Proposed changes o he compilaion of zone aggregaes for moneary aggregaes and oher variables in he OECD Main Economic Indicaors publicaion Summary The Main Economic Indicaors (MEI) has radiionally published a range of OECD zone aggregaes for he narrow money and broad money conceps of moneary aggregaes. These zones have been compiled on a fixed weigh basis, wih he weighs being updaed every five years coinciding wih he change in OECD base year, he las updae being in he summer of 2004 wih he inroducion of base year 2000=100. The curren variable used o weigh counries daa consiss of he underlying series value for he counry in he year 2000 adjused for purchasing power pariy. There are a number of deficiencies in his curren mehodology for compiling zone aggregaes of moneary aggregaes which have lead he OECD o sudy a new approach. Some of he key problems wih he curren approach are: 1) Fixed weigh zone aggregae indexes are biased for nominal variables such as moneary aggregaes. Ideally, a process of annual chain linking should be used for compiling zone aggregaes for moneary aggregaes which is he curren pracice for oher nominal variables published in he par 1 subjec ables of he MEI, namely he consumer price index, producer price index, hourly earnings in manufacuring and uni labour coss. 2) The definiion of he weighing variable used for each counry should be he same. However, he weighing variable currenly used is no he same across all counries. This leads o some counries having eiher a higher or lower weigh han appropriae. 3) For he OECD zone aggregaes of OECD-oal excluding high inflaion counries and OECD-Europe excluding high inflaion counries, he only counry currenly excluded is Turkey. However, oher counries which experienced high inflaion for long ime periods in he pas (e.g. Mexico, Greece, Porugal) should be excluded from conribuing o he hisorical par of he ime series when hey experienced high inflaion. However, under a fixed weigh approach his is no possible, hus significanly reducing he analyical value of hese zone aggregaes. This aricle provides more deail on each of he above problems, and hen presens a se of indicaive series as proposed new zone aggregaes for moneary aggregaes. New series for OECD-oal excluding high inflaion counries are also proposed for consumer prices indices and producer price indices, ogeher wih he inroducion of chain-linked zone aggregaes (wih re-weighing and linking every five years) for he index of indusrial producion and reail rade volume. All (indicaive1) proposed new series can be reviewed in he Excel spreadshee available a hp://www.oecd.org/daaoecd/58/20/39733396.xls. User feedback on hese proposed changes is now invied by emailing sa.conac@oecd.org. The curren inenion is o inroduce he new series in he February 2008 ediion of he MEI. 1. Fixed weigh vs chain indexes A fixed weigh zone aggregae index uses he same se of counry weighs for he enire ime series. Conversely, chain indexes allow for differen weighs o be used for differen segmens of he ime series which are hen chained ogeher. This chaining may be done frequenly, e.g. annually, a regular inervals (e.g. five yearly), or a some oher irregular inerval. I seems inuiive ha a key advanage of chaining over using fixed weighs would be ha he weighs used o aggregae counry daa relae o a ime period which accords wih he daa being aggregaed and hus allows for changes in he relaive imporance of counries over ime. Algebraically, one can show ha fixed weigh series are also significanly biased for nominal variables such as consumer prices and moneary aggregaes. The same can be rue for volume based variables such as he index of indusrial producion, alhough his can depend on he weighing variable used. Inernaional manuals which provide guidelines for compiling ime series a naional level for boh volume variables (e.g. for componens of he naional accouns, see IMF 2001) and nominal variables (e.g. producer prices, see IMF 2004) srongly recommend he use of chaining a annual frequency. OECD research has shown ha he bias associaed wih no chaining nominal variables such as consumer prices and moneary aggregaes when compiling zone aggregaes is significan, alhough i is less eviden for volume series such as he index of indusrial producion. 1 The proposed series are referred o as indicaive, reflecing he fac ha hey may undergo sligh revisions when implemened wihin he MEI producion sysem. 1
The following simplified algebra is presened o demonsrae he bias associaed wih fixed weigh zone aggregae indexes for nominal variables and how frequen chaining can effecively overcome his. Noaion = Linking poin Z -1 = Zone esimae before linking poin, based on old weighs (series of chain links) Z = Zone esimae for a daa iem a ime C = Value of daa iem repored for counry i a ime (as an index) R = Se of counries belonging o zone Z w i = Weigh of counry i in he laes weighing period relevan for he curren link Chain linked formula for zone aggregaes of indices Z Z w 1 * i * ir Ci, C 1...(1) How does his differ from a fixed weigh formula? Z w * C i ir...(2) Here he value of he zone is recalculaed for all ime poins prior o (i.e. from he sar of he series) whenever he weighing base year is changed Why chain link? Inuiion, does i make sense o combine counries daa for 1960 wih weighs of 2000 or vice versa? Wha does i really mean algebraically? If we sar wih he sandard fixed weigh formula: Z w * C i ir and hen muliply by C -1 /C -1 we obain: Z ir ( w * C ) *( C / C ) i 1 1...(3) Expressing he formula in his way we see ha he weigh of counry i s conribuion o he index value is w i *C -1 and is hus affeced by he level of he series i.e. C -1. Thus counries whose index values have grown faser since he base year (where all counries sar wih he same value, i.e. 100) will have a higher impac on he curren value of he series and hus he zone will be biased upwards. Equivalenly, prior o he base year, for hose counries wih higher growh raes in he pas he C -1 will decrease faser and he zone esimae will be biased upwards. This means zone aggregae growh raes are acually biased upwards afer he base year and downwards before he base year. Hence he formula and degree of bias is dependen on boh he base year and he values of C -1. The difference wih he chain linking formula is ha i only depends on he counries growh rae since he las link. Given ha nominal series (e.g. consumer prices, moneary aggregaes, hourly earnings) are srongly influenced by inflaion which can vary considerably across counries (paricularly in he pas for OECD counries), he poenial for bias in zone aggregaes for hese series is considerable. This concep is perhaps easier o undersand if we look a he formula for zone growh raes. PRINCIPAUX INDICATEURS ÉCONOMIQUES, DÉCEMBRE 2007 - ISSN 0474-5523 - OCDE 2007 IX
Formula for zone growh raes Fixed weigh formula for zone growh raes: ( 1 w * C ) * ( C / C ) i 1 w i * C 1 Z /Z -1 = /...(4) ir ir (In he above formula we have used he expanded version of he fixed weigh zone formula, i.e. equaion (3) as he numeraor and he simple version, i.e. equaion (2) as he denominaor) Equaion 4 can be viewed as a sandard weighed average formula where he weigh of counry i is: w i *C -1. Tha is he weigh of counry i is dependen on he value of he series C -1 (whose level is dependen on he base year). Hence he zone growh raes are biased upwards afer he base year where he value of C -1 ges very high for counries wih high growh raes, and biased downwards prior o he base year where he value of C -1 ges very low for counries wih high growh raes (e.g. he index value for Turkey in broad money (M3) wih a base year of 2000=100 has been rounded o zero by abou 1983). whereas for he chain link formula: C Z /Z -1 =( Z 1 * * wi )/ Z -1 ir C 1 C = w * i...(5) ir C 1 Equaion 5 can be viewed as a sandard weighed average formula where he weigh of counry i is w i (as he sum of he w i is 1). Tha is he weigh of counry i is independen of he value of he series C -1 and hus independen of he choice of base year also. Main conclusion from he heory If we reweigh our zone aggregaes and do no chain link, we risk causing a large bias, in paricular for nominal series where annual linking is recommended o minimise he impac of such poenial bias. As a consequence, a se of experimenal zone aggregae indexes for narrow money and broad money have been compiled on his basis as oulined in Secion 4. In addiion, i is proposed o use his approach for compiling zone aggregaes for reail rade volume and he index of indusry producion. However, empirical research has shown ha five yearly chain-linking is sufficien for hese subjecs o couner he effecs of poenial bias. 2. Weighs used for compiling zone aggregaes for moneary aggregaes For moneary aggregaes he MEI compiles zone aggregaes for: 1) Narrow money, for which he arge variable is M1, defined as currency in circulaion (noes, coins and overnigh deposis). 2) Broad money, for which he arge variable is M3, defined as M1 plus deposis and deb securiies wih a mauriy of up o wo years or redeemable wih a noice up o hree monhs, repurchase agreemens and money marke fund shares / unis. The curren counry weighs used are he value of he above series for each counry in he year 2000, adjused for purchasing power pariy (PPP) using gross domesic produc (GDP) based PPPs. Despie he problems wih he use of fixed weighs as oulined above, oher problems wih his approach include: No all counries have available he arge series of M1 and M3 and hus he naional series which are closes o hese definiions are used as proxies. This is generally accepable for he purpose of assessing he evoluion over ime of moneary aggregaes and making comparisons across counries. However, when used as weighs, counries where proxy series are used will be over or underweighed depending on he inclusions or exclusions of he proxy series relaive o he arge. In some cases his can resul in a significan over or under represenaion of a counry in he relevan zone aggregae. The curren approach uses fixed weighs based on counry daa in 2000 a he ime he weighs are changed. However, counries may ofen subsequenly revise heir ime series and such changes are no refleced in he weighs. X MAIN ECONOMIC INDICATORS, DECEMBER 2007 - ISSN 0474-5523 - OECD 2007
Because moneary aggregaes are nominal series in naional currency, hey mus be convered o a common currency for he purpose of using hem as weighs. The sandard approach used by he OECD is o conver series o a common currency using PPPs. However, PPPs are based on he relaive prices across counries for componens of GDP, and herefore GDP based PPPs may no be enirely appropriae for adjusing moneary aggregae series o a common currency. An alernaive could be o use exchange raes; however he volailiy of exchange raes may lead o large flucuaions in weighs beween years which is also undesirable. A soluion o he above issue would be o use a weighing variable which is uniform across counries, reflecs changes in counries relaive economic srucures and has a relaively sable developmen over ime. The suggesed variable which is commonly used for he purposes of compiling weighs for zone aggregaes is GDP in naional currency a curren prices convered o a common currency using GDP PPPs. This is he weighing variable used o compile he experimenal indexes as presened in Secion 4. 3. Producing zone aggregaes excluding high inflaion counries The Main Economic Indicaors has hisorically published a ime series for he OECD-oal and OECD-Europe zone aggregaes which exclude high inflaion counries for he key nominal series of consumer prices, producer prices, hourly earnings in manufacuring and moneary aggregaes. The definiion of high inflaion counries was previously based on a loosely defined definiion of hose counries having inflaion raes of greaer han 10% per annum on average in he previous en year period where inflaion is measured using he GDP implici price deflaor. A he ime of he change in OECD base year o 2000=100 for MEI in he summer of 2004 which included he updaing of weighs for OECD zones for all subjecs, a decision was made ha only Turkey me he crierion of a high inflaion counry, whereas immediaely prior Mexico, Poland and Hungary had also been included in his group. As a resul, he enire ime series for OECD zone aggregaes excluding high inflaion counries were recompiled only excluding Turkey which lead o subsanial upward revisions o he ime series of year-on-year growh raes for hese series. I would clearly have been more appropriae o produce a hisorical series where he composiion of he group of high inflaion counries is allowed o change over ime, which would naurally require he use of chain-linking raher han fixed weigh series as a pre-requisie. However, sric applicaion of he above definiion of a high inflaion counry would have lead o many counries being included or excluded for differen pars of he ime series. I would also have resuled in large counries such as France, Ialy, Spain and he Unied Kingdom being excluded from he zone aggregae for large pars of he hisorical series which would herefore have diminished he significance of he zone as a saisical benchmark. Consequenly, i became apparen ha he OECD needed o develop a hisorical reference chronology o define periods whereby OECD Member counries would be regarded as high inflaion counries and use his chronology as he basis for compiling he hisorical ime series for zone aggregaes which exclude hese high inflaion counries. The chronology which has been developed as shown in Table 1 below is derived from he following assessmen framework which aimed o be sufficienly robus saisically and also logical from a common sense / pracical perspecive. Saisical assessmens The following wo broad saisical assessmens were considered in regards o deciding wheher a counry should be classified as having high inflaion over a paricular period of ime. (1) A counry should exceed some average rae of inflaion This is a difficul crieria o apply on is own because he resuls depend on he ype of average used (e.g. if a moving average is used should i be over he las 10 years, 5 years, 3 years?). However he main problem is ha average raes of inflaion are dependen on he relaive economic siuaion hroughou ime. For example 10% inflaion would be considered very high in 2000 bu quie moderae in 1975. (2) A counry should exceed some benchmark rae of inflaion A logical poin of comparison would be o compare a counries rae of inflaion as defined by he GDP deflaor o he OECD-oal. A difficul issue hen is o decide how much above he benchmark is olerable (e.g. 10%?) and wheher his should be averaged over some period of ime (e.g. 3, 5 or 10 years?). This measure would also be affeced by he relaive economic siuaion hroughou ime; alhough no o he same exen as in opion 1 above (e.g. 20% inflaion compared o a benchmark rae of 10% migh be considered less significan han 12% compared o a benchmark rae of 2%). Pracical issues / inerpreabiliy for users Only a small number of counries should be classified as high inflaion, oherwise he scope of he zone loses relevance. A counry needs o break a saisical crierion for a cerain lengh of ime o qualify as a high inflaion counry, o avoid is definiion changing ofen hroughou ime. Also, he number of imes a counry changes is definiion of being a high inflaion counry or no should be minimised, ideally no more han once in he ime series. PRINCIPAUX INDICATEURS ÉCONOMIQUES, DÉCEMBRE 2007 - ISSN 0474-5523 - OCDE 2007 XI
A corollary o he above poin is ha if a counry changes from one definiion o anoher on he basis of he saisical crieria his should be ignored where hese changes are of shor duraion. Changes o definiions of a counries inflaion saus should be made a fixed and raher infrequen inervals, and revisions o hisorical ime series as a resul of changes should be minimized. A hisorical chronology of high inflaion counries over periods in he pas should be esablished and say fixed for he consrucion of he hisorical segmen of all OECD zones excluding high inflaion counries which are only compiled for nominal variables. Combining saisical assessmen and pracical consideraions Clearly a degree of subjeciviy was required in applying he above saisical assessmens wih due consideraion for he pracical issues oulined above. From a saisical perspecive he idea of looking a counries raes of inflaion relaive o a benchmark was given more weigh han considering absolue levels of inflaion. A value of 10% more han he OECD-oal for he GDP deflaor was used as a saring poin, bu consideraion of he economic condiions ha exised over he differen periods as shown by he general level of inflaion was also aken ino accoun. The approach aken was o observe he run of years ha a counry exceeded he benchmark by 10% or more, ignoring years in beween ha may be below his if hey are irregular (e.g. a year or wo here or here) or sill relaively high (e.g. exceeding he benchmark by greaer han 7%). Shor erm moving averages above he benchmark (i.e. 3 year or 5 year) were also used as a guide in performing his assessmen. Wih regards o he pracical consideraions, i seemed sensible ha changes o he composiion of excluding high inflaion counries zone aggregaes be made a 5 year inervals for easy inerpreaion by users, and also allowing fuure reviews o ake place in conjuncion wih he five yearly rebase program. If a counries definiion were changed, his would be back daed o he laes rebase period and he series linked a his poin. The hisorical chronology of high inflaion counries which resuled from his exercise is oulined in Table 1. Table 1: Counries o be declared as high inflaion when compiling hisorical "OECD excluding high inflaion counries" zones Counry Exclusion Period Commens Czech Republic Unil 1995 Time series only considered from 1990 Greece Unil 1995 High inflaion from mid 70 s o early 90 s Hungary Unil 2000 Time series only considered from 1990 Iceland Unil 1990 Very high inflaion in he 70 s and 80 s Mexico Unil 2000 Very high inflaion unil recen years Poland Unil 2000 Time series only considered from 1990 Porugal Unil 1990 High inflaion from mid 70 s o lae 80 s Slovak Republic Unil 1995 Time series only considered from 1990 Turkey All periods Repea offender The ime series for he former easern European counries are only considered for inclusion in OECD zone aggregaes from he period where hey become a marke economy, which is approximaed as 1990. I should also be noed ha he above chronology assumes a saring poin for he hisorical ime series of zone aggregaes a 1970 a he earlies, which is he case for consumer prices which has he longes ime series of zone aggregaes. XII MAIN ECONOMIC INDICATORS, DECEMBER 2007 - ISSN 0474-5523 - OECD 2007
4. New series proposed for he MEI In accordance wih he issues oulined in his documen, he OECD proposes o revise a number of is currenly published zone aggregaes as follows: Inroducion of annual chain linked indexes for all moneary aggregae zone aggregaes wih ime series back o 1970 in mos cases, o replace he curren fixed weigh series where ime series commence in 1990. The weighs used will be GDP a curren prices adjused for PPPs. Inroducion of five yearly chain linked indexes for indusrial producion and reail rade volume zone aggregaes, o replace he curren fixed weigh series (noe five yearly chain linked indexes were inroduced for he OECD Composie Leading Indicaors in July 2007). Compilaion of zone aggregaes for OECD-oal excluding high inflaion counries and OECD-Europe excluding high inflaion counries based on he hisorical chronology presened in Table 1. These new series would replace he exising excluding high inflaion counries zone aggregaes series which only exclude Turkey and are compiled for he nominal variables: consumer price index (all-iems, energy, food, and all-iems less food less energy); producer price index; hourly earnings in manufacuring and; moneary aggregaes (narrow money and broad money). All newly proposed series in comparison o hose currenly published which hey are inended o replace are shown in he Excel spreadshee "MEI new zone aggregaes.xls" available a hp://www.oecd.org/daaoecd/58/20/39733396.xls. The chars below show he difference in he ime series of year-on-year growh raes beween he currenly published and proposed series for moneary aggregaes, narrow money, OECD-oal and he consumer price index, all-iems, OECD-oal excluding high inflaion counries. In he firs, he bias in he currenly published series is apparen, which is downwards prior o he curren base year of 2000 and upwards aferwards as explained heoreically in Secion 1. For he second, we see he impac of excluding counries oher han Turkey in he hisorical pars of he ime series as discussed in Secion 3. PRINCIPAUX INDICATEURS ÉCONOMIQUES, DÉCEMBRE 2007 - ISSN 0474-5523 - OCDE 2007 XIII
References IMF, 2001. Quarerly Naional Accouns Manual Conceps, Daa sources, and Compilaion: Chaper 9 Price and Volume measures: Specific QNA/ANA issues hp://www.imf.org/exernal/pubs/f/qna/2000/texbook/ch9.pdf IMF, 2004. Producer Price Index Manual: Theory and Pracice. Inernaional Moneary Fund, Washingon D.C. hp://www.imf.org/exernal/np/sa/egppi/index.hm XIV MAIN ECONOMIC INDICATORS, DECEMBER 2007 - ISSN 0474-5523 - OECD 2007