Chapter 1: Introduction, Overview, and Basic Steps for CPI Development

Transcription

1 Chapter 1: Introducton, Overvew, and Basc Steps for CPI Development I. INTRODUCTION TO DEVELOPING THE CPI 1.1 Consumer prce ndces (CPIs) measure changes over tme n the general level of prces of goods and servces that households acqure, (use or pay for) for the purpose of consumpton. In many countres they were orgnally ntroduced to provde a measure of the changes n the lvng costs faced by workers, so that wage ncreases could be related to the changng levels of prces. However, over the years, CPIs have wdened ther scope, and nowadays are wdely used as a macroeconomc ndcator of nflaton, as a tool by governments and central banks for nflaton targetng and for montorng prce stablty, and as deflators n the natonal accounts. Wth the globalzaton of trade and producton and the lberalzaton of the markets, natonal governments, central banks and nternatonal organzatons place great mportance on the qualty and accuracy of natonal CPIs, and n ther nternatonal comparablty. 1.2 Dfferent conceptual frameworks can be used to address fundamental ssues relatng to the nature of the ndex. For example, whether the CPI should measure the change n cost of a fxed-weght basket of goods and servces or whether t should measure the change n the cost of lvng,.e., the cost of mantanng a gven standard of lvng, takng nto account the fact that when prces change consumers change ther expendture patterns. The use and conceptual bass of the ndex wll determne the method of constructon, ncludng the formulae used for the aggregaton of prces. 1.3 The method of constructon also allows (or should allow) CPIs to be adapted for a wde range of specfc uses. For example, they can be adapted to calculate specfc nflaton rates for socal groups such as pensoners or low-ncome households. Ther product coverage can be adapted so as to show what the rate of nflaton s n partcular sectors such as energy or food, or excludng partcular sectors such as alcohol and tobacco. They can shed lght on the effect of tax changes or government-regulated prce changes on the rate of nflaton. They can be compled on a regonal bass, showng dfferent nflaton rates wthn dfferent parts of a country or between urban and rural areas. 1.4 In short, CPIs are now consdered as one of the most mportant economc and socal ndcators produced by Natonal Statstcal Insttutes (NSIs) throughout the world. Aganst ths background, the challenge of statstcal offces s fourfold: to dentfy user needs; to conceptualze user needs n terms of economc theory; to translate the underlyng concept nto statstcal measurement terms followng the fundamental prncples of prce ndex measurement; to construct the ndces so defned and evaluate them aganst purpose. Systems of Prce Indces and supportng frameworks Hgher-level frameworks 1.5 The System of Natonal Accounts (SNA) provdes the basc gudelnes for buldng approprate frameworks for CPIs and other prce ndces n the famly of prce ndces, coverng all major economc actvty ncludng, most partcularly, producton and consumpton and ntermedate outputs. In theory the SNA also provdes a methodology for developng and extendng the core system of prce ndces to meet specalsed needs whle mantanng consstency of approach, both n concepts and practcal measurement, and coherence n terms of the defnton, classfcaton and measurement of flows and stocks of goods and servces. The use of such frameworks also facltates the examnaton of the relatonshps between dfferent prce nflaton measures and can help analysts n ther understandng and nterpretaton of nflatonary sgnals. Wth ther ablty to dentfy

2 statstcal gaps, these frameworks can also facltate future developments n the feld of prce statstcs. One such framework s the Stage of Processng Framework (see Appendx 1.1). Stage of Processng Framework 1.6 A Stage of Processng framework has a useful role to play n nformng polcy formaton. It can also provde the bass for systematcally analysng the buld-up of nflatonary pressures n the economy and for trackng relatonshps between prce developments n partcular sectors. From a compler s pont of vew t also provdes a useful tool for dentfyng gaps n avalable prce ndex seres and a focal pont for nvestgatng coherence n ndex constructon. 1.7 The frst step n constructng such a framework s to dvde the entre economc process nto dstnct stages: Rest of the world Producton: splt nto broad categores such as agrculture and manufacturng (and further sub-dvded nto prmary, ntermedate and fnal producton). Fnal demand: separated nto prvate fnal consumpton, government consumpton, captal purchases and exports. 1.8 The man seres of prce ndces are then grouped accordng to coverage. For nstance, ths mght be as follows: Consumers expendture (Consumer Prces Index). Other elements of Fnal Demand (an Index of Government Prces [IGP] and an Index of Investment Prces [IIP]. Inputs nto and outputs from the producng sectors (the Producer Prce Index [PPI]). Trade prce ndces 1.9 In prncple, for all sgnfcant transfers of goods and servces from one part of the economy to another, there should be a representatve prce ndex (based on drectly-measured prces) that reflects the changng level of prces for all such goods (or servces). 1.1 The Stage of Processng Framework s shown n Appendx 1.1. The source of the goods/servces s shown down the left sde of the table; the destnaton s shown along the top of the table. For nstance, the prces charged by the manufacturng sector for consumer goods sold to the retal/wholesale sector are reflected n the PPIs (Producer Prce Indces). The prces charged by the retal sector for consumer goods sold to prvate consumers are reflected n the CPI. Across the top of the table are the buyers, classfed nto two broad groups: ntermedate demand and fnal demand. Most sectors of producton are nvolved n both buyng nputs and sellng output-so they appear n the table both as buyers and as sellers. In practce, n some cells transactons from seller to buyer wll be nsgnfcant and these are denoted by a dash. For nstance, there are unlkely to be any mported goods feedng nto the constructon sector as ntermedate demand. In the example n Appendx 1.1, whch for llustratve purposes has been populated on the bass of a typcal stuaton, the shaded cells represent sgnfcant transactons n the economy for whch there are no relevant prce ndces. Where publshed ndces do exst, the name (or acronym) of the ndex s gven. For nstance, the sale of fnshed goods by retalers to prvate fnal consumpton s measured by the CPI The practcal advantages of usng a Stage of Processng Framework can clearly be llustrated: The assocated table of nflaton rates can be used for analytcal purposes and help nform economc polcy. Aggregate prce ndces (and nflaton rates) for each row and for each column may be computed to produce nflaton rates for dfferent sectors. Possble enhancements to currently publshed ndces can be dentfed. Major gaps can be dentfed.

3 Issues relatng to statstcal ntegraton and coherence can also be dentfed. For example, ssues relatng to coherence n concepts and practcal measurement arse when combnng rows and columns. Lower-level Frameworks 1.12 Hgher-level frameworks can be supplemented by lower-level frameworks such as the Stage of Producton Framework. The latter, whch can be used for n-depth nvestgaton of the Producer Prce Index, s not expanded upon here The lower-level theoretcal framework for CPIs s essentally based on economc theory relatng to consumer behavour. In ths context the Schultze Panel on Conceptual, Measurement, and Other Statstcal Issues n Developng Cost-of-Lvng Indexes (1964) makes a dstncton between the tradtonal CPI based on a fxedweght basket, as beng a cost-of-goods ndex (COGI), and a cost-of-lvng ndex, whch measures the change n expendtures a household would have to make n order to mantan a gven standard of lvng or utlty. Many references n the plentful lterature on CPIs argue that an advantage of a COLI over a COGI s that the former s supported by economc theory (.e. the behavor of the ndvdual n the market place), mplyng that the latter s not. But economc theory can be used to support both a COGI and a COLI, one at a macro-level and one at a mcro-level, and both complement one another. Thus: 1.14 A COGI measures the nflatonary pressures n the economy from prce developments n the retal sector. It represents one of many sectors, albet an mportant one, n the Stage of Processng Framework A COLI measures the expendture requred by a household to mantan ther standard of lvng or utlty. Puttng asde the defnton of utlty and whether the COLI s unconstraned or constraned, t s an ndex constructed from the vewpont of the ndvdual consumer and has ts foundaton n mcro-economcs and the theory of ndvdual consumer behavour. Unlke a COGI, t takes nto account the substtutons consumers make when faced by relatve changes n prces ether between dfferent goods and servces or between dfferent outlets and supplers The dfferent uses of a CPI set wthn these frameworks determne a number of measurement ssues ncludng the choce of ndex number formula. Choce of ndex number formula 1.17 Most countres state that they use a Laspeyres ndex or a Laspeyres-type ndex for ther natonal CPI whch, n practce, s somewhere along the contnuum between a cost-of-goods ndex (COGI) and a cost-oflvng ndex (COLI) Experts generally agree that the deal type of ndex for a CPI would be a superlatve ndex such as the Fsher ndex, whch wll be dscussed later. Superlatve ndces make equal use of the prces and quanttes n both of the perods beng compared (such as the base perod and the latest month). In realty, quanttes for recent perods are usually not known, so that n practce nearly all statstcal offces produce CPIs whch rely on quanttes relatng to a base perod some tme earler. An excepton to ths s actual transactons that can be captured at the ponts of purchase through the use of scanner data Some countres am to produce a cost of lvng ndex (COLI). But such an ndex s n fact a type of superlatve ndex and suffers from the same practcal defect as mentoned above. 1.2 It s mportant, nevertheless, for statstcal offces to be able to state publcly what type of ndex s beng calculated n ther CPI. A true Laspeyres ndex uses quantty data whch relate to exactly the same perod as the prce reference perod. Ths s rarely the case. Most statstcal offces have a prce reference perod whch s later than the perod to whch the quantty data (.e. the weghts) relate and wll span, say, a year rather than a pont n tme. Ths s because one of the man source of weghts data s a Household Expendture Survey (HES) whch typcally produces usable results a year or more after the end of a survey perod. In these crcumstances, ether the HES perod weghts are used wthout adjustment, or they are prce-updated to the prce reference perod. Ether of these methods results n a Lowe ndex, but s often referred to as a Laspeyres-type ndex.

4 Index Formula at Lower (Elementary Aggregate) Level 1.21 The frst stage n the calculaton of CPIs s the calculaton of elementary prce ndces, whch are then aggregated to obtan hgher-level prce ndces. Expendture weghts are not usually avalable below the elementary aggregate level. The three most wdely known elementary ndex formulae are the Carl, the Dutot and the Jevons and each s assocated wth a number of assumptons whch wll mpact on measured nflaton. The Carl (a smple arthmetc average of prce relatves) and Dutot (the rato of smple arthmetc averages of prces) formulae have a number of problems assocated wth ther use partcularly the Carl, whch s postvely dscouraged as t s partcularly assocated wth some bad characterstcs. Because of ths the Jevons formula (the rato of smple geometrc averages or the geometrc average of prce relatves) s ncreasngly used. It should be noted that an arthmetc average s always greater than or equal to a geometrc average and that the dfference wll be greater the greater the varance n the prce relatves. The choce of formula becomes more mportant the greater the dversty of prce movements whch s one argument for ensurng that elementary aggregates are as homogeneous as possble. These ssues are dscussed n more detal later. Index Formula at Hgher Level 1.22 The hgher-level ndces are calculated smply as weghted averages of the elementary prce ndces. The weghts typcally reman fxed for a sequence of at least 12 months. Some countres revse ther weghts at the begnnng of each year n order to try to approxmate as closely as possble to current consumpton patterns, but many countres contnue to use the same weghts for several years. The weghts may be changed only every fve years or so. The use of fxed weghts has the consderable practcal advantage that the ndex can make repeated use of the same weghts. Ths saves both tme and money. Revsng the weghts can be both tme-consumng and costly f t requres new HES to be carred out. However, the longer the perod between weght updates, the less relevant the CPI becomes. Many countres are movng toward annual or bannual weght updates In Chapters the superlatve ndces Walsh, Fsher and Törnqvst show up as beng best n all the approaches to ndex number theory. These three ndces gve very smlar results so that for any practcal reason t wll not make any dfference whch one s chosen as the preferred target ndex. The theoretcal target ndex s a matter of choce. In practce, a statstcal offce may prefer to desgnate a basket ndex that uses the actual basket n the earler of the two perods as ts target ndex on grounds of smplcty and practcalty. In other words, the Laspeyres ndex may be the target ndex. Acquston, use, or payment approach 1.24 A CPI s based on the measurement of the change n prces of the goods and servces ncluded n the basket. The vast majorty of goods (but not necessarly of total values) are prced n the retal outlets sellng them. It should be noted that most often the prces recorded are the labelled prces, whch are assumed to be the prces actually pad by consumers. It s also generally assumed that payment for the goods s made at the tme of purchase ndeed the consumer would regard the two events as dentcal. However, payment can be n cash or on credt, ncludng credt cards for whch the due date of payment may be several weeks after the actual purchase The tme factor s mportant n other ways too. A shopper may decde to buy a larger than normal quantty of a partcular good f there s a specal prce reducton. The product may then be stored at home and consumed (.e. used) over a relatvely long perod. Cans of food, for example, offered cheaply for a lmted perod, may be stored at home wthout deteroraton for months and consumed at the usual frequency Another ssue concerns the defnton of usage. A bottle of mlk wll typcally be consumed wthn a few days of purchase. Consumpton s lkely to take place n the month for whch the CPI s calculated. But a sem-durable such as a shrt wll be worn many tmes over a perod of probably several years. A durable such as a televson set may be used for a decade or more and a house over an even longer perod of tme. The queston arses as to whch CPI month (or months) should the purchase be allocated Wth servces these questons can be even more complex. Take, for example, the purchase of a season tcket for a bus servce. Ths may be a sngle payment for a pass whch gves free bus transport for a year. It

5 can be seen that although ths example s clearly a servce (the use of bus transport over a perod of tme) t has much n common wth the purchase of a durable such as a televson or shrt whch provdes a type of servce over a long perod. A servce such as a medcal operaton can also be regarded as durable, snce t s lkely to gve long-term health benefts to the patent CPI theory devotes much thought to these ssues, whch can have mportant mplcatons not only for how a CPI s compled but for the results themselves. Three dfferent approaches can be dentfed: The acqustons approach relates to when the good or servce s acqured, rrespectve of when t s actually used or consumed. The tme of acquston of a good s the moment at whch the legal ownershp of the good passes to the consumer. Ths s usually the pont at whch the purchaser ncurs a lablty to pay. On the other hand, wth a servce there s no change n ownershp; t s acqured at the tme the producer provdes t (e.g. the bus journey or arlne flghts). A CPI based on ths approach measures the change n the cost of acqurng a product. The tmng of the recorded prces should be consstent wth the way n whch the value would be recorded n the expendture data used for the CPI weghts. The use approach relates to the perod over whch the product s consumed or used; a CPI based on ths approach measures the change n the cost of usng the product over tme; n other words the cost of the good s dstrbuted over ts useful lfe. Expendtures on durable goods and servces are lable to fluctuate. The payments approach relates to the perod of tme when the actual perod-to-perod payments for the product are made. Ths can dffer from the perod when t s acqured and when t s used. When payments are not made n cash, there may be a long perod before the purchase s pad for, whether by cheque, credt card or other method. The tme at whch these debts are made s rrelevant for the recordng of the prce. The prce to be recorded s the prce payable at the tme of acquston (though sometmes the method of payment may tself affect the prce) Statstcal offces need to have a clear polcy on whch of these approaches s used n ts CPI. In practce, the choce between the three approaches s an ssue relatng to durable goods and ts mpact s lkely to affect the weght gven to owner-occuper housng costs. The latter s dscussed n Chapters 11 and 24. In countres where food expendtures and other expendtures on non-durables, sem-durables and even servces account for a sgnfcant share of the CPI basket and where credt fnancng s rarely used, the acquston, use and payment approaches wll gve very smlar results and hence the CPI can satsfy many uses equally well. Ths s the prncpal reason why most countres use, ether mplctly or explctly, the acqustons approach to defne what consttutes consumpton expendture. II. OVERVIEW OF THE CPI USES AND NEEDS 1.3 A CPI can be used for a varety of purposes, the more common ones beng: the ndexaton of wages, rents, contracts and socal securty payments; the deflaton of household consumpton n the natonal accounts; as a general macroeconomc ndcator, especally for nflaton targetng and for settng nterest rates. Elements of a CPI are also often used n the calculaton of purchasng power partes (PPPs) requred n the Internatonal Comparson Program (ICP) Gven the many uses of CPIs, t s unlkely that one ndex can perform equally satsfactorly n all applcatons. Some countres therefore construct several CPI varants for specfc purposes. Each ndex should be properly defned and named to avod confuson and a headlne CPI measure should be explctly dentfed. Where only one CPI s publshed, t s the man use that should determne ts type and scope. If there are several major uses, compromses may have to be made wth regard to how t s constructed. The purpose of a CPI should nfluence all aspects of ts constructon. CPI producers need to know how ther ndex s beng used f they are to ensure that t s ft for purpose. In ths connecton user consultaton s mportant. 1 See Chapter 17.

6 1.32 Ths secton revews some of the ssues confronted by the ndex compler relatng to the scope of the ndex and the practcal measurement and complaton decsons whch have to be made. But before dong so t revews the varous uses of a CPI to put the dscusson n context. The dfferent uses of a CPI 1.33 CPIs have three man uses: Indexaton A CPI used for wage or contract ndexaton of any specfc group, whether of populaton or of products, should represent the coverage of the group concerned. For nstance, t can be argued that the weghts of a CPI used for ndexaton of pensons should cover only the expendture of the pensoner populaton. The product and outlet lst could also be more approprately targeted, f the data exst. Ths means, for example, that a CPI used for ndexng pensons may use weghts relatng to pensoner households and may exclude products whch may be thought largely rrelevant to, say, poorer households, such as luxury tems or brands. Smlarly, for domestc ndexaton, the CPI should cover only the expendture of the resdent populaton (see secton below on Geographcal coverage ). More generally, t has to be decded whether the CPI should be, n prncple, a cost-of-lvng ndex or a pure prce ndex these two very dfferent concepts are dscussed n Chapter 1. For certan very specfc types of ndexaton, such as for rents, users may prefer to use just the sub-ndex for rents. In such cases, the sub-ndex should be of a statstcal qualty suffcent for that purpose. Natonal accounts deflaton Ths use requres consstency between the prces data used for the CPI and the expendture data used n the natonal accounts. Both data sets should cover the same set of goods and servces and use the same concepts and same classfcaton, whch for natonal accounts would be the Classfcaton of Indvdual Consumpton Accordng to Purpose (COICOP). For example, the natonal accounts requre the valuaton of own-account consumpton, whereas ths s sometmes excluded from the CPI ether as a matter of prncple or for pragmatc reasons. Ths apples to the valuaton of the servces of owner-occuped housng as well as the consumpton of own-produced food. Inflaton targetng It can be argued that central banks deally need a tmely ndex relatng to total nflaton, not just consumer nflaton. But statstcal offces generally are unable to construct such ndces, n part because of the measurement ssues relatng to government consumpton. In the absence of such an ndex most central banks rely on a CPI, usng the domestc concept but measured on as wde a bass as possble, n terms of both products and geographcal coverage. The same apples to the use of the CPI as a general macroeconomc ndcator. Geographcal coverage 1.34 There are two dstnct aspects to the queston of the geographcal coverage of a CPI. The frst relates to the country as a whole, the second to ts regons. The natonal versus domestc concept 1.35 Followng the SNA, conventons, household consumpton expendtures may nclude all purchases n the natonal economy or be lmted to domestc economy purchases only. Thus, a CPI can have natonal or domestc coverage. Natonal coverage means that the CPI should cover the non-busness expendture (and prces) relatng to the resdent populaton of the country, regardless of where the expendture takes place. The natonal concept s approprate when the CPI s beng used for ndexaton of ncomes and cost of lvng measures. The weghts

7 for expendture abroad can be ncluded n the HES, but measurng prces pad abroad poses problems. The natonal concept thus poses a measurement problem for collectng prces abroad. Domestc coverage means that the CPI should cover all the expendture made wthn the domestc terrtory of the country, ncludng the non-busness expendture made by foregn vstors. It s approprate where the CPI s used for natonal nflaton analyss and nflaton targetng. Many countres carry out surveys of the expendtures of foregn vstors, for example, va Internatonal Passenger Surveys conducted at major border crossngs and arports. Ths s partcularly mportant for those countres whch have a large number of foregn toursts, or a hgh level of cross-border shoppers. The domestc coverage follows the SNA. Foregn vstors wll generally have very dfferent expendture patterns from those of natonal resdents (e.g., they wll spend more on hotels and restaurants) and to omt them could ntroduce serous dstortons nto a CPI amng to follow the domestc concept, especally f the man purpose of the ndex s to measure the nflatonary trends n the economy Purchases made va the nternet from foregn webstes or webstes of retalers based abroad should be ncluded f a CPI s based on the natonal concept. So also should purchases made abroad more generally, ncludng such tems as fees for foregn boardng schools, even f the tem, n ths case educaton, s consumed outsde the country. Where such purchases are made n the foregn currency, they should be converted to the domestc currency at the relevant exchange rate. Clearly t would be mpractcable to collect prces drectly n foregn countres on a contnuous bass although surveys of prces done n order to compute purchasng power partes may provde an occasonal benchmark. Where the regular collecton of the relevant prces s not practcal, t may be possble to obtan a reasonable proxy for prce movements usng publshed sub-ndces of the other countres CPIs Although no frm rules exst about the treatment of nternet shoppng n respect to the domestc concept, most countres whch have examned the ssue have concluded that nternet shoppng from home whether from domestc or foregn webstes, and pad n the natonal currency, should be ncluded. Regonal coverage 1.38 As far as regonal coverage of the CPI s concerned, the general rule s that a natonal CPI should cover expendtures and prces throughout the country. However, comprehensve coverage s not always necessary, especally f regonal CPIs are not publshed and the samplng scheme ensures that the ndex s representatve of the whole country. In such stuatons, CPI complers should collect evdence from tme to tme on the trends n prces n dfferent regons over perods of tme whch cover dfferences n seasonal varatons etc., to ensure that the sample remans representatve. Any regon whch shows prce trends sgnfcantly dfferent from the others should be covered by the CPI f ts ncluson s lkely to have a sgnfcant effect on the natonal CPI and wll mprove representveness. But there s lttle pont n spendng scarce resources collectng prces n sparsely populated regons f to do so would have lttle or no mpact on the natonal CPI. When carryng out such senstvty tests regonal weghts can often be an ssue. In ths case populaton may sometmes be used as a proxy for regonal consumer expendture. However, where regonal CPIs are aggregated to compute the natonal CPI, weghts should be based on regonal expendture rather than populaton data Another dffculty regardng regonal CPIs s related to the natonal" versus "domestc concept. It can sometmes be the case that a household lves n one regon but does most of ts shoppng n an adjacent regon, partcularly when a household lves close to a regonal border. The queston of whether the expendture weghts and the prces should be allocated to the regon of expendture or the regon of resdence s usually dctated by practcal ssues. As wth the natonal concept dscussed above, f the regon of expendture s used (equvalent to expendture abroad by a domestc resdent) some means has to be found of estmatng the proportons of expendture made by vstng consumers n the varous regons so that ths can be reflected n the prces. 1.4 Fnally, the queston often arses as to whether a CPI can be lmted to urban areas or f rural areas should also be covered. Agan, n prncple the whole terrtory should be covered, but clearly the mpact on the natonal CPI of ncludng rural areas where relatvely few monetary transactons take place wll often argue aganst ther ncluson on grounds of cost. But the vew taken wll depend, at least n part, on the sze and treatment of ownaccount producton. If own-account consumpton s ncluded n the CPI, the weghts should nclude a valuaton

8 of the physcal quanttes of such products, the latter often derved from the HES. The prces wll normally be the same as those used for actual transactons for the same goods sold n the same localty Where the weghts derved from a HES are avalable for rural as well as urban households t s generally better to use the weghts for urban and rural households combned, even f prce collecton s lmted to urban areas, as ths wll normally mprove the representveness of the ndex. But, where feasble, prce statstcans should undertake plot calculatons to test whether ths s the case. Reference populaton for the CPI 1.42 Households comprse one or more ndvduals lvng together and sharng resources. The offcal UN defnton also ncludes people lvng n mltary accommodaton, convalescent homes and boardng schools. Such people are treated as belongng to prvate households. But for the purposes of CPI constructon consderaton must also be gven to the ncluson of other nsttutonal households, such as those lvng n relgous nsttutes, hosptals, prsons, retrement homes etc. Temporary foregn workers may lve together n specal housng blocks, whch may also be treated as nsttutonal households n the populaton census. Expendture on accommodaton and lvng costs, such as lodgng fees and charges for meals, mposed by the nsttuton, as well as personal expendture by the ndvdual on, for example, clothes and toletres, should be ncluded n the CPI. However, care should be taken to ensure there s no double countng where, for nstance, a famly rather than an ndvdual pays the accommodaton costs charged to a patent n a hosptal. A data source lke the HES should be desgned to pck up the amount spent on such charges just once and the standard conventon s to record t aganst the household whch ncurs the costs. In the above example, the accommodaton costs should be ncluded under the expendture of the famly, and not the ndvdual. If ndvduals spend ther own money on clothes and other ncdental expendture then the HES should record ths expendture as beng ncurred by the ndvduals. In realty, many HESs do not cover nsttutonal households and, where ths expendture s consdered to be sgnfcant, estmates wll need to be made from, for example, specal surveys of people lvng n nsttutons or by reference to the expendture patterns of smlar people, say, the same sex, age and soco-economc group, lvng n non-nsttutonal households In realty, n consderng the practcal ssues relatng to the ncluson of nsttutonal households n a CPI, two questons need to be asked. Frst, s the expendture pattern of nsttutonal resdents lkely to be sgnfcantly dfferent from household resdents? Secondly, even f the answer s yes, would ther excluson from the CPI be lkely to sgnfcantly affect the natonal (or regonal) CPI? To answer these questons, some research should be carred out on a sample bass Some countres exclude certan household types from the CPI, such as the very wealthy or the very poor. Such exclusons may be on theoretcal grounds (for example, usng the argument that the expendture of the wealthy, who are relatvely few n number, should not be allowed to affect a CPI whch may be used for ndexaton of wages of ordnary workers) or on practcal grounds (for example, usng the argument that wealthy households tend to have low response rates to HESs, and ther ncluson can lower the qualty of expendture weghts). For a CPI whch s used for ndexaton of wages, the excluson of pensoners and wealthy households may be justfed on conceptual grounds. For example, t may be consdered that such households are lkely to spend ther money on atypcal thngs and ncludng them would dstort the relevant overall average. It s also argued by some that the ncluson of pensoner households should be excluded n prncple from an ndex used for the up-ratng of state pensons because of the crcularty nvolved (the level of state penson nfluences expendture patterns whch are then used n the up-ratng calculaton) whle others would argue that t s logcal that ndexaton should be based on an ndex reflectng the expendture of pensoner households and ther specfc nflatonary experence. Note that f wealthy households are excluded, the CPI basket should not nclude products lkely to be bought only by the excluded group, nor should outlets specalsng n such products be ncluded n the sample. Conversely, f the wealthy are ncluded, some luxury products and outlets should also be ncluded n the sample. For the analyss of natonal nflaton, t s consdered that the more comprehensve the CPI the better.

9 Democratc versus plutocratc weghts 1.45 A democratc CPI uses weghts whch average unweghted expendture proportons over the whole populaton, rather than dvdng total aggregated expendture by the populaton. The latter method gves more weght to the hgh-spendng households, and s referred to as a plutocratc ndex. It s argued that a democratc ndex s more sutable for showng the mpact of nflaton on the average household but n realty s very rarely computed by statstcal offces. There s a general consensus that a plutocratc ndex s the approprate ndex to use for natonal accounts deflaton and for a general measure of nflaton Many countres publsh a range of CPIs relatng to sub-sectors of the populaton such as: all households, low-ncome households, pensoner households, etc. but a CPI based on democratc weghts s very rare. Product exclusons 1.47 In ts role as an ndcator of total consumer nflaton, the CPI should n prncple cover all types of goods and servces whch are consumed n the natonal retal market. In practce, some types of product may be excluded for polcy reasons whle other exclusons are unavodable n practce. These may nclude: goods sold llegally, such as narcotcs; black market sales; gamblng; and prosttuton. In most of these cases, except perhaps for legal lotteres, there wll be no expendture data from the HES or natonal accounts and prces are dffcult f not mpossble to collect. Thus, n prncple, estmates of weghts and prces wll need to be made for the purpose of producng deflators, even f the expendture s not covered n the CPI. Solutons wll need to be found to the practcal measurement ssues. For nstance, f a CPI covers gamblng, t s not the gross stakes whch should be ncluded n the weght, but the net stakes, whch s broadly equvalent to the margn taken by the gamblng operator. As ths s not lkely to be measurable, one soluton may be to dstrbute the weght for gamblng across other classes n COICOP group 9.4 (Recreatonal and sportng servces) The treatment of second-hand goods s often found to be problematc. As far as transactons wthn the household sector are concerned, sales wll balance purchases (apart from any dealer charges), so the effect on the CPI s close to zero, and they may be excluded. But n many developng countres there are sgnfcant sales of mported second-hand goods, such as cars and clothng. Where sales of mported second hand goods are sgnfcant relatve to sales of new goods of the same product, such sales should be ncluded, both n weghts and prces CPI complers sometmes face proposals from governments or pressure groups to exclude certan categores of product for non-statstcal reasons. Common examples are alcohol and tobacco where n some countres ther consumpton s assocated wth socal stgma or t can only be purchased llegally. Whle t s acceptable to produce a varant of the general CPI excludng such products, the man CPI should nclude them, where practcal, to ensure that the ndex presents a true and accurate pcture of natonal nflaton. 1.5 Chapter 3 covers n some depth the treatment of other excluded or partly-excluded products, ncludng: taxes and lcenses, subscrptons, nsurance, gamblng, fnancal transactons, hre purchase and nterest payments. Imputed transactons and mputed prces 1.51 A dstncton can usefully be made between mputed transactons and actual transactons where a prce s mputed. In the second category would be a prescrbed medcne provded free as part of a natonal health servce. There s a transacton n the sense that a product changes hands but at zero prce so that t doesn t consttute a monetary transacton. The conventons for a CPI constructed for the purposes of ndexaton or the measurement of nflaton as a macroeconomc ndcator exclude from coverage ths category of non-monetary transacton, so no prce should be mputed A CPI should measure the prces of fnal consumpton by a household. In prncple, the frst category, mputed transactons where households do not ncur a fnancal lablty but bear the costs of acqurng the good or servce n another way, should be ncluded n a CPI where used for GDP deflaton and, n prncple, can be ncluded n a CPI compled for other purposes. Perhaps the most mportant example s the consumpton of ownproduced products such as food. Here, there s no actual transacton at all, and thus no prce. If the transacton

10 s to be valued a prce has to be mputed. Ths would usually be done by reference to actual purchases of the same product n, say, nearby markets. But even when ths s done for the purpose of estmatng GDP (whch values consumpton regardless of ts market mechansm) t s not necessarly approprate to nclude t n the general CPI or n a CPI used for ndexaton where the narrowest concept of consumpton that can be used s one based on monetary expendtures only. From the pont of vew of measurng nflaton and also for the purposes of ncome ndexaton, the most common vew s that t s best to omt t on pragmatc grounds although goods and servces purchased by households whch are then used as nputs nto own account producton are normally treated as f they themselves were consumpton goods and servces, and should therefore be ncluded n the CPI. Some countres may fnd t useful to produce two versons of the CPI: one ncludng and the other excludng own-account consumpton. III. DECIDING ON THE INDEX COVERAGE AND CLASSIFICATION STRUCTURE 1.53 The mportance of the product classfcaton system cannot be overstated. Choosng a classfcaton system s the frst step n complng the CPI because ts sub-aggregates must be defned n such a way that the expendture weghts and prces wll relate precsely to the coverage of the sub-aggregates. The classfcaton s mportant also because t establshes a framework from whose boundares the representatve tems for ncluson n the ndex (and sometmes the outlets) wll be defned and drawn. Fnally, the classfcaton system helps n defnng whch level of the herarchy wll be sutable for publcaton Classfcaton s a central theme n the complaton of the CPI. In years past, countres used ther own dstnct systems for classfyng the range of products covered by ther CPI. Most countres have now, however, moved to the nternatonal standard classfcaton COICOP COICOP was frst developed for the UN System of Natonal Accounts (SNA 1993) to provde the structure for classfyng household consumpton expendture. Expendtures on the varous components of household consumpton are often used as the bass for the weghts n the CPI. The 23 ILO Resoluton on CPIs requres that natonal CPI classfcatons should be reconclable wth COICOP at least at ts hgher aggregaton levels. Most countres have adopted COICOP n ther economc statstcs (for example, n the CPI, natonal accounts, Internatonal Comparson Program (ICP), and household expendture surveys), wth a clear advantage for ntegraton of data-sets and enhanced analytcal capabltes. Classfcaton systems: the general case 1.56 In ts broadest sense a classfcaton s a procedure n whch ndvdual tems are organzed nto categores based on nformaton on one or more characterstcs nherent to the tems. A classfcaton scheme wll usually have these same tems (or elements) arranged n a herarchcal ordered system based on category-subcategory relatonshps where the subcategory has the same descrpton as the assocated class n addton to one or more descrptons. For example, an apple s a subclass of frut. So any apple s a frut, but not every frut s an apple. A product needs to have a more detaled descrpton to be an apple than to be a frut In prncple, a classfcaton system can be based on any attrbute of the objects beng classfed. Normally, organsng a populaton of tems nto categores must leave no two categores wth any tem n common, n other words the categores must be mutually exclusve. Also, the categores must collectvely nclude all of the tems whch are n the populaton the categores must be exhaustve. For example, n the case of the CPI, ts classfcaton should nclude the entre unverse of goods and servces that are covered by the ndex (for example, fresh food purchased n a store by a consumer s part of the CPI, whle heavy machnery such as a tractor s not) and no product should be ncluded n two dfferent classes n the structure. The CPI classfcaton system 1.58 In practce, the prncples used to classfy a doman of objects wll depend upon the nature of the objects themselves. For example: ndvduals or ndvdual households mght decde to classfy ther expendture nto expensve tems and cheaper tems as an ad to budgetng; polcymakers mght need products grouped by country

11 of orgn or accordng to ther contrbuton to the nflatonary process; and researchers mght fnd tbenefcal to have goods dentfed as durable or non-durable. In the end, there s a multtude of dfferent approaches that can be used for classfyng the expendtures and prces on goods and servces that are wthn the doman of the CPI. Three of the more popular approaches for aggregatng goods and servces are: Accordng to the purpose they serve. Bread can be classfed accordng to ts purpose n provdng nourshment and thus be located n a category called Food. Ths s the approach used n all levels of the COICOP classfcaton. Accordng to product type. Products can be grouped accordng ether to ther physcal characterstcs, the nature of the economc actvty that produced them, the producton process from whch they orgnated, the type of outlet they are sold from, or where they orgnated. Take for example rolls of alumnum fol and alumnum fol contaners, both of whch could be classfed under a headng Household alumnum products. Clearly these two products share many smlartes whle ther purposes could be qute dfferent. Accordng to economc crtera where such aspects as: degree of substtutablty, degree of complementarty, sales tax applcablty, and country of orgn could be the defnng factors behnd the classfcaton. For example, dfferent varetes of apple (because they are close substtutes) could be classfed nto a category called Apples whereas moble phones and ther accessores (because they are complements) could be classfed nto a group called Moble phones and related products COICOP, as ts name mples, s founded on the prncple of purpose. It s a purpose-type classfcaton because throughout the aggregaton scheme the products are grouped accordng to the purpose (or functon) they usually fulfl such as transport, nourshment, shelter, etc. Most natonal CPIs am at measurng the change of the cost of a basket of goods and servces, whch s consumed for the purpose of satsfyng certan needs. A purposebased classfcaton would therefore appear to be the logcal classfcaton system for a CPI. 1.6 The offcal COICOP s only a 4-dgt classfcaton. NSIs wll expand the COICOP to 5 and 6 dgts to obtan more detal for ther use. At the hgher level of the classfcaton, the products are grouped accordng to purpose. Households wll select varous goods and servces n order to satsfy ther consumpton objectves (.e., rentng an apartment for the provson of shelter or eatng an apple for the purpose of nourshment). These goods and servces are dsaggregated further nto varous groups not based on the prncple of purpose but accordng to product-type. For example, oranges and apples are ncluded n the Frut category. The more detaled breakdown s often a product-type classfcaton because these tems share a smlar producton process and are certanly sold at frut stands or the same locaton n the supermarket. IV. DERIVING THE WEIGHTING PATTERN A CPI measures changes n the cost of a representatve basket of goods and servces. Ths nvolves weghtng together aggregated prces for dfferent categores of goods and servces so that each takes an approprate share to reflect the budgets of the household covered by the ndex. For nstance, f most people spend far more on fresh vegetables than on electrcty then a prce rse for fresh vegetables must have more effect on overall prce rses than a smlar-szed ncrease for electrcty. At the lowest level therefore, each elementary aggregate should receve a weght equal to the rato of expendture by ndex households on goods and servces represented by that aggregate to all expendture by ndex households on tems wthn the scope of the CPI. Chapter 3 dscusses the dervaton and sources of the expendture weghts and provdes detaled gudance on specfc ssues The 23 ILO Resoluton on CPIs makes the obvous but mportant pont that the weghts follow drectly from the scope of the ndex as well as from the choce between the acquston, use or payment approach. It also states that there are two basc sources of nformaton: HESs and natonal accounts and that the weghts should be revewed at least every fve years. Addtonally, new sources of weght nformaton are beng developed such as actual expendtures on varous types of transactons through the use of scanner data and other electronc meda. Such sources are beng evaluated and exploted for use n development of weghts now and n the future.

12 1.62 The use of expendture weghts s consstent n concept wth a CPI based on the acquston, payment and user cost approaches although the treatment of major durable goods and housng can present a problem, partcularly the housng costs of owner-occupers. The use of expendture weghts n a CPI s often referred to as a CPI based on the concept of plutocratc weghts and beng based on total expendture ths concept gves more weght mplctly to the expendture patterns of hgh-spendng households (whch wll also tend to be those wth hgher ncomes) The goods and servces consumed by the households can n prncple be acqured n four ways: Purchase n monetary transactons. From own producton. As payment n knd. As transfers or gfts from other economc unts The weghts themselves are determned by the scope of the CPI and should be derved on bass of the relevant coverage and types of consumpton and wth reference to SNA concepts. The broadest possble scope for goods and servces would cover all four of the above categores. It would nclude all socal transfers n knd n the form of educaton, health and housng and other goods or servces provded free of charge or at nomnal prces. The total acquston of goods and servces thus descrbed s equvalent to total actual consumpton of households n the SNA. For the CPI as a general measure of nflaton the more relevant would be to nclude only goods and servces purchased n monetary transactons by the households. The latter excludes the second and thrd categores. Only monetary expendture generates prces that can be observed for the CPI but ths then leaves outstandng the ssue of owner-occuper housng whch s consdered n detal n Chapter It s aganst ths background, that a CPI often follows the concept of household fnal monetary consumpton expendture (HFMCE), as lad down n the System of Natonal Accounts (SNA 28), the approach often recommended for a CPI beng used as a macroeconomc ndcator, restrcted to the approprate reference populaton, or Index Households, where the CPI s beng used as a compensaton ndex. The latter mght, for example, exclude the very rch But t should be noted that the SNA operates wth many delneatons of consumpton. One of them s household fnal consumpton expendture (HFCE) whch ncludes non-monetary consumpton (such as for owner-occuper housng and consumpton of own producton (food)). HFMCE s certanly a very useful concept but many countres prefer to also nclude some non-monetary expendture (n partcular owner-occuper housng) n ther CPI, sometmes usng mputed costs thus movng the coverage of the ndex closer to HFCE The conceptual ssues relatng to the constructon of weghts are dscussed n detal n Chapter 4. Weghtng structure 1.68 The weghtng structure should follow the aggregaton structure of the CPI. For nstance, f the latter s based on COICOP, then ths s the structure whch should be used for the weghts Addtonal sub-dvsons can be ntroduced where there s further stratfcaton of the sample to nclude geographcal locaton, outlet type or a more detaled product level classfcaton. Thus the weghtng structure wll depend on the sample desgn for prce collecton and complaton and n partcularly the need for more detaled weghts whch may be generated by addtonal sample stratfcaton. In general, NSIs wll collect some prces centrally and adopt up to four levels of samplng stratfcaton for local prce collecton: locatons; outlets wthn locatons; tems wthn dfferent sectons of expendture; and product varetes. The latter s normally conducted n the feld by prce collectors and does not normally nvolve explct weghts (see Chapter 5). Stratfcaton s frequently used to ncrease samplng effcency, especally where the retal market s heterogeneous (see chapter 5), and also operatonal effcency. Some levels are: Central shops weghts to represent a small number of large super-market chans or chan stores whch have unform prces across branches and prces are provded by the shops Head Offces.

13 Stratum weghts. For some types of expendture, purchasng patterns may dffer markedly by regon or type of outlet and n these cases stratfcaton wll mprove the accuracy of tem ndces. For example, each locally collected tem n the ndex could be allocated to one of the dfferent stratum types to allow the best avalable nformaton about purchasng patterns to be ncorporated n the ndex calculaton. Dependng on the structure of the retal market, the stratum types could be: regon and shop type; regon only; shop type only; and no stratfcaton. The assgnment of stratum type wll depend on the nformaton avalable for constructng the weghts for each tem and the number of prces collected per tem. In prncple, all locally collected tems would be stratfed by both regon and shop type, but f the weghts data are unrelable or non-exstent, then the tem may be allocated to another stratum type. Allocaton also partly depends on whch shop types are specfed for the collecton of prces and the number of prces collected. If the rules for the choce of outlets dd not specfy that both a multple and an ndependent should be chosen for an tem, there may be too few prces collected n one of these shop types to make stratfcaton by shop type meanngful. In some nstances, there may be no stratfcaton because research has shown that stratfcaton has lttle effect. The weght of an elementary aggregate,.e. the stratum weght, should reflect the expendture on the entre elementary aggregate and not the weghts of the outlets and tems that have been chosen to represent t. Therefore, f spaghett s chosen as the representatve product under the elementary aggregate wth the headng of pasta products, then the weght of ths category should reflect expendtures for all pasta products and not solely the lower weght of spaghett,.e. the weght of the pasta category wll be represented entrely by spaghett. Smlarly, f an expendture category s dvded nto two elementary aggregates accordng to outlet type, say, open markets and supermarkets, wth correspondng market shares of food sales, 7 percent and 3 percent respectvely, then the same rule as above should apply. For nstance, suppose a sngle store s selected as the representatve outlet for a partcular food tem sold n supermarkets n a country where two supermarket chans have equal sales, then the sales from the sampled store wll account for the total value of the expendture weght of 3 percent; the weght of the elementary aggregate for ths food tem sold n supermarkets should not be 15 percent (.3 x.5).e. a weght based only on the sales of the selected supermarket. Product or tem weghts (n the current context the terms can be nterchangeable). Some products or tems may be ntended only to represent themselves; others represent a subclass of expendture wthn a secton. For nstance, wthn electrcal applances, an electrc cooker may represent only tself and not any other knds of electrcal applances. However, other products or tems wll represent prce changes for a set of products or tems, whch are not all prced, so for these the weght reflects total expendture on all products or tems n the set. For example, a screwdrver may be one of several tems representng all spendng on small tools wthn home mprovement and mantenance materals, and there are other tems wthn the secton representng all spendng on pant, tmber, fttngs and so on. Some tems, such as fresh frut and vegetables, have seasonal weghts that vary over the year. These are covered n Chapter 8 of ths Handbook. Secton weghts. It s common practce to gve each secton an nteger weght n parts per thousand or per hundred so that the sum of the secton weghts s 1 or 1. It s lkely that most of these weghts wll be based on the HES results. The man exceptons wll be some housng sectons, ncludng (where applcable) mortgage nterest payments and deprecaton, where weghts are estmated from other sources (agan see Chapter 11), and for certan other sectons (tobacco, confectonery, soft drnks and alcoholc drnks) where the HES may be thought to under-record expendture and better data are avalable elsewhere. Many countres may also use natonal accounts ndvdual household consumpton estmates where avalable at the COICOP group or class level and the HES expendture dstrbuton at the lower levels. Ths s also an area where scanner data can be used to more accurately reflect the expendture dstrbuton. Implct weghts wthn elementary aggregates 1.7 An un-weghted formula (e.g. Jevons or Dutot see secton on elementary prce ndces) s usually used when aggregatng the ndvdual prce relatves of the sample of products at the elementary aggregate level. Ths practce s usually justfed on the grounds that the requred nformaton such as market shares s smply not