National Income Estimates

Transcription

1 THE ECONOMIC WEEKLY ANNUAL February National Income Estimates Why Not Discontinue Them? A Rudra The national income estimates that are now compiled can well he dispensed with. The purpose of planning served by them can all be served just as well, or even better, by other statistical measures which, while lacking in elegance and theoretical finesse, would call {or less statistical work. These estimates should be discarded not because of their poor quality this the authors themselves fully admit or of their large and unknown margin of error, but because they deceive the unwary by their external appearance. Even those who are knowledgeable go on using them as if these estimates contained information which they certainly do not. It would be better to change over to some other estimates which hare less redundant stuff and, therefore, communicate 'Information with a higher degree of 'Efficiency'. MY plea for discontinuing national income calculations following the currently accepted method and definition is based on the following principles : First, the principle of economising effort; and Second, the principle of maximising The Amount of Information' contained in any series of economic statistics. This second principle I am building up here is an application unorthodox, but not illegitimate, I hope of a principle of the well-known Theory of Information of Mathematical Statistics and Communications Engineering. The contention is that much of the statistical material that goes into the calculation of national income estimates is useless stuff, without any significance; so it would not make any difference if they were to be discarded. If this waste matter was acknowledged to be mere ornamentation, decorating a solid core of useful Information, there would have been cause for irritation. But no harm would have been done if it were retained. It happens, however, that the national income estimates in India pass for containing much more Information than they actually do. Resembling as they do the national income estimates of more developed countries in external appearance, they are taken by the unwary to have as much significance and to carry as much Information as the estimates for those countries. But why only the unwary? Even those who are in a position to know the truth about economic statistics in the country, go on pretending, deliberately or through some kind of absentmindedness that these estimates contain Information that they certainly do not. Under the circumstances, it is desirable to change over to some estimate or estimates which would contain much less of redundant stuff and therefore communicate Information with a higher degree of 'Efficiency 1 (to borrow once more a term from the Theory of Information). A Mixed Bag Statistical material used in the construction of the national income estimates in India can be classified into three types which we shall name as : i) Informative Statistics ii) Derived Statistics, and iii) Baseless Statistics. Informative Statistics are those that are based on. direct data collection about the object of measurement, in this case the economy during a certain year. Derived statistics are those that are derived from other Informative Statistics, other Derived Statistics and other Baseless Statistics. Baseless Statistics are those that have no objective basis at all, either in actual measurement or in any other statistics. We shall give illustrations taken from the agricultural sector for the three types of statistics just defined. Statistics for the outturn of the principal crops in the so-called reporting areas of India are based on some direct data collection. These are Informative Statistics (usually with low Efficiency due to the large margin of error to which they are subject). The statistics for the actual cost incurred in agricultural 209 operations are. however, not based on any direct data collection: they arc derived from the statistics for the values of outturn of crops with the help of some assumptions of proportionality between outturn and cost. These are examples of Derived Statistics. The statistics for the nonreporting areas and for the minor crops arc not based on any direct measurement at all or on any other statistics for that matter. These are examples of Baseless Statistics. The accompanying table presents schematically the occurrence of Informative. Derived and Baseless Statistics in the national income. Derived Statistics Valueless Baseless Statistics obviously do not contain any Information at all. for it has no relation with the object of measurement either directly or indirectly. Derived Statistics do contain Information, but can never contain more Information than the Informative Statistics from which they are derived. The Information contained in Derived Statistics is either entirely contained in other elements belonging to the estimate, so that it does not supply any additional Information, or the Information contained in it is of no relevance to the time point concerned. so that as far as national income estimate for that particular point is concerned, it docs not contain any Information. One will have the first situation when the Derived Statistics are derived from other statistics which refer to the time point under consideration and are used in other parts of the national income estimate. The second situation obtains when the Derived Statistics are derived from Informative Statistic^

2 February 4, 1961 THE ECONOMIC WEEKLY ANNUAL 210

3 THE ECONOMIC WEEKLY ANNUAL February referring to time points other than the point under consideration. An example will make the position clear. Let US consider the way the estimate of income from smallscale enterprises is derived. We find the following explanation in the Final Report of the National Income Committee: "In many cases, however, the data do not relate to the period under review and it becomes necessary to adjust these figures on some arbitrary basis. Tims, when We have data on average earnings for an earlier period, we have inflated it by an index that has been calculated on the basis of rise in average earnings of the factory workers during the corresponding period"' (p 69). Thus the Derived Statistics re pre. senting income from a particular small-scale enterprise during the year, say makes use of the following three ingredients: (a) statistics regarding per head income from that industry in a past year, say 1918; (b) census estimate of the working force engaged in that industry in the year 1951 : and le) an estimate of the rise in average earnings between 1918 and 1050 of factory workers. Thus, the Informative Statistics made use of in this piece of Derived Statistics either do not refer to the \ear in question at all or if they refer to that year, they do so to a sector other than the one to which the estimate refers, namely factory industries. That is to say. given the estimate of income generated by factory industries, the estimate of income generated bv small-scale enterprises is redundant. It may be said that, after all. the working force engaged in small-scale enterprises in 1955 will not be very different from that in 1951 and that the per head incomes in 1955 as compared to that in the base year 1918 may have changed in the same proportion as in factory industries between these two years. Hut this would amount to saying that the Information in regard to the change in the working force engaged in small-scale enterprises and changes in their per capita average earnings are just not important. It is not important to know to what extent the working force engaged in small-scale enterprises have actually changed or in what way the movement of per capita income of workers in smallscale industries has differed from that of workers in large-scalp factories, If that be the attitude, it only follows that there would have been no harm if the small-scale enterprises sector were altogether left out from the national income estimate! It is. of course, true that if there are some Derived Statistics, derived from other Informative Statistics which refer to the same lime period but which are not made use of in other parts of the national income calculation, they would indeed con tribute additional Information to the estimate. Hut in that case we would not call it Derived Statistics: it would be Informative Statistics. Better Alternative Available Thus Derived Statistics cannot. any more than Baseless Statistics, provide am useful Information over and above that supplied by the Informative Statistics used in preparing the estimator From this, one can draw the conclusion that the present national income estimates do not tell us anything more about the state of the economy during a Year than what would be done by suitable measure or measures that one can build up by utilising only the Informathe Statistics used in the estimates. It will be seen from the table that if the principle of rejecting everything but Informative Statistics is to be followed, we would have to do without calculations of depreciation for all the sectors and cost of production for almost all the sectors. We would, therefore, have to give up the attempt of estimating the net value added in any sector. One would also have to leave out of con sideration Animal Husbandry. Fishery. Small-Scale Industries. large parts of Trade and Transport. Pro fessions and Liberal Arts. Domestic Service and House Property, retain ing only production of major agricultural crops. Forestry. Large-Scale Manufacturing Industries. Mining, some parts of Trade. Commerce and Transport. Government Services and Communication. What can we do within the limits of this rump of a national income framework? Obviously, all that one can do is to construct some index numbers with the statistics of gross outputs of the sectors retained or with that of the final products of these sectors: consumer goods, capital goods and such 211 intermediate goods as are exported. These index numbers would serve to measure changes in the levels of production in real and money terms in these sectors. They would do so very partially and very insensitively, but as We have just seen in terms of the notion of Information. the degree of partiality and the degree of insensitivity will not be any higher than that of the current national income estimates. The national income estimates are used to measure riot only the level of overall production, they are also used in preparing national accounts. If they are discontinued, it would not be possible to prepare national accounts. It may he complained that in that case, it would he difficult to take decisions about the size of the Plan and the modes of mobilising resources if these cannot be placed in the accounting picture. The argument would have force if it were true that decisions are being taken at present on the basis of any dependable and complete accounting picture. Hut the estimates of Savings and Investment currently made use of are. statistically speaking, of the same or even poorer quality than the national income estimates. As a result, whatever national accounts there are in India provide the same sorry picture as the national income estimates: a few islands of sound and solid ground in a treacherous marsh of Derived and Baseless Statistics. The islands, in the case of national accounts, it is easy to see. have the same base of Informative Statistics as those of national income estmates So. it would be possible to bring in relation the dependable elements in the national accounts, with the measures which we are suggesting for replacing the national income estimates. Thai national income estimates and national accounts do not really provide any sound basi- tor building up plan estimates is proved bv ii comparison between the estimates of investment in the private sector in the Second Plan, as made at the beginning and at the end of the five-year period. Thus, for the Agriculture sector, the original estimate of Rs 275 crores has changed to Rs 675 crores; for Transport, the corresponding estimates are Rs 85 crores and Rs 135 crores: for Village and Small Industries, they are Rs 100 crores and Rs 225 crores. Larger errors in the estimates can

4 February 4, 1961 THE ECONOMIC WEEKLY ANNUAL 212

5 THE ECONOMIC WEEKLY ANNUAL February 4, 1961 hardly be imagined, even if they are to be made without any national accounting framework! The national income estimates are also regarded as providing a measure of the changes taking place in the standard of living of the people. But standard of living is much hotter measured by consumption statistics for which we have an extensive, regular and direct collection of data by a widespread and elaborate organisation : the National Sample Survey. The National Sample Survey estimates of consumption are. of course, subject to fluctuations due to sampling, and are probably subject to many other non-sampling errors. They do not, however, contain any redundant material. They do not incorporate any Derived Statistics or Baseless Statistics. Therefore, there will be no loss from the point of view of the measurement of level of living, if national income estimates are abandoned. Uptil now. we have been concentrating attention on the changes taking place in the different parts of the economy. But it may serve some purpose to ignore the changes that take place within a few years and have a broad statistical picture of the economy in terms of an average national account valid over a period- For this, it is undeniable that Derived Statistics can also be informative, as long as they are derived from Informative Statistics referring to time points spread over that period. Thus, according to the official estimates, national income in was Rs 09.8 abjas and Rs 98.2 abjas in ; income from Professions and Liberal Arts was Rs 5,6 abjas in and Rs 5.2 abjas in , We have argued before that Rs 0.1 abjas, representing the increment in income from Professions and Liberal Arts between and , is a Baseless Statistics containing zero Information; also that the difference of Rs 1.6 abjas in the national income between these two years contain much of redundant stuff in the form of Derived Statistics and Baseless Statistics, But the fact that in both and the proportion of the estimated income from Professions and Liberal Arts in the national income is of the order of 5 per cent is a piece of Information which is communicated by the figures Rs 5.6 abjas and Rs 52 abjas which happen to be Derived Statistics, for this type of Information which does not relate to any specific year but to the average situation holding over a number of years, it is quite unnecessary to make calculations for individual years. Such calculations need be made only at such intervals over which we aim at obtaining Information about annual changes. Statistics purporting to communicate Information about a longer period will have to be calculated only once, for the entire period, and published as such, i e. as referring to the entire 213 period. For this purpose, more "complete" estimates, involving the use of Derived Statistics, may indeed be published, but only at. say, quinquennial intervals. The additional advantage of allowing such a five-year interval to pass between two estimates would be that it would be possible to devote much larger resources to the collection of direct statistics and, therefore, to extend the coverage and improve the quality of these statistics.

6 February 4, 1961 THE ECONOMIC WEEKLY ANNUAL 214