higher figure for the product of banks and a sm aller figure for the product of other enterp rises than does the A ustralian practice. Their relativ e effects on total national product depend on how much of the $28 million of imputed charges is attributed to enterp rises and how much to persons and governm ents, as the United Nations practice adds the la tte r component to the m arket price estim ate of national product, while the A ustralian practice adds the net expenses of banks (total expenses less actual charges, amounting to $8 million in the example above) to the m arket price estim ate. 7 G ross and net concepts The estim ates of product and of capital expenditure may be m easured either before or after deducting depreciation as a cost. In the form er case gross figures are obtained, in the latter net figures. D epreciation is considered further in Chapter 6, and exam ination of the nature of the concept is left until then. The gross and net concepts are mentioned here solely for the purpose of noting th eir relevance to two of the principal applications of social accounting data - w elfare analysis and Keynesian analysis. The net product is the value of goods and serv ices left after maintaining capital intact, and it m easures the nation s income, in the sense of the amount which it could consume or tra n sfe r overseas without living on its capital. The net concept appears to be the one m ore suited to w elfare applications such as studies of income per head and the study of income sh ares. In Keynesian applications involving study of the level of demand the gross concepts are m ore appropriate, because the construction of new 7 N atio n al A ccounting P ractices in Sixty C ountries (U.N. 1964a) states th a t th e A ustralian procedure 'results in higher incom e, product and expenditure aggregates fhan in the U nited N ations system )to the ex ten t th a t service charges are im p u ted to enterprises under the U nited N ations sy stem '. This co m m e n t overlooks the fa c t th a t the U nited N ations p ractice values the output of banks a t a h igher figure th a n does the A ustralian p ractice (cost plus p ro fit in the form er case, co st only in the la tte r).

25 P roduction in Social Accounting 17 assets calls forth a demand for labour and other re so u rces, w hereas the wearing out of old a sse ts does not involve a diminution of that demand. For exam ple, if 100 new buildings w ere erected, partly for replacem ent and partly for expansion of the stock of buildings, there would be a demand for labour and other reso u rces used in th eir construction, and that demand would be the sam e irresp ectiv e of whether old buildings w ere w earing out at a rate of 10 per y ear, 20 per y ear, or some other rate. The 100 corresponds to the g ross product, which is therefore the m ore im portant concept for studies of aggregate dem and.

26 2 Classification of Transactions by Sectors and by Accounts In Chapter 1 the definition of the national product and the corresponding expenditures and income sh ares w ere considered. The national totals may be classified by industry, by type of incom e, or by type of final expenditure. Many applications of social accounting, however, require m ore inform ation than we can obtain from such sum m aries. Accounts of separate secto rs The national expenditure estim ates provide figures for some of the im portant Keynesian aggregates, such as personal consumption, private investm ent, and public authority expenditure. In o rd er to explain the variations in these expenditures it is often n ecessary to consider other related aggregates which are not provided in sum m ary tables of national product and expenditure. For exam ple, personal consumption is influenced by the level of personal incom e, and the la tte r figure is not contained in the national product or expenditure sum m aries. An in crease in final expenditure by public authorities may be financed by taxation or by borrow ing, but the fo rm er is likely to re s tric t private expenditure to a g reater extent than the latter; as the national product and expenditure sum m aries do not indicate the method of financing p articular expenditures, some supplem entary inform ation is again needed. In ord er to throw light on these m atters the statistical data m ust show not only the amount produced and the uses to which the product is put, but also the sources of funds used by the p u rch asers to finance each of the main types of final expenditure. This involves presenting details of tra n sfe rs from those who initially receive the proceeds of production to those who make the final expenditures. The two groups overlap to a considerable extent (e. g. wages and sala rie s a re im m ediate proceeds of production, and the wage and salary earn ers them selves devote a large p art of th eir earnings to final expenditure 18

27 C lassification of T ransactions 19 on goods and serv ices), but th ere rem ains a substantial amount of expenditure financed from cu rrent tra n sfe rs such as tax revenue, pensions, property incom e, or from borrowing, ra th e r than directly fro m the im m ediate p ro ceed s of production. F or this reason official statistician s supplem ent the national product and expenditure estim ates with sets of social accounts showing transactions between the m ajor secto rs of the economy. T ran sacto rs are grouped into a sm all num ber of secto rs according to the nature of th e ir activities". Each se c to r s share of the im m ediate proceeds of production is credited to its account, and so also are tra n sfe rs from other secto rs. On the other side of the account is shown the use made of the funds - tra n sfe rs to other secto rs and final expenditure by the secto r in question are the m ajor u ses, and each is broken up into suitable item s. The num ber of secto rs for which accounts are shown can be large or sm all. The minimum detail needed for Keynesian analysis req u ires separate accounts for perso n s, governm ents, en terp rises, and overseas tra n sacto rs, because of the very different influences affecting the expenditures by those four groups. F u rth er break-dow ns of secto rs can be useful if the transactions of the separate groups are d issim ilar and substantial. In that case significant inform ation on p articular expenditures and their determ inants is revealed by the m ore detailed classification. In accordance with this principle the tran sactio n s of the Commonwealth governm ent and of the state and local governm ents are shown separately in the A ustralian estim ates, because of th eir different functions and sources of revenue and because of the im portance of the financial relations between them. Financial e n terp rises have also been separated from other en terp rises because of the special nature of th eir business (which is p rim arily concerned with borrowing and lending and tra n sfe rs of in terest ra th e r than with production) and because of the special method of valuing th eir product. Sub-division of accounts within s e c to rs Accounts derived in the m anner described above would each contain details of many different kinds of transaction relating, for exam ple, to production of goods and serv ices, to tra n sfe rs of incom e, to current and capital expenditure on goods and serv ices, and to borrowing and lending. A c le a re r picture of each se c to r's business could be obtained

28 20 National Income Estimates if some of these activities were presented separately; this could be done by breaking up the sector account into a number of accounts, each of which covered a general class of transactions. For example, each sector account could be broken into accounts, such as the following three, dealing respectively with production, other current transactions, and capital transactions: Production Account Expenditure $ Receipts $ Purchases of intermediate output of other sectors Wages, salaries, and supplements Gross operating surplus Indirect taxes less subsidies Total Sales of intermediate output 1o other sectors Sales of final output - current - fixed capital - increase in value of stocks Total Current Account Expenditure $ Receipts $ Current expenditure on final output Current transfers to other sectors Gross saving Total Wages, salaries and supplements (personal sector only) Gross operating surplus Indirect taxes less subsidies (public authority sector only) Current transfers from other sectors Total

29 C apital Account C lassificatio n of T ran sactio n s 21 E xpenditure $ R eceipts $ C apital expenditure on fin al output - fixed capital - in c re a se in value of stocks P u rc h a ses of existing cap ital a s s e ts G ro ss saving Sales of existing cap ital a s s e ts T ra n s fe rs of saving from o th er s e c to rs T ra n s fe rs of saving to o th e r s e c to rs T otal Total The value of the s e c to r's production is set out in its production account. That account also shows the use m ade of its output, and the m anner in w hich the incom e produced by the s e c to r is d istrib u ted am ong its im m ediate re c ip ie n ts. The incom e is taken to the c u rre n t accounts of the v ario u s s e c to rs, w here it is supplem ented by incom e re d istrib u te d from other s e c to rs in the form of c u rre n t tra n s fe rs such as ta x e s, in te re s t, dividends, and pensions. P a rt of the incom e is used fo r c u rre n t expenditure on goods and s e rv ic e s, p a rt of it m ay be re d istrib u te d to other se c to rs by m eans of c u rre n t tra n s fe rs, and the balance is the s e c to r's g ro ss saving. The g ro ss saving is taken to the s e c to r's capital account, which shows its final cap ital expenditure on new a s s e ts and sto ck s, and the so u rc e s of funds used to finance it. T hese funds include, in addition to the g ro ss saving of the se c to r, its borrow ing from other se c to rs and its sale s of second-hand a s s e ts to o th er s e c to rs, le ss its lending to other s e c to rs and p u rch ases of second-hand a s s e ts from o th er se c to rs. The purpose to be serv ed by the set of accounts re q u ire s that they include som e im puted tra n sa c tio n s. F o r exam ple, if a b u sin ess does not se ll its e n tire output its stocks of finished goods in c re a se. In the social accounts the estim ate of production should include the value of all goods and se rv ic e s produced, w hether sold or not, and the u ses to which the production is put a re shown as re c e ip ts of the production account. In th is exam ple the p ro d u cer adds the unsold production to

30 22 National Income E stim ates his stocks, and this use of the output is rep resen ted as an imputed sale from the b u s in e s s's production account to its cap ital account. The conventional definition of government output and the A ustralian and the United Nations definitions of the output of financial e n terp rises (described in Chapter 1) also req u ire the recording of imputed sales in ord er to rep resen t the use made of certain outputs. O ther fa c to rs affecting the accounts If the form of accounts shown abovel w ere adopted as the basic stru ctu re, different choices about a num ber of other m atters could still lead to differences of detail in the contents of the accounts. For exam ple, some of the headings could be fu rth er divided. Production item s could be broken up by industry or by type of e n terp rise, expenditures by function, and tra n sfe rs by type (e.g. into in terest, dividends, taxes, pensions, etc., in the case of curren t tra n sfe rs, and into various form s of borrowing and lending in the case of capital tra n sfe rs). However, the various choices which could be made regarding such dissections would m erely split the totals in various ways without altering the totals them selves. More fundamental changes, which would alte r some of the to tals, would be brought about by choices between alternative groupings of tra n sacto rs into secto rs and between alte rn ativ e concepts of production. Different groupings of tra n sacto rs into secto rs generally require different sets of basic data to prep are estim ates of item s in the accounts, and the lack of certain types of data may therefore influence the choice of secto r groupings. A sector account is a consolidation of the accounts of tra n sacto rs included in the secto r, and, like all consolidated accounts, shows transactions between the group which it covers and those outside the group, but excludes details of transactions between different m em bers within the group. Consequently, if data are lacking on transactions between two groups of tra n sacto rs it may still be possible to prep are estim ates of the item s in a consolidated account of the combined group, as the m issing details would relate to internal transactions which are not shown in the consolidated account. The treatm ent of non-profit-m aking organisations such as clubs, churches, and private schools illu strates this point. Data on m ost transactions between these organisations and persons are not available A lternative analyses of each sector's transactions among different accounts are described in Karmel (1963: Ch. 11) and Mathews (1962: Ch. 18).

31 C lassification of T ransactions 23 but some of the basic statistic s, such as retail sales figures, include transactions of the combined group (persons and the clubs, etc.) with other secto rs. If such organisations a re grouped with persons it is not n ecessary to estim ate the transactions between them, and the consolidated account shows th eir combined transactions with other secto rs. Some alternative concepts of production w ere considered in Chapter 1. The choice between them obviously affects the contents of one or m ore of the production accounts in the com plete set of sector accounts. In addition, as the choice of a concept of production involves the choice of corresponding income sh ares and final expenditures, it also affects other accounts in the set. Consider the valuation of government output as it affects the following exam ple. Suppose that the governm ent's cu rrent purchases of interm ediate output from other secto rs are $100 m illion, that it pays wages of $50 m illion and indirect taxes of $5 m illion, that fees and charges to other sectors are $10 million and taxation is $150 m illion, that $20 m illion of the interm ediate purchases and wages go into governm ent construction, that a fu rth er $50 m illion of capital goods are purchased from other secto rs, and that $45 m illion is borrow ed. Using the usual convention of valuing governm ent output at cost, the accounts of the government sector would be as follows: Production Account Expenditure $m R eceipts $m Interm ediate purchases 100 Sales: Value added: wages 50 indirect taxes 5 55 to other secto rs to governm ent - cu rren t - cap ital Output (cost) 155 Output (cost) 155

32 24 National Income E stim a te s C urrent Account Expenditure $m R eceipts $m C u rren t expenditure on final Taxation 150 output 125 Saving 25 Total 150 Total 150 C apital Account Expenditure $m R eceipts $m C apital expenditure on final Saving 25 output B orrow ing 45 - of governm ent 20 - of other secto rs 50 Total 70 Total 70 If expenditure on cu rren t operations w ere valued at m arket price instead of co st, the accounts would take the following form: P roduction Account Expenditure $m R eceipts $m Interm ediate purchases 100 Sales: to other secto rs 10 Value added: to governm ent (capital) 20 wages 50 operating surplus -125 in d irect taxes 5-70 Output 30 Output 30

33 C lassification of T ransactions 25 C urrent Account E xpenditure $m R eceip ts $m Saving 25 O perating su rp lu s -125 Taxation 150 Total 25 Total 25 C apital Account E xpenditure $m R eceip ts $m C apital expenditure on final Saving 25 output - of governm ent 20 B orrow ing 45 - of other secto rs 50 Total 70 Total 70 If the sets of accounts corresponding to the alternative valuations of financial e n te rp rise s' output described in Chapter 1 w ere considered, those corresponding to the m arket price basis and the cost basis would differ in th eir classification of transactions of the financial e n terp rises secto r. The differences between them would be form ally sim ilar to those between the two sets of governm ent accounts derived from those bases and set out in the last paragraph. The differences between either of those treatm en ts of financial enterp rises and the treatm ent recom m ended by the United Nations S tatistical Office would, however, extend beyond the financial e n terp rises secto r. The United Nations p ractice involves imputing sales by financial e n terp rises to trading e n terp rises and persons. When com pared with the m arket p rice or cost b a sis, th erefo re, this treatm ent re c lassifie s not only the financial e n terp rises se c to r's accounts, but also the production and appropriation accounts of the trading e n terp rises sector and the curren t account of the personal sector. A large num ber of accounts (three for each sector) would be required to set out the details described in the preceding paragraphs. It may be possible to sim plify the p resen tatio n without sacrificin g

34 26 National Income E stim ates essential inform ation if som e, but not all, of the accounts are consolidated. The reason given at the beginning of this chapter for presenting sep arate accounts for each secto r was that they showed the tra n sfe rs of income and savings from which some final expenditure is financed. Although separate curren t and capital accounts for each sector are needed for that purpose, separate production accounts are not. Separate production accounts are not shown for each secto r in A ustralian National A ccounts, but only the national production account obtained by consolidation. This leaves final, but not interm ediate, expenditures in the national production account. F u rth er developm ents Two specialised developm ents are re fe rre d to h ere because of th eir relationship to the secto r accounts, although they are not pursued at any length in this book. Input-output tables aim to show the relationship between the output of each industry and its demand for the output of other industries; this a ssists in tracing the ultim ate im pact on different industries of changes in the com position of final demand. Such tables may be prepared in various form s - the figures may be expressed in physical o r in money te rm s, and the la tte r may be constructed on sev eral different bases. Tables expressed in money term s are obtained by breaking up the national production account in the social accounts into separate production accounts for a num ber of in d ustries, and showing not only final expenditure on the output of each industry but also the interm ediate purchases and sales between each pair of in d u stries. 2 The re su lts may be presented m ore conveniently in the form of a m atrix than a set of accounts. The capital account of each sector shows the se c to r s saving (i.e. income le ss cu rren t expenditure and cu rrent tra n sfe rs to other secto rs), its capital expenditure, and its borrowing and lending. For some purposes it may be n ecessary to concentrate on the borrowing and lending item s and to provide m ore detail of them. One method of doing this, used in flow-of-funds data, is to break each secto r's borrowing and lending up among the classes of financial a sse t involved in the transactions. 3 The one lis t of classes of financial a sse t is used in the 2 For an exposition of the relationship between the social accounts and the input-output matrix see Stone and Croft-Murray (1959: Ch. 2). The Commonwealth Statistician has published tables for Australia for 1958/59, and the introduction to the tables describes their relationship to the social accounting data published in the Australian National Accounts. 3 For Australian estimates see Reserve Bank of Australia (1965, 1966).

35 C lassification of T ransactions 27 analysis of all sectors; consequently, the sources and uses of funds shown under each class of asset balance when totalled over all secto rs, and an analysis of transactions in each class of financial asset throughout the economy can be derived from the statistics. Because of the magnitude of transactions of this type by financial e n terp rises, the financial enterp rises secto r can, with advantage, be broken into a num ber of p arts corresponding to p articu lar types of en terp rise (e.g. banks, life insurance com panies, pension funds, instalm ent credit com panies, etc.) in flow-of-funds statistic s, w hereas the magnitude of the secto r's production is not great enough to w arrant a break-up in that d etail for social accounting p u rposes in g en eral.

36 3 Estimation and Reliability of Social Accounting Statistics Concepts Before attem pting to p rep are estim ates of item s in the social accounts, decisions must be reached on the definition of production and on the general scheme of accounts (including the sector groupings, the types of accounts, and the transaction headings within the accounts). These m a tte rs have been considered in C hapters 1 and 2. P rin cip al steps in estim ation Given the definition of production and the general schem e of accounts, the principal steps involved in preparing estim ates are: (i) Examination of the accounting relations between the item s listed in the accounts, to see in what ways the choice of a p articular definition for one item determ ines the definition which m ust be used for other item s. F ailure to c arry out this step could lead to inconsistencies (and thus to e rro rs ) in the estim ates. After com pleting this step we could still be free to choose between several alternative sets of consistent concepts. (ii) Examination of the basic statistic s available, and assessm en t of the extent to which the concepts used in them m atch the concepts used in the social accounts. When the accounts a re confronted by the data, it may prove im practicable to use some of the alternative sets of consistent concepts which are still open to us at the end of step (i). (iii) The reliability of the data and the com pleteness of its coverage m ust also be considered. Adjustm ents m ust be made to the estim ates to c o rre ct known differences of coverage between the data and the accounts. In practice these three steps a re usually c a rrie d out together, but as they involve different aspects of the analysis they are here considered separately. This chapter is concerned p rim arily with steps (i) and (ii), while Chapter 4 is concerned with step (iii) and with the assessm ent of the reliability of the estim ates of the social accounts. A com prehensive survey of the e stim a te s could not be attem pted in the

37 E stim ation - Concepts 29 space available. The exam ples given in these two chapters are therefore illustrative only, although p articu lar attention is given to one group of item s in the A ustralian estim ates (the business income item s) in order to illu strate how the estim ates for one section of the accounts are affected by the th ree steps in the procedure. Accounting re la tio n s and consistency The social accounts are a consolidation of accounts of individual tran sacto rs. The nature of the restric tio n s im posed by the accounting relations can be seen by considering the derivation of item s in the accounts of the tra n sacto rs. Several of the exam ples below re fe r to the production account of a trading en terp rise in one of the following equivalent form s: Production Account of a Trading E nterp rise F irst Form Expenditure $ R eceip ts $ Opening stocks C urrent purchases Wages and salaries Indirect taxes G ro ss operating su rp lu s Total Sales Closing stocks Subsidies Total Second Form Expenditure $ R eceip ts $ C urrent purchases Wages and sala rie s Indirect taxes less subsidies G ro ss operating su rp lu s Total Sales Increase in value of stocks Total

38 30 National Income E stim ates Consider the relation between the gross operating surplus and the increase in value of stocks. The gross operating surplus is the balance of the production account. D ifferent methods of valuing stocks therefore lead to different gross operating surpluses. A sim ilar relation applies to consolidated figures for all trading e n terp rises. If the gross operating surplus of all trading e n terp rises, as shown in th eir consolidated production account, is estim ated from one of the available sources of data, certain opening and closing stocks figures w ill have gone into the derivation of that operating su rp lu s. Consistency req u ires that those opening and closing stocks figures also be used to estim ate the in crease in the value of stocks in the consolidated production account, otherw ise the account does not balance. A consequence of the point just noted is that unless the gross operating surplus has been calculated in a special way, the associated increase in the value of stocks contains elem ents of p rice change as well as quantity change. The data used as a basis for estim ates of operating surplus would not usually be derived from opening and closing stocks figures valued at identical p rices, and therefore the increase in the value of stocks would not be equal to the physical increase in stocks m ultiplied by the average price for the year. Except when the g ro ss operating surplus has been calculated in a special way, th erefo re, the in crease in value of stocks is different in nature from the other item s of final expenditure (personal consumption, governm ent expenditure on goods and serv ices, private expenditure on fixed capital a sse ts, exports, etc.) which can be thought of as physical flows of goods and serv ices m ultiplied by the average price. There is no objection, of course, to breaking the in crease in value of stocks into two components (the value of the physical in crease and the change in valuation) if the data perm it it. The official A ustralian estim ates attem pt such a break-up as a step in the calculation of constant price figures, but, as the com m ents in A ustralian National Accounts indicate, it is based on m eagre data. The gross operating surplus can be broken into a num ber of components, for example depreciation; net in terest, rent and royalty payments: and the net income of the trading en terp rise. The net income then becom es the balance of the production account. If an estim ate of net income of a group of enterp rises is made from one of the available sources of data (e.g. income tax statistic s), the e stim ate of the depreciation item for the production account m ust equal

39 E stim ation - Concepts 31 the depreciation allow ances deducted in the data, otherw ise the production account would not balance. The provisions of the income tax legislation regarding amounts deductible for depreciation may be amended from tim e to tim e. Amendments to those provisions affect the estim ates of both depreciation and net income if the la tte r is based on income tax statistic s (as it is for companies and for non-farm unincorporated businesses in the A ustralian official estim ates). The A ustralian income tax statistic s do not tabulate figures of depreciation every y ear, and for some y ears it is therefore necessary to make e stim a te s of the am ounts allow ed as deductions in the incom e tax data. The curren t p u rch ases', shown in the production account, do not include purchases of fixed capital a sse ts. Consequently the concept of gro ss operating surplus cannot be selected independently of the distinction drawn between c u rrent and capital expenditure, as the la tte r distinction determ ines the item s which are deducted to a rriv e at the gross operating surplus. In the A ustralian income tax legislation certain types of capital expenditure by farm ers (e.g. on dam s and on som e kinds of fencing) are deductible in the sam e way as operating expenses. If we wished to include the cost of dam s and fences in capital expenditure and if we w ere relying on income tax data as the b asis of the estim ate of gross operating surplus, we would be obliged to adjust the income figures upwards by the amount of the expenditure on dam s and fences, otherw ise the production account would not balance. C urrent expenses paid by a trading en terp rise are included in the expenditures shown in the production account, except for tra n sfe rs such as in terest paym ents which, in the schem e of accounts considered h ere, are regarded as a disposition of the operating surplus ra th e r than as a cost deducted to a rriv e at it. Some cu rren t expenditures, such as em ployers' contributions to superannuation funds on behalf of em ployees, do not appear to be covered by any of the headings listed in the production account above. It is necessary for these transactions to be accom m odated in the schem e of accounts, and further exam ination of the headings or th eir definition is required. Superannuation contributions are an allowable deduction by the em ployer for income tax purposes. This im plies that if the operating surplus is based on income tax data the contributions are deducted as a cost before arriv in g at the operating surplus. In the A ustralian official estim ates the contributions are treated as supplem ents to wages and sala rie s and they a re added to that item - a consequence of

40 32 N ational Income E stim ates this is that the contributions a re included in personal income even though they are paid by the em ployer d irect to the superannuation fund without passing through the em ployees' hands. The Commonwealth governm ent m akes fimds available to finance the distribution to persons of certain benefits in kind (e.g. free m ilk for school children, pharm aceutical benefits). Because the benefits are received by persons, the sum s involved a re included in personal income under the item 'Cash benefits from public auth o rities'. Consequently the purchases of goods (milk and pharm aceutical products) out of the funds are shown in the personal account ra th e r than the public authorities account, and a re included in personal consum ption expenditure. The farm income component of personal income could be defined as the produced incom e, including stocks and other a sse ts still held by m arketing authorities, or as the cash incom e, excluding assets still held by m arketing authorities. In the fo rm er case the increase in a sse ts held by m arketing authorities would be a disposition of personal income and should be included in personal savings. In the la tte r case the in crease in a ssets held by m arketing authorities would not be a disposition of personal incom e, and it should not be included in personal savings but in institutional savings. The form er treatm ent is followed in A u stralian N ational A ccounts. Some headings in the accounts are estim ated as balancing item s ra th e r than directly from other data. The only way of defining what is included in such balancing item s is to have re s o rt to accounting relations. For exam ple, personal saving is a balancing item in the personal curren t account in the A ustralian official estim ates, and it follows from the stru ctu re of the account that personal saving consists of all personal income which is not used to pay taxes, consumption, in terest, or overseas rem ittan ces. P ersonal income includes em ployers contributions to superannuation funds; the disposal of the contributions is not in taxes, consum ption, in te re st or overseas rem ittances, and it is th erefo re included in personal saving even though the money has not actually passed through the hands of the em ployees concerned. In some applications of social accounting data we may be interested in using (and estim ating) only the main aggregates from the sum m ary national production account (e.g. national incom e, gross national product). D etailed co n sideration of s e c to r accounts may seem

41 E stim atio n - Concepts 33 superfluous for that purpose. However, proper definition and estim ation of the main aggregates, as well as the sector accounts, requires that sev eral of the points noted above receive attention. C onsideration of the accounting relations helps in the system atic exam ination of the relationships between concepts, and the estim ation of the broad aggregates as well as the detailed accounts is improved by the clarification which this can bring. For exam ple, to bring out th eir significance for estim ation of the main aggregates in the sum m ary table, a few of the points noted in the preceding paragraphs could be restated as follows: (i) If we have estim ated g ross national product at m arket p ric e sl by totalling incom es, it is n ecessary to be aw are of the nature of the stocks component which is built into the estim ate via the business income figures - unless the la tte r have been calculated in a special way, the stocks component of the estim ate of production is not the physical in crease in stocks valued at curren t m arket p rices. (ii) If income tax data are used as the basis of som e of the business income estim ates included in the gross national product, we m ust ensure that superannuation contributions by em ployers are included som ew here in the list of income item s if the gross national product is to equal the value added by producers (i.e. sales plus increase in stocks le ss interm ediate purchases). (iii) Depending on how the final expenditure by public authorities is defined, it may be n ecessary to include in personal consumption some imputed purchases financed by public authorities (e.g. free m ilk for school children, pharm aceutical benefits) if the estim ates of gross n ational expenditure a re to be com plete. Basic statistical data - concepts The item s listed in the social accounts include a wide range of tran sactio n s concerned with production and with the distribution and expenditure of the proceeds of production. The estim ates aim to provide com prehensive totals for transactions of each type. Because of the extensive coverage of the accounts it is necessary to use a large p a rt of the available statistical data to prep are the estim ates. These data consist of a num ber of collections and other sources (e.g. * One of the im p o rtan t aggregates in the so cial accounts is com m only referred to as 'G.N. P. a t m ark et p rices', even though it includes govern m en t output v alu ed at cost. A ctu al m a rk e t prices are used only for the o utput of enterprises, but it has b ecom e co n v en tio n al usage to describe G.N. P. a t facto r cost plus in d ire c t taxes (n et of subsidies) as G.N.P, a t m ark et prices.

42 34 N ational Incom e E stim ates governm ent accounts) w hich a re p re p a re d m ore o r le ss independently of one another and independently of the social accounts. It can th e re fo re be expected that they w ill not p re c ise ly m atch the concepts in the so cial accounts, and w ill not be co n siste n t with one another but w ill overlap in som e places and have incom plete co verage in o th e rs. When p re p a rin g e stim a te s of the so cial accounts, a carefu l co m parison m ust be m ade of the concepts in the data and the concepts re q u ire d for the accounts, and the two m u st be brought together by m eans of adjustm ents to the data. The basic s ta tis tic a l data fall into two m ain c a te g o rie s. The firs t category c o n sists of co llections in itiated solely for sta tistic a l p u rp o ses (e.g. the cen su s, p rim a ry and secondary production s ta tis tic s, re ta il sale s s ta tis tic s). The second c o n sists of data derived from official re tu rn s supplied for a d m in istrativ e ra th e r than sta tis tic a l p u rp o ses (e.g. incom e tax and p ayro ll tax d ata, o v e rseas tra d e fig u res). The la tte r category often has special a sp e c ts, and perh ap s sudden changes of concept, which a ris e from the legislatio n governing them 2and from the step s which the re tu rn s p ass through before being totalled for sta tis tic a l purposes. ^ in p a rtic u la r, th e re a re special a sp e c ts of th is type affecting the incom e tax d ata, which play an im p o rtan t ro le in the p re p a ra tio n of e stim a te s for the A u stralian so cial accounts. B efore exam ining the incom e tax data in d e ta il, let us co n sid er in g en eral te rm s the estim atio n of the g ro ss operating su rp lu s from any of the available so u rc e s of data (e.g. production d ata, incom e tax data). The g ro ss o perating su rp lu s is the balance of the production account and, a s a p e ru sal of that account re v e a ls, is equal to the total,of c e rta in specified c la s s e s of re c e ip ts, nam ely: (i) all s a le s of the p ro d u c e r's output, and (ii) the in c re a s e in the value of h is sto ck s, le ss the to tal of c e rta in specified c la s s e s of expenditure, nam ely: (i) in term ed iate p u rc h a se s, w hich do not include p u rch ases of fixed a s s e ts, (ii) w ages and s a la rie s (including supplem ents), and (iii) in d ire c t taxes le s s su b sid ies. To p re p a re an estim a te of the g ro ss operating su rp lu s based on a 2 See, for exam p le, the co m m en t1' on d ep reciatio n in the final paragraph of this ch ap te r. 3 See, for ex am p le, the co m m e n t on the relatio n sh ip betw een incom e ta x statistics for partnerships and trusts and those for individuals in footnote 7 of this ch a p te r. See also the co m m e n t on th e cla ssific a tio n of tax ab le incom e of individuals by type of incom e on p. 40.

43 E stim atio n - Concepts 35 given type of data it is n e c e ssa ry to check the lis t of re c e ip ts and expen d itu res which have gone into the data again st the lis t set out in the production account. If the two lis ts do not a g ree p re c ise ly, som e adju stm en t of the data is re q u ire d to obtain an e stim a te of the re q u ire d concept. If one w ished to use facto ry cen su s data as the b a sis of an estim ate of g ro ss o perating su rp lu s of m anufacturing in d u stry, one would begin with the fig u res of 'value of production' and deduct factory w ages and s a la rie s from them. Value of production, as defined in the A u stralian facto ry c e n su s, is equal to the value of output le ss the co st of re p a irs and specified c la s s e s of goods used in production (the specified c la sse s a re pow er, fuel and light; w ater; oil; co n tain ers; and m a te ria ls used, tre a te d, operated upon, o r w orked up). C om parison of th ese re c e ip ts and ex penditures with those liste d in the production account indicates som e of the m ore obvious ad ju stm en ts re q u ire d to convert value of production le ss w ages and s a la rie s to g ro ss operating su rp lu s; for exam ple: (i) C osts of se rv ic e s (e.g. in su ran ce, ad v ertising) a re not deducted. An adjustm ent would be re q u ire d to avoid duplicating the value added in the se rv ic e in d u strie s. (ii) In d irect ta x es le s s su b sid ies a re not am ong the lis t of item s deducted from the value of output to a rriv e at value of production. Some of them have alread y been deducted before a rriv in g at the value of output, as that concept excludes excise and sale s ta x e s, and includes bounties and su b sid ies. A part from those specific exceptions, the value of production includes in d ire c t taxes w h ereas the g ro ss o p eratin g su rp lu s does not. (iii) C osts of selling and d istrib u tio n a re deducted from the selling value to d eriv e the value of output, and so also a re ch arg es for tra n s p o rt and in stallatio n. H ow ever, the w ages and s a la rie s and the m a te ria ls used also exclude m ost am ounts re la tin g to those a c tiv itie s, and consequently no m ajo r adjustm ent to the value of production le ss w ages and s a la rie s is needed on account of th e se item s. It could be noted in p assin g, how ever, th at if the wage and sa la ry fig u re,a s w ell as the g ro ss operating su rp lu s, for the production account w ere being d eriv ed from the factory c e n su s, and if such activ itie s as sellin g, d istrib u tio n, and in sta lla tio n, c a rrie d out by m a n u fa c tu re rs, w ere to be c la ssifie d as a c tiv itie s of m anufacturing in d u stry (as they a re in A u stra lia n N ational A ccounts), an upw ard adjustm ent to the factory

44 36 National Income E stim ates census wage and salary figure would be required in order to give it the required coverage. (iv) The factory census wage and salary figure does not include em ployers' contributions to superannuation schem es on behalf of em ployees. The value of production le ss wages and sala rie s therefore includes contributions on behalf of factory em ployees, w hereas the g ro ss operating surplus does not. Although the factory census is not used as the basis of the official estim ates of gross operating surplus, its relationship to the official estim ates becom es im portant if one w ishes to link the two sets of data, as one does when preparing input-output tables. Income tax statistic s are the m ajor source of data for the estim ates of g ross operating surplus of com panies and of non-farm unincorporated businesses. E stim ates are first made of net income (i.e. gross operating surplus less depreciation and net in terest, rent and royalties paid) and these are based entirely on income tax statistics in the case of com panies, and principally on them in the case of non-farm unincorporated businesses. 4 D epreciation and net in terest, rent and royalty payments are estim ated partly from other sources (principally from other sources in the case of net in terest, e tc.) and are added to the net income to obtain the g ross operating surplus. In the following paragraphs the matching of the income tax data to the concept of net income is considered. The basic income figures which a re obtained from income tax retu rn s are 'taxable incom es', as defined by the legislation which applies to the y ear under consideration. Taxable income is not a satisfactory concept for economic analysis in general or social accounting in p articu lar. The following a re some of the principal reasons: (i) Taxable income excludes certain types of income which are exempt from tax. The entire income of certain types of organisation is exempt (e.g. superannuation funds; state and tru stee savings banks; m unicipal corporations; religious, charitable and certain other nonprofit organisations). Income derived from some classes of mining and dividends paid out of such income are also exempt. The Act also contains a large num ber of other exem ptions, many of which relate to p articu lar kinds of personal incom e. As there is a tim e lag of two years in preparing the incom e tax statistics, this co m m en t does not apply to figures for the last two c o m p leted years.

45 E stim ation - Concepts 37 (ii) C ertain item s deducted from actual income to obtain taxable income a re not operating expenses during the y ear in question, but are m erely allowable deductions of a concessional nature. Many of these are of a personal nature (e.g. deductions for dependants, life insurance, m edical and dental expenses). O thers are business deductions (e.g. expenditure by fa rm ers on dam s and certain types of fencing mentioned on page 31; investm ent allowances; lo sses made during the previous seven y e a rs). In periods when there are fluctuations in incom e, the deduction of previous y e a rs' lo sses can lead to m arked differences between movements in taxable and actual income. 5 (iii) If a business m akes a loss during the y ear, its taxable income is nil, not a negative amount. This also can cause substantial differences between the m ovem ents in taxable and actual income. 5 (iv) T here is a good deal of double counting of company income, 5 The n o rm al relatio n sh ip betw een a c tu a l incom e and tax ab le incom e is disturbed in years when losses are incurred and in years "when th ey are w ritten off. The disturbed relatio n sh ip m ay re m a in for som e years after the a c tu a l incom e shows a return from losses to profits, as the ex am p le represented in the follow ing graph shows. In this ex am p le it is assum ed th a t the co m p an y receiv es no dividends or ex em p t incom e and has no deductions of a concessional nature apart from previous losses; consequently a c tu a l incom e and taxable incom e should n o rm ally be eq u al. INCOME 7 YEAR In year 3 losses are incurred; som e of th em are w ritten off in y ear 4 and the b alan ce in year 5. A lthough profits are earn ed ag ain in y ear 4, the norm al relatio n sh ip betw een a ctu a l in co m e and taxab le incom e does n o t return u n til y ear 6, and the y e a r-to -year m ovem ents are n o t back in step u n til y ear 7 (i. e. u n til two years after the last of the losses are w ritte n off).

46 38 N ational Income E stim ates because dividends received by com panies (except for a few classes of exempt dividends) a re included in th eir taxable incom es.. The one original source of income may be tra n sfe rre d from company to company as dividends, and may thus be included in the taxable incom es of two or m ore com panies. R esident com panies receive a rebate on dividends included in th eir taxable incom e, and as a resu lt of this they do not actually pay tax on the dividends (apart from sm all amounts arisin g from differences between the rate of rebate and the rate used for the initial assessm ent). N evertheless, the taxable income includes the dividends; tax is firs t a ssessed on that figure, and the rebate is deducted at a la te r stage in the calculation. It is clear from this that the statistical im plications of rebates are quite different from those of exem ptions, how ever s im ila r th e ir effects on tax paym ents. (v) Amounts deductible for depreciation differ from ordinary usage in many resp ects (e.g. m ost buildings are not depreciable). (vi) The provisions of the legislation a re amended from tim e to tim e.6 In p articular, the list of exemptions and allowable deductions has been subject to change. In the income tax tabulations relating to com panies, the item s m entioned in sections (i) to (iv) of the preceding paragraph a re tabulated each year. Net income (i.e. gross operating surplus less depreciation less net in terest, e tc.) can be derived from the following calculation: plus plus less less taxable incom e exempt income (other than dividends) concessional deductions lo sses incurred dividends included in assessab le income equals net incom e To a rriv e at the net income of trading com panies it is necessary to deduct amounts classified as financial en terp rise income in the so cial accounts. To obtain income tax data relating to unincorporated businesses 5 For a list of the p rin cip al changes in the legislatio n w hich affect the d ep reciatio n estim ates, see the definitions and descriptions of the re le v a n t tab les in A ustralian N ational A ccounts (e.g. p. 97 of A ustralian N atio n al A ccounts - N ational Incom e and Expenditure to ).

47 E stim atio n - Concepts 39 one m ust re fe r to the tabulations of assessm en ts of individuals'. 7 They contain sta tistic s of actual incom e as well as taxable income. Actual income is defined as 'g ro ss income (including exempt income) other than social service benefits and w ar and service pensions, less expenses incurred in gaining that incom e'. It is also equivalent, in the case of taxable assessm en ts, to taxable income plus exempt income plus concessional deductions; in the case of non-taxable assessm en ts the definition would appear to be equivalent to the sum of those three item s le ss losses incurred, but the annual income tax tabulations do not cover non-taxable a ss e ss m e n ts of individuals. Three main components of the incomes of individuals are distinguished: wages and sala rie s, other income from personal exertion, and property income. The other personal exertion component of actual income (i. e. gross income from personal exertion other than wages and sa la rie s, less expenses incurred in gaining that income) m atches the concept of net income of unincorporated businesses. The detail contained in the tabulations of individuals is less com plete than that contained in the tabulations of com panies, and it is n ecessary to estim ate some of the item s involved in the calculation of unincorporated business income. In p articular, estim ation is 7 T he relatio n sh ip of the 'partnerships and trusts' tab u latio n s to the 'in d iv id u als' tab u latio n s calls for som e co m m en t. Each partnership and trust subm its a ta x a tio n return, but does not itself pay tax; the assessed incom e is divided am ong the partners or b eneficiaries and in clu d ed in th e ir personal returns. This ensures th a t statistics of to ta l tax able incom e d erived from the personal returns include the incom e re ceiv ed from partnerships or trusts. H ow ever, the sam e conclusion does n ot necessarily apply to statistics of p a rtic u la r com ponents of tax ab le incom e or to other item s such as stocks or d epreciatio n, e.g. dividends receiv ed by a partnership or trust and divided am ong the partners or b eneficiaries are not en tered under the heading 'd ividends' on the personal returns - th ey are en tered under the headings 'share in partnership in co m e' or 'in co m e as b eneficiary under w ill, se ttle m e n t, deed of g ift or in stru m en t of tru s t'. S pecial arrangem ents w ould therefore be necessary if the statistics of item s other than to ta l tax ab le incom e were to include am ounts en tered in the first instance on partnership or trust returns; such arrangem ents w ould involve ascertaining from the partnership or trust return the am o u n t of, say, dividends transferred to the personal return of a partner or b en eficiary, and adding it to the figure en tered under the heading dividends' on his personal return. For assessm ent purposes such arrangem ents have to be m ade for item s of incom e su b ject to a d iffe re n tia l rate of tax (though there have been none since 30 June 1954) or subject to rebates ( e.g. in te re st on C om m onw ealth g overnm ent secu rities). For sta tistic a l purposes it is also the p ractice to m ake such arrangem ents for the property incom e and the other personal ex ertio n incom e com ponents of tax ab le incom e, and for to ta l a c tu a l incom e and the item s required to derive it (ex em p t incom e and business deductions of a concessional nature such as expenditure on dam s and fences d ed u ctib le under Sections 75 and 76). It is n o t the p ractice to m ake such arrangem ents for statistics of other item s of incom e (although they have been m ade on som e, b ut not all, occasions for dividends), or for item s such as d e p re c ia tio n or stocks. Figures for these item s can be taken from the partnership and trust tab u latio n s and added to figures from the individuals tab u latio n s to o b tain the to tals.

48 40 National Income E stim ates necessary at three main steps: (i) The annual tabulations do not cover non-taxable assessm en ts of individuals, although some details have been tabulated for y ears adjacent to Commonwealth census y ears. This group includes persons who have made business lo sses during the year and persons who have w ritten off sufficient previous lo sses during the y ear to bring th eir taxable income below the exemption lim it. It also includes other persons receiving significant amounts of actual income but who do not incur tax either because th eir actual income itself is below the exemption lim it or because, though th eir actual income is higher, th e ir exem pt incom e and concessional deductions fo r dependants, life insurance, etc., bring th eir taxable income below the exemption lim it. To a rriv e at a business income figure for this group the num ber of non-taxable assessm en ts is estim ated and is m ultiplied by an assum ed average incom e. (ii) The figure we wish to estim ate is the other personal exertion component of actual income. The statistic s of actual income are not classified by type of incom e, but taxable income is broken into three components (wages and sala rie s, other personal exertion, property) every third year, and in the other two y ears out of every three it is broken into two components (wages and sa la rie s, other). The classification of taxable income is p repared by offsetting concessional deductions against each type of actual income in a definite o rder. The deductions are offset first against wage and salary incom e, if there are still some deductions left they are then offset against other personal exertion income, and if there are still some left after that they are offset against property income. The re tu rn s are then classified into seven 'grades of occupation, partly on the b asis of the types of taxable income received. Given this data, if we s ta rt with the classification of taxable income for the grades of occupation which receive some 'other personal exertion incom e, it is then n ecessary to re tra c e approxim ately the steps which went into deriving the components of taxable incom e, and thus to a rriv e at an estim ate of the other personal exertion component of actual income (in the y ears when all three components of taxable income are tabulated). (iii) In the two y ears out of every three when taxable income is broken into only two com ponents, the retracin g of steps noted in (ii) above leads to an estim ate, not of the other personal exertion component of actual income, but of that component combined with the property

49 E stim ation - C oncepts 41 incom e receiv ed by the sam e group of ta x p ay ers. It is th e re fo re n e c e ssa ry to elim inate p ro p e rty incom e from th at fig u re. In the so cial accounts the e stim a te s of p ro p erty incom e receiv ed by p erso n s a re not based on the incom e tax tabulations fo r individuals, but a re d eriv ed from a num ber of o th er s ta tis tic a l so u rc e s, m o s t of w hich a re available annually. The annual e stim a te s reco n cile with the incom e tax data well enough to s e rv e as a b a sis for e stim a te s of m ovem ents in the p ro p erty incom e to be elim inated. The farm incom e fig u res used in A u stralian N ational A ccounts a re based on p rim a ry production s ta tis tic s ra th e r than on incom e tax data. T his autom atically m akes som e allow ance for u n d erstatem e n t and noncoverage which would have to be independently estim ated if incom e tax data w ere used. B asing the estim a te on that so u rc e, how ever, m akes it n e c e ssary to elim in ate the farm incom e com ponent of the incom e tax data from the calcu latio n of u n in corporated b u sin ess incom e d escrib e d above, leaving fig u res for 'o th e r1 unincorporated b u sin e sses. Although the incom e tax s ta tis tic s a re c la ssifie d by in d u stry, a ta x p a y e r's e n tire incom e is c la ssifie d to the industry of h is m ain activity. Many ta x p ay ers c la ssifie d as p rim a ry p ro d u cers e a rn p a rt of th e ir incom e from o th er so u rc e s, and many ta x p ay ers c la ssifie d to other in d u strie s earn p a rt of th e ir incom e from farm in g. The farm incom e figure b ased on p rim a ry production s ta tis tic s, how ever, re la te s solely to incom e from farm ing. Some adjustm ent has to be m ade to the incom e tax industry cla ssific a tio n to deal with m inor so u rces of incom e, in o rd e r to en su re that the farm incom e elim inated from the tax data has the ap p ro p ria te industry coverage. Some points of concept can be m o re read ily analysed by exam ining the rela tio n sh ip s betw een the accounts of different p ro d u cers and the consolidation of those accounts. C onsider, for exam ple, the tra n sactio n s of general in su ran ce com panies (excluding life in su ran ce com panies). T h eir p ro fit, fo r incom e tax p u rp o ses, is equal to p rem iu m s le ss claim s and other deductible ex p en ses, plus investm ent incom e. P rem iu m s le ss c la im s is, in effect, the charge to the r e s t of the econom y for th e ir output. In so fa r as the se rv ic e s a re sold to o th e r b u sin e sses, consistency re q u ire s that the sam e ch arg e be deducted as in term ed iate p u rc h a ses by other b u sin esses to a rriv e at th e ir operating su rp lu s. Upon consolidating the production accounts of the in su ran ce com panies w ith those of other p ro d u c e rs, the p rem iu m s le ss claim s tra n s fe rre d betw een them would then be o ffset, like o th er in term ed iate p u rc h a se s and sale s. H ow ever, the tre a tm e n t

50 42 N ational Income E stim ates of those item s in the accounts of other producers is not uniform, as th,e provisions of the income tax legislation vary with the type of insurance. P rem ium s are always deductible, but claim s are included in taxable income in some cases (loss of profits insurance, loss of stock insurance, and a few other types) and excluded in others (e.g. claim s for the loss of fixed assets). As suggested above, prem ium s less claim s should be treated as interm ediate expenditure of all businesses. This im plies that prem ium s should be deducted to a rriv e at the gross operating surplus of the other producers, and claim s should be added. The income tax legislation ensures that this is done in the case of loss of profits and loss of stock insurance, as the claim s are included in net income; the treatm en t of insurance com panies and other producers is th erefo re consistent in this case. In the case of insurance of fixed a sse ts, the claim s paid by the insurance com panies still have to be accom m odated as receipts som ewhere in the accounts of the other producers. A special adjustm ent of the tax data is therefore needed to obtain a consistent treatm ent, but as these claim s m erely reim b u rse an amount by which fixed capital a sse ts are depleted they are added to depreciation (and thus to gross income) ra th e r than to net incom e. As the preceding paragraphs suggest, a significant p art of the A ustralian social accounts has an income tax basis. This does not mean, of course, that the income figures are taxable incom es, or that there is no attem pt to allow for understatem ent or non-coverage in the income tax data. It means firstly, that som e of the im portant figures are derived from income tax statistic s (supplemented, in some instances, by other data), and secondly, that other item s related to them are defined in such a way that they will be consistent with income tax concepts. For exam ple, company income and the income of unincorporated businesses other than farm s are both derived from taxation statistic s. Farm incom e, though not estim ated from tax data, is in certain resp ects defined sim ilarly to the other unincorporated business income figures (e. g. farm depreciation is estim ated on a basis consistent with tax legislation). The depreciation and stocks figures also reflect th eir tax origin - they a re, for the m ost p art, those used in deriving taxable incom e, and they reflect any peculiarities of the legislation and any am endm ents which are made to it. F or exam ple, buildings a re not depreciable unless they are farm buildings, or a re an integral p art of m anufacturing plant, o r a re

51 E stim atio n - Concepts 43 buildings in which scientific re searc h related to the taxpayer's business is c a rrie d on. The depreciation estim ates reflect th is,^ and also any changes in the ra te s allowed by variations in the legislation. These peculiarities of the depreciation estim ates are also reflected in the net, but not the g ro ss, national product figures. In the case of stocks figures, adherence to the tax basis involves including certain types of work in pro g ress on contracts which might not be thought of as stocks in less specialised investigations. Work in p ro g ress on buildings is included in stocks, except in so far as it is already covered by p ro g ress paym ents and included in the capital expenditure of the purchaser. A sim ilar point apparently applies to less tangible item s, such as expenditure by an excavating contractor. However, the stocks figures specifically depart from the income tax basis by excluding land, sh ares, debentures, and other secu rities held by businesses which trad e in those a sse ts, because an increase in the holding of them does not rep resen t a final use of production or im ports. 8 However, the depreciation estimates include property insurance claim s on buildings, as shown in the preceding paragraph.

52 4 Estimation and Reliability of Social Accounting Statistics Coverage and Accuracy In Chapter 3 the m atching of concepts with one another was considered. In this chapter questions of accuracy a re examined. They are considered under three general headings: the coverage of the basic data, the reliability of the basic data, and the reliability of the estim ates of the social accounts derived from the basic data. B asic s ta tistic a l data - coverage Social accounting data are intended to provide com prehensive estim ates of the c lasses of transactions listed in the accounts, showing th eir total value over the entire economy. The principal collections of basic statistical data which are available annually, or at m ore frequent intervals, often do not have the com prehensive coverage required. When they are used as the b asis of estim ates of social accounting item s it is necessary to find ways to estim ate the m issing tran sactions. Even in the case of data collected solely for statistical purposes, there may be incom plete coverage because it has frequently been found necessary to exclude sm all item s from the collections. For exam ple, for the purpose of the A ustralian factory census a factory is defined as an establishm ent in which four or m ore persons are employed or where power (other than manual) is used in any m anufacturing pro cess. The figures therefore exclude p articulars of some sm all m anufacturing establishm ents - the non-statistical fa c to rie s. The A ustralian prim ary production statistic s a re derived from a census of ru ra l holdings, for which purpose a 'ru ra l holding' is defined as a piece of land of one acre or m ore in extent, used for the production of agricultural products or for the raisin g of livestock and the production of livestock products. Again, some production on sm aller holdings is not covered. In the case of data collected p rim arily for adm inistrative purposes, but subsequently classified and totalled to provide statistic s, the problem s of coverage can be g re a ter. This is because the m issing 44

53 E stim ation - C overage and A ccuracy 45 transactions can be substantial, because the non-cove rage can arise in a variety of ways, and because its magnitude may change suddenly as a re su lt of am endm ents to the legislation or the adm inistrative arrangem ents governing the collection of the data. The following exam ples illu s tra te th ese points: (i) F igures derived from taxation data (whether income tax or other form s of tax, such as payroll tax) may exclude cases which fall below the exemption lim it. F u rth erm o re, the exemption lim its can be amended and thus the coverage of the data can change suddenly. For exam ple, the exemption lim it for payroll tax has been raised on a few occasions, and w hereas the exemption applied to payrolls of up to $2, 080 p er y ear when the tax was introduced, it now (1967) extends to payrolls of up to $20,800 p er year. (ii) The legislation also grants total exemptions to certain types of organisation, irresp ectiv e of th eir size. The Commonwealth governm ent is exempt from payroll tax, and so also are religious and charitable organisations and m ost hospitals. Section 23 of the Income Tax and Social Services Contribution A ssessm ent Act exem pts many types of o rg an isatio n from income tax. (iii) The collection and assessm en t of income tax retu rn s take tim e, and when additional tim e spent on appeals and litigation is added it may be sev eral y ears before the outcome is known for all taxpayers. As m ost of the assessm en ts a re com pleted in a sh o rter tim e, it is not worthwhile holding up the tabulations until all assessm en ts are available. In the case of com panies the p ractice is to tabulate all assessm en ts com pleted up to eighteen months after the end of the y ear in which the incom es w ere earned. 1 The tabulations are therefore incom plete, and 1 Some confusion is occasionally caused by the term inology regarding the year to which the incom e tax data relate. Three different years are distinguished: (i) The 'incom e year1 is the year in w hich the incom e under consideration was earned. (ii) The 'assessm ent year' is the year in which assessment of the returns com m ences, and is the year follow ing the incom e year. (iii) The 'financial year' or 'year of tax' is the year during w hich tax on the incom e is payable. Com panies pay tax on a given year's incom e during the follow ing year, but because of the p ay-as-you-earn and provisional tax arrangements tax is payable by individuals during the year in w hich the incom e is earned. The 'year of tax' therefore coincides with the 'assessment year' for com panies and w ith the 'incom e year' for individuals. In the case of provisional tax, however, the position is com plicated by the fact that the tax is paid in two steps; tax on an estim ated incom e is paid during the incom e year, and tax on the balance of actual incom e is paid (or refunded) during the follow ing year. The preceding year's incom e is usually taken as the estim ated incom e for the purpose of the first paym ent, excep t where the self-assessm ent provisions of the legislation are applied. C onsequently, although provisional tax is payable in the year in w hich the incom e is earned, the estim ation arrangements have the e ffect of relating actual payments more c lo sely to the preceding year's in com e.

54 46 National Income E stim ates th eir coverage is not the sam e from y ear to year. However, the total income involved in the m issing assessm en ts is thought to be relatively sm all and not to change substantially from year to y ear, unless the assessm ent of som e larg e com panies is subject to extended litigation. In the case of individuals the p ractice is to p rep are two sets of tabulations, the firs t set (known as the 'm ain' tabulations) covering assessm en ts up to fifteen months after the end of the income y ear, and the second set (known as the supplem entary tabulations) covering the additional assessm en ts up to four y ears after the end of the income year. The supplem entary tabulations include significant amounts of incom e, and the proportion of the total tabulations covered by them has often varied considerably from y ear to year (with equivalent variations in the proportion covered by the main tabulations). (iv) Evasion of tax by failing to subm it retu rn s may lead to incom plete coverage, both of incom es and of num bers of taxpayers, while evasion by understating incom es or by overstating expenses may lead to incom plete coverage of incom es. 2 The C om m issioner of Taxation publishes in his Annual R eport p articu lars of cases detected. The figures published th ere are not easy to use in conjunction with the other income tax tabulations, because if the evasion is detected and dealt with before the ordinary tabulations for the y ear concerned are closed off the income is included in them, w hereas if it is detected after they are closed off it is not covered by them. (v) Income tax data relating to non-taxable assessm en ts of individuals are not prepared each y e ar, although som e details have been tabulated for y ears adjacent to Commonwealth census y e ars. As som e persons whose incom es are not taxable a re not required to subm it re tu rn s, a tabulation of the non-taxable assessm en ts would not cover all nontaxable incom es. N evertheless, the lack of such a tabulation leaves a gap in the data. F or social accounting purposes, and also for many other purposes such as preparation of statistic s of employment and of wages and sa la rie s, it is n ecessary to fill in som e of the gaps in the basic collections. If we can obtain m ore com plete data for a p articular item during one y ear, it may be possible to estim ate the m issing components in the other y ears by starting from the benchm ark figure 2 One form of evasion of incom e tax w ould not affect the estim ates of incom e item s in the social accounts, nam ely, the overstatem ent of concessional deductions such as m ed ical expenses, e tc. Although this would lead to the understatem ent of taxable incom e, it would not introduce any error into 'actual incom e', and the latter is the basis of the estim ates of incom es in the social accounts.

55 E stim ation - C overage and A ccuracy 47 and basing estim ates of the y e a r-to -y e a r movements on closely related serie s which are available each y ear. The m ost com prehensive benchm ark for a num ber of item s is the Commonwealth census. Some of the figures from the census which a re p articularly im portant in the estim ation of social accounting aggregates a re those relating to num bers of em ployees, num bers of em ployers and self-em ployed p ersons, and num bers and^rents of dwellings. In addition to the Commonwealth census th ere a re, of course, many other statistical sources which are available occasionally but not every y ear, and which may serve as benchm arks for estim ates. Exam ples are censuses of re ta il establishm ents, business surveys on special topics (e.g. in terest, ren t and royalties) which have been conducted occasionally, and item s contained in the income tax statistics but not tabulated every y ear. If we have annual data covering part of an item to be estim ated, and benchm ark data covering the whole of it (or at least a g re a ter p a rt of it than is covered by the annual data) at some p articu lar tim e, a com parison of the two sets of data can yield an estim ate of the m issing component (or p art of it). In general, however, it is not sufficient m erely to com pare the totals from each set of data, without any adjustm ent, and take the difference. F or exam ple, by com paring the employment figures from the census with those from the payroll tax tabulations we would hope to obtain an estim ate of the num ber of em ployees outside the scope of the payroll tax data. However, a person who has a p art-tim e job with a second em ployer in addition to his principal job may be counted twice in the payroll tax data but only once in the census. The duplication would have to be rem oved before the difference between census and payroll tax figures could be accepted as an estim ate of the num ber of em ployees outside the scope of the payroll tax data. Such differences of concept m ust be identified and allowed for before any conclusions about coverage can be draw n. Other adjustm ents to the two sets of employment figures may also be required before they are com pared. F or exam ple, census employment includes crew s of foreign ships w hereas the payroll tax data do not, and in ord er to estim ate the num ber of A ustralian em ployees outside the scope of the payroll tax data allowance m ust be made for this. Another illustration is provided by the treatm ent of persons tem porarily absent from work - the census emploj^ment figures include th em, w hereas some of them a re excluded from the payroll

56 48 National Income E stim ates tax figures. 3 If the data a re used as the basis of wage and salary estim ates, the com parison of the two sources needs to yield an estim ate of the num ber of em ployees receiving earnings which a re not covered by the payroll tax d ata. Many of the tem porary absentees who are not covered by the payroll tax employment figures may not receive any earnings during th eir absence, and an adjustm ent to the com parison of the two sets of employment figures is required in order to avoid estim ating an income for them. It may not always be possible to obtain a benchm ark from a single statistical collection. F or exam ple, the census, the income tax statistics and the surveys of in te re st, rent, and royalties conducted by the Bureau of Census and S tatistics all contain some inform ation on ren ts. It may be n ecessary to use the three of them in conjunction in order to obtain a benchm ark for certain types of ren tal paym ents. A fter a com parison of the annual and the benchm ark data has yielded an estim ate of the m issing component for one year or point of tim e, it may be desirable to break it up into a num ber of p arts to sim plify the estim ation of corresponding figures for other y e ars. For exam ple, the num ber of em ployees outside the scope of the payroll tax data could be broken into separate categories which could be covered by special collections (e.g. of Commonwealth governm ent employees) or which could be related to other basic statistical collections (e.g. total census employment in p rim ary production and m anufacturing, whether covered by the payroll tax data or not, might be related to the employment figures obtained from the prim ary production and the factory cen su ses). That would then leave a sm aller hard core of unrecorded employment to be estim ated by less d irect methods in the y ears between the cen su ses. Although th ere may not be any benchm ark data directly available for an item we wish to estim ate, it may be possible to derive it indirectly via a related serie s. Thus the payroll tax data contain wage and salary figures, while the Commonwealth census does not. But both sets of data contain employment figures, and an estim ate of the unrecorded wages (i. e. the amounts not recorded in the payroll tax data) can be based on the estim ate of unrecorded employment. When 3 The census em p lo y m en t figures include 'persons who are not out of a job b ut are only tem p o rarily absent from th e ir jobs through illness, a ccid en t, annual holidays, lo n g -serv ice leav e, etc.the e m p lo y m e n t figures shown on the m onthly p ayroll ta x returns are the 'n um ber of em ployees on payroll a t concluding date of last p erio d ' of the return, ascertain ed by count from pay sheets - it appears th a t th ey m ig h t include som e, b ut n o t all, of the persons tem p o rarily absent, depending on the precise arrangem ents w hich apply during tem porary absence of p articu lar em ployees.

57 E stim ation - C overage and A ccuracy 49 estim ates a re prep ared in this indirect m anner, it is n ecessary to allow for the possibility that a p articular set of data may not have the sam e coverage of all item s derived from it. For exam ple, the percentage of wages and sala rie s covered by payroll tax data would differ from the percentage of employment covered if the average wage of the unrecorded em ployees differed from that of re c o rd ed em ployees. Basic statistical data - reliability E rro rs may enter into statistical data in a num ber of ways. The m ost obvious way is through inaccurate answ ers to the questions asked. An answ er may be wrong because the question is not clear and is m isunderstood, because the inform ant does not know the c o rre ct answ er, o r because he does know it but p refers to e rr in p articu lar directions (e.g. it is natural to suspect som e understatem ent of incom es in the income tax data). Questions which ask the inform ant for a description ra th e r than a figure (e.g. questions about industry o r occupation ) may easily produce answ ers which are vague and are th erefo re difficult to classify accurately. In som e cases the definition of the figures requested from inform ants may only be a second-best approxim ation to the definition we would have p re fe rre d to use. This may be n ecessary because it is unlikely that the figures could be supplied in any other form, or because statistic s obtained as a by-product from retu rn s supplied for adm inistrative purposes may be affected by special arrangem ents for reporting figures by p articu lar inform ants. One resp ect in which we som etim es have to accept the second-best alternative is in the choice between figures on a cash or an accrual b asis. P a rticu la rs of governm ent receip ts and expenditure a re, for the m ost p a rt, recorded on a cash b asis. Wage and salary figures taken from payroll tax re tu rn s a re also on a cash basis; consequently if the la st day of the y ear happens to be a pay day for a very larg e proportion of em ployees the wage and salary figure will ris e in that year for a reason of no special economic consequence (e.g. in 1948/49 and 1959/60 there w ere tw enty-seven fortnightly pays by the Commonwealth governm ent instead of the usual tw enty-six). A sim ilar re su lt occurs in y ears when th ere is an extra payment of pensions or other cash benefits by the Commonwealth governm ent. Child endowment paym ents, if lodged directly in payees bank accounts, a re made at twelve-weekly intervals and may th erefo re ris e by a q u arter in y ears when five payments occur.

58 50 National Income E stim ates Another resp ect in which we a re som etim es obliged to accept the second-best definition concerns the choice of balance date on which the figures are based. F or exam ple, com panies may obtain perm ission to base th eir income tax retu rn s on accounting y e a rs ending on dates other than 30 June. If they balance, say, at the end of D ecem ber, th eir income tax retu rn for a p articu lar income y e ar will show th eir income for the year ended in the previous D ecem ber. The trend shown by the se rie s may not depart much from that of the d esired series in m ost y e a rs, but in y ears when incom es change rapidly the difference can be of som e consequence. Companies operating in A ustralia and balancing in D ecem ber might have shown b etter re su lts in th e ir 1960/61 re tu rn s (based on calendar year 1960) than they actually experienced between July 1960 and June 1961 because economic conditions w ere m ore depressed in the first half of 1961 than in the firs t half of They might then have reported poorer resu lts in th eir 1961/62 re tu rn s (based on calendar y ear 1961) than they actually experienced between July 1961 and June 1962 because economic conditions w ere b etter in the first half of 1962 than in the first half of P rocessing e rro rs can occur in a variety of w ays, and a num ber of m echanical checks to detect them a re a norm al p art of the processing routine. It is advisable for people using the data to apply some checks of th eir own and to be on th eir guard for changes of definition o r coverage. It is difficult to g eneralise about the nature of such checks. Some of them might m erely involve investigating the causes of substantial changes in the figures, so as to check the plausibility of the changes shown. One obvious check is the exam ination of m ovem ents in categories such as industry not stated ; substantial changes in such a classification certainly indicate that changes in other industries a re suspect. Statistics may be collected by m eans of sam ples ra th e r than com plete censuses. The coverage of the sam ple estim ates is, of course, the sam e as that of the fram ew ork or the population from which the sam ple was selected, but the figures derived are n ecessarily estim ates which contain sam ple e rro rs. By using la rg e r sam ples and m ore efficient sam ple designs we can reduce the e rro rs 4 As a result of a recent change of procedure, the Australian estimates of company income from 1960/61 onwards now incorporate an adjustment to put them on a June year basis. See Australian National Accounts - National Income and Expenditure, to , p. 13.

59 E stim ation - C overage and A ccuracy 51 but not elim inate them. An assessm en t of the probable size of the e rro rs (but not of th eir direction) can be based on sam pling theory. In som e c ase s, com parisons with other data may provide an assessm en t of both the size and direction of the e rro rs. Some idea of when they a re likely to be g reatest might also be obtained if we know the general sam pling procedure, for example after the com position of the sam ple is changed (perhaps as a re su lt of a periodical overhaul to keep it rep resen tativ e of a constantly changing 'population ) the size of the movement from the previous level of the figures may be m ore suspect than usual. R eliability of social accounting estim ates Social accounting estim ates are based on a larg e num ber of statistic a l s e rie s, some of which becom e available prom ptly and regularly, others m ore slowly and at le ss frequent intervals. The first estim ates of many of the item s in the accounts a re based on a lim ited amount of inform ation which is available prom ptly, and as additional data come to hand the estim ates a re rev ised to take them into account. Such revisions n ecessarily continue until the la st of the regular collections and occasional benchm arks which affect a p articular y e a r's estim ates has become available. If the concepts and the methods of calculation w ere held fixed the estim ates might then settle down to th eir final form. That would not im ply, of course, that the figures w ere perfectly accurate. F u rth er revisions could th erefo re come la te r still, as the re su lts of additional re searc h into concepts, data, and estim ating procedures w ere taken into account. The accuracy of the estim ates v aries from item to item, and because of this it is frequently suggested that an indication of the accuracy of each item should be given when official estim ates are published. However, it is difficult to find an objective b asis for assessin g th eir accuracy. T here is no objective basis for deriving num erical m easures to be used in the sam e way as standard e rro rs of sam ple estim ates. One could average the past revisions to various item s in the accounts, but although one could derive som e useful im pressions from such figures the following qualifications would have to be allowed for: (i) The averages would m erely indicate the amount of revision between the e arliest estim ates and the final e stim a te s. They would not indicate the reliability of the final figures which a re calculated

60 52 National Income E stim ates when the last of the benchm ark data has become available. The final figures a re still only estim ates, and the statistical discrepancy does not vanish when all of the benchm arks have been incorporated in the estim ates. (ii) The availability of new statistical collections, and research into existing data and methods of estim ation, may im prove the accuracy of the estim ates. To this extent an average of past revisions would exaggerate the inaccuracy of the la te st methods of estim ation. (iii) In some circum stances the methods of estim ating p articu lar item s can dim inish in accurac}' as a re su lt of relatively rapid in creases in components the estim ates of which are relatively weak. Such developments can be brought about by changes in economic stru ctu re, changes in business p ractice, and changes in legislation, as the following exam ples show: (a) the rapid increase of h ire-p u rch ase and general finance business in A ustralia in the post-w ar period has involved a larg e in crease in certain types of in terest payment which are m ore difficult to estim ate precisely from the available data than are the m ore slowly increasing components of in terest; (b) a growing p ractice of leasing business p rem ises or equipment from specialist property-ow ning o r leasing com panies rath er than owning it has led to a rapid increase in a component of the rent estim ates for which annual data are scarce; (c) legislation which in creases the exemption lim it for various types of taxation leads to an increase in the unrecorded component of item s estim ated from the corresponding taxation statistics - thus a ris e in the exemption lim it for payroll tax adds to the unrecorded component of the wage and salary estim ates. As a re su lt of changes of the types noted in (iii), the w eaker components of an item can becom e a la rg e r proportion of its total. Consequently an average of a num ber of past revisions could underestim ate the e rro rs in the latest estim ates. When allowance is made for these qualifications, how ever, some useful im pressions can be obtained from a study of past revisions. Official publications in many countries give a broad indication, based on such im pressions, of the reliab ility of prelim inary estim ates of social accounting item s, for example estim ates based on inadequate data may be enclosed in b rack ets, or a code of le tte rs (A, B, C, e tc.) may be used to indicate different degrees of reliability.

61 E stim ation - C overage and A ccuracy 53 In many applications of social accounting data the y e ar-to -y e ar changes are of m ore im portance than the levels of the item s. It is th erefo re im portant to inquire how the revisions to the y e ar-to -y e ar changes com pare with the revisions to the levels of the figures. If the rev isio n s to the levels of successive y e a rs' figures w ere independent, the average revision to the y e ar-to -y e ar changes would exceed the average revision to the levels. 5 However, the revisions to successive y e a rs' levels are not, in general, independent. Quite often the nature of the am endm ent im plies that the estim ates for a num ber of consecutive y ears have to be revised in the one direction, and in these circum stances the revisions to the y e a r-to -y e a r changes are le ss than the revisions to the levels. For exam ple, when some benchm ark data which a re not collected annually (e.g. the census) becom e available, estim ates for all subsequent y ears and also for all y ears between that and the preceding benchm ark may be revised in the one direction to bring them into line with the new data. Also, if some of the annual data become available only after a considerable tim e lag (e.g. income tax statistics) and if estim ates for succeeding y ears are based on lim ited inform ation about y e a r-to -y e a r changes since the la st available tabulations, the receip t of a later set of annual tabulations may lead to revisions to the year which they cover and also to revisions in the sam e direction to succeeding y e ars. Other kinds of revision may, of course, have different consequences. The relativ e im portance of the various kinds of revision v aries from item to item, depending on the data and procedures used to estim ate them. Some exam ples from the A ustralian e stim a te s a re given: (i) Net national product. For each y ear from 1943/44 to 1962/63 the range between the highest and lowest estim ate has been calculated to m easure the extent of the revisions to the levels of the figures. Over the tw enty-year period the ranges averaged 81 m illion. A sim ilar calculation applied to the y e a r-to -y e a r changes in the estim ates leads to an average range of revision to the changes of 66 m illion. The y e a r-to -y e a r changes a re somewhat m ore accurate than the levels. (ii) Sim ilar calculations derived from the estim ates of another three im portant item s yield the following resu lts: 5 If X i and X2 v ary in d ep en d ently w ith standard d ev iatio n S, th en (X 2 -X ]) varies w ith standard d ev iatio n s/t " S

62 54 National Income E stim ates Item Average range between highest and low est: level change m m P ersonal consumption * Wages, sa la rie s, and supplem ents Company incom e * * The averages for personal consumption and company income are based on figures up to 1959/60 only, as too few com parable figures are available for la te r y ears because of changes in concept. The re su lts for personal consumption, like those for net national product, indicate that the changes a re significantly m ore accurate than the levels. The difference is much m ore m arked in the case of wages, sa la rie s, and supplem ents. This might be expected from our knowledge of the nature of the estim ates, because a substantial p art of the payroll tax and other annual data used in the final estim ate is available when the figure for a p articular year is firs t published. Revisions following the receip t of fu rth er annual data a re therefore relatively m inor by com parison with revisions following the receipt of the m ore infrequent benchm ark data such as census figures, and, as stated above, these benchm arks lead to g reater revisions to the levels than to the y e a r-to -y e a r changes.. The estim ates of company income do not depend on any infrequent benchm arks, but a re virtually finalised when the annual company tax tabulations are com pleted. As there is a tim e lag of two y ears in the completion of the tabulations, there a re revisions to the first two estim ates of company incom e, but, as the figures above show, the average revisions to the levels are practically the sam e as the average revisions to the y e a r-to -y e a r changes. In his discussion of the accuracy of national income statistic s O skar M orgenstern (1963: 266-9) devotes som e space to a consideration of the accuracy of changes as distinct from levels, and he reaches some very gloomy conclusions. However, an exam ination of his argum ent shows that his method of m easuring the e rro rs in the changes greatly exaggerates the e rro rs which a re actually present in the data. In the

63 E stim ation - C overage and A ccuracy 55 following graph the different estim ates of U. S. national income for 1948 and 1949 are shown. 6 F igures reported in the one national income publication are joined by fine lines, so that the different estim ates of the change between 1948 and 1949 can be read from the g raph. $ billion The changes reported vary from -$7* 2 billion to -$4-7 billion, a range of $2-5 billion. M orgenstern's m easures imply that the actual change between these two y ears could be anything from -$10-0 billion (corresponding to a movement from A to D) to -$() 1 billion (corresponding to a movement from C to B), a range of $9-9 billion. The la tte r m easure only rep resen ts the range of uncertainty on the assum ption that revisions to successive y e a rs' estim ates are independent, and that one is free to believe any of the figures for 1948 in conjunction with any of those for However, such an assum ption is untenable, for the reasons given in the previous paragraph. If some additional data with im plications for both the 1948 and the 1949 estim ates come to hand, it is not legitim ate to derive a y e a r-to -y e a r change by com paring the revised figure for one year with the unrevised figure for the other y ear. Yet that is precisely what is done to a rriv e at M orgenstern's m easure of uncertainty, and it greatly exaggerates the e rro rs in the data. In the two y ears considered above, the exaggeration is fourfold. The following table sets out the com parison for the y ears considered by M orgenstern, and it shows that the 1948 to 1949 com parison is not exceptional. ^ The figures are taken from Table 23, Morgenstern (1963:249-50).

64 56 National Income E stim ates Years Range from lowest to highest annual change shown by: actual data M orgenstern's m easu re* $ billion $ billion 1947 to to to to * to to to to to to to to to * Range between low er and upper points graphed for each y ear in Fig. 9, M orgenstern, p The objection noted in the la st paragraph also applies to the conclusions which M orgenstern (1963:268-9) has drawn from his Fig. 10. In Fig. 10 the quarterly national income se rie s is graphed, and a band extending from 10 per cent below to 10 per cent above each figure is also graphed to indicate 'the zone of uncertainty surrounding the d a ta '. M orgenstern notes that a serie s which is very different from the published one could be graphed within that band of uncertainty, and concludes that the quarter-to- q u arter changes are not significant. However, the zone of uncertainty indicates the extent of possible e rro rs in each level of the s e rie s, and M orgenstern derives his conclusions about e rro rs in the q u a rte r-to -q u a rte r changes and about the alternative path the graph could take within the zone of uncertainty on the assum ption that the e rro rs in successive q u arters' estim ates are independent. Without that assum ption, one's ability to draw serie s with alternative paths within the zone of uncertainty would be somewhat re stric te d. It has been argued above that such an assum ption is untenable. In p a rtic u la r, it could be expected that a

65 E stim ation - C overage and A ccuracy 57 revision to a p articu lar y e a r s figure would commonly lead to revisions in the one direction to the quarterly figures which make up the annual to tal. An exam ination of past revisions to the estim ates can give an indication of the tim e needed for the estim ates to approach th eir final form, subject to the qualifications noted on pp and subject to the possibility of further revisions following re searc h into methods of estim ation. It has been noted that revisions to the y e ar-to -y e ar changes in net national product in the A ustralian estim ates for the twenty y e ars ending in 1962/63 extended over an average range of 66 m illion. If the first published estim ates for each year are excluded the average range is reduced to 36 m illion, and if the first two estim ates for each y ear a re excluded the average range is reduced to 26 m illion. The additional data which come to hand during the y ear after the first publication lead to a substantial im provem ent in accuracy. The extent of the im provem ent up to any stage v aries from item to item. For exam ple, the estim ates of compan}^ income are virtually finalised after two y e a rs, which is the tim e lag in preparation of the annual company tax sta tistic s, w hereas many other series cannot be finalised until a la te r stage, pending receip t of benchm ark data such as the census. There rem ains the question of the accuracy of the final estim ates, as calculated when the la st of the benchm arks affecting a particu lar year has been taken into account. It is difficult to find objective bases for such an assessm en t. Some details which are relevant may be ascertained; for example people who prep are the estim ates may derive from the w orksheets the proportion of each item which is based on annual data and the proportion based in interpolations between and projections of benchm arks. That is one of the facts relevant to an assessm en t of accuracy, but a com plete assessm en t would also have to consider the accuracy of the annual data for each item and the accuracy of the benchm arks and the methods of interpolation and projection. The early p art of this chapter has dealt with some aspects of these questions to a lim ited extent. Although broad judgments might be made, they would n ecessarily be somewhat subjective. For som e of the item s it may be possible to p rep are two or m ore estim ates from independent sources of data. In such cases the alternative estim ates provide a check on one another. For example both farm income and property income received by persons could be

66 58 National Income E stim a te s calculated from income tax data. The official estim ates of these two item s in the A ustralian accounts are actually derived from other sources, but m ovem ents in the figures can be com pared with those based on incom e tax data. C ertain item s in the accounts are derived as balancing item s, and could not be expected to be as accurate as the directly estim ated item s on which they depend. The final balancing item in the accounts, the statistical discrepancy, is the net effect of all of the e rro rs and inconsistencies in the estim ates. Although it reflects the inaccuracies in the complete set of accounts, it does not tell us anything specific about the reliability of any p art of the accounts. The statistical discrepancy consists partly of inconsistencies in the concepts in various p arts of the accounts. For exam ple, as some item s are estim ated on a cash b asis and some on an accrual basis it would be possible for each of them to be estim ated accurately on those bases and yet give rise to a discrepancy, if a p articu lar transaction were recorded on a cash b asis in the re c e iv e r's account and on an accrual b asis in the p ay er s account.

67 5 Quarterly Estimates Q uarterly estim ates of national income and expenditure and supporting details a re compiled and published in a num ber of countries, including A ustralia. The A ustralian estim ates a re published about seven or eight weeks after the end of each q u arter and serve to provide a m ore up-to-date picture of economic activity than is shown by the annual estim ates. They are widely used in discussions of cu rren t economic trends and provide a basis for projections of activity. The concepts and methods used in the quarterly accounts are broadly sim ila r to'those used in the annual estim ates and much of the discussion in previous chapters about the meaning and reliability of national income estim ates is relevant also to the interpretation of the quarterly figures. T here a re, how ever, some im portant differences in types of data used, p articularly in the much g re a ter use of sam ple surveys to provide estim ates, and th ere are also additional problem s in the interpretation of changes from one quarter to the next as a re su lt of differences in the tim ing of item s and the influence of seasonal variations. This chapter deals mainly with the effect of these facto rs on the meaning of the quarterly estim ates; as in previous chapters no attem pt is made to describe in detail the sources and methods used in compiling estim ates fo r each item. The form and purpose of the quarterly estim ates The q uarterly estim ates are set out partly in the form of accounts and partly in the form of tables, but both are related to the fram ew ork of the annual accounts. However, because of difficulties in estim ating item s of income and outlay of trading and financial enterp rises by q u a rte rs, no accounts a re shown for these secto rs in the quarterly estim ates. In addition, the capital account shows le ss detail of the undistributed income of com panies. The main purpose of the quarterly estim ates is to provide a m ore up-to-date sum m ary of economic activity than is given by the latest 59

68 60 N ational Income E stim ates annual accounts. They provide, as it w ere, a sum m ary of individual indicators which is designed to help assessm en t of the undertying economic situation. They also fill in gaps in available indicators in ord er to show the total demands for goods and serv ices, the incom es originating from curren t production, and a division of expenditure and incom e into the m ain com ponents. The use of the estim ates may be com pared to that of an index of industrial production. A good deal of detail of the production of individual com m odities is available but for the purpose of gauging changes in the overall level of production the indicators need to be aggregated and weighted by th eir im portance and estim ates are required of the production of item s not shown in individual se rie s. An index of industrial production is thus a sum m ary of the available statistic s of production with an allowance for the likely effect for production of other goods. In the sam e way, the quarterly estim ates bring together details of income and expenditure given by the available economic indicators, weight these by th e ir relative im portance and allow for item s not covered by these indicators. While the main use of the estim ates is in cu rrent economic analysis and forecasting of economic activity, estim ates for past q u arters are also of use in supplem enting the annual estim ates, particularly for the purpose of identifying turning points in economic activity. Annual estim ates tend to smooth out fluctuations in economic activity during the y ear and provide only a very rough m easure of the timing of changes in levels of demand and incom e. Thus, for exam ple, a fall in national income from one year to the next may originate in a decline in activity in the la tte r p art of the firs t year or in the early p a rt of the second y ear, or it could be the re su lt of a very sharp fall in the la tte r p art of the second year. The tim ing of changes, as well as th eir magnitude, is m asked in the annual se rie s. It should be noted, however, that q uarterly estim ates are not p rim arily compiled as h isto rical se rie s and the reliability of changes in past figures is not much g re a ter than that of the current estim ates. Little inform ation is available after one quarter to improve the accuracy of the figures and such revisions as a re made a rise from the need to fit the quarterly se rie s into annual benchm ark totals and annual concepts. In addition, of co u rse, som e revisions will be made when additional retu rn s a re received to quarterly surveys. There are, however, few new q uarterly benchm ark se rie s available to replace the first estim ates, as is the case with the subsequent estim ates of annual

69 Q uarterly E stim ates 61 national incom e fig u res. A fu rth er use of the quarterly estim ates is in up-dating annual figures. F or exam ple, just before an annual figure is firs t published, quarterly estim ates will be available for the first three q u arters of the y ear. In assessin g the long-term ra te s of growth it may be useful to bring the annual figures up to date and this can be done approxim ately by the use of the estim ates for the first three q u arters of the curren t y e a r.1 F o r the assessm en t of curren t economic trends it is not necessary that the quarterly accounts should follow the form of the annual estim ates o r that the supporting tables should be in the sam e amount of detail. The main use of the quarterly accounts appears to have been in te rm s of the changes shown by total national product and the m ain aggregates. As explained in a later section of this chapter, discrepancies in the b asis of tim ing of transactions also make it difficult to re late item s and virtually im possible to in terp ret the m eaning of balancing item s of savings. In addition, a good deal of the detail of the annual estim ates is required for long-term com parisons of the use of reso u rces to m eet various needs, but is not required in the assessm en t of curren t levels of economic activity. This includes, for exam ple, the use of public authority expenditure for such purposes as law, o rd e r, public safety, and education. On the other hand, other detail may be particu larly useful to show by q u a rte rs, even though it is not included in the annual estim ates. This may be the case, for exam ple, with detail of investm ent in stocks by industry where the usefulness of the figures for cu rrent analysis m ore than offsets the com parative unreliability of the estim ates. T here is also a m ore general point involved. The definition of national product used in the annual accounts is at least partly based on the use of the figures for com parisons of real output and income over tim e. For this purpose, as explained in the following chapter, certain imputed values are included for non-m arketed output, ^ The preparation of the quarterly estimates did, as well, lead to a major improvement in the array and reliability of short-term economic indicators. Until the time of publication of the quarterly accounts, virtually the only comprehensive income and expenditure data available by quarters were imports and exports, wages and salaries, and retail sales (the retail sales figures were, however, subject to a considerable lag). The preparation of the quarterly estimates led to new quarterly surveys of expenditure by private businesses on equipment, buildings, and stocks; revenue, expenditure, and capital outlay of public enterprises; profits of companies, as well as leading to an improvement in the timing of retail sales surveys and a good deal of new analysis resulting in new series of expenditure and income of public authorities, farms, personal consumption expenditure, etc.

70 62 National Income E stim ates principally the ren t on ow ner-occupied dwellings. These im putations a re not n ecessarily appropriate in the use of the accounts for cu rren t analysis. There is also the possibility that a different arrangem ent and subdivision of item s may be m ore useful for this purpose. The main reasons for the adoption of the form of the annual accounts in presenting the quarterly estim ates w ere the practical ones that the concepts used in the annual estim ates w ere probably fairly well understood by u sers and adoption of a different fram ew ork could lead to confusion, and that it is sim p ler in compiling the estim ates to use the sam e concepts for the annual as well as for the quarterly data. Timing discrepancies One of the main difficulties in interpreting the changes shown by the quarterly figures is caused by the different points of tim e at which transactions are m easured. Some transactions are m easured at the tim e of the cash receip ts and paym ents, some when goods change hands, and others on a proper accrual b asis, that is at the tim e when obligations and rights over ownership are settled. It would be possible in principle to use any of these methods of recording transactions and there are some purposes for which the flow of cash2 or goods may be appropriate. But for the main purpose of the quarterly estim ates, which is to analyse the circum stances affecting the demand and supply of goods and serv ices, there is little doubt that the appropriate point of tim e at which transactions should be recorded is at the date of receivables and payables. In fact, however, only a few transactions are m easured on this b asis, the m ost im portant being those connected with business income. This leads to two problem s. F irstly, estim ates which a re not on an accrual b asis may not accurately show the demands and supplies of goods and serv ices. Secondly, the inconsistencies of tim ing make it difficult to relate tran sactio n s, that is to com pare the movement in different item s and to divide a total into component item s which have the sam e meaning. T here is also the difficulty that estim ates derived as a balance between several item s are affected by the way these item s are m easured. The inconsistencies in tim ing would not m atter very much if there w ere a fairly constant difference between the flows of cash or goods and the receivables and payables in each q u arter. In this case the differences in tim ing would not affect the com parison of item s between 2 This m ethod was adopted, for ex am p le, in a set of so cial accounts designed to show the flow of funds in the econom y, co m p iled by M.A. C opeland (1952).

71 Q uarterly E stim ates 63 the sam e q u arters of different y ears or estim ates co rrected for seasonal influences. In p ractice, how ever, the difference between transactio n s m easured on different bases is likely to vary in different periods. This is shown up, for exam ple, in big swings in the balance of paym ents item of discrepancy, caused mainly by inconsistencies in the tim ing of paym ents for im ports and receipts from exports of goods and serv ices and other balance of paym ents item s. The difference between a cash and an accrual recording of transactions will vary with, am ongst other things, conditions of credit; in tim es of cred it re stra in t outstanding liab ilities tend to ris e relative to the value of transactions. The difference between the flow of goods and an accrual recording w ill depend largely on factors affecting the supply of goods. A dock strik e in A ustralia will lead to a g re a ter difference between the im port of goods and the incurring of the obligation for payment. Since these differences are partly associated with fluctuations in economic activ ities, they are an im portant lim itation on the utility of the accounts. The method of recording transactions may be sum m arised briefly as follows: the principal item s which are on a cash basis a re m ost of the capital and curren t item s of public authorities, m ost of the capital expenditure and the surpluses of some public trading e n terp rises, wages and sa la rie s, and rent and in terest (other than dwelling rent); the chief item s m easured at the tim e of the flow of goods and services a re im ports and exports and the gross output of prim ary production; other item s are m ore or less on an accrual basis. The difference between the flows of cash or goods and changes in receivables and payables v aries for individual item s. Exports of goods, for exam ple, are likely to be m ore closely related to the change in ow nership than im ports. The m ajor discrepancy between cash flows and an accrual recording of transactions is in item s of public authority receip ts and paym ents. The cash payments of personal income tax by provisional tax p ayers, for exam ple, are m ainly concentrated in the la st quarter of each y ear, but each of the q uarterly figures for a p articu lar y ear will relate largely to taxes payable out of the income earned in the previous y ear. Public authority cash expenditure tends to be concentrated in the last quarter of each y ear owing to the need to settle outstanding obligations before the end of the fiscal year. The cash paym ents of wages and sala rie s vary slightly between q u arters from paym ents due, largely owing to the differing num ber of pay periods and the irre g u la r payment of annual

72 64 N ational Income E stim ates bonuses. The effect of changes in economic conditions on the difference between the flows of goods or cash and changes in receivables and payables also v aries between item s. Fluctuations in the level of economic activity have little effect on the difference between cash and accrual paym ents of wages and s a la rie s, but budgetary considerations and the need by public authorities to accelerate or re ta rd the flow of cash in the private secto r, depending on w hether or not economic conditions a re generally deflationary o r inflationary, may influence the tim ing of public authority cash paym ents. A part from the effect of a dock strik e, mentioned above, a num ber of other factors could affect the tim ing of im ports of goods. Anticipated changes in im port re stric tio n s or ta riff regulations, for exam ple, have led in recent y ears in A ustralia to abnorm al placem ents of o rd e rs for im ports, and changes in the origin of im ports may a lte r the period of tim e between the purchase of im ports and the a rriv a l of the goods. Balancing item s derived as the sum or difference of several estim ates are the savings of secto rs, the statistical discrepancy, fa rm incom e and fa rm product. The savings of public authorities, or the net deficit on current account, is largely on the b asis of cash transactions. The meaning of these figures is th erefo re c le a r, although they may need careful interpretation in making com parisons between the figures and other estim ates in the social accounts. P ersonal savings, how ever, a re obtained as a balance between item s of receip ts and paym ents which are based on a m ixture of different methods of m easurem ent. Wages and sa la rie s, cash social service benefits, in te re st and dividends, and personal rem ittances to and from overseas a re on a cash basis. Supplements to wages and sala rie s, the income of unincorporated business (other than farm s), dwelling ren t, and virtually all of personal consumption expenditure are recorded as receivables and payables, while farm income and product are partly on an accural basis (costs) and partly on the basis of flows of goods (farm output). 3 Changes in personal saving from one q u arter to the next will therefore reflect differences between some ^ Gross v alu e of production is defined as 'the estim a te d v alu e of produce d elivered for sale during the quarter to g eth er w ith the e stim a te d v alu e of changes in undelivered stocks other th an w o o l1 (Q uarterly E stim ates of N atio n al Incom e and E xpenditure, C om m onw ealth Bureau of Census and S tatistics B ulletin No. 26, D ecem b er quarter, 1966, p. 41). Inform ation about d eliv eries is a v a ila b le for som e item s, as for w heat and wool w hich are recorded a t the tim e the products are d eliv ered to w h eat silos or w oolbuying brokers. For som e item s production is a llo c a te d on th e basis of harvesting periods.

73 Q uarterly E stim ates 65 item s m easured as receivables and payables and some item s of cash re c e ip ts and expenditure. B usiness saving is included in other savings, the main components of which a re undistributed income accruing to resid en ts, retained investm ent income of life assurance funds, and the increase in dividend and income tax provisions. While the first item is on an accrual b a sis, the inclusion of dividend and income tax provisions means that the tax and dividend paym ents by businesses are on a cash b asis (this amount includes, as w ell, the increase in tax re se rv e s of persons) and the figure of other savings is therefore a m ixture of tran sactio n s m easured in different ways. 4 The statistical discrepancy, because it is the balance between item s of national product and expenditure, is obviously subject to a very large extent to factors of tim ing. Changes in the discrepancy between one q uarter and the next are due partly to tim ing differences and partly to e r ro rs of estim ation of item s on the b asis on which they purport to be recorded. The fo rm er elem ent is undoubtedly the main cause of variations in the statistical discrepancy from one quarter to the next. Probably the m ost difficult item s to m easure consistently are farm income and product. As has been stated, these are a balance between farm output, m easured at the tim e the produce is delivered, and co sts, m easured when the expenditure is incurred. This m eans that costs are not m atched against production, and that some of the costs in a quarter properly relate to production in a subsequent q u arter. Some item s of costs, such as harvesting and m arketing expenditure, are closely related to output, but a wide range, including seed and fe rtilis e r, occur in different periods. There a re several ways in which costs could conceivably be m atched with production and which have been used in som e overseas estim ates. One approach would be to impute values for growing crops and increasing num bers of livestock. This could be done either by d irect valuation or by fairly a rb itra ry allocation of the estim ated annual values of production between q u arters. The la tte r procedure involves making a fo recast of production for the cu rren t year and leads to the im putation of a profit on production before it is earned. An alternative approach would be to distribute costs to the periods when the ^ In effect, the increase in provisions for tax and dividend payments is written back into undistributed income of businesses; and this item of business saving is therefore on a cash basis.

74 66 National Income E stim a te s production takes place, with unallocated costs shown as a w ork-inp ro g ress stock. E ither method would lead undoubtedly to statistical problem s, although they would reduce the present very large fluctuations in farm income and product between successive q u arters. It should be noticed that these types of tim ing difference occur also in some item s in the annual national income estim ates. But they are much m ore im portant in the case of the quarterly estim ates, for two main reasons. F irstly, the difference between the flow of cash or goods and an accrual recording does not depend on the length of the period over which item s are m easured and accordingly it is relatively m ore im portant in relation to the sm aller quarterly totals. Secondly, for some item s the difference tends to cancel out over the year. For public authority expenditure, for exam ple, purchases tend to be above cash paym ents in the early p art of the y ear, but this is made up for by the re v e rse situation in the la st quarter of the year. Sim ilarly, income tax paym ents by provisional taxpayers run behind accruals in the first th ree q u arters of the y ear but a re ahead of accruals in the la st quarter. The question a ris e s, in view of the difficulties in interpreting quarterly changes as a re su lt of tim ing discrepancies, whether or not it is possible to adjust the basic data to put them m ore on a consistent accrual b asis. One possibility in the case of farm income and product has been discussed above. In estim ates fo r some overseas countries, including the United States, B ritain, and Canada, an adjustm ent is made to public authority transactions to convert them from a cash to an accrual b asis. In A ustralia very little work has been done on this question, and the only adjustm ents made to elim inate timing discrepancies are for several item s in the balance of paym ents, stocks, and company income. The change in the value of stocks includes some stocks (e. g. wool and sugar) owned by non-residents and held in A ustralia. This m eans that stocks include goods owned by non-residents which a re not included in statistic s of exports until they are shipped and the change serv es to im prove the consistency of stocks and exports. An adjustm ent is also made to figures of public authority expenditure for purchases of large item s of defence equipment to im prove the consistency of this item with the im ports. The public authority expenditure, for exam ple, shows p ro g ress payments made for w arships bought from o v erseas, while the recorded trade figures include im ports of w arsh ip s when they a re handed over to the

75 Q uarterly E stim ates 67 A ustralian governm ent. The public authority expenditure is now adjusted to include deliveries of overseas equipment. E stim ates of the operating surplus of com panies for early years also included a tim ing e r r o r which has now been elim inated in figures for m ore recent y ears. Annual figures of company income derived from taxation re tu rn s re p re se n t a m ixture of different balancing dates. Until the institution of a q uarterly survey of company profits in 1961, quarterly figures of profit w ere obtained simply by distributing the annual total to q u arters using various indicators. But because of the m ixture of balance dates the annual figures represent only p art of company income in a p articu lar y ear and include som e of the income earned in the previous and succeeding y ears. Now, using the resu lts of the survey, annual figures of company income a re adjusted to show only the profit earned in the y ear. Q uarterly figures for recen t y ears now also show only the profit earned in the q uarter. Seasonal adjustm ent A num ber of the item s in the quarterly estim ates are affected by seasonal factors of production or demand. Thus farm income and product vary according to the effect of the seasons on values of production, personal consumption expenditure is la rg e r in the D ecem ber q u arters as a re su lt of p re-c h ristm as trading, and capital expenditure on buildings is la rg e r in d rie r periods of the year. Changes in the figures from one quarter to another are therefore likely to reflect not only underlying trends in production or demand but also the effect in the q u arters of reg u lar seasonal variations. In ord er to obtain a serie s which shows underlying trends it is n ecessary to elim inate the influence of seasonal factors. One way of doing th is, which is used by the Commonwealth Statistician, is to relate figures for each quarter to those for the sam e q u arter of the previous year. In this approach a s e rie s of annual percentage changes is derived, leading to a m ore uniform sequence than if the percentage changes in successive q u arters had been taken. This method assum es, however, that the figures of the previous year show a norm al trend and a re unaffected by either cyclical changes or e rra tic variations. If, for exam ple, a serie s was showing abnorm al in creases or declines twelve months e a rlie r, the sequence of annual percentage changes would tend to reflect this by showing abnorm al declines o r in creases. As w ell, the influence of a p articu lar e rra tic

76 68 National Income E stim ates event twelve months e a rlie r, such as a strik e (reducing output) or an increase in ra te s of indirect taxes (leading to higher values of expenditure) would also be re fle c ted in the annual percen tag e changes. A second approach to derive seasonally corrected figures is to divide estim ates for each quarter by factors designed to elim inate norm al seasonal change. Since the essence of seasonal variation is that it is a regularly re c u rrin g change, the factors can be derived by inspection of past changes. 5 This method has been used by the Statistician to provide seasonally adjusted estim ates for the chief item s of national income and expenditure from the Septem ber 1967 issue of Q uarterly E stim ates of National Income and Expenditure. The approach used is based on a method developed by the U. S. Bureau of the Census and applied by the Statistician to seasonally adjust a num ber of other economic indicators. The main features of the approach are set out in Seasonally Adjusted In dicators, 1967, published by the Commonwealth B ureau of C ensus and S tatistics in July It essentially involves smoothing past quarterly figures by a fiveterm moving average, averaging the ratio s of the actual figures to the smoothed series in each equivalent q u arter, and using these ratio s to adjust each figure. This is described by the Statistician as the m ultiplicative variant of the U.S. Bureau of the Census method. For several item s, how ever, this approach could not be used or is thought by the Statistician to produce m isleading re su lts, and alternative approaches are used. The main item s w here different m ethods a re used are farm income and product, changes in stocks, and the statistical discrepancy. F arm income and product a re derived as balances between farm output and farm co sts. Item s of production and costs have different seasonal patterns and the im portance of each series v aries m arkedly in different y ears. Hence it is desirab le to seasonally adjust the individual item s separately and to derive farm income and product as the balance of these item s; if the net figure only w ere adjusted the seasonal factors would reflect not only regularly re c u rrin g seasonal patterns but also the changing com position of the production and costs. This problem runs through a lot of the item s of national income and expenditure, and the S tatistician has used g re a te r detail than is ^ 'S easonally adjusted d ata are w hat re m a in after rem o v al of regularly re c u rre n t ebbs and flows in econom ic a c tiv ity due to th e w eath er, crop-grow ing cycles, holidays, vacations, reg u lar industry m odel c h a n g e-o v er periods, and the lik e ' (U.S. P resident's C o m m ittee 1962:65).

77 Q uarterly E stim ates 69 published in o rder to derive the seasonal adjustm ent facto rs. But farm income and product a re the chief exam ples of a serie s where the component item s exhibit quite different seasonal variations. The seasonal adjustm ent of farm income and product is further com plicated by the fact that the seasonal pattern of individual series is unstable. The great im portance of seasonal factors in affecting the distribution of production between q u arters m eans that even a sm all change in seasonal conditions could lead to a quite noticeable difference in the distribution of production. As a consequence, seasonal adjustm ent using the ratio-to-m oving-average method would lead to considerable fluctuations in these series between one quarter and the next. In o rd er to avoid this resu lt the Statistician has adopted the approach of distributing m ost item s of production and some item s of costs in a fairly a rb itra ry m anner designed to elim inate as far as p o ssib le fluctuations in farm incom e and product. The methods used a re se t out by the S tatistician as follows: G ross value of production of wool. The seasonally adjusted series has been obtained by multiplying a smoothed quantity series by a seasonally adjusted price serie s. The smoothed q u a n ta serie s was obtained from a moving average of the estim ated production in two half y ears, ending December and June. The figures for September and Decem ber q u arters are the sam e, nam ely one-quarter of the estim ated production in the y ear ended Decem ber. Those for M arch and June q u arters a re one-quarter of the estim ated production in the y ear ended June. The price serie s was seasonally adjusted (m ultiplicatively) in o rder to elim inate the seasonal effect of quality variation in quarterly sales of wool. G ross value of production of grain crops and sugar cane. The estim ated annual value for each crop has been distributed equally among four q u arters commencing with that q u arter in which production com m ences each y ear in the original serie s. Special adjustm ent has also been used for m arketing costs of wheat, the annual total being spread over four q u a rte rs, in line with the production figures. A sim ilar approach was not practicable for the o th er com m odities, but would have made little d ifference. This method, although reducing the magnitude of the fluctuations in farm product, does have a num ber of im plications for the interpretation of the estim ates.

78 70 National Income E stim a te s F irstly, changes in the annual levels of production of different item s tend to be reflected in the seasonally adjusted series in the change between the last and first periods of successive periods of production. Thus, for exam ple, the change in annual figures of wheat production appears in the D ecem ber q u arter, and for sugar in the September q uarter. This discontinuity in the estim ates is in p ractice likely to be somewhat offset by differences in the production seasons for different ite m s. Secondly, the apportionm ent of the annual figures fairly evenly between quarters leads to a somewhat different concept of farm product from that underlying the method of presenting the unadjusted se rie s. As has been pointed out, in the unadjusted se rie s farm income and product a re m easured as the difference between output of farm products and costs incurred each q uarter. A lower level of production in one quarter due, for exam ple, to the effects of a drought, is reflected in the values of farm income and product in that quarter. The seasonally adjusted s e rie s, however, would not show a sim ilar fall in the serie s in that q u arter. The seasonally adjusted serie s, obtained by the spreading of farm production fairly evenly over all q u arters of each y e ar, are in effect a form of accrual recording. The meaning of this difference in interpretation on the estim ates can be very roughly shown by the following table, which com pares the changes in the seasonally adjusted figures of farm product during 1965/66 with those obtained by a sim ple com parison of the figures in com parable q u a rte rs of 1964/65 and 1965/66. Seasonally adjusted F arm product, 1965/66 Changes over 12 months June 1964/65 to Sept. 1965/ /65 to 1965/66 Sept /66 Sept, to Dec. -28 Dec. -32 Dec. to M arch +2 M arch -126 M arch to June +3 June -13 The method of com paring differences over tim e in the change in item s between com parable q u arters of successive y ears is, for reasons mentioned, an im perfect one of allowing for seasonal changes, and it is quite likely that som e of the difference in the pattern of farm product in 1965/66 shown by the figures would have no economic significance to fa rm e rs. But the drought of that year did occur mainly in the m iddle of the y ear and it did affect mainly wheat, which

79 Q uarterly E stim ates 71 is harvested in the m iddle of the year. On an interpretation of farm product as used in the unadjusted series it would seem that there w as, in fact, a sharp fall in production in the middle two q u arters of the y ear, 1965/66. The apparent stability in farm product (and income) shown in the seasonally adjusted series seem s to involve a difference in definitions which m ay, in some resp ects, be le ss satisfactory than those used in the unadjusted serie s. The Statistician has made com parable adjustm ents to the change in farm stocks as to farm product. This sim ilarity of treatm ent is im portant in view of the fact that the seasonally adjusted item s a re set out in the form of a balancing account of national product and expenditure. The seasonally adjusted figures of the changes in farm stocks a re derived by adding the difference between the seasonally adjusted and unadjusted value of farm output to the unadjusted value of the change in farm stocks and seasonally adjusting the revised stocks s e rie s. In the case of wool, seasonally adjusted changes in the value of stocks a re derived by deducting seasonally adjusted w ithdrawals out of stock from seasonally adjusted values of production. These m ethods of adjusting changes in farm stocks lead to a consistent treatm en t of this item with other seasonally adjusted series of national income and expenditure but the interpretation of the figures is not straig h tfo rw ard. A different treatm ent from the standard approach in deriving seasonally adjusted figures is also used for item s of the change in the value of non-farm stocks and the statistical discrepancy. Because of the negative figures in non-farm stocks the m ultiplicative method cannot be applied and the S tatistician has used the additive ra th e r than 1 the m ultiplicative version of the U. S. Bureau of the Census method, which involves calculating the seasonal adjustm ent from the actual changes in stocks ra th e r than the ratio of the actual changes to the moving average. An alternative method would have been to calculate total values of stocks held each q u arter, adjust these figures by the m ultiplicative method, and then derive the seasonally adjusted series of changes in stocks as the difference between the adjusted totals at the end of successive q u a rte rs. The item of statistical discrepancy which in the unadjusted series is a balance between item s of income and expenditure making up national product is also treated as a balancing item in the seasonally adjusted estim ates.

80 72 N ational Incom e E stim ates R eliability of q u a rte rly e stim a te s The usefulness of the q u a rte rly e stim a te s for the in te rp re ta tio n of c u rre n t events depends on the accuracy of the fig u res and how quickly they a re available a fte r the end of each q u a rte r. T hese two conditions conflict and le s s re lia b le fig u res could be published e a r lie r. An ideal approach to solving th is conflict would be to d eterm in e how the fig u res a re used and the accuracy needed for th ese different p u rp o ses. In p ra c tic e, how ever, the fig u res m eet a wide range of different u ses and it is not p o ssib le to su m m a rise even the chief u ses of the q u a rte rly accounts, o r to provide objective m e a su re s of re lia b ility. When the e stim a te s w ere f ir s t p rep ared it was p o ssib le to s trik e only a rough balance betw een re lia b ility and p ro m p tn ess. E stim a te s of the m ain item s - such as w ages and s a la rie s, im p o rts and ex p o rts, and som e public auth o rity tra n sa c tio n s - w ere availab le w ithin about four w eeks of the end of each q u a rte r. To obtain inform ation about o th er item s new su rv ey s w ere needed and these w ere designed to obtain the re q u ire d inform ation w ithin five w eeks. O riginally, for exam ple, q u a rte rly su rv ey s of p riv ate investm ent expenditure on buildings, equipm ent, and changes in n o n-farm stocks w ere lim ited to la rg e b u sin e sse s, and expenditure by other b u sin e sse s w as assu m ed to v ary in p ro p o rtio n to th is expenditure. The adoption of a five w eeks tim etable also affected v ario u s in d ire c t m ethods of estim atio n which could be used. It is difficult to be su re how re lia b le the q u a rte rly e stim a te s a re and, as with the case of the annual fig u res, it is not p o ssib le to e x p re ss m a rg in s of e r r o r in a sim ple fashion, such as sam pling e r r o r s. F o r the re a so n s given above it is also not p o ssib le to draw any conclusions about the re lia b ility of the fig u res from re v isio n s m ade. A ssessm e n t of accuracy can only be m ade by co n sid erin g the so u rc e s and m ethods used fo r each item. The following g en era l account of so u rc e s of e r r o r m ay, how ever, give som e indication of the re lia b ility of the fig u res as w ell as illu stra tin g the difficulty of providing a sim p le answ er to the question of how re lia b le the q u a rte rly e stim a te s a re. In the f ir s t p lace it m ay be noted that for a num ber of im p o rtan t item s in the accounts q u a rte rly e stim a te s a re b ased v irtu ally on the sam e so u rc e s of inform ation as the annual fig u re s, o r from so u rc e s which provide v irtu a lly a com plete enum eration. T hese include im p o rts and ex p o rts of goods, w ages and s a la rie s, a la rg e p a rt of the

81 Q uarterly E stim ates 73 public authority tran sactio n s, some item s of personal consumption expenditure, private investm ent expenditure on dwellings and other new buildings, and investm ent in farm stocks. As explained previously, however, even though the sources may provide estim ates which are reliab le, they are not always on the co rre ct b asis of tim ing. E rro rs in concept due to tim ing discrepancies affect p articu larly item s of public authority receip ts and expenditure and various derived estim ates, such as savings. For other item s the main sources of e rro r are due to th ree factors: the use of sam ple surveys, differences in concepts, and indirect methods of estim ation. Sample su rv eys. E rro rs due to the use of sam ple surveys affect p articularly expenditure by private businesses on equipment, building and stocks, and company income. To som e extent it would be possible to calculate standard e rro rs for these figures, although the population from which the sam ples are drawn covers only business firm s subm itting payroll tax retu rn s (in the case of the expenditure surveys) and a fu rth er adjustm ent is needed to estim ate the rem aining (about 25 p er cent) expenditure. Q uarterly sam ple surveys are the basis of estim ates of private business income and fixed investm ent expenditure on stocks and capital expenditure (other capital equipment as well as other new buildings). These surveys vary in num bers and coverage, although in all cases the retu rn s account for m ore than 50 per cent of the estim ated total figure. The survey of company profit covers m ore than 500 of the la rg e r company groups, which account for som e 60 p er cent of total company income. The surveys both of stocks and of capital expenditure are circulated to about 3,300 and 4,200 businesses respectively, subm itting payroll tax re tu rn s. P relim in ary figures are usually based on about 70 p er cent of re tu rn s and final estim ates are made when la te r re tu rn s are received. D ifferences in concepts. In a num ber of cases the basic figures differ in concept from that ideally required for the quarterly estim ates, and additions or adjustm ents are made to im prove th e ir usefulness for this purpose. This may reduce the accuracy of the basic figures. E stim ates of wages and sala rie s are available, based on payroll tax retu rn s (and other inform ation), which are reasonably reliable. For the purpose of the quarterly estim ates, how ever, an addition is made

82 74 National Income E stim ates to the figures for supplem ents (com prising largely em ployers' contributions to pensions and superannuation funds and d irect payments of pensions), which are estim ated by a rb itra ry methods. The resulting figures are clo ser to what the estim ates purport to m easure, that is the total rem uneration of wages and sa la rie s, but the statistical accuracy of the basic wages and sala rie s figures is g reater than that of the combined serie s. Indirect m ethods of estim ation. Where no d irect inform ation is ivailable it is n ecessary to estim ate item s by the use of related, or indirect, data. In the case of early estim ates made of the profit of com panies an attem pt was made to construct production accounts leading to a derived figure of profit based from indicators of output and costs. Again, as has been mentioned, som e item s of farm costs are based on known or assum ed pattern of production, such as harvesting dates for crops. The main item s which are indirectly estim ated, however, are the income of unincorporated businesses other than farm s, ren t and in terest paid by trading e n terp rises, and depreciation. N on-farm income is divided, in preparing the estim ates, into broad industry groups, the annual figures of which (from taxation tabulations) are interpolated and projected by q u arters using production and other indicators appropriate to each industry. Rent and in terest paid is the sum of a num ber of separate serie s which are estim ated by the use of indicators, and depreciation is interpolated and projected on trend.

83 6 National Product as a Measure of Total Output Most of the development in national income methods in the la st two decades has concerned the social accounting form s of presenting these estim ates. As explained in Chapter 1, this is a reflection of the in terest of econom ists in cu rren t economic analysis involving the in terrelatio n of various p a rts of the economy and of the usefulness of the accounts in providing the inform ation re q u ire d fo r th is purpose. At the sam e tim e, how ever, th ere has been a continuing in terest in the m ore traditional purposes of national income estim ating, that is of providing m easures of the aggregate output and income of countries needed to com pare the levels of economic activity and living standards between countries and over tim e. Quite recently the in te re st in aggregate estim ates has increased as a re su lt of the development in quantitative methods of economic history and the concern of econom ists with problem s of the economic growth rate of countries. A good deal of w ork has been done, and is continuing, in the com pilation of longterm se rie s of the real national product of countries. F or A ustralia, estim ates of national product, investm ent expenditure, and the balance of paym ents from 1860 to 1938/39 have been compiled by N.G. Butlin (1962). These serie s, used in intertem poral and international com parisons, are based on concepts proposed by the Statistical Office of the United Nations. Although these w ere drawn up as p art of the definitional fram ew ork of social accounts, the concepts are relevant also to the use of estim ates of national product as a m easure of the aggregate output and income of co u n tries.1 In fact, there is not a g reat deal of There is no necessary reason why the social accounting tables should include an entry for national product. The main object is to measure economic transactions of types of goods and services and particularly those which lead to a demand for employment. This would strictly confine the coverage of the accounts to market transactions. However, since the definition of national product as a measure of total output or income of countries is based largely on market transactions, there is not a great deal of difference between this definition and the sum of individual items shown in the accounts, and there are certain advantages in maintaining thic consistency. 75

84 76 National Income E stim ates difference between the presen t concepts and those used in p re-w ar histo rical series; they a re still based on the criterio n set out by A. C. Pigou in 1928, that national product (or income) should include the values of production of goods and serv ices that can be brought directly or indirectly into relation with the m easuring rod of m oney as interpreted narrow ly by Pigou (1932:34) as covering 'everything that people buy with money income, together with the serv ices that a man obtains from a house owned and inhabited by h im se lf'. The use of the standard definition in com piling the estim ates does ensure a broad m easure of com parability in the figures for different countries, thus facilitating com parisons between these estim ates. However, the definition is based on institutional c h aracte ristic s of econom ies and th ere still rem ains the problem of allowing for the effect on the figures of any variation in the coverage of these institutions. T here are differences in the scope of the m arket economy between countries and over tim e, which could mean that a differing proportion of economic activity is included. As w ell, the use of standard definitions has tended to lim it the experim ents with alternative m easures made in p re-w ar y e ars. For exam ple, Swedish estim ates published in 1936 covering the period 1860 to 1930 included in addition to the values of production sold for money, estim ates of the unpaid dom estic work of housewives and the serv ices of household goods and m otor c a rs. One of the featu res of the work in national income estim ation in recent decades has been the lack of any development in methods of allowing for the effect of different institutional c h aracte ristic s of econom ies over tim e or between countries. There have, however, been substantial im provem ents in one aspect of the s e rie s, that is in the methods of allowing for differences in levels of p rices. Although th ere are still many unresolved problem s in this a re a, th ere is a fair m easure of agreem ent about the appropriate methods of deflation in com parisons both over tim e and between countries. This chapter and the following two deal chiefly with problem s of definition and deflation which a rise in using national product as a m easure of the aggregate output and income of countries. This chapter discusses questions of concept and the methods of valuing the goods and serv ices in cu rren t money p rices; the following two chapters deal with the procedures in adjusting money values for differences in levels of p rices in com parisons over tim e and between countries.

85 M easu res of Total Output 77 Concepts of national product A detailed discussion of the definitions of national product is provided in the United Nations document, A System of National Accounts and Supporting T ables. The definitions and methods set out in this document are generally followed by m ost W estern countries; th ere are some differences in the treatm ent of p articu lar item s, largely as a resu lt of differences in the type of basic data available, but these do not lead to any substantial difference in the total fig u res. National product com prises the total value of goods and serv ices currently produced and available for consumption, investm ent, government use or net export. It does not include the addition to wealth as a re su lt of new discoveries of natural re so u rces, or the increase in natural a sse ts, such as fo rest tim ber or livestock, which does not resu lt from productive activity. 2 it may (and invariably will) differ from capacity o r potential output, and generally will m easure changes in the productiveness of an economy only to the extent that this is reflected in goods and services produced. There are several definitions of national product. Three of the m ost common v arian ts, which differ by the treatm ent of overseas tran sactions, are: (i) G ross dom estic product. The m arket value of the gross product of establishm ents located within a country, including exports and excluding im p o rts. (ii) G ross national product. The m arket value of the gross product of establishm ents owned by residents. This is equivalent to gross dom estic product plus net incom e from abroad. (iii) G ross national product, defined as dom estic expenditure (on consum ption and fo r capital form ation) plus net foreign investm ent. At curren t p ric e s, this is equivalent to (ii), but will differ from (ii) at constant p rices if th ere is a change in the term s of trade. The p articu lar form ula used will depend partly on statistical considerations, partly on the nature of the problem and probably 2 A System of N ational Accounts and Supporting Tables (U.N. 1964b) recom m ends that expenditure on forest clearance, land reclam ation and im provem ent, e t c., and on the developm ent of plantations, orchards, vineyards, forests, e t c., be included but that increases in natural resources due to growth (as w ith forests and standing crops) be excluded. Practice w ith regard to natural increase in livestock varies; för exam ple, this item is included in the British officia l estim ates of national incom e and in the Australian long-run series of national product com piled by Professor N.G. Butlin, but om itted from Australian o fficia l estim ates from 1938/39. The treatm ent of this and other item s in o fficia l estim ates of various countries is set out in detail in N ational A ccounting Practices in Sixty Countries (U.N. 1964a).

86 78 National Income E stim a te s partly also sim ply on the preference of the u ser. It is usual to favour (i) over (ii) o r (iii) because of the statistical difficulties involved in revaluing to constant p rices the net income from overseas. Form ula (ii) may be sim pler to m easure than (iii) but has the disadvantage for som e purposes of analysis that it ignores the effect of changes in te rm s of trade. F or exam ple, a fall in im port p rices followed by an increase in the consumption of im ports would not affect the estim ate of national product (at constant p rices). These effects may be quite im portant if the purpose of the analysis is related to a w elfare approach, since they do lead to changes in quantities of goods and serv ices available for consumption and investm ent. On the other hand, (ii) may be m ore appropriate for purposes of production or productivity com parison, w here it is desirable to elim inate the effects of changes in external p rices in the m easurem ent of local production. In some circum stances, also, (iii) may be appropriate for this purpose - for exam ple, if in the long run the changes in the term s of trad e reflect stru ctu ral changes in an economy leading to g re a ter productiveness' of the foreign trade secto r, and it is d esired to take th is into account. In addition, if a deduction is made for depreciation, the concepts are described as n e t. They may also be expressed at either m arket or factor p ric e s. Thus national income is equivalent to net national product at factor p ric e s. A ustralian official term inology is different from the above accepted international usage. Following a suggestion by Mr H. P. Brown, 'product' re fe rs to geographic concepts, while 'incom e' is used to denote residence 'N ational' has no p articu lar significance except perhaps to denote an aggregate. Hence 'g ro ss dom estic product' as defined above is described in the A ustralian estim ates as 'g ro ss national product'. T here is no A ustralian equivalent to 'gross national product' but it could be described as 'g ro ss national income at m arket p ric e s'. Selection National product is confined largely to those goods and serv ices which pass through the m arket exchange, although som e allowance is made for a few specific non-m arket activities. The United Nations standard system of accounts includes the comment that 'in a m onetary economy all goods and serv ices are included which are exchanged for m oney'.

87 M easures of Total Output 79 As w ell, in the case of p rim ary producers...all prim ary production whether exchanged or not and all other goods and services produced and exchanged are included in production. In the case of other producers... the total of their prim ary production is included as for prim ary producers. In addition there is included the total of th eir other production which is exchanged together with the unexchanged p art of th eir production in th eir own trad e. (U.N. 1964b:5) This m eans that in principle all p rim ary production is included, together with a p art of the rem aining production. It is obvious that such a definition excludes a good deal of activity which could possibly be described as econom ic. The classic definition of econom ics as the study of human behaviour in disposing of scarce m eans which have alternative uses, to satisfy ends of varying im portance (Robbins 1935:15-16) would bring into the coverage of production a whole range of non-m arket activities. But attem pts to provide a concept of national product which takes these activities into account have invariably foundered on the difficulty of distinguishing between that activity which is undertaken for its own sake and that which involves consciously the use of scarce reso u rces to provide goods and serv ices. Some activities of households may properly be classed as le isu re, while other activities may reflect the use of reso u rces which could be available for other purposes. The m otives of households are so entangled that it is im possible in practice to separate clearly these two types of activity without having reco u rse to some fairly a rb itrary c rite ria, such as to include in national product only those transactions which represent cash outlays. The rule to include all production which is akin to the output produced by the m ajor activity of m em bers of the household does, how ever, have som e justificatio n in a p rim a rily exchange economy. It could be assum ed that people are m ost efficient in th eir norm al occupation and that other work has som e, and probably a large elem ent of. non-econom ic justification. Thus, for exam ple, a builder may construct his own house because it is a m ore efficient use of his tim e to do it him self than to employ another builder or buy a house, but if he uses his spare tim e to m aintain his car it could be assum ed that this activity is not undertaken as the m ost efficient m eans of getting his c a r fixed. N evertheless, it does appear that some spare-tim e household

88 80 National Income E stim ates activity which is not related to the main occupation does have an economic m otivation. This is reflected in the fact that an increasing proportion of goods and serv ices produced by unpaid household activity tends to be purchased through the m arket as real national product per head ris e s over tim e. These include, for exam ple, h air-cu ttin g, the preparation and serving of m eals, laundry and dry cleaning and am usem ent serv ices of various kinds. This has led to the belief that the adoption of the definition resu lts in an understatem ent of the national product of le ss developed econom ies and an upward bias in the ra te of growth of national product over tim e. In addition, the definition is not fully applied in p ractice and a good deal of activity which is sim ilar to the main occupation of the household is not included in the m easurem ent of national product. No amount is included, for exam ple, for the unpaid serv ices of housewives and few attem pts have been made in h isto rical serie s to m easure the household production of goods and serv ices in p re-in d u strial econom ies which are la te r produced in facto ries, such as weaving and knitting, and the m anufacture of household uten sils. There have been substantial changes in the proportion of housewives who work for w ages, not only over long periods of tim e but also during relatively short periods, such as in w artim e. As housewives enter the m arket, an increasing proportion of national product com prises goods and serv ices previously perform ed in the home. The om ission of the la tte r type of production could lead to a further overstatem ent in the growth rate of national product, although at the tim e when this is likely to be m ost im portant, that is during the early stages of industrialisation of an economy, real national product is likely to be risin g at such a fast rate that the relative e rro r in neglecting this activity is not likely to be significant. Finally, even w here an attem pt is made to m easure non-m arket activities, the estim ates are likely to have a large m argin of e rro r. This applies particu larly to the m easurem ent of subsistence agriculture in le ss developed econom ies which accounts for as much as or m ore than 80 p er cent of the national product of these countries. Difficulties a ris e in m easuring this production as a re su lt of the frequent lack of census or other statistical reg istratio n of this production, differences in the quality of products consumed at home and supplied to the m arket, and the lack of any readily available p rices to value th is output. A further om ission from the value of national product is the serv ices

89 M easu res of T otal Output 81 of a s s e ts owned by p e rso n s and governm ents. E xpenditure on th ese a s s e ts is included in national p ro d u ct at the tim e they a re acquired but th e ir serv ic e s m ay extend over a long perio d of tim e and m ay be g re a te r than the in itial outlay. T his m eans not only th at the tim ing of the incom e is likely to be d isto rte d, but that any su rp lu s of incom e over th e ir co st p ric e is om itted. D uplication N ational product c o m p rises the value of goods and se rv ic e s c u rre n tly produced and available for the final p u rp o ses of consum ption, investm ent, and net ex p o rts. Output used in fu rth e r production is excluded since th is is duplicated in c u rre n t production. The duplicated output co m p rises th re e m ain types of p ro d u cts. F irs tly, th e re a re b u sin ess co sts deducted in a rriv in g at accounting fig u res of p ro fit. Secondly, th e re a re a num ber of item s w hich, w hile not a p a rt of the accounting co st of goods sold, m ay in fact fa c ilita te fu rth e r production o r not p ro p erly be a p a rt of final output used fo r consum ption, in vestm ent, o r net ex p o rts. T h ird ly, th e re is the co st of a s s e ts used up in c u rre n t production, o r dep reciatio n. The allow ance fo r b u siness c o sts of production lead s to no sp ecial p roblem s since the m ain so u rc e s of inform ation on b u sin ess p ro fits a re accounting re c o rd s which do provide d etails of th ese expenses. It has been m entioned e a rlie r th at in som e in sta n c e s, w here taxation s ta tis tic s a re used as a b a sis for estim atin g com pany p ro fits, som e adjustm ent may be needed to th e se data to allow for om itted or im p ro p erly included item s, but th e se adju stm en ts a re usually of m inor im portance and do not lead to any p a rtic u la r p ro b lem s. They concern, in p a rtic u la r, adjusting for the d ifference betw een actual and a sse ssa b le incom e. D ifficulties a r is e, how ever, in defining and m easu rin g the o th er two types of duplication. T hese have led to a com m on adoption in national accounting w ork of lim iting the deductions to the actual reco rd ed o ffsets to b u sin ess p ro fits only and of ignoring le s s c ertain deductions. In som e u ses of the re s u lts and in com parison betw een p a rtic u la r p erio d s or econom ies th is could d isto rt the m eaning of the fig u res. The difficulty about the second type of duplication is th at the am ounts depend on the p urpose for w hich national product e stim a te s a re being used and on an a ssessm e n t of the usefu ln ess of individual goods and se rv ic e s. Let us take firs tly those products w hich, w hile not a p a rt

90 82 National Income E stim ates of norm al business costs of production, may not rep resen t a final use of goods and serv ices. The expenditure by public authorities and consum ers is sim ply the total value of goods and serv ices purchased. Some of these products may not n ecessarily add to real incom es. Item s such as the cost of travelling to work, types of governm ent regulatory or adm inistrative functions necessary to m aintain peace and o rd er, and special clothing required for a p articular type of job could m ore properly be regarded as a cost of living, or earning incom e, than as contributing to real incom e. Sim ilarly, some of the costs charged by businesses against sales may provide satisfactions to em ployees and thus add to th eir re a l income. These include not only m edical serv ices, clothing, and other am enities provided by businesses for th eir em ployees, but also various 'fringe b en efits', such as lavish expense accounts, and personal telephone calls on business phones. Up to the p resent tim e, however, there has been no agreem ent as to the way in which the item s which are truly final output can be distinguished from the interm ediate costs. P a rt of the trouble lies in deciding on the final goal of economic activity. If the em phasis is on the benefits to consum ers, then the problem is one of lim iting the item s of personal consumption expenditure to elim inate those which a re p rim arily costs of earning incom e, such as travelling to work, and of including in the public authority expenditure only the item s which have a counterpart in the residual consumption expenditure. For practical purposes, this approach com es down to making adequate allowance in com parison between two econom ies for the effect of institutional o r geographic d ifferen ces on the utility of item s of p erso n al consum ption expenditure. Other end goals may, of course, be p referred. Thus, for exam ple, Eckstein (1961:20) drew a distinction between m ilitary expenditure in the W estern and in the Soviet type econom ies. In the la tte r case m ilitary production and expansion of industrial capacity for m ilitary production are accorded a central place in the planning p ro cess. In an extrem e statem ent of the case, one could say that the central task of a Soviet-type economy is to allocate reso u rces in such a way as to m axim ise the ra te of growth in m ilitary production. In these te rm s, the appropriate definition of final product might be m ilitary consumption and capital form ation, while household consumption could be considered an interm ediate product.

91 M easu res of Total Output 83 He com m ents la te r (p. 21) that it would seem that in m easuring the com parative perform ance of different types of econom ies, much can be said for using three alternative concepts of final product, each directed tow ards a different aspect of com parison. These could be designated as the w elfare, the power, and the com prehensive definitions respectively: the firs t refe rrin g to household consumption and capital form ation that may contribute to it, the second focusing on governm ent expenditure and capital form ation associated with it, and the third encom passing both. F or the purposes of this study, the third concept is m ost relevant since our prim ary in terest is in the stru ctu re of China's Mainland economy, in the prevailing pattern of reso u rce allocation, and in the magnitude and interrelationships of the broad financial flows. If, on the other hgnd, it is assum ed that it is not possible to consider the end goals of economic activity other than in te rm s of the types of output produced, the problem then is sim ply one of making proper allowance for the costs of producing goods and serv ices. However, this ra ises the second difficulty that some of final output as conventionally m easured may m ore properly be a p art of the cost of producing business incom e. This output would include, for exam ple, some expenditure on ro ad s, expenditure on the regulatory activities of public authorities for m aintenance of law and o rd er, negotiation of international trad e agreem ents, and so on. If not undertaken by the public auth o rities, this expenditure would be incurred by busm ess firm s and charged as a cost to the value of sales. But there is no clear way of distinguishing between this expenditure, which facilitates business production, and that which is available for final use. T here seem s little doubt, how ever, that some item s of national product a re properly a cost of production and that this expenditure is relatively m ore im portant the m ore developed the economy. As a re su lt the conventional m easure of national product is likely to lead to som e overstatem ent of the real product of countries with high incom es per head and to an upward bias in the rate of growth over tim e of real product. In discussions of the duplication in national product arisin g from these products there appears to be a genuine difference of opinion as to the amounts involved and the way this affects the use of national product figures for different purposes and in com parisons between d ifferen t econom ies. One conclusion which has em erged from th is

92 84 National Income E stim ates discussion is the need for a detailed classification of national product by function to show separately, for exam ple, the goods and serv ices which could represent costs of production and 'n ecessary ev ils' in government and personal consumption expenditure. Thus, interpretation of the meaning of figures of national product will vary according to the purpose of the com parison and the significance of the output used for different purposes. The third type of deduction to be considered is the depreciation of capital a sse ts as a re su lt of th eir use in curren t production. Since capital a sse ts are p a rt of final output (as investm ent expenditure) when produced, allowance is needed for th eir depreciation in producing other output in o rd er to avoid double-counting. To the extent that p art of investm ent expenditure is required sim ply to replace a ssets used up in curren t production, including the full expenditure in national product would overstate the level of economic output. The depreciation charge is not an identifiable value of output or cost and is imputed from the value of capital a sse ts and the estim ated loss in value due to production in each y ear. It is an attem pt to estim ate what would be the value of output needed in each y ear to replace the depreciation of capital. Because of the difficulties in defining what is meant by capital used in cu rren t production and in m easuring the amount, national product is often used in the gross sense as an approxim ation to net national product. The official A ustralian estim ates of national product at constant p ric e s, for exam ple, are shown only on a gro ss b asis and are used to serve purposes w here a net concept would in principle be m ore appropriate. It has also been argued that the true value of depreciation is just as likely to differ m ore from any statistical estim ate made than from zero, and that it is th erefo re m ore reliable to adopt a value for depreciation of zero. D epreciation appears to be a fairly constant proportion of national product and the use of the gross concept is not likely to lead to large e rro r in co m p ariso n s of national product. The depreciation estim ates should show the value of capital used up at cu rrent p rices. It is to be distinguished from financial provisions for depreciation made in arriv in g at business profit and from the change in total capital stock from one y ear to the next. The financial provision is a way of w riting off past capital expenditure against income in o rd er to recoup the original expenditure on the a sse ts.

93 M easu res of Total Output 85 A ssets are usually w ritten off at th eir original cost p rice, instead of at cu rren t replacem ent p ric e s, and the allocation of depreciation charges is usually made in a fairly a rb itra ry fashion, generally by a method which does not lead to any big distortions in profits. Financial provisions are often also affected by taxation requirem ents which may have little relationship to economic considerations. D epreciation is an elem ent of the change in capital stock, but in addition capital stock will a lte r as a re su lt of new investm ent and capital lo sses. The la tte r item is not deducted in the m easurem ent of national product on the grounds that it is not a cost incurred in productive activities. Item s of norm al re p a ir and m aintenance a re usually excluded from investm ent and hence also from depreciation allow ances. Thus depreciation is the using up of a sse ts after allowance has been made to m aintain them in a reasonable state of efficiency. In national income estim ates of Scandinavian countries, however, all re p a irs are capitalised and then w ritten off as depreciation, leading to a m ore gross (or g ro s s-g ro s s ) concept of national product and investm ent expenditure. Abnormal re p a irs which a lte r the nature of the capital asset are usually capitalised but the distinction between norm al and abnorm al re p a irs and m aintenance is based on accounting p ractice, which leads to some uncertainty about what re p a irs and maintenance are in fact charged against the cost of sales. The national income account is com pelled to rely on the available figures of business profit, and practice in dealing with re p a irs and m aintenance may vary between different businesses. The main elem ents of depreciation a re w ear and te a r, the decline in value owing to foreseen obsolescence both as a re su lt of the introduction of new and m ore efficient m achines and changes in demand, and accidental damage (U.N. 1964b:32). All of these elem ents contribute to the decline in the value of a sse ts but it is not c lear how far each could be attributed to loss of productiveness due to the use of the asset in cu rren t production. Strictly speaking, only w ear and te ar could be directly attributable to the use of the a sse t in production and part of this could also a rise even if a sse ts w ere not used. By including all of these elem ents in depreciation, th erefo re, a somewhat wider concept is obtained than would be strictly relevant to depreciation as an elem ent of duplication of cu rrent output. The depreciation allowance in national income accounting m ore properly rep resents the amount of cu rrent output needed to m aintain the stock of a sse ts, irresp ectiv e of the causes leading to decline in values. It

94 86 National Income E stim ates has been argued that w ear and te a r is likely to be a m inor p art of the depreciation charge, and to this extent the usual allowances contain only a sm all elem ent of decline in value as a resu lt of cu rren t production. 3 D epreciation is th erefo re largely the re su lt of foreseen changes in conditions of supply and dem and, to g eth er with accidental dam age. The point a ris e s, however, that the changes in conditions of supply and demand could affect values of other a sse ts and that this p a rt of depreciation could in fact be attributed to im proved efficiency of new a ssets (which contribute to obsolescence of existing assets) or increased demand for other products. Thus the effect of including obsolescence as an elem ent of depreciation is to w rite off the fall in the value of a sse ts of business or industry which a ris e s from increased values of a sse ts in other businesses or industries. To a larg e extent, th erefo re, the depreciation estim ates do not m easure a net decline in the productiveness of a sse ts. It may also be queried whether it is c o rre c t to make a deduction from national product for the im proved productiveness of new equipment; on the conventional m easure depreciation will be la rg e r the fa ster the rate of technical im provem ent in investm ent goods. The conventional m easure of depreciation involves estim ating the life of the a sse t, deducting the value for scrap at the end of its life, and w riting off the rem aining value using som e standard method of apportioning annual charges. This may be on the straig h t-lin e method (in which case the sam e amount is w ritten off each year) or the method of reducing balance (in which case the sam e percentage of the value at the beginning of each y ear is w ritten off). An amount is then added in each y ear for accidental damage as it occurs. 4 The valuation of national product The goods and serv ices com prising national product are weighted by prices as the m ost satisfacto ry available m easure of the economic significance of different item s. These may be either the m arket prices or the imputed 'factor co st' p ric e s, that is, m arket p rices 3 Sir Robert Shone (1957) has commented that 'I do not really think it right to think of depreciation occurring because efficiency falls. The efficiency of a plant may well remain substantially constant throughout its life, or indeed increase if managers find better ways of operating it. 1 ^ For a discussion of the effect of using different methods of calculating depreciation, see U.S. Department of Commerce (1963) and Redfern (1955).

95 M easu res of Total Output 87 excluding net indirect taxes. As first proposed by Hicks (1940), the choice between these alternative methods of valuing output is determ ined by the purpose for which the estim ates are to be used. M arket p rices m easure the m arginal utility of goods and services used for final purposes of consumption, investm ent, and net exports, while facto r c o st shows the m arginal productivity of fa c to rs used. The form er method of valuation is appropriate for com parisons of national product used for different purposes, including com parison of levels of economic w elfare, while the la tte r method of valuation is appropriate for studies of levels of production, or the productivity of factors used. However, p rices have only a vague so rt of relevance for these purposes. For exam ple, m arket p rices m easure the utility derived from only the la st unit of p articu lar goods and serv ices acquired and thus leave out the co nsum ers1 su rp lu s or additional utility derived from other purchases. This m eans that com parisons of the total values of expenditure on each product provide only a rough m easure of the relativ e utility obtained. As w ell, the use of p rices to com pare utilities of goods and services assum es that p u rch asers adapt th eir expenditure to equate the satisfactions derived from the goods and serv ices (which involves perfect knowledge of the p rices of different products), that wants are constant between individuals and at different periods of tim e, and that income is fairly evenly distributed. ^ The la st qualification is probably the m ost im portant in practice. It m eans that p rices can only be in terp reted as relating to a p articu lar income distribution, and a substantial change in income distribution will alte r com pletely the significance of the p rices. In this case, com parison using a set of p rices will need to be confined to a p articular group the real income of which is approxim ately unchanged. At the sam e tim e, the relevance of any set of p rices as weights w here income is distributed unequally may be doubted. It m eans, for exam ple, that the rich are given g re a ter weight sim ply because they are rich. Sim ilar difficulties apply in the interpretation of factor cost as a m easure of the m arginal productivity of reso u rces used in production, as a re su lt of the lack of perfect com petition between business firm s and in the supply of factors of production. But in addition there is ^ There is an extensive literature on this subject. A useful summary of the significance of market and factor prices is given in Ohlsson (1953).

96 88 National Income E stim ates some doubt as to the extent to which, even if in practice factor costs did m easure the m arginal productivity of factors used, it is appropriate to deduct the full amount of indirect taxes in m easuring national product at factor cost. It has been argued that the distinction between indirect taxes as a cost to businesses and other taxes as tra n sfe r paym ents is by no m eans c lear-cu t. This distinction is based on the assum ption that indirect taxes are shifted forw ard and are thus properly a cost of goods sold while other taxes are an allocation out of profits or personal incom e. In fact, however, some types of indirect taxes may m ore properly be regarded as a tra n sfe r paym ent by businesses or p ersons, including, for exam ple, the paym ent by persons of various types of licences and fees usually included in indirect taxes while income taxes paid by businesses may in some circum stances be passed on. The problem is that it is alm ost im possible to determ ine in practice how far different types of business taxes are in fact passed on. But it is clear that while all of personal income tax and indirect taxes paid by persons are tra n sfe r paym ents and not a cost of business production, the great bulk of business paym ents of indirect taxes is a cost of production. The position is less certain with reg ard to business paym ents of income taxes, but even here the weight of theoretical arid statistical evidence is that changes in ra te s of tax affect company profits after taxes m ore significantly than before tax profits (National Bureau of Economic R esearch 1958:52). There is also som e doubt as to w hether the factor cost concept does really show the value of output in te rm s of the m arginal productivity of reso u rces used. From the point of view of the individual business the costs which are relevant to business decisions are the payments made directly for factors used, plus the total cost of m aterials. The value of m aterials used is made up of the cost of factors used in producing the m aterials and also the indirect taxes levied at previous stages of production. Thus it has been argued that expenditure on final output at factor cost should exclude only those indirect taxes levied after the final stages of production; other indirect taxes have the sam e relevance in producing final output as the accum ulated sum of paym ents to facto rs of production at interm ediate stages of production. Since these account for the large bulk of indirect taxes it has been further argued that a m ore accurate m easure of national product at factor cost is in fact obtained by including instead of

97 M easu res of Total Output 89 excluding all of in d irec t taxes (Nicholson 1955). Because of these difficulties in both defining and m easuring national product at factor cost there are some grounds for accepting national product at m arket p rices as the m ore useful m easure for production and productivity com parisons. In p ractice there is not a large difference between the changes in the two aggregates and the choice between these aggregates has probably m ore theoretical than practical significance. As well as the difficulty of interpreting the meaning of national product valued at m arket prices or factor cost, there are practical difficulties in estim ating national product in both valuations. In the case of the m arket price valuation, these concern mainly the value of the change in stocks and depreciation and the valuation of non-m arketed production and government expenditure; in the case of the factor cost valuation th ere is the additional problem of estim ating the net indirect tax es on individual item s of final expenditure. In cu rrent p rices for m ost item s of national product the basic sources used show the quantities of goods or serv ices com prising expenditure or output valued at an average price in the period. For the two item s of investm ent on stocks and depreciation, however, the basic data a re not in this form and com plicated methods of adjustm ent a re needed to obtain the equivalent v alues. The main source of information of investm ent in stocks is balance sheet value of stocks held by businesses. The difference between successive balance sheet figures shows the change in the value of stocks during the year. This includes, however, not only the change in the quantity of stocks held, valued at the p rices at which they are acquired (the value of the change in stocks) but also the change in stocks as a re su lt of any variation in the p rices at which opening and closing stocks are valued (stock appreciation). Even if the physical level of stocks was unchanged during the y ear, any variations in p rices would affect the values shown in balance sheets and thus lead to a value of stock change. In o rd er to obtain a figure of the in crease in stocks which corresponds to that of other item s of expenditure it is n ecessary to elim inate this elem ent of stock appreciation from the change in the book value of stocks. The procedures of doing this are described in the next chapter. E stim ates of depreciation shown in business reco rd s are sim ilarly not in the form required for national income estim ates. As has been

98 90 National Income E stim ates pointed out, business provisions for depreciation are usually calculated by w riting off the original cost at which assets are acquired, ra th e r than a cost m easured at curren t p rices. The valuation of some non-m arketed production may be based on the actual p rices of equivalent production sold through the m arket. This is the case, for exam ple, w here the non-m arketed production accounts for only a sm all proportion of total production and it could be assum ed m arket p rices would be unaltered o r only slightly changed if all of the output w ere sold on the m arket. The consumption of farm produce by p rim ary producers in a predom inantly exchange economy is likely to be only a sm all p art of total value of farm production and hence this is usually valued by referen ce to the price of that production which is sold. W here the non-m arketed production accounts for a larg e proportion of total output, however, m arket p rices may be an im perfect guide to the relative values of goods and serv ices. It may then be n ecessary to derive p rices from actual or imputed costs of production, or from the p rices of equivalent goods im ported, or from those of the sam e products in another country. The la tte r procedure is quite often used but it involves the assum ption that the conditions of supply and demand in one country a re equally applicable in the other country. Actual circum stances may in fact be quite different, leading to an artificial portray al of economic activity. M arket p rices are also of doubtful significance in valuing the cu rren t purchases by public authorities (other than by business undertakings, in which case cu rrent purchases are treated as a norm al cost of business and deducted from sales revenue), since in effect the price of this output to final consum ers is nil. This point has supported a good deal of the argum ent as to w hether public authority expenditure should be treated as final or interm ediate product. It is argued, for exam ple, that the absence of any relevant price valuation m eans that this expenditure cannot be com pared with, o r related to, other final expenditure. A way of overcom ing this difficulty has been suggested by Kuznets (1941), and was adopted by him in some early calculations of the national product of the United S tates, in which the expenditure of the public authorities is treated as interm ediate outlay financed by fines and taxes, with the difference between the revenue and expenditure of public authorities, roughly equivalent to the budgetary surplus, being regarded as the final expenditure of public a u th o ritie s. A nother suggestion has been made

99 M easu res of Total Output 91 that the governm ent could somehow be viewed as an ultim ate consum er, the satisfaction of whose needs can be equated with those of the population. The conventional practice, however, is to tre a t the governm ent as a final purchaser of goods and serv ices justifying this on the grounds that the bulk of public authority expenditure does, in fact, rep resen t final output and that no other method of valuation is possible except in term s of the costs of the goods and services acquired. The main problem in the m easurem ent of factor cost is that while m arket p rices are the values at which actual transactions are recorded, factor costs are an imputed price obtained by deducting one elem ent of cost, net indirect taxes, from m arket p rices. Because of the statistical difficulties in distributing the indirect taxes on interm ediate transactions to categories of final expenditure few attem pts have been made to estim ate national product as the sum of item s of final expenditure at factor cost valuations. Instead, estim ates of national product at factor cost are usually derived by deducting the total indirect taxes from the sum of final expenditure valued at m arket p rices o r, m ore directly, by simply adding the paym ents made to factors of production.

100 7 Deflation of National Product P rev io u s ch a p te rs have d iscu ssed the p ro b lem s involved in m e asu rin g national product in c u rre n t money p ric e s. F o r many p u rp o ses, including the com pariso n of levels of national product over tim e and betw een c o u n trie s, e stim a te s a re also needed of the value of n ational product a fte r allow ing fo r d ifferences in p ric e s. This leads to a co m p ariso n in re a l v alu es, o r in te rm s of the quantities of the goods and s e rv ic e s co m p risin g national product. T hese e stim a te s a re often d e scrib e d a s quantity indexes but since they involve the valuation of item s at a co n siste n t set of p ric e s of one b ase y e a r or in one co u n try, and thus include both a p ric e and quantity com ponent, they a re m o re a c c u ra te ly d escrib e d as quantum s e rie s. The c o rre c tio n of e stim a te s in c u rre n t money values for d iffere n ces in the level of p ric e s is re fe rre d to as the deflation of national p roduct. This c h a p te r deals with the question of deflation in in tertem p o ral co m p ariso n s; the following chap ter d iscu sses the revaluation of money values of national product of d ifferent co u n tries to an equivalent level of p ric e s. M ethods of deflation It is u su al to m ake a distinction betw een the m ethod of deflating national product by su b tractin g the quantum of im p o rts from the quantum of final output on the one hand, and by sum m ing the quantum of product of in d u strie s on the other hand. T hese altern ativ e ap p ro aches sim ply follow two of the d ifferent m ethods by which national product is estim a te d in c u rre n t p ric e s, and hence the re s u lts should be the sam e, provided the data used a re co m parable, and a p p ro p ria te techniques of deflation a re used. In p ra c tic e, s im ila r data a re not used and the re s u lts a re d ifferent. The difference is often supposed to indicate in som e way or other the possible e r r o r of the fig u re s. The re la tio n sh ip betw een the two m ethods m ay be indicated as 92

101 D eflation of N ational P ro d u ct 93 follow s. N ational product for the econom y is defined as the final expenditure on goods and se rv ic e s produced, le ss im p o rts of goods and s e rv ic e s. It is a difference betw een the output of the econom y and the input (im p o rts). P ro d u ct of individual in d u strie s is sim ila rly defined a s output le ss input, but in the case of each in d u stry the output c o n sists of the final expenditure of the econom y w hich is a ttrib u tab le to the products of the in d u stry plus sa le s to o th er in d u strie s, and the input includes im p o rts used by the in d u stry plus p u rc h a ses from o th er in d u strie s. Thus the production of each in d u stry c o m p rises com ponents of the fig u res used in p re p a rin g the e stim a te of national production fo r the econom y plus in te r-in d u stry tra n sactio n s. The quantum e stim a te s of national product should be the sam e, w hether d eriv ed by deducting from the quantum of final output the quantum of im p o rts o r by sum m ing the quantum of product of in d u strie s. If the detail of final expenditure and im p o rts used in the two m ethods is id entical, the m ethods used fo r deflation a re the sam e, and th e e stim a te s of the in te r-in d u stry tra n sactio n s cancel out, this lead s to the sam e estim ate of re a l product. Thus the difference betw een e stim a te s derived from th e se a lte rn a tiv e appro aches m ay be due to the use of d ifferent e stim a te s of final expenditure o r to inconsistency in dealing with in te r-in d u stry tra n sactio n s. One lim itatio n of the m ethod of aggregation of industry product is that the av ailab le data of input and output a re generally incom plete. The m ethod of approach should entail deflation of the detail of the output and input of each in d u stry, the difference betw een the output and input re p re sen tin g the in d u stry product. In p ra c tic e, lim itatio n s of data m ake it n e c e ssa ry to use only one in d icato r, usually of output but so m etim es of input, to re p re s e n t changes in industry product, in which case the aggreg ate of quantum of industry products w ill not total to the quantum of national product obtained by the deflation of expenditure e s tim a te s. T his could lead also to e r r o r s in individual in d u stry e stim a te s of quantum of production. An exam ple of the m agnitude of e r r o r s which do a ris e in p ra c tic e from this so u rce in h is to ric a l s e r ie s is provided by a co m p ariso n of the tre n d in product of 'f a r m s in the United States over a fo rty -th re e -y e a r p erio d from 1910 to D uring th is perio d, the quantum of output of 'fa rm s ' in c re a sed about 81 p e r cent, the quantum of input by 324 p e r cent, and the quantum of product by 39 p e r cent (K endrick 1961:347). The rela tio n sh ip betw een input and output s e rie s se e m s, how ever,

102 94 National Income E stim ates to vary between industries and there is some evidence that the two serie s are likely to move closely together in the case of m anufacturing industry. As w ell, the deflation of inputs and outputs could in some cases produce strange re su lts, owing to changes in relative p rices and quantities. Negative real product could be obtained in extrem e cases. For this reason, and because of lim itations of data, it is generally preferable to rely on single indicators to represent changes in industry product except w here, as in the m easurem ent of farm product in the United States, there are quite substantial differences in the ratio of m aterials used to output. A deflated serie s shows the quantities of national product in each y ear, expressed in the p rices of a given base y ear. Ideally, each item of output is taken to com prise a quantity and price component, and the process of deflation involves substituting the base y ear prices for cu rren t p rices. This can be done either directly, by multiplying quantity data by the base y ear p ric e s, o r indirectly, by dividing the value figures by a price index. Where properly constructed price series exist, it is usually the practice to p refer the indirect approach of price deflation to the d irect revaluation of quantity data. There are two main advantages in using price deflation com pared with d irect revaluation. In the firs t place, the revaluation of components of output is norm ally based on price or quantity indexes covering only a p art of expenditure or product; that is, a sam ple of prices or quantities is used and it is assum ed that this sam ple is representative of all item s in the p articular component of output. In p ractice, there is usually less variation in the movement of p rices than of output of sim ilar products. In the second place, the specification of p rices is likely to be m ore detailed than that of quantities, with the resu lt that a price se rie s is m ore likely to relate to sim ilar item s than quantity data. F or exam ple, in dealing with the revaluation of output of flour, a price serie s used to deflate value of output may be specified as a grade of flour, thus providing an answ er in term s of quantity of the specific grade, but a quantity serie s used in d irect revaluation may com prise the total output irresp ectiv e of grade, and the deflated series will accordingly not reflect change in the grade of output. One difficulty in using p rice se rie s in deflation a ris e s, however, from the fact that norm ally the quantum series are needed with a base y ear price weighting, and this is obtained in deflation only by the use of a currently weighted price index. These a re ra re ly com piled, with the re s u lt th at, in the case

103 D eflation of N ational P ro d u ct 95 of those com ponents of national output which a re deflated, the w eighting of quantities is in te rm s of c u rre n t p ric e s. A m ore su b stan tial problem is th at fo r a g re a t m any of the item s in national product, the values cannot be divided betw een a p ric e and quantity com ponent. In these c a ses it is n e c e ssary to im pute in som e way a quantity equivalent to the value. T his is the position fo r much of the output of se rv ic e in d u strie s and for cu sto m -b u ilt pro d u cts. In the case of the cu sto m -b u ilt pro d u cts it is so m etim es p ossible to im pute p ric e s from those of stan d ard ise d products of the sam e industry. In som e in d u strie s (for exam ple, shipbuilding and a irc ra ft m anufacture), how ever, alm o st all of the output co m p rise s unique products and in th is case the usual p ro ced u re is to com bine co st item s in o rd e r to re c o n stru c t the probable p ric e change of a p a rtic u la r m odel if it had continued to ex ist. An index of co st is usually com piled by com bining ap p ro p riately w eighted m a te ria l c o sts and earn in g s of the types of m a te ria ls and labour used in production. An adjustm ent m ay be m ade also for changing p ro fit m a rg in s. A m ain defect of th is approach is the neglect of productivity change. F o r the sam e reaso n the physical output cannot be identified and m e asu re d, n eith er can output p e r unit of input. The re s u lt of th is om ission is to lead to a downward b ias in the quantum indexes. M ethods have been proposed to overcom e th is difficulty, but they all su ffer from som e lim itation o r o th e r. F o r exam ple, an attem p t m ay be m ade to use d ire c t m e a su re s of p hysical output - such as tonnage in the case of ships o r square feet of space in the case of buildings. H ow ever, the value p er unit m ay v ary consid erably accord in g to the type of product (coastal or p a ssen g e r ship, factory o r office building, e t c.), so that a change in the com position of output m ay cause a co n sid erable change in unit value. In the official e stim a te s of real product of the building industry in B rita in, to overcom e the lack of any allow ance for productivity change a p ro ced u re is used which involves adjusting labour input by a change in the ra tio of m a te ria l to labour c o sts. If th is ratio fa lls, it is assu m ed th at productivity has in c re a sed and the index of labour input is adjusted upw ards. This p ro ced u re ignores ra te s of re tu rn to capital and involves som e s ta tis tic a l im p recisio n. The deflation of expenditure on output of the se rv ic e in d u strie s r a is e s p a rtic u la rly difficult p ro b lem s. In the f ir s t place, for a la rg e p a rt of th is output it is difficult to envisage any re a lis tic unit of

104 96 National Income E stim ates transaction. That is, it is not clear just what is being acquired. This is the case with many of the serv ices of public authorities which are supplied without charge to the final consum er. In this instance the usual practice is to revalue the expenditure by the use of quantity or price series of inputs used. But even where there is a charge for the service th ere are difficulties in dividing the values of expenditure into a price and quantity component. This is basically a problem of identifying the attrib u tes of the service provided. In the case of expenditure on goods the attributes are largely (but not entirely) composed of a tangible product, and quantity com parisons can be obtained by simply com paring the num bers of these products. (Where specifications change a difficulty is encountered in com paring the old with the new product, and the method of doing this - of allowing for quality change - is discussed in a la te r section of this ch ap ter.) For item s of serv ices, however, the attrib u tes com prise different types of services which somehow need to be combined into a unit of quantity'. In p ractice, quantum serie s of expenditure on the output of service industries are frequently based on a main attribute of the serv ice. F or exam ple, the real expenditure on transp o rt serv ices may m easure only the num ber of passen g ers c a rrie d, on education the num ber of pupils enrolled, on h aird ressin g the num ber of h aircu ts, and on funerals the num ber of deaths. It is obvious that changes in the circum stances under which these serv ices are provided could lead to an e rro r in the quantum se rie s. Finally, reco u rse may be made to some general m easure of p rice change, such as a p rice index of consum ption goods, to rev alu e expenditure to constant p ric e s. The use of these methods of m easuring the quantum of expenditure on serv ices is generally thought to lead to some understatem ent. Thus com parisons of the real expenditure on goods and serv ices may be affected by the procedures of m easurem ent used. However, few attem pts have been made to calculate the magnitude of the e rro r involved and such attem pts as have been made have been n ecessarily based to a large extent on subjective judgment. 1 Investm ent in stocks and d epreciation Some components of the investm ent in stocks at current p rices are e stim ate d by m ultiplying the change in the quantities of goods in stock l See, for ex am p le, H aig (1966).

105 D eflation of N ational P roduct 97 during the year by an estim ated average price. This is the case, for exam ple, with a large part of the investm ent in farm stocks shown in the official A ustralian national income estim ates. These item s of stocks are revalued to constant p rices by simply substituting base y ear p rices for the cu rren t p rices. But for the m ajor p art of investm ent in stocks the current value estim ates a re based on annual balance sheet values of stocks held. In estim ating the change in stocks at constant p rices from this inform ation it is first necessary to revalue the book values at each balance date to show the quantities of stocks held at base y ear prices; the estim ates of investm ent in stocks are then obtained as the difference between successive balance sheet values at constant p rices. While the method of making the revaluation of balance sheet figures is straightforw ard, there are im portant statistical lim itations in carrying this out in practice. The revaluation of balance sheet values to constant p rices involves firstly, identifying the goods held in stock, and secondly, applying appropriate price indexes to the values of the goods. The identification of goods in stock is by no m eans c lear-cu t. In the case of m anufacturing in d u stries, for exam ple, goods held in stock com prise raw m aterials, work in p ro g ress, and finished goods. While finished goods will be sim ilar to the output produced by each industry, stocks of raw m aterials and work in p ro g ress will reflect the goods purchased from other industries. For both m anufacturing and trading industries the flow of goods as sales or purchases will not n ecessarily rep resen t the type of goods held in stock, owing to variations in the ra te s of stock turnover. The methods of price deflation are sim ilar to those used to revalue other item s of final expenditure, although the available price serie s are not usually sufficiently detailed to enable a proper weighting to be made for the com position of the goods held in stock. The m ajor problem is encountered, however, in determ ining the p rices at which goods are valued. Balance sheet values are usually at the lower level of cost or m arket p rices. Where p rices are falling, values will accordingly tend to be w ritten down to m arket p rices, in which case values will be those at the date of the balance sheet. But where valuation is at cost, p rices will depend partly on the method of charging goods against sales and partly on the rate of stock turnover. If, for exam ple, goods are charged against sales using the method of first-in first-o u t, stocks of raw m aterials will be valued at the p rices of the m ost recent acquisitions. In this case, raw m aterials stock will be valued at p rices of purchases in the num ber of months p rio r to

106 98 National Income E stim a te s the balance date given by the ratio of annual purchases to the value of stocks held, and work in p ro g ress and finished goods will be valued at p rices during e a rlie r months. The slow er the rate of stock turnover, the longer will be the period p rio r to balance date for which p rices will be relevant to the revaluation of the goods held in stock. Other methods of charging goods to stock will lead to a different method of valuing stocks and thus to the selection of different p rices needed for the revaluation. The inform ation needed to c a rry out these steps is generally not available and much sim pler procedures a re usually applied. In the revaluation of non-farm stocks in the official A ustralian estim ates of national product, for exam ple, p rice indexes which correspond only approxim ately to the types of goods in stock are applied to the total balance sheet values of stocks held by m anufacturers, tra d e rs, and other industries separately, and the period of turnover is roughly estim ated from the values of total stocks held and annual output or sales. There is alm ost no inform ation available about methods of charging goods to production, balance date revaluations, or the com position of goods held in stock, and the wholesale and re ta il price indexes used are not strictly appropriate for the purpose of revaluing the goods in stock. As in the case of investm ent in stocks, estim ates of depreciation at constant p rices cannot be derived by sim ply applying a price index to figures of depreciation shown in business reco rd s. This is chiefly because business figures of depreciation represent the value of asse ts w ritten off at h isto rical cost and in any one y ear a sse ts will be valued at a num ber of different levels of p rices depending on the date of acquisition. It is n ecessary, th erefo re, to build up a serie s of estim ates of capital stock valued at equivalent p rices. This may be done either by accum ulating past investm ent outlays at constant p ric e s, allowing for past depreciation and capital changes, or by directly revaluing the stock of a sse ts to constant p rices. Both methods have been used in obtaining estim ates of depreciation at constant p rices in overseas calculations of national product at constant p rices. A difficulty a rise s over the interpretation of the values of capital stock at constant p rices and the meaning of the derived depreciation figures as a resu lt of deficiencies in the methods of adjusting investm ent or capital stock to constant p rices. Strictly speaking, the price serie s used to deflate the curren t values of investm ent expenditure or capital stock should take into account any change in the

107 D eflation of N ational P ro d u ct 99 p roductiveness of equipm ent as a re s u lt of tech n ical change. In p ra c tic e, how ever, it is difficult if not im possible to allow for changes in the productiv en ess of equipm ent o r even in the technical m ethods of producing identical equipm ent, and the p rice s e rie s used for deflation a re g en erally b ased on data of c o sts of production. This m eans th at the d eriv ed d epreciation fig u res sim ply show the capital used up. valued not at the co st of rep lacin g a s s e ts of equivalent efficiency but at the co st of rep lacin g s im ila r a s s e ts, often without allow ing for any tech n ical im provem ents in th e ir m ethods of production. This leads to an o v erstatem e n t in the annual e stim a te s of d epreciation calculated from capital stock values up to the date of the b ase y e a r used in the constant p rice e stim a te s, and, because future cap ital stock fig u res w ill be rela tiv e ly undervalued, to an u n d erstatem ent of d epreciation calcu lated from capital stock a fte r the date of the b ase y e a r. It h as been m entioned previously th at th e re is som e doubt as to w hether obsolescence should p ro p erly be re g a rd e d as an elem ent of d epreciation charged ag ain st c u rre n t output. To the extent th at th e re is a tech n ical reaso n for thinking th at dep reciatio n is o v e rsta te d even on the conventional m e asu re m en t, th is could tend to support the contention th at in p ra c tic e national product m e asu re d before any deduction is m ade for dep reciatio n is, in fact, a m o re useful m e a su re of net national product. The base period A fu rth e r point in revaluation m ethods is the choice of the base y e a r. The change shown by a quantum s e rie s will usually differ according to the base y e a r used to value the quantities of goods and s e rv ic e s, owing to the fact that re la tiv e p ric e s change over tim e. In in te rte m p o ra l co m parison the e a r lie r the base perio d the la rg e r is the change in the quantum of national product; the item s of product which in c re a se m ost tend to show below av erage p ric e in c re a s e s, while those that in c re a se le ss tend to show above av erage p ric e change, w ith the re s u lt that the item s which show the la rg e st in c re a se have a h ig h er re la tiv e p ric e w eighting the e a r lie r the base y e a r. T his is shown in the official A u stralian e stim a te s of national product which a re com piled using two a lte rn a tiv e base y e a rs /54 for a s e rie s from 1948/49 to 1959/60 and 1959/60 for a s e rie s com m encing at 1953/54. In the s ix -y e a r overlap period the fig u res b ased on the e a r lie r y e a r in c re a se by 2-1 p e r cent, o r an annual ra te of 0-3 p e r

108 100 N ational Incom e E stim ates cen t, fa s te r than those based on the la te r y e a r. T his d ifference is not, how ever, a tru e reflectio n of the effect of using different base p erio d s; only m a jo r item s a re affected in the reb asin g and m uch of the individual detail is com bined using p ric e s of a s im ila r p erio d in the two s e rie s. It is likely that the tru e d ifference, if it could be calcu lated, would be som ew hat la rg e r than that shown by th is com parison. In p rin cip le, th e re is no way of overcom ing the effect of using d ifferent base y e a rs, e ith e r by com piling one s e r ie s (obtained by averaging changes re su ltin g from the use of different base periods) o r by re fe re n c e to the n eeds of a p a rtic u la r problem. It does not seem c o rre c t, as so m etim es proposed, that v ery often a re c e n t valuation b a se, which re fle c ts the technological re la tio n s and b u y e rs' p re fe re n c e s th at obtain in the actual econom y, w ill seem m o re rele v an t than som e rem o te base which re fle c ts econom ic conditions that have long sin ce ceased to ex ist. (Jaszi 1961:325) R a th e r, the p a tte rn of p ric e s of one y e a r used in w eighting quantities w ill re la te only to the econom ic conditions in th at y e a r and th e ir use in other p erio d s im p lies that b asic fa c to rs of supply and dem and - which affect re la tiv e p ric e s - have rem ain ed unchanged. T here may be a good deal of econom ic significance in changes in re la tiv e p ric e s betw een p e rio d s, which could affect the in terp reta tio n of the m eaning of s e rie s b ased on different p erio d s. Hov/ever, for p ra c tic a l p u rp o ses, the a lte rn a tiv e s e rie s may be thought of as showing one so u rce of e r r o r (that due to the m ethods of su m m a risin g the basic data into to tal quantum s e rie s) which m ay or m ay not lim it the usefu ln ess of the re s u lts fo r any p a rtic u la r problem. Q uality change The problem of quality change a r is e s w henever th e re is a change in specification of goods included in a p ric e index used fo r deflation, o r in a quantity s e rie s used in the d ire c t rev alu atio n. In o rd e r to have a continuous s e rie s a m ethod has to be devised to allow for the change in specificatio n. The p ro c e d u re s adopted affect the m agnitude of the resu ltin g quantum s e r ie s. While th e re is no ag reem en t about the way in which quality change should be allow ed fo r, th e re is fa irly gen eral ag reem en t th at the p re sen t p ro c e d u re s do not n o rm ally m ake full allow ance; th at is, th e re is an u n d erstatem ent in existing s e rie s of national product. The extent of the u n d erstatem e n t and its re la tiv e

109 Deflation of National Product 101 im portance at different tim es and between different countries is just not known. It may be suggested that the existence of quality change will not invalidate com parison between periods of sim ilar economic conditions for one country or between countries at approxim ately the sam e stage of economic developm ent, provided that there a re no obvious reasons why the allowance made for quality change in the p rice or quantum indexes should vary. But even in these circum stances there is cause for some reservation. The nature of quality change may be a sm all im provem ent in a product or the introduction of a com pletely new product. Although the theoretical im plications are sim ila r, the main difficulties in practice in allowing for quality change are due m ore to the gradual but w idespread m odifications to existing products than to the occasional introduction of a new product. Whenever a change occurs in a product which is included in a price o r quantum se rie s, it is necessary to devise a way of linking the price or quantity of the two products. In o rd er to do th is, one requirem ent is detailed specification of the attributes of the products. This is norm ally easier to obtain in the case of price data than quantity data, and it is mainly in the case of price serie s that the question really a ris e s as to the allowance for quality change. In quantum series obtained by d irect revaluation it is usual to ignore the effects of quality change. It should be noted also that inadequate allowance for changes in the proportions in which existing products are supplied may lead to o v eror understatem ent of im provem ent in the quality of goods and serv ices. But this is a different problem from that which a rise s from the change in qualit}^ of an individual item ; it is a question, in the case of price indexes, of sampling and of the index num ber problem, and in the case of quantum se rie s obtained by d irect revaluation, of the detail of specification and the index num ber problem. T here are three ways in which item s which differ in specification may be related. F irstly, they may be com pared in term s of relative costs of production. Secondly, an attem pt may be made to allow for the effect of changes in specifications on the utility or efficiency of the product. Thirdly, they may be related by th eir relative prices. C o sts. Com parison in term s of relativ e costs is an attem pt to provide an objective basis for the m easurem ent of changes in real product. Any other method of com paring products involves in some

110 102 N ational Income E stim a te s way or other a subjective assessm en t of the attributes which affect the usefulness of products. 2 D ifferences in costs can be m easured in term s of quantities of factors of production used, thus avoiding the need for subjective evaluation. The p ractical effect of this method of com paring products is to exclude from the m easurem ent of re a l product changes in quality which do not resu lt from altered methods of production. This approach has been advocated by Milton G ilbert and adopted in com parisons of the real product of m em ber countries of the Organisation for European Economic Co-operation (OEEC). This method is often the p ractical re su lt of the usual methods of allowing for quality change, and G ilb ert's argum ent provides a justification for these estim ates. N evertheless, for most purposes of com parison of real product it does seem desirable to attem pt to make allowance, however im perfectly, for the change in the quality of goods and serv ices. Over tim e, this is an im portant aspect of economic p ro g ress and between countries it could lead to an im portant difference in the levels of real output and income. Equivalent quality. An attem pt may be made to allow for changes in specifications of products as they affect the attributes of the goods and serv ices. This may be done either in a subjective fashion o r, as has been proposed recently, by use of m ore objective c rite ria. In the form er case, rough weights a re attached to various selected specifications (such as durability of rubber ty re s, h orse-pow er, and m echanical im provem ent of m otor vehicles), and changes in these are allowed for according to the relative im portance of the weights. In the second case, valuation of selected specifications is made either by the tim e se rie s analysis or by applying objective te sts to products to see how w ell they m eet the purpose for which they are used. Once the valuation has been made for each specification, new m odels can be directly related in term s of these given specifications. There a re, in practice, three lim itations to this approach. F irstly, the method cannot be used to relate an entirely new product to an existing one. Secondly, only a lim ited num ber of specifications can be handled in this way and th eir selection is based on subjective facto rs. Thirdly, 2 'It is essen tial to draw a d istin ctio n betw een w hat we have c a lle d eco n o m ic and noneconom ic differen ces in q u ality. S ta tistic a l adjustm ents can be m ade for th e form er w hereas none c a n be m ade for th e la tte r since th ey are essen tial n o n -q u a n tita tiv e in c h a ra c te r. We c a n define a n o n -eco n o m ic change in q u ality as one w hich com es about through the g en eral im p ro v em en t of scien ce and the arts but w hich does n o t involve any increase in production costs' (N ational Bureau of E conom ic R esearch 1961:260).

111 Deflation of National P roduct 103 the valuations derived in this way will represent relative costs only, since consum ers' valuation will be reflected in quantity purchased ra th e r than price. P rices com parison. If a change is made in the specification of a product without a corresponding price change, the new product is sim ply treated as equivalent to the old. If there is a price change, the assum ption is made that the difference in price does or does not mainly rep resen t a quality change. In the fo rm er case the new and old products are equated in term s of the respective p rices, in the la tte r case the new and old products a re treated as sim ilar and the p ric e difference as a genuine p ric e change. If there is a period of overlap, when both the old and the new com m odities are available, it is usual to tre a t the price difference as a m easure of quality change, since th ere is a presum ption that the valuations of the m arket do adequately reflect quality change in the sense that p rices are proportional to m arginal u tilities. If there is not a period of overlap, there is no p articular a p rio ri reason for choosing either method, but it may be noted that w here prices are changing a bias is introduced into the index by eith er of the alternatives. Thus, in tim es of risin g p rices, if the difference is treated as a quality change the price change will be biased downwards, and if regarded as a price change, there will tend to be an upward bias in the index. The assum ed understatem ent of quality change in the m easurem ent of national output is due to three main reasons. F irstly, there is the w idespread use of the sim ple price relative method of incorporating altered products into price indexes. This is likely to lead to some overstatem ent in the price indexes used for deflation because it may be assum ed that the new product would not have been introduced unless th ere had been an im provem ent in product. Secondly, new products are likely to be introduced into a price index at a relatively late stage, and p art of the benefit of the early price fall is likely to be om itted. P ric e s of new models tend to fall until they obtain m arket acceptance, and an upward bias may be introduced into the index unless they are priced at an early stage. The re v e rse bias may exist in other circum stances w here, for exam ple, products a re introduced at artificially low levels to obtain m arket acceptance, and are then raised. Thirdly, if an attem pt is made to allow for quality im provem ent it is likely that the allowance will be made on the basis

112 104 National Income E stim a te s of some only of the c h a ra c te ristic s of the product. Offsetting these effects, there are also reasons why quality change can be overstated in price indexes, p articu larly, for exam ple, in w artim e or where there are special reaso n s for m aintaining fixed p rices. Thus, while it may be presum ed that in practice there is inadequate allowance for quality change in preparing estim ates of national output, the amount of understatem ent is hard to determ ine and it will vary according to both the extent of quality change o ccu rrin g and the m ethods of allow ance in p rice and quality indexes.

113 8 Comparison of the National Product and Income of Countries The international com parison of le v els of national product and incom e lead s to much the sam e so rt of problem s as a re encountered in com paring the re a l national product of one country in d ifferent y e a rs. The two m ain considerations are firs tly, to com pile estim a te s of the m oney values of national product of c o u n tries e x p re sse d in the p ric e s of cu rre n c y units of each country, and secondly, to rev alu e these e stim a te s to an equivalent level of p ric e s. The th e o re tic a l issu e s, involving the concepts of national product and incom e and the techniques used to com pare rela tiv e p ric e le v e ls, a re sim ila r in in ternational and in tertem p o ral co m p ariso n s. T h ere a re few p ractical problem s in obtaining com parable money values of national product, since official e stim a te s b ased on concepts and m ethods proposed by the S tatistical Office of the United N ations a re now available for m ost W estern c o u n trie s, and th ese a re reaso n ab ly consistent. Some m inor d ifferences do ex ist but they can be read ily allowed for. H ow ever, acute p ra c tic a l problem s a re encountered in intern atio n al com p ariso n s in assem bling the sta tistic a l e stim a te s needed for the p ric e revaluation, and these have so f a r la rg e ly lim ited such studies to se v e ra l com parisons made by the OEEC of the re a l product and incom e of m em ber countries for se v e ra l recent y e a rs. 1 T here a re, as w ell, g re a te r difficulties in in te rp re tin g the re s u lts, owing to v a ria tio n s in the econom ic stru ctu re and institutional conditions of the econom ies being com pared. Instead of applying the detailed m ethods of p ric e revaluation now used in in tertem p o ral com parisons, an a ltern ativ e approach is often used to com pare the national product of d ifferent c o u n trie s, based on the revaluation of money figures of national product of co u n tries by the official exchange ra te s. This m ethod of com parison w as used by * G ilb ert and Kravis (1954); G ilbert and A ssociates (1958); Paige and B om bach (1959)- See also H aig (1967). 105

114 106 National Income E stim ates the OEEC in the early years after World War II and it is still one of the approaches used by the Statistical Office of the United Nations to com pare the real product of a num ber of countries in te rm s of U. S. dollars. 2 This method of exchange rate conversion could, however, involve a large m argin of e r r o r, and studies made by the OEEC have shown that in com parisons between the United States and European countries the e rro r could be as large as 50 per cent. The main lim itations of the method a re firstly, that exchange ra te s are strictly only relevant to com pare p rices of goods and serv ices which enter international trad e, and secondly, even for these item s the actual price relationships will differ from the exchange rate as a re su lt of restric tio n s on im ports, the costs of transporting products between countries, and custom s duties and other charges on im ports. 3 The w idespread use in the post-w ar years of artificial exchange ra te s not determ ined by m arket forces and of disequilibrium in the balance of paym ents on cu rren t account are further lim itations of this method. The la tte r difficulty can be overcom e to som e extent by imputing m ore re a listic exchange ra te s, for exam ple, by bringing up to date ra te s applying in a norm al y ear using the m ovem ents in internal indexes of p rice changes. The United Nations Statistical Office uses this technique to estim ate the real product of some underdeveloped countries where exchange rates and international transactions are distorted by governm ent action or chronic balance of payments disequilibrium. Using exchange ra te s based on a m ore norm al period overcom es these difficulties but it does not avoid the m ore fundamental lim itations of any rate of exchange to rep resen t relative p rices. It also involves a fu rth er source of statistical e rro r as a resu lt of deficiencies in available indexes of p ric e change. A lternative approaches As in the case of intertem poral com parisons, the revaluation of national product of countries to one equivalent price level could be made by deflating either figures of expenditure and im ports or values of the product of in d u strie s. These two approaches should give the 2 The estimates are published in the U.N. Statistical Office Yearbooks of National Accounts Statistics. The limitations of the method of exchange rate conversion are described in more detail in, for example, Gilbert and Kravis (1954:14-17).

115 International Com paris on of National P ro d u ct 107 sam e resu lt if they are properly applied. Only one international com parison (made by the OEEC) using both methods is available and these show a slightly different answer.4 The chief reason for this is the use of a single indicator to compare industry products in real term s instead of the deflation of the values of inputs and outputs of industries. The use of the expenditure method by the OEEC resu lted in a relatively higher level of real product for the United States, com pared with that for B ritain, than that obtained by the method of revaluation of industry product. It is not possible, how ever, to isolate the reasons which led to this difference, even though in the OEEC study, using the product approach, it was suggested that there are facto rs in the methods used in the industry (product) study which seem likely to favour the relatively low er-incom e, lower productivity country, and factors in the expenditure study which may favour the United States both as the higher income country, and as the country with m ore com prehensive p rice and quantity data on the consum ption sid e. (Paige and Bombach 1959:16) It was concluded (p. 17) that 'the differences between the two studies do not seem...to be larg e, considering the wide scope of the estim ates and th eir experim ental c h a ra c te r'.5 ^ The expenditure approach has been applied in two studies published in 1954 and 1958 which compared real product and income of European countries and the United States in 1950 and 1955, and the product approach was used in a study published in 1959 of the real output and productivity of Britain and the United States in 1954 and The results of the latter study were extrapolated to 1950, thus enabling a comparison to be made of the results of the product method with the expenditure method, obtained from the earliest OEEC study, of the real product of Britain and the United States. 5 The comparison of the two results is shown in the following table: Indices of Gross National Product of Britain and the United States in 1950 Total and per head (Britain = 100) Total Per head at British atu.s. at British at U.S. prices prices prices prices Industry method Expenditure method Source: Paige and Bombach (1959:15).

116 108 National Income E stim ates The choice between the two approaches depends partly on the availability of data. In principle, as pointed out, the product approach req u ires not only details of final expenditure and im ports used in the expenditure method, but also data of in ter-in d u stry transactions. This leads to the need for a much g re a ter volume of inform ation. In p ractice, however, the com parison of industry product in the OEEC study was made largely by using single indicators of goods produced. N evertheless, while expenditure details and p rices of item s of expenditure are generally available annually, com prehensive production figures are usually available only in occasional censuses of production, which lim its the use of the method to p articu lar y e ars. For other y ears some details of production are available but th ere are many gaps in the figures. In A ustralia annual censuses of production are conducted, but these show le ss detail of individual com m odities than the work sheets underlying the expenditure estim ates published in the official national incom e publication. There are also likely to be g re a ter differences in the composition of production than of final expenditure, and this makes it m ore difficult to com pare item s of production. A la rg e r num ber of item s of production than of expenditure are not likely to be common in two countries; this leads to la rg e r problem s in the imputation of p rices or the g reater use of prices of substitute products. In p articu lar, expenditure patterns are likely to be fairly com parable in the case of countries at sim ilar levels of real income per head while the composition of production could differ substantially as the re su lt of different reso u rces and skills available. To some extent this is offset by the fact that indicators of output are m ore likely to com prise standard types of products than item s of expenditure, thus leading to fewer difficulties in m atching specifications of item s. The two resu lts also provide different break-dow ns of the total national product, both of which a re useful for different purposes of analysis, and im portant in international com parisons. In the expenditure approach re su lts show the use of national product for various purposes. These data a re useful in studies of the use of output to m eet various political o r social objectives, such as defence or health serv ices, for adding to the productive capacity of industry, o r for providing various goods and serv ices related to living standards. Thus the expenditure approach provides a com parison of the differences in the end re su lts of economic p ro cesses of countries.

117 International Com parison of N ational P ro d u ct 109 The product approach divides national product according to the industry of origin and this information is useful in studies of the com parative level of production of industries; divided by the quantity of labour and other inputs used the resu lts provide a com parison of to ta l and labour productivity of indu stries. Up to date several studies have been c arrie d out in the use of the expenditure method and resu lts are available using this approach for a num ber of countries and for several y e ars. The only fu ll-scale com parison using the method of comparison of industry products has been m ade by the OEEC for B ritain and the United States. Interpretation of international comparisons A m ajor problem in interpreting the resu lts of international com parisons is caused by the big variations in relative p rices in different countries, leading to large differences in com parisons based on the country's pattern of prices. This problem is sim ilar to that which a ris e s from the use of different base periods in the case of intertem poral com parisons. But while in com parisons over tim e the different re su lts are not too great to invalidate the use of the re su lts for m ost purposes, in international com parisons the differences have a much g re a ter effect on the meaning of the re su lts. Thus in the previous chapter it was shown that the use of different base periods at an interval of eleven years leads to a difference in A u stra lia's post-w ar growth rate of 2-3 per cent. In international com parisons, the re su lts could be up to 50 p er cent different, or m ore, by using the p rices of different countries to weight expenditure, as shown by the following table.

118 110 N ational Incom e E stim ates Indices of G ross N ational P ro d u ct p e r head of Eight E uropean C ountries (U.S. A. = 100) Country U.S. p ric e w eights E uropean p ric e w eights Exchange ra te s 1950 at 1955 at 1950 a t 1955 a t 1950 a t 1955 at p ric e s p ric e s p ric e s p ric e s p ric e s p ric e s D en m ark B rita in N orw ay B elgium F ra n c e N etherlands G erm an y Italy S o u rce: G ilb ert and A sso ciates (1958:28). The national product p e r head of these co u n tries is 25 to 50 p e r cent hig h er re la tiv e to that of the United States if national product item s a re w eighted by United States p ric e s ra th e r than th e ir own p ric e s. The f ir s t of the OEEC studies included a d etailed com pariso n of the p ric e and quantity ra tio s fo r individual p ro d u cts, and it w as found th at the in v e rse rela tio n sh ip betw een p ric e and quantity applied to all goods, consum ption, p riv ate investm ent, and governm ent p u rc h a ses. With rela tiy e ly few exceptions the p roducts with high purchasing pow er equivalents have a low ranking am ong the quantity ra tio s, while low -p riced products have a high ranking. The m edium purchasing pow er equivalents a re m ore sc a tte re d o v er the range of quantity ra tio s, but even in this case they tend to be concentrated in the m iddle th ird of the product c la sses. In the low er th ird of the quantity ra tio s th e re a re alm o st no product c la s s e s with low re la tiv e p ric e s, while in the upper th ird th e re a re alm o st no product c la sses with high re la tiv e p ric e s. Some of the exceptions to the rela tio n sh ip between p ric e and quantity ra tio s re fle c t m e re ly the national p referen c e for clo se su b stitu te s. (G ilbert and K ravis 1954:52) T h ere a re th re e im plications of the re s u lts shown by th is table. F irs tly, the re s u lts of using different p a tte rn s of p ric e s bring out

119 International C om parison of National Product 111 the fact of negative correlation between p rices and quantities; item s which a re m ore expensive in one country are relatively le ss consumed and thus item s which are relatively less im portant in one country s national product tend to have a higher price weight. This leads to the position that countries are relatively b etter off where the pattern of p rices used to weight quantities of goods and serv ices is that of another country. This implication is sim ilar to that of using different base periods in intertem poral com parisons w here national product tends to show a faster increase the e a rlie r the period used as the base year. Secondly, it also shows that the g reater the difference in income per head the la rg e r is the variation between estim ates based on one or another country's se^ of relative p rices. Thirdly, the e rro rs in using the method of exchange rate conversion a re also strikingly brought out. Again, the method of exchange rate conversion leads to la rg e r e rro rs in com parisons between countries with g re a ter differences in income levels p er head. One reason for this seem s to be that exchange ra te s tend to be m ore closely related to relative p rices of goods than serv ices and serv ices tend to be m ore highly priced than goods in countries at higher income levels. Since capital is relatively m ore im portant the m ore developed the economy, higher income countries tend to have a higher capital-labour ratio and thus a lower price of capital relative to labour. This leads to a lower price of goods and services which involve m ore capital intensive methods of production. The OEEC studies, for exam ple, showed prices of durable consum er goods to be low er relative to the p rices of services in the United States. This dispersion of p rices leads to the question of how the resu lts can be interpreted when m arkedly different answ ers are obtained using p rices of different countries as weights. The OEEC studies have produced single re su lts by geom etrical averaging of relative p rices. But as suggested in the previous chapter th ere is little economic meaning in such a method; m ore properly the reasons for differences in relative p rices are likely to be an im portant factor in assessin g the meaning of com parisons. Little work has, how ever, been done to isolate the causes of differences in relative p rices due to either factor of demand or supply, and the OEEC studies sim ply comment that the im portance of the factors affecting relative p rices cannot be isolated in any system atic fashion', and add that bearing these factors in mind

120 112 N ational Incom e E stim ates w ill help the re a d e r to understand m any of the d iffe re n c e s in the com p ariso n s of re a l product of co u n tries (G ilbert and K ravis 1954:7). ^ The second m ain problem in in te rp re tin g the re s u lts is due to the possib ility that d ifferences in quantities of goods and se rv ic e s in national product m ay be the re s u lt of v a ria tio n s in ta s te s and needs. To the extent that th is is the c a se, it is v irtu ally im p o ssib le to draw any conclusion about levels of re a l incom e from quantities of goods and se rv ic e s consum ed; the populations have d ifferent ta s te s and the significance of products is accordingly d ifferent. It has often been assum ed that d ifferences in ta ste s and needs a re m o re im p o rtan t in in ternational than in te rte m p o ra l co m p ariso n s and th is is an im p o rtan t lim itation in using th e se re s u lts. But re c e n t stu d ies have shown that in p ra c tic e th ese d ifferences a re not likely to be im p o rtan t. The second study by the OEEC attem pted to explain how fa r d ifferences in quantities of goods and se rv ic e s co m p risin g p erso n al consum ption expenditure could be explained by d ifferences in level of incom e o r re la tiv e p ric e s, leaving a re sid u a l to be accounted for by o th er fa c to rs, m ainty ta s te s and needs. M ultiple c o rre la tio n coefficients w ere obtained using as data the quantities consum ed and the p ric e s and incom e le v e ls of the co u n tries included in the study. The re su ltin g co efficients w ere found to be high and not d is s im ila r to those found in dem and studies based on tim e s e rie s in individual co u n tries. O ther stu d ies have also tended to the sam e conclusion - that in fact th e re is little evidence to suggest th at at le a s t in re la tiv e ly developed econom ies th e re is likely to be a su b stan tial difference in ta ste s of the populations (B eckerm an 1966:9). N ev erth e less, th e re a re co m p ariso n s w here ta s te s and needs could obviously vary as a re s u lt of physical d iffere n ces in the clim ate o r o th er geographic fa c to rs. A m o re se v e re w in ter, for exam ple, may lead to higher expenditure on fuel, clothing, and c e rta in types of food, and a g re a te r d isp ersio n of the population m ay lead to la rg e r expenditure on tra n sp o rt s e rv ic e s. In using in tern atio n al co m p ariso n s to m e a su re re la tiv e re a l incom e, th e se item s m ay m ore p ro p erly be tre a te d as c o sts of living than as com ponents of incom e, and adjusting fo r d ifferences in levels of expenditure m ay im prove the e stim a te s fo r th is p urpose. ^ It h as, how ever, also been argued th at fo r every 6 It has been argued by N icholson (1955) th a t these differences in re la tiv e prices co m p le te ly in v alid ate in tern atio n al com parison of re a l product. ' This is sim ilar to the p oint m ade in C h ap ter 6 about the e ffe c t on n a tio n a l product of including item s of cost of living and of using the results for com parisons of re a l incom e over tim e in one country.

121 International Com parison of National Product 113 adjustm ent made which serves to low er the re a l product of one country a corresponding need for adjustm ent could be found in the national product of another country. While estim ates of the real product of B ritain could, for exam ple, exclude the higher expenditure on fuel when used to compare levels of real income in B ritain and A ustralia, there would be a case for excluding from the A ustralian figures expenditure such as that on a ir conditioning, to offset the w arm er clim ate. For this reason it has been suggested that adjustm ents for higher expenditure needed to offset disadvantages of living in one country could lead to an e rro r in the com parison if it is not m atched by corresponding adjustm ents to expenditure in another country. Since these adjustm ents involve a good deal of subjective interpretation it has been further argued that they do not n ecessarily im prove the estim ates for com parisons of real income (Beckerm an 1966:15). The final problem in using international com parisons of re a l product a ris e s from the use of national product, com prising goods and serv ices which are exchanged for money, as a m easure of total economic activity. This problem is quite com parable to that discussed in Chapter 6, and while it exists as a general lim itation of the usefulness of the totals, it is not likely to be a serious one in com paring econom ies having broadly sim ilar levels of real income per head and institutions. In com parisons between countries with widely differing levels of income per head they may, how ever, be im portant. In this circum stance the wide dispersion in relative p rices leading to two substantially different com parisons is likely to be an equally or m ore im portant lim itation of the re su lts. Sources of inform ation It has previously been stated that relative p rices of goods and serv ices are likely to be widely dispersed in different countries. This can be illustrated by the resu lts of a com parison made of real product, income, and relative prices in A ustralia and B ritain in Using quantities of item s consumed in B ritain to weight relative p ric e s, the overall price level of consumption goods (after allowing for the ra te of exchange between the two countries) was 8 per cent higher in A ustralia than in B ritain. But if expenditure is divided into twenty-eight main categories of goods and services it is found that for thirteen of the categories A ustralian prices differ by m ore than 20 per cent from

122 114 National Income E stim ates this average. For some item s relative p rices are very different; p rices of hardw are in A ustralia, for exam ple, are some 70 per cent above those in B ritain and prices of ch em ists' goods are nearly 50 per cent higher, while p rices of cigarettes and tobacco are nearty 40 per cent lower in A ustralia. In order to obtain a reliable m easure of national product it is im portant that the products are broken down into considerable detail before being weighted by the relativ e.p rices. This involves the prelim inary step of dividing national product up to show expenditure (or product) on individual goods and services or groups of item s. Published estim ates of national product of countries show a good deal of this information but not in the detail needed for the weighting of p rices of individual products. Thus it is necessary to supplem ent the published figures; this can be done by using unpublished sources based on working sheets used in compiling official national income figures, but in many instances new calculations are required based on data of production, im ports, and exports, o r from surveys of expenditure of various costs including, for exam ple, household expenditure in q u iries. The next step involved is to assem ble the data of price or quantity needed to revalue the components of national product to equivalent p rices. This is the main problem in international com parison. Not only do data of price or quantity need to cover the main item s of product, but in addition care is needed to allow for differences in sp ecificatio n s of the p ric e o r quantity data available. In the OEEC com parisons the main sources of price and quantity data w ere the collections made by national statistical offices and p rices obtained directly from re ta il sto res and other businesses. F or both sources detailed specifications could be obtained. In the com parison of B ritish and A ustralian re a l product and p rices m ost reliance was placed on p rices collected by national statistical offices and on price catalogues of business firm s. B ecause of the lim ited reso u rces available for the la tte r study it was not possible to undertake the extensive field inquiries c arrie d out by the OEEC, and this procedure could lead to e rro r in com paring p rices of item s, particularly of investm ent goods, which a re not collected to any extent by national statistical offices and which could vary considerably in specifications. It has been mentioned that a problem is encountered in comparing quality of item s w here the available price data do not relate to identical products. The methods of doing this are sim ilar to those

123 International C om parison of N ational P ro d u ct 115 used in in tertem p o ral com parisons in allowing for the effect of quality change. H ow ever, while the conceptual approach is the sam e, p ra c tic a l difficu lties a re likely to be m ore im p o rtan t as a re s u lt both of g re a te r d ifferences in the quality of products betw een co u n tries than over tim e (at le a st between periods which a re not too distant), and to m ore consid erable p roblem s in collecting p rice inform ation. It is m uch e a s ie r to continue the collection of p ric e data from one period to the next than to obtain corresponding inform ation for different co u n tries. As w ell, it is likely that a la rg e proportion of goods and se rv ic e s produced and consum ed in countries w ill be unique p ro d u cts, th at is, products which a re not common to the different co u n tries. In the com parison betw een B ritain and A u stra lia, fo r exam ple, it w as found th at coal is a com m on item of household fuel in B ritain but it is of negligible im portance in A ustralia, and accordingly no equivalent p rice com p ariso n s for th is product could be m ade in the two co u n tries. In handling unique products the OEEC studies deduce the re la tiv e p ric e from that of a different product req u irin g s im ila r c o sts of production, o r obtain com petent opinion from production sp e c ia lists about the equivalent cost (G ilbert and K ravis 1954:86). An altern ativ e approach, which w as adopted in the com parison between B ritain and A u stra lia, is sim ply to use the rela tiv e p ric e s of item s m eeting s im ila r needs as the price com parison. The fo rm e r m ethod follows from the gen eral approach developed by the OEEC of re la tin g p ric e s of products which differ in specification in te rm s of re la tiv e c o sts of production (see C hapter 7). In the la tte r m ethod unique p roducts a re b asically those for which no sensible com parison of quality can be m ade; if co sts of production are not accepted as providing the b a sis for p ric e co m p ariso n s th e re is no a ltern ativ e way of com paring p ric e s except by re fe re n c e to p rices of som e other pro d u cts. Unique products a re not of sufficient im portance as a proportion of national product, how ever, for the method used in th e ir p ric e or quantity co m parison to lead to any significant e r r o r in the o v erall re s u lts. The tre a tm e n t adopted could to a noticeable extent affect the co m parison of som e item s of product.

124 Further Reading The publications noted below a re no m o re than a selection from the available lite ra tu re, but they a re d ire c tly re la te d to the to p ics covered in th is book and provide som e additional discu ssio n of them. C hapters 1 and 2 Brow n (1949). The A u stralian tre a tm e n t of financial e n te rp ris e s is explained in th is a rtic le. Com m onw ealth B ureau of C ensus and S ta tistic s, A u stralian N ational A ccounts - N ational Incom e and Expenditure Downing (1966). K arm el (1963). N ational B ureau of Econom ic R e searc h (1958), p ap er by J a s z i, pp T his p ap er d isc u sse s a num ber of the m ore c o n tro v e rsia l points of concept w hich a r is e in social accounting. Pigou (1946). Stone and Stone (1961). U. N. S tatistical O ffice (1964b). C hapters 3 and 4 C o m m issioner of T axation, Com m onwealth of A u stra lia, Annual R e p o rts, and Supplem ents on Taxation S ta tis tic s. Com m onwealth B ureau of C ensus and S ta tistic s publications, esp ecially the follow ing: A u stralian N ational A ccounts - N ational Incom e and E x p en d itu re. Com m onwealth T axation A s s e s s m e n ts. L abour R e p o rts. M anufacturing In d u stry. Y ear Books of the Com m onw ealth of A u s tra lia. Com m onw ealth of A u stra lia, Incom e Tax and Social S ervices C ontribution A ssessm e n t A c t. U. N. S tatistical Office (1955). C onsiderable d etail on the co n cepts, so u rc e s, and m ethods used in the 116

125 F u rth e r R eading 117 national incom e e stim a te s of B ritain and the U.S.A. is available in th e follow ing: U. K. C entral S tatistic al Office (1956). N otes which help to brin g th is publication up to date a re published in the annual Blue Books on national incom e and expenditure. U. S. D epartm ent of C om m erce (1958). C om m ents on som e asp ects of the accounts also ap p ear from tim e to tim e in the re g u la r issu e s of the Survey of C u rren t B u sin ess. Some of tne m o re im p o rtan t d ifferences between the concepts used in the e stim a te s of a num ber of co u n tries a re set out in U.N. S tatistical Office (1964a). C hapter 5 Com m onwealth B ureau of C ensus and S ta tistic s, Q u arterly E stim a te s of N ational Incom e and E xpenditure. International A ssociation fo r R e searc h in Incom e and W ealth (1964). A sh o rt account of so u rc e s and m ethods used in the A u stralian official q u a rte rly e stim a te s is given in H aig (1961). Some changes have been m ade in m ethods of estim atio n se t out in th is p ap er, p rin cip ally in the use of a new su rv ey of com pany p ro fits to estim a te com pany incom e. C hapters 6 and 7 G ilbert (1961). J a s z i (1961). K uznets (1951, 1961). N ational B ureau of Econom ic R e searc h (1958, 1961). Studensky (1958). Includes a com prehensive account of the developm ent of national incom e e stim a te s since the seventeenth century. U.N. S tatistical Office (1964b). Von H ofsten (1952). C hapter 8 B eckerm an (1966). G ilbert and K ravis (1954). Haig (1967). P aig e and B om bach (1959).

126 List of References cited B eckerm an, W. (1966). International C om parisons of R eal Incom e. P a r i s. Brow n, H. P. (1949). Some asp ects of social accounting - in te re st and b an k s', Econom ic R eco rd (August supplem ent) 25: B utlin, N.G. (1962). A u stralian D om estic P ro d u ct, Investm ent and F oreign B orrow ing, /39, C am bridge. C o m m issioner of T axation, Com m onwealth of A u stra lia. Annual R ep o rts, and Supplem ents on Taxation S ta tis tic s, C an b erra. Com m onwealth B ureau of C ensus and S ta tistic s, Com m onwealth of A u stralia. A u stralian N ational A ccounts - N ational Incom e and E xpenditure (annually). C ensus of R etail E sta b lish m e n ts and O ther S erv ices (1962). C ensus of the Com m onwealth of A u stralia (1961). Com m onwealth T axation A sse ssm e n ts (annually). Input-output T a b le s. 1958/59 (1963). Labour R ep o rts (annually). M anufacturing In d u stry (annually). Q u arterly E stim ates of N ational Income and E xpenditure (q u a rte rly ). R u ral In d u stries (annually). S easonally A djusted In d ic a to rs (annually). Y ear Books of the Com m onw ealth of A u stra lia (annually). Com m onwealth of A u stra lia. Incom e Tax and Social S ervices C ontribution A ssessm e n t A ct, C an b erra. (1965). R ep o rt of the C om m ittee of Econom ic E n q u iry, C an b erra. Copeland, M. A. (1952). A Study of Money Flow s in the United States (N ational B ureau of Econom ic R e searc h ), New York. Downing, R. I. (1966). N ational Income and Social A ccounts, M elbourne E ckstein, A. (1961). The N ational Incom e of C om m unist China, New York. G ilb ert, M. (1961). The problem of quality changes and index num bers Monthly Labour Review 84: and A sso ciates (1958). C om parative N ational P ro d u cts and P ric e L e v e ls, P a ris. 118

127 L ist of R eferences Cited 119 and K ravis, I. B. (1954). An International Com parison of National Products and the Purchasing Pow er of C u rrencies, P a ris. Haig, B.D. (1961). 'Q uarterly estim ates of national income and expenditure', paper delivered to Conference of A ustralian and New Zealand A ssociation for the Advancement of Science, B risbane. Haig, B.D. (1966). 'The m easurem ent of re a l expenditure and product of goods and se rv ic e s', Economic R ecord 42: (1968). Real Product. Income and R elative P rices in A ustralia and the United K ingdom, C anberra. Hicks, J.R. (1940). 'The valuation of the social incom e', Economica (New S er.) 7: H ofsten, E. von (1952). P ric e Indexes and Quality Change, Stockholm. International A ssociation for R esearch in Income and Wealth (1964). Incom e and W ealth, Ser. XI, London, P t I. Jaszi, G. (1961). 'The m easurem ent of aggregate economic grow th', R eview of Econom ics and S tatistics 43: K arm el, P. H. (1963). Applied S tatistics for E conom ists, M elbourne, Ch. 11. Kendrick, J. (1961). Productivity Trends in the United States (National Bureau of Economic R esearch), P rinceton, p Keynes, J. M. (1936). The G eneral Theory of Employment, Interest and M oney, London. Kuznets, S. (1941). National Income and its Composition, (National Bureau of Economic R esearch), New York, (1951). 'Governm ent product and national incom e', in International A ssociation for R esearch in Income and Wealth, Income and W ealth, S er. I, C am bridge, pp (1961). 'Some conceptual problem s in m easurem ent', in B. Okum a n d R. R ichardson (ed s.), Studies in Economic Developm ent, New York. M athews, R. L. (1962). Apcounting for E conom ists, M elbourne, Chs. 18, 19. M orgenstern, O. (1963). On the A ccuracy of Economic O bservations, P rinceton, Ch. 14. National Bureau of Economic R esearch (1958). Studies in Income and W ealth, Vol. 22, A Critique of the United States Income and Product A ccounts, Princeton,, Pt I. (1961). Studies in Income and W ealth, Vol. 25, Output, Input and Productivity M easurem ent, P rinceton, P t II.

128 vc 120 L ist of R e fe re n c e s C ited N icholson, J. L. (1955). 'N ational incom e at facto r c o st o r m a rk e t p r ic e s? ', Econom ic Jo u rn al 65: O hlsson. I. (1953). On N ational A ccounting, Stockholm. P aig e, D. and B om bach, G. (1959). A C om parison of N ational Output and P ro d u ctiv ity of the United S tates and the United K ingdom, P a r is. P igou, A. C. (1932). The E conom ics of W elfare, London (first published 1928). (1946). Incom e, London, Ch. 1. R ed fem, P. (1955). 'N et investm ent in fixed a s s e ts in the United Kingdom, ', Jo u rn al of the R oyal S tatistic al Society (S er. A) 118 (P t II): R e serv e Bank of A u stra lia (1965). F low -of-funds, A u stra lia, 1953/54 to 1 961/62, Sydney. (1966). F low -of-funds, Supplem ent to S tatistic al B u lletin, Sydney. R obbins, L.C. (1935). An E ssay on the N ature and Significance of Econom ic S cience, London. Shone, R. (1957). C om m ents on T. B arn a. 'T he rep la c e m e n t co st of fixed a s s e ts in B ritish m anufacturing in d u stry in 1955', Jo u rn al of the R oyal S ta tistic a l Society (Ser. A) 120 (P t I):39. Stone, R. and C ro ft-m u rra y, G. (1959). Social A ccounting and E conom ic M o d els. London, Ch. 2. and Stone, G. (1961). N ational Incom e and E x p en d itu re, London. Studensky, P. (1958). The Incom e of N atio n s, New York. U. K. C en tral S tatistic a l O ffice (1956). N ational Incom e S tatistics: S ources and M ethods, London. N ational Incom e and E xpenditure (annually). U. N. S tatistic a l O ffice (1955). M ethods of N ational Incom e E stim ation (Studies in M ethods, S er. F, No. 8), New York. - - (1964a). N ational A ccounting P ra c tic e s in Sixty C ountries (Studies in M ethods, S er. F, No. 11), New Y ork. (1964b). A System of N ational A ccounts and Supporting T ables (Studies in M ethods, S er. F, No. 2), New York.. Y earbooks of N ational A ccounts S ta tis tic s, New York (annually). U. S. D epartm ent of C om m erce (1958). U. S. Incom e and Output (Supplem ent to the Survey of C u rre n t B u sin e ss), W ashington. --( ). Survey of C u rre n t B u sin ess 43(10): Survey of C u rre n t B u sin ess. W ashington (m onthly). U. S. P re s id e n t's C om m ittee to A p p raise Em ploym ent and U nem ploym ent S ta tistic s (1962). M easuring Em ploym ent and U nem ploym ent, W ashington, p.6 5.

129 B. D. Haig, at present a Senior Fellow in the Department of Economic History, Australian National University, was Director of the National Income and Expenditure Division of the Commonwealth Bureau of Census and Statistics. He is the author of several books and a number of articles on national income and related topics in journals of economics and economic history. S. S. McBurney has lectured in economic statistics at Australian universities. He was on the staff of the Commonwealth Bureau of Census and Statistics for some years, working on national income estimates and other economic statistics. He is at present at the Commonwealth Treasury.