The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
J U L Y / A U G U S T 1 9 9 5 • V O L U M E 77, N U M B ER 4 E M U : W ill It Fly? C hanges in In v e n to ry M a n a g e m e n t a n d the Business Cycle Is Th e re a Case fo r " M o d e r a t e " In fla tio n ? E va lu a tin g the Efficiency of C o m m ercial B anks: Does O u r V ie w of W h a t Banks Do M a tte r? President Th om a s C. M e lz e r Director o f Research W illia m G . D e w a ld A ssociate Director o f Research Cletus C. C o u g h lin Research Coordinator and Review Editor W illia m T. G a v in Banking R. A lto n G ilb e rt D a v id C. W h e e lo c k International C h ristoph e r J . N e e ly M ic h a e l R. P a k k o P a tricia S. P o lla rd M acroeconom ics D o n a ld S. A lle n R ichard G . A n d e rso n Ja m es B. B u lla rd M ic h a e l J . D u e k e r Joseph A . R itter D a n ie l L. Th o rn to n P eter Yoo Regional M ic h e lle A . C la rk K e v in L. K liesen A d a m M . Z a re ts k y Director o f Editorial Services D a n ie l P. B re n nan Managing Editor Charles B . H enderson G raphic Designers B ria n D . E b ert Inocencio P. Boc Review is published six times per year by the Research Department of the Federal Reserve Bank o f St. Louis. Single-copy subscriptions are available free o f charge. Send requests to: Federal Reserve Bank o f St. Louis, Pu blic A ffairs D epartm ent, RO. B ox 4 4 2 , St. Louis, Missouri 6 3 166-0442. The views expressed are those o f the individual authors and do not necessarily reflect official positions o f the Federal Reserve Bank of St. Louis, the Federal Reserve System, or the Board o f G overnors. A rticles may be reprinted or reproduced if the source is credited. Please provide the Public Affairs Department with a copy o f the reprinted materials. REVIEW JULY/AUGUST 1995 27 Is Th e re a Case fo r " M o d e r a t e " In fla tio n ? A lv in L. M a rty a n d D a n ie l L. T h o rn to n V o lu m e 7 7 , N u m b e r 4 3 E M U : W ill It Fly? P a tricia S . P o lla rd W ill a monetary union be established in Europe before the end of this century? Prior to the establishment of a mone tary union within the European Union, countries are expected to meet desig nated criteria regarding their interest rates, inflation rates, government finances and exchange rates. These criteria were designed to ensure a com m itm ent to price stability and economic convergence among potential entrants. Patricia S. Pollard examines the progress that has been made in fulfilling these criteria and the countries’ prospects for meeting the entry criteria. Furthermore, she examines the options available to the European Union if there aren’t enough countries to fulfill the criteria before the target date for monetary union. 17 C hanges in In v e n to ry M a n a g e m e n t a n d the Business Cycle D o n a ld S. A lle n The change in business inventory has always played a major role during down turns in the business cycle. Innovations in inventory management, such as “justin-tim e” methods and computerized stock management using bar code scan ning, have become more popular over the last decade. Donald S. Allen looks at whether these changes are expected to have a major impact on the amplitude or duration o f the business cycle. Two objections to making price stability the primary objective of monetary policy are: (1) Moderate inflation is good for the economy and (2) It is less costly to live with moderate inflation than to eliminate it. Reviewing the first objec tion, Alvin L. Marty and Daniel L. Thornton consider four arguments, namely, that moderate inflation enhances econom ic stability, increases output per person, increases the efficiency of inter industry wage adjustments, and enhances the efficacy of countercyclical monetary policy. Considering the second point, they argue that concern for transitional unemployment is a frail foundation for a policy of moderate inflation. 39 E v a lu a tin g the Efficiency of Com m ercial B anks: Does O u r V ie w of W h a t B anks Do M a tte r? D a v id C. W h e e lo c k a n d Pawl W . W ils o n An inefficient business wastes resources, either by producing less than the feasible level of output from a given amount of input or by using excessive input to produce a given amount of output. Researchers often find that banks are quite inefficient, but don’t agree on how best to measure that inefficiency, or even how to measure bank produc tion. David C. W heelock and Paul W. W ilson show that the average estimated inefficiency, and even the ranking of banks by their inefficiency, is sensitive to whether bank loans and deposits are measured in dollar amounts, or the number of loans and accounts. Their research indicates that in the absence o f a standard view o f how to measure bank production, the extent to which banks are inefficient will remain an open question. REVIEW JULY/AUGUST 1995 Patricia S. Pollard is an economist at the Federal Reserve Bank of St. Louis. Jerram C. Betts provided research assistance. EMU: W ill It Fly? treaty creates a series of criteria which coun tries must meet to jo in the monetary union. These criteria are designed to ensure that potential entrants share a commitment to that union. Much has been written critiquing the usefulness of econom ic convergence prior to monetary union.2 Some papers, such as De Grauwe (1 9 9 4 ), focus on whether the convergence indicators detailed in the treaty are the proper indicators to ensure a well-functioning monetary union. This article does not enter this discussion; rather, given the criteria established by the M aastricht Treaty, it assesses the progress of the members of the European Union in meeting these criteria. After illustrating the lack of progress of the EU in meeting them, I consider the two main alternatives available to the member states that hope to achieve monetary union in the near future. One is to allow latitude in the application of the convergence criteria and the other is to view the starting date for monetary union as flexible. Patricia S. Pollard n December 1991, the leaders of the member states of the European Union met in Maastricht, the Netherlands, to conclude the negotiations on a Treaty on European Union. The M aastricht Treaty, as it is commonly known, encompasses a wide range of issues, from foreign affairs and secu rity policy to citizenship, health and tourism. Primarily, however, the M aastricht Treaty is known for formalizing the intentions of the member states of the European Union to cre ate an economic and monetary union (EMU) by the end of this century. The main features of EMU are the creation of a single monetary policymaking body and a single currency for the European Union. W hile EMU seemed certain in December 1991, within a year the outlook had turned much bleaker. In a referendum in June 1992, Danish voters rejected the treaty. This was followed by a series of exchange rate crises affecting the European Union in 1992 and 1993. Despite these setbacks, the Maastricht Treaty was ultimately approved by all member states (a second referendum passed in Denmark in 1993) and the treaty entered into force on November 1, 1993. In accordance with the treaty, the European Union is laying the groundwork for monetary union: creating the institutions and studying the technical details necessary to meld as many as 14 independent monetary policy making bodies into one cohesive system.1 Furthermore, to make themselves eligible for entry into EMU, countries are undertaking policies aimed at achieving econom ic con vergence across the European Union. This economic conversion is seen as an integral part of the process toward monetary union. Indeed, the M aastricht Treaty is based on the idea that econom ic convergence is a prerequisite for monetary union. The I BACKGROUND Serious discussion in Europe of a move to monetary union began in 1988 with the decision of the European Council to create a Committee for the Study of Econom ic and Monetary Union. This committee was chaired by Jacques Delors, the president of the European Commission.3 The Delors Committee, as it was commonly known, was given a mandate to examine the issue of EMU and to develop a program aimed at its imple mentation. In 1989, the committee issued a report stating: “Econom ic and monetary union in Europe would imply complete freedom of movement for persons, goods, services and capital, as well as irrevocably fixed exchange rates between national currencies and 3 1 Belgium and Luxembourg already operate in a monetary union. 2 See, for example, De Grauwe (1 9 9 2 ) and Portes (19 9 3 ). 3 See the shaded insert, "Institutions of the European Union" on page 2 for an explanation of the institutional structure of the European Union. REVIEW JULY/AUGUST 1995 INSTITUTIONS OF THE EUROPEAN UNION The European Com m ission is the executive branch of the European Union government. The president of the commission, who serves a two-year renewable term, is chosen by the European Council. The other 19 commissioners are appointed by their national governments for four-year renewable terms. France, Germany, Italy and the United Kingdom each appoint two commissioners and the remaining 11 EU coun tries each appoint one commissioner. Although the president of the commission has no control over the selection of commissioners, he does control the selection of the portfolios assigned to each commissioner. During their term in office, the commissioners are expected to repre sent the interests of the European Union, not those of their home countries. The Council o f Ministers consists of the representatives of the national governments. The composition of the Council of Ministers depends on the issue being considered. For example, issues related to the Common Agricultural Policy are addressed by the agri cultural ministers of the member states, where as finance matters are addressed by the finance ministers. W ithin the Council of Ministers, each country is allocated a number of votes based loosely on the size of its population. France, Germany, Italy and the United Kingdom have 10 votes each. Spain has eight. Belgium, Greece, the Netherlands, Portugal and Sweden have five votes each. The remaining countries, Austria, Denmark, Finland and Ireland, have three votes each. In sum, there are 85 votes. To pass by qualified majority, a measure must receive at least 61 votes. Thus, two large states and two small states can form a blocking coalition. The European Council consists of the heads of state or government of the member coun tries. The president of the European Commission is a non-voting member o f the European Council. The presidency of the European Council rotates among the member states on a six-m onth basis. The European Council holds a meeting at the end of the six-month period (in December and Jun e). The European Parliam ent is the legislative branch of the European Union. The 626 members o f Parliament are elected in national elections and serve renewable five-year terms. In the Parliament, members are grouped according to their party affiliation, not their nationality. The European Parliament is the weakest institution within the European Union, having mainly consultative powers. The exception to this weakness is in budgetary issues, over which it has considerable control. The European Parliament may dismiss the European Commission en masse, but cannot dismiss individual members of the Commission. In the plan suggested by the Delors Report, and incorporated in the Maastricht Treaty, EMU was to be achieved in three stages. Broadly speaking, stage one would emphasize econom ic convergence and stage two would emphasize institutional conver gence. The final steps to full EMU would occur during stage three. During stage one, which began in July 1990, the member countries of the European Union were to achieve greater convergence in econom ic performance through increased policy coordination. Stage one was also to be characterized by the completion of the single internal market and removal of all finally, a single currency. This, in turn, would imply a common monetary policy and require a high degree of compatibility of economic policies and consistency in a number of other policy areas, particularly the fiscal field. These poli cies should be geared to price stability, balanced growth, converging standards of living, high employment and external equilibrium” (Committee for the Study of Econom ic and Monetary Union, 1989, p. 17). The recommendations of the Delors Committee formed the basis for the negotia tions on EMU in the Maastricht Treaty. NK OF ST . L OU I S 4 REVIEW JULY/AUGUST the monetary union, acting in consultation with the European Commission and the European Central Bank, will determine the exchange rates at which currencies are to be fixed. The determination of these fixed exchange rates requires the unanimous con sent of the member states. As the final step to EMU, individual currencies will be replaced with a common currency. Monetary policy decisions will be made by the independent, supranational European Central Bank. According to the Maastricht Treaty, stage three must start by January 1, 1999. The exact starting date will be deter mined as follows. By December 1996, an inter-governmental conference comprised of the leaders of the European Union countries must meet to determine if EMU is ready to commence. Prior to this meeting, the European Commission and the EMI are to issue reports detailing the progress made by each country in meeting the convergence criteria. These reports will be sent to the Council of Ministers. The Council of Ministers will use these reports to determine: capital controls.4 In addition, all currencies would be linked in the Exchange Rate Mechanism (ERM ), and procedures would be established for budgetary policy coordina tion.3 The goals for the completion of stage one have yet to be met because the currencies of five countries do not participate in the ERM. In accordance with the Maastricht Treaty stage two began o n janu ary 1, 1994. During this stage, the member states are to make their central banks independent. As part of the steps toward independence, central banks are prohibited from providing overdraft facilities to their governments and from directly financing the government debt. The European Monetary Institute (EMI) began operations at the start of stage two. It is charged with ensuring cooperation between national central banks and strength ening the coordination of national monetary policies. The EMI is also to begin prepara tions for a single currency and the conduct of a single monetary policy. Perhaps most importantly in this regard, it is to create the instruments and procedures necessary for the operation of a single European monetary policy. Also, during stage two, countries are to achieve further economic convergence, as detailed by the criteria in the Maastricht Treaty. The most important role of the EMI is to ensure that the technical barriers to EMU are removed prior to the start of stage three. These barriers include cross-country differ ences in the conduct of monetary policy, financial regulations, payments systems and currencies. The EMI is studying issues related to the conduct of monetary policy. For example, should the future European Central Bank target the money supply as the German Bundesbank does, or should it target infla tion, as the Bank of England does? Another issue being studied by the EMI is the design and implementation of the single currency system. This is a politically volatile issue because each country has an interest in hav ing the new currency resemble its own. Stage three will mark the final transition to a full-fledged monetary union. At the start of stage three, exchange rates between member countries will be permanently fixed. The governments o f the member countries of 1995 • whether each member state fulfills the necessary conditions for the adoption of a single currency; and • whether a majority of the member states fulfill the necessary conditions for the adoption of a single currency ( Treaty on European Union, Article 109j.2). The decisions of the Council of Ministers will be made on the basis of a “qualified” majority vote. The determinations of the Council of Ministers will be forwarded to the European Parliament, which will make its own recommendation on the readiness of the member states to move to the final stage of monetary union. Taking into account the decisions of the Council of Ministers and the European Parliament, the European Council at the inter-governmental conference must then decide, again by qualified majority: FEDERAL RESERVE B A N K OF ST. L OU I S 5 4 In accordance with the Maastricht Treaty, Greece was allowed to maintain capital controls until the end of June 1994. 5 The Exchange Rate Mechanism, created in 1979, set narrow margins for exchange rate fluctua tions between member countries. Normally, each currency was allowed to fluctuate by ± 2.25 percentage points agoinst any other member currency. Some currencies, however, were given wider margins of fluctuation ( ± 6 percentage points) to smooth their transition upon entering the ERM. REVIEW JULY/AUGUST 1995 according to the treaty, EMU will commence for those countries (however few) that meet Progress in M e e tin g C o nve rgence C rite ria the entry conditions. The countries that do not meet the entry Number of Criteria Met conditions and are excluded from EMU will, 1990 1991 1992 1993 1994 according to the Treaty, be referred to as “member states with a derogation” ( Treaty on Belgium 2 3 33 3 European Union, Article 109k.2). This exclu Denm ark 5 4 43 3 sion, however, need not be permanent. At France 5 5 4 4 4 least once every two years, following the Germ any 5 4 4 3 5 guidelines outlined above, the European Greece 0 0 0 0 0 Council will decide by qualified majority Ireland 4 4 4 3 3 which member states with a derogation have Italy 0 00 0 0 fulfilled the entry criteria and admit them to Luxembourg 5 5 54 5 the monetary union. T a b le 1 Netherlands 3 34 Portugal Spain 11 3 00 United Kingdom 3 Austria 4 3 0 0 1 1 3 4 0 CONVERGENCE CRITERIA 0 2 2 3 2 Finland 2 2 1 Sweden 2 3 21 Num ber meeting criteria 4 2 1 0 2 • whether a majority of the member states meet the necessary conditions for monetary union; • whether it is appropriate ... to enter the third stage; and if so, • set the date for the beginning of the third stage ( Treaty on European Union, Article 109j.3). s See Protocolon the Convergence Criteria referred to in Article 109j of the Treaty Establishing the European Community (1 9 9 2 ) and Protocol on the Excessive Deficit Procedure (19 9 2 ). If no date for the start of monetary union has been set by the end of 1997, the treaty obligates the leaders of the European Union countries to meet by July 1, 1998, to determine, based on the same procedure outlined above, which member states fulfill the conditions for monetary union. These states are then to enter the third stage o n janu ary 1, 1999. For m one tary union to begin prior to 1999, a majority of countries must meet the criteria established by the M aastricht Treaty. However, in 1999, As noted above, entry into EMU is 3 dependent upon the fulfillment of what the M aastricht Treaty calls “necessary condi 3 tions.” W hat are these conditions? First, to 1 2 facilitate the common monetary policy, each 1 member must guarantee the independence of its central bank and pass national legislation in accordance with the protocol establishing the European Central Bank. Second, in mak ing their reports on the progress of countries in meeting the necessary conditions, the European Commission and the EMI are to consider the progress made in developing a comm on currency, “the results of the integra tion of markets, the situation and development of the balances of payments on account and an examination of the development of unit labour costs and other price indices” (Treaty on European Union, Article 1 0 9 j.l). Most attention, however, has been focused on the conditions that the Maastricht Treaty says are designed to ensure “the achievement of a high degree of sustainable convergence” ( Treaty on European Union, Article 1 0 9 j.l). Convergence must be achieved in exchange rates, inflation rates, long-term interest rates and government finances. The treaty and two separate protocols detail these convergence criteria as follows:6 • The currency of each member state must have remained within the normal fluctuation margins of the ERM for a least two years prior NK OF ST. 6 LOUIS REVIEW JULY/AUGUST Union improved slightly with Germany and Luxembourg meeting all five criteria. As the performance of the countries in 1990 and 1994 is compared, only Belgium improved its overall performance on the cri teria. In contrast, six countries met fewer criteria in 1994 than they met in 1990. This worsening performance reflects the crises in the ERM and a deterioration in the public finances of many countries. to the examination. Specifically, a member state may not have devalued its currency against any other currency within the ERM on its own initiative. • The average inflation rate for any member state during the year prior to the examination by the European Commission must have been no more than 1.5 percentage points above the average rate of inflation in the three best-performing countries during this same period. Exchange Rate Criterion Although the ERM had functioned smoothly since 1987, it was beset by a series of crises during 1992 and 1993. These crises resulted in the September 1992 withdrawal of the British pound and the Italian lira from the ERM, and the February 1993 devaluation of the Irish pound. The Portuguese escudo and the Spanish peseta were devalued several times throughout 1992 and 1993. As a result of these crises, fewer countries met the exchange rate convergence criterion in 1994 than in 1990 (see Table 2). The exchange rate crises ended in August 1993 with the expansion of the bilat eral bands from ± 2 .2 5 percent to ±15 percent for all pairs of currencies with the exception of the Dutch krona/Deutsche mark. The consensus within the European Union is that these wider bands have reduced currency speculation and thus have lessened the prospects for exchange rate crises within the ERM. Thus, no return to the narrow margins is likely. The maintenance of the expanded margins presents no problem for the fulfillment of the convergence criteria as long as the European Commission and the European Council agree that the treaty’s ref erence to “normal fluctuating margins” means margins of ±15 percent. In March 1995, the currencies within the ERM again experienced sharp fluctuations. The movement in the exchange markets away from dollars and into Deutsche marks caused problems for weaker currencies within the ERM. As a result o f this turbulence, the escudo and the peseta were both devalued. In the absence of any further devaluations, only eight of the 15 member countries of the European Union would meet the exchange • The long-term interest rate (on government bonds or comparable securities) of any member state during the year prior to the examination by the European Commission must have been no more than 2 percentage points above the average long-term interest rate of the three countries with the lowest inflation rates during this same period. • The government budget deficit of any member state may not exceed 3 percent of that country’s GDP at the time of the examination. • The government debt of any member state may not exceed 60 percent of the country’s GDP at the time of the examination.7 Table 1 summarizes the performance of each current EU member state in fulfilling the convergence criteria during the years 1990-94. As this table shows, the path toward convergence has not been smooth. On the basis of these five criteria, more countries met the eligibility requirement in 1990, the year before the treaty was conclud ed, than in any subsequent year. Denmark, France, Germany and Luxembourg met all five convergence criteria in 1990.8 The number of countries fulfilling the criteria declined in each following year, reaching a low of zero in 1993. In 1994, the perfor mance of the members of the European 1995 7 7 As discussed in ttie Protocol on the Excessive Deficit Procedure, the deficit and debt ratios ate based on general government budgets, that is, the central government, regional or local governments and social security funds. Commercial opera tions of the public sector are excluded. The deficit is defined as net borrowing by the government. Net borrowing excludes any portion of the deficit that is used for "the acquisition of loans or other finan cial assets" by the government. Thus, for example, the funds bor rowed by the German government that were in turn lent to agencies in eastern Germany do not show up in these deficit figures (Collignon ond others, 1 994). Privatization pro ceeds connot be used to reduce the deficit, although some countries ore trying to change this provision. Whereas the deficit ratio is based on net borrowing, the debt ratio is based on gross debt. 8 If Austria had been o member of the European Union, it too would have met all five convergence crite ria in 1990. Although it was not a member of the ERM, its currency has shadowed the Deutsche mork. REVIEW JULY/AUGUST countries met the inflation criterion in 1990. This number fell to five in 1993, but rebounded strongly with 11 countries meet ing the criterion in 1994. Greece, Italy, Portugal and Spain were the countries with yes inflation rates exceeding the criterion in 1994. yes Although these four countries have not met the criterion in any year, each yes country has made progress in lowering its yes inflation rate over the period in question. n.m The econom ic recovery currently under no way in Europe is expected to lead to a slight n.m. increase in inflation in most member countries yes by 1996. Because the criterion is based on yes the performance of the three countries with no the lowest inflation, a general increase in the no rate of inflation will not affect the overall n.m. performance o f countries. As shown in Table 3, the increase in the inflation forecast for n.m. is not expected to reduce the num 1996 ber n.m. countries satisfying the inflation of criterion. Moreover, the inflation perfor n.m. mance of the countries not currently meeting the criterion is expected to improve over the 6 next two years. Ta b le 2 C o nve rgence In dicators: Exchange Rate 1990 1991 1992 1993 1994 Belgium yes yes yes yes Denm ark yes yes yes yes France yes yes yes yes Germ any yes yes yes Greece n.m. n.m. Ireland yes yes yes n.m. n.m. yes no Italy no no no Luxembourg yes yes yes yes yes yes yes n.m. n.m. no no Netherlands Portugal n.m. yes Spain no no no no United Kingdom no no no n.m. Austria n.m. n.m. n.m. n.m. Finland n.m. n.m. n.m. n.m. Sweden n.m. n.m. n.m. n.m. Num ber meeting criterion 7 7 7 6 Notes: n.m. indicates that the country w as not a member of the ERM during any part of the relevant year. The Irish pound wos devalued by 1 0 percent in February 1 9 9 3 . The Italian lira wos devalued by 3.7 percent in January 1 9 9 0 when it was incorporated into the narrow (2 .2 5 percent) bands. The lira left the ERM in September 19 9 2 . The Portuguese escudo w as devalued by 6 percent in November 1 9 9 2 and by 6.5 percent in M o y 19 9 3 . The Spanish peseta w as devalued by 5 percent in September 1 9 9 2 , by 6 percent in November 1 9 9 2 and by 8 percent in M a y 19 9 3 . 1995 Interest Rate Criterion The interest rate criterion has been the one that countries have usually found easiest to meet. Furthermore, the member coun tries showed steady improvement over the period 1990-94. In 1990, as shown in Table 4, nine countries had long-term interest rates within the lim it set forth in the Maastricht Treaty. This number rose to 10 in 1991 and increased to 11 in 1993. In 1994, however, the number of countries meeting the interest rate criterion slipped back to 10. In 1994, Greece, Italy, Portugal, Spain and Sweden did not meet this criterion. The former four have never met the criterion. rate criterion at the end of 1996. The currencies of five countries — Finland, Greece, Italy, Sweden and the United Kingdom— are not participating in the ERM and therefore will not meet the two-year rule by the end of 1996. As a result of the recent devaluations of their currencies, Portugal and Spain will not meet the criterion by the end of 1996. Public Finance Criteria Inflation Criterion The two public finance criteria have caused the biggest problems for countries in their quest to jo in the EMU. In 1990, nine Comparing 1990 to 1994, the perfor mance of the EU countries with regard to the inflation criterion has improved. As shown in Table 3, seven of the present 15 EU of the current 15 EU countries met the deficit criterion while only three met it in 1994. Similarly, nine countries met the gov ernment debt criterion in 1990 but only four NK OF ST. L O U I S 8 JU1Y/ AUGUST 1995 Ta b le 3 : Inflatio n C onvergence P e rc e n t 1990 1991 Belgium 3.7 2.5 Denm ark 2.7 2.4 1.8 3.2 2.4 France 2.8 1992 1994 1993 2.4 2.6 2.1 1.9 2.2 1.7 2.3 1.8 1.0 2.8 4 .0 4.7 3.8 2.7 18 .8 15.1 10.9 13 .6 Ireland 1.4 2.5 Italy 5.9 6.9 Luxembourg 3.6 2.9 2.8 Netherlands 2.2 3 .2 3 .0 11 .7 12 .5 10 .0 Spain 6.5 6.4 6.4 United Kingdom 5.5 7.4 4.7 Austria 3.1 3.4 Finland 6.0 5.6 Sweden 9.6 10 .2 2.2 5.8 .0 3 Convergence criterion 3.6 4.0 3.6 3.13.4 Number meeting criterion 7 8 7 5 11 Portugal 2.8 1.6 3.6 2.1 3.9 2.1 2.3 5.6 2.5 8.9 9.6 3.0 2.9 2.7 5.2 2.2 2.2 7.9 5.1 4.5 4.5 4.5 3 .0 3.3 1.6 2.5 2.2 4.9 3.4 2.5 3.5 4.5 2.3 1.8 5.1 3.9 4.1 2.7 5.14.7 5.2 2.4 1.9 19 .2 Germany Greece 1996 1995 2.8 1.7 3.0 2.9 3.3 3.2 3.2 3 .3 3.7 11 11 Notes: Prior to 1 9 9 2 , data for Germany is for western Germany only. Data for 1 9 9 5 and 1 9 9 6 are forecasts Convergence criterion is based on data for the 12 member states prior to 1 9 9 5 and the 15 states thereafter. SOURCE: European Economy (April/May 1 9 9 5 , Supplement A, Table 1 0 ) did in 1994. Much of this decline can be attributed to the expansionary nature of fiscal policies in reaction to the recession of the early 1990s, from which Europe is ju st beginning to recover. The effect of the reces sion on public finances can be seen by considering the example of Finland. Output growth in Finland fell from 5.7 percent in 1989 to -7 .1 percent in 1991. Consequently, Finland’s government budget balance declined from 5.4 percent of GDP in 1990 to a low of -7 .8 percent in 1993. The govern ment budget deficit shrank in 1994 as its economy moved out of recession. The econom ic recovery currently under way in Europe is expected to lead to a grad ual improvement in the budget balances of the EU countries. Nevertheless, only six of the 15 countries are expected to meet the budget deficit criterion in 1996. The recovery is expected to have less of an effect on countries’ performance with respect to the debt criterion. The ratio of debt to GDP is expected to increase through 1996 in m ost countries. The criterion limiting the government debt to 60 percent of GDP has been the most difficult for countries to meet. Only Luxembourg has a debt ratio well below that level. The other three countries that met this criterion in 1994 (France, Germany and the United Kingdom) all have debt-to-GDP ratios close to 50 percent. Among those countries not meeting the criterion, some have debt ratios so high that they would have to run substantial budget surpluses for a number of years to meet it. For example, Buiter, Cosetti and Roubini (1 993) calculat ed that based on the 1991 debt levels and assuming a 5 percent nominal GDP growth 9 m i n i m JULY/AUGUST Ta b le 4 C o nve rg ence In dicators: Lo n g -Te rm Interest Rates P e rc e n t 1990 Belgium 10.1 Denm ark 1991 11.0 1992 9.3 8.9 8.6 18.5 Ireland 10.1 Italy Luxembourg 8.8 9.5 7.0 9.0 8.0 18.8 9.2 18.2 9.1 7.7 13.0 8.2 7.5 6.3 17.7 13.4 8.6 7.2 10.1 10.4 Germany 1994 8.6 10.1 France Greece 1993 n.a. 1995 most member states are expected to improve through 1996, the debt ratios are unlikely to show significant improvement. Turning to the exchange rate criterion, five countries are not members of the ERM and thus do not meet the convergence criterion. For the remaining 10 counties, although the wider 7.8 bands eliminated tensions within the ERM 8.5 between August 1993 and March 1995, there is now evidence that even these bands can not 6.7 prevent pressure from accumulating on weak currencies. 8.1 13.7 7.9 11.3 6.9 10.6 PROSPECTS FOR EMU 6.4 6.7For the 1996 inter-governmental confer 7.2 ence to set a date for monetary union, eight Portugal 15.4 16.8 18.3 12.5 10.0 countries must fulfill all of the convergence 12.4 Spain 14.7 12.2 10.2 9.7 criteria. If there are no further devaluations United Kingdom 11.8 9.9 9.1 7.8 8.2 within the ERM, eight countries— Austria, Belgium, Denmark, France, Germany, Austria 8.7 8.6 8.3 6.6 6.7 Ireland, Luxembourg and the Netherlands— Finland 13.2 11.9 12.1 8.4 8.2 will fulfill the exchange rate criterion in Sweden 13.6 10.9 10.4 8.5 1996. Thus, if9.5 EMU is to get off the ground prior to 1999, all eight of these countries Convergence criterion 12.0 11.5 11.2 9.7 9.4 must meet the other four convergence crite Num ber meeting criterion 9 10 10 11 ria. However, the debt/GDP ratios of four 10 of these countries— Belgium, Denmark, Ireland and the Netherlands— are not SOURCES: European Economy (1 9 9 5 , Number 59, Table 54 ) expected to be close to the 60 percent refer and OECD Economic Outlook (June 1 9 9 5 , Number 57, Annex Table 3 6 ) ence value by the end o f 1996. Thus, based on the five convergence criteria, it is almost certain that a majority rate, Belgium needs a government surplus of of the EU countries will not be ready for more than 9 percent of GDP a year for each monetary union when the inter-govemmental year through 1996 to meet the convergence conference is held in 1996. If EMU is criteria. To meet the criteria by the end of postponed, the next issue is: How many 1998, Belgium would need an annual gov countries will be eligible at the start of ernment surplus greater than 5 percent of 1999, the last possible date for monetary GDP. union in accordance with the treaty? Barring unforeseen econom ic shocks, Germany and Luxembourg should both be eligible Summary on Convergence for monetary union. The eligibility of the To summarize, the data indicate that remaining 13 countries is less certain, even inflation and interest rate convergence are leaving aside the uncertain future o f the ERM. Austria and France are the most likely taking place in the European Union. The outlook for the next two years anticipates additional candidates. Both, however, could run into problems meeting the government further convergence with respect to these budget requirement, and Austria is not two criteria. In contrast, the public finances of the EU members have worsened since the expected to meet the debt criterion. Belgium and Italy have public debts establishment of the convergence criteria. Although the government budget balances of totaling more than 100 percent of their Netherlands 9.0 8.7 8.1 N K OF ST. L OU I S 10 REVIEW JULY/AUGUST Responses to the Lack of Progress in Meeting the Convergence Criteria respective GDPs. It will be many years before these debt ratios come close to meeting the 60 percent limit. Denmark, Finland, Ireland, the Netherlands and Sweden also have high debt ratios unlikely to fall within the target range by the end of the century. The Dutch central bank last year calculated that if the Netherlands limited its annual public sector deficit to 1 percent of GDP, and achieved an average nominal GDP growth of 4 percent a year, it would still take 10 years to reach the 60 per cent public debt target (Financial Times, January 17, 1995). W hile 4 percent was the average nominal GDP growth for the Netherlands during 1985-94, its average yearly budget deficit has been more than double 1 percent of GDP over the last 10 years.9 Portugal and Spain are likely to have difficulty meeting several of the criteria. Although they both have substantially low ered their inflation rates in recent years, the 5.1 percent Portuguese and Spanish inflation rates remain outside the ceiling. The debt ratios of both countries also have grown recently and that of Spain is likely to remain a problem as long as it maintains its high unemployment rate (estimated at more than 22 percent in 1994). No one expects that Greece will be a candidate for monetary union for many years to come. It alone among the EU countries still has double digit inflation. The remaining country, the United Kingdom, is a good candidate for meeting all of the eligibility requirements for monetary union, except the exchange rate criterion. The United Kingdom is unlikely to rejoin the ERM in the next few years. Even ignoring this problem, opposition to EMU is strong within the British govern ment and Britain is one of two European Union countries that have the right to refuse entry into the monetary union.10 A change in the government from the ruling Conservative party to the opposition Labour party is likely to increase the prospects for Britain joining EMU simply because the latter is much more amenable to the idea o f monetary union than the former. 1995 The reality that a m ajority o f countries will not meet the convergence criteria in 1996, and that most, including some key countries, are unlikely to meet the criteria in 1998, has generated three responses within the European Union. One reaction has been to label the idea of monetary union impractical. A second suggests that the pub lic finance criteria for monetary union can be and should be interpreted with some leeway. A third reply suggests that the timetable for monetary union should be interpreted with some flexibility. Abandoning EMU Those who have reacted to the difficulty in meeting the convergence criteria by label ing EMU impractical are basically opposed to the idea of monetary union. They see the lack of progress in meeting the criteria as a means to gain support for the idea of aban doning the treaty. Proponents of this view, most notably some members of the British Parliament, have reacted to each crisis within the ERM with predictions of the demise of monetary union. For example, British Prime Minister Joh n Major responded to the August 1993 widening of the bands of the ERM with the statement that the Maastricht timetable for monetary union was now “totally unrealis tic.” The reaction of Norman Lamont, the former chancellor of the exchequer in Britain, was even more pointed. He claimed that the crisis in the ERM meant “the end of monetary union in Europe” (Financial Times, August 3, 1993). In practice, this group sup ports strict adherence to the convergence criteria, since this will delay the starting date for monetary union. Flexibility in Interpreting the Convergence Criteria In opposition to this group are those who not only support EMU but believe that the earlier the starting date the better. This latter group favors a liberal interpretation of the convergence criteria. One reason for 11 ’ A reduction in public debt cun occur through several meons besides a government surplus. Both nominal GDP growth and a reduction in interest rates on government debt will reduce the debt/GDP ratio. Nominal GDP growth may be achieved through growth in output or inflation. This might lead one to think thot inflating oway the debt would be a compelling option. Such o strategy, however, will only work in the short run. An increase in inflation raises the interest rote at which the government must bor row to finance its debt. The shorter the maturity of the outstanding debt, the shorter the period of time before which the engineered infla tion will affect the interest rate on the debt. Furthermore, any such attempt by the government to meet the debt convergence criterion through inflation is likely to have long-term repercussions for the interest rate ot which the govern ment borrows by reducing the government's credibility. 10 In Maastricht, the United Kingdom refused to conclude negotiations on the treaty unless it was given the right to opt-out of EMU. Denmark is the other country with the right to opt-out of monetary union. It negotiated this right following the rejection of a referendum on the treaty. After securing the opt-out provision, a new referendum approved the treaty. REVIEW J u ly / a u g u s t supporting a quick move to monetary union is the belief that a long transition period may itself be the source of instability. A proponent of this view is Portes (1993). In addition to arguing that a long transition period creates instability, Portes contends that the convergence criteria are unnecessary because “monetary union will deliver convergence— at least the extent required to maintain it.” De Grauwe (1994) takes this argument one step further by claiming that the convergence criteria cannot be met prior to EMU. Although support for a quick move to monetary union is generally tied to the belief that convergence is not a necessary prerequi site for EMU, support for a flexible approach to the criteria is based on additional reasons. One is to provide a wide participation in EMU. Another is the fear among countries that have little chance of meeting the requirements that non-participation in EMU will be cosdy both politically and econom i cally. In the political sphere, countries are afraid that remaining outside EMU will reduce their political power within the EU, particularly as the inner core of countries (the members of EMU) become more inter dependent. In econom ic terms, countries are concerned that exclusion from EMU may be viewed as a mark against them, and result in a higher interest rate premium and a weak ness in their currencies. Supporters of a flexible approach to the convergence criteria make reference to the M aastricht Treaty to bolster their case. The treaty provides an opening for a relaxation of both the deficit and the debt criteria. The 3 percent deficit/GDP ratio and the 60 percent debt/GDP ratio are referred to in the treaty as reference values, not fixed limits as are the criteria for inflation and interest rates. The treaty says that these reference values must be met unless, in the case of the deficit: " During the Maastricht negotiations, several countries proposed adopting a concept of cyclically adjusted deficits. The proposal was rejected because of measurement problems (Bini-Smaghi and others, 1994). 1995 and temporary and the ratio remains close to the reference value ( Treaty on European Union, Article 104c.2.a). In addition, in preparing its report on whether an excessive deficit exists, the Commission is to take into account: • whether the government deficit exceeds government investment expenditure (gross fixed capital formation); and • all other relevant factors, including the medium-term economic and budgetary position of the Member State ( Treaty on European Union, Article 104c.3). These clauses provide the commission a means by w hich to relax the deficit require ment. As noted by Collignon and others (1 9 9 4 ), the treaty could be interpreted as applying the deficit criterion to only the part of the deficit not accounted for by govern ment investment, and only requiring the 3 percent ratio to be met “when the economy was near full capacity” Looking at the data in Table 5, one could argue that Austria, Denmark and the Netherlands all meet the deficit criterion since their budget deficits remain close to the reference level, and that the elevated levels are merely temporary — caused by the recession.1 1 W ith respect to the debt criterion, the Maastricht Treaty states that the reference level (60 percent debt/GDP) is binding “unless the ratio is sufficiently diminishing and approaching the reference value at a satisfactory pace” ( Treaty on European Union, Article 104c.2.b). The debt levels of all the countries, with the exception of Ireland and the Netherlands, have increased between 1990 and 1994, as shown in Table 6 . In Ireland’s case, substan tial progress has been made in reducing its debt ratio. Ireland has m et the deficit con vergence criterion in every year and has reduced its debt ratio from 97 percent of GDP in 1990 to 90 percent in 1994. In the fall of 1994, the European Council, assessing the progress of countries toward the • either the ratio has declined substantially and continuously and reached a level that comes close to the reference value; or • alternatively, the excess over the reference value is only exceptional N K OF S T . L O U I S 12 REVIEW Y/AUGUST 1995 Ta b le 5 C onvergence Indicators: G o v e rn m e n t B udg et Balance 1990 1991 -6.5 -5.4 Denm ark -1.5 France -1.6 Germany -2.1 -3.9 -3.3 -14 .0 -6.7 -2.2 Belgium -11.6 Greece Ireland Italy luxem bourg Percent of GDP 1993 1992 -2.1 5.9 2.3 Netherlands -5.1 -2.4 0 .8 -3.3 -5.5 -3.9 United Kingdom -1.5 -2.6 Austria -2.2 -2.4 -2.0 Finland 5.4 -1.5 Sweden 4.2 Convergence criterion -3.0 Number meeting criterion 9 -3.3 -7.5 -6.1 -3.0 -4.63.9 2 -1.1 -9.1 -3.0 -3.0 5 view. In May 1994, the then-commissioner for econom ic and monetary affairs stated that it had “always been understood that the judgem ent on whether a member state fulfills the conditions for participation in stage 3 would be based on an assessment, and not on a mechanical application of the convergence criteria” (Financial Times, May 16, 1994). In contrast, the president of the current commission, Jacques Santer, in his first speech before the European Parliament, pledged that the commission would insist on strict application of the criteria (Financial Times, January 18, 1995). Flexibility in the Starting Date for Monetary Union In contrast to those who have responded to the lack of progress in meeting the con vergence criteria by suggesting that the F EDERAL RESERVE B A N K OF ST. L OUI S 13 -5.8 -3.0 34 Notes: Prior to 19 9 1 the data for Germany are for western Germany only. Data for 1 9 9 5 and 1 9 9 6 are forecasts. SOURCE: European Economy ( April/M ay 1 9 9 5 , Supplement A, Table 21). M aastricht criteria, accepted the recommendation of the commission and determined that Ireland met the debt criterion. This decision indicates some willingness on the part of the European Union to reward countries that are making efforts to control public deficits yet remain outside the numerical targets. However, it does not mean that such a policy will be followed at the inter-govemmental conference in 1996. The decision that Ireland met the debt convergence require ment was not without controversy. Germany, in particular, had severe reserva tions about the exemption. Furthermore, while the previous European Commission, the term of which ended in December 1994, supported a flexible interpretation of the convergence criteria, it is not clear that the present commission also supports this -4.8 -2.9 -5.6 5.0 -10-13.4 .4 -7.8 -3.0 -6.6 -4.0 -7.8 -3.2 -5.64.7 -6.0 -4.8-6.9 -7.8 -4.1 -1.1 1.5 -2.5 -5.8 -4.2 -2.6 -8.1 -9.0 -3.1 -7.0 -5.9 -4.9 -2.8 -7.9 1 2.3 .4 2.1 -6.5 Portugal Spain -10 .2 -2.3 -9.6 -3.9 9 -11.3 -9.5 -2.9 -3.9 -2.4 -2.1 -13.212 .5 -2.4 -10.2 -1.2 -2.5 -3.3 -12 .3 -2.2 -1.9 -6.04.9 -6.1 -2.9 -3.9 -4.2 -5.3 -4.54.0 -2.9 1996 1995 -6.6 -2.1 -10.9 1994 -3.0 6 UGU T a b le 6 C o nve rgence Indicators: G o v e rn m e n t D e b t P e rc e n t o f G D P 1990 1991 1992 Belgium 13 0 .8 59 .6 35 .4 3 5 .7 1 3 7 .2 6 4 .6 France 131.1 130.1 Denmark 1993 1994 8 0 .3 4 1 .5 Greece 8 2 .6 86.1 Ireland 9 696 .9 .8 9 7 .9 9 4 .2 Luxembourg 5.4 Netherlands 78 .8 Spain 45.1 5.7 41 .9 6.9 6 2 .3 6 2 .8 7.2 78.1 6 70 .59 .2 48 .5 58 .4 8 6 .8 1 2 4 .4 8 1 .4 66 .6 5 48 .39 .9 4 5 .8 84 .6 1 2 4 .9 1 2 5 .4 79 .9 61 .7 3 5 .0 3 5 .7 United Kingdom 114.1 1 1 6 .2 89 .8 1 1 9 .4 52 50.1 1 1 5 .2 11 5 .3 9 7 .0 4.9 76.1 51 .2 4 8 .2 9 2 .3 78 .9 6 68 .69 .3 Portugal 75 .6 4 8 .5 44.1 1 0 8 .4 10 1 .3 1 3 4 .33 2 .3 1 4 5 .8 4 3 .8 1996 136.1 6 9 .0 39 .6 Germany Italy 1995 70 .7 6 4 .6 50.1 6 5 .2 51 .5 Austria 5858 .7 .3 Finland 3 142.5 .0 4 1 .5 57.1 60.1 64 .4 Sweden 4 3 .5 67.15 3 .0 76 .2 79.1 8 8 4 .65 .7 Convergence criterion 6 0 .0 Number meeting criterion 9 6 0 .0 8 6 4 .5 6 0 .0 7 66 .2 51 .5 6 0 .0 4 6 6 7 .4 6 0 .0 4 Notes: Data for the United Kingdom in 1 9 9 0 ore based on OECD calculations of general government gross financial liabilities. All other data are based on the Maastricht Treaty's definition of public debt. Prior to 19 91 the dota for Germany are for western Germany only. Data for 1 9 9 5 and 1 9 9 6 are forecasts. SOURCES: European Economy (April/May 1 9 9 5 , Supplement A, Table 2 2 ) and OECD Economic Outlook (June 1 9 9 5 , Number 57, Table 34 ). convergence criteria be treated with flexibility are those who believe that the 1999 deadline should be viewed as flexible. The propo nents of a flexible timetable believe that strict adherence to the convergence criteria is a necessary condition for a well-functioning monetary union. Thus, rather than relaxing the criteria to guarantee that an optimal number of countries will participate in EMU, they suggest that the date for monetary union be delayed if the criteria are not met by a sufficient number of countries. German Chancellor Helmut Kohl was the first leader to publicly address this issue. In 1993, he stated that strict adherence to the convergence criteria might delay monetary union beyond 1999. The October 1993 ruling of the German Constitutional Court supported those who argue that the timetable for monetary union is more flexible than the criteria. The court, in ruling on the constitutionality of the M aastricht Treaty, wrote that strict adherence to the convergence criteria was essential to Germany’s participation in EMU. In other words, the criteria could not be weakened without the consent of the German parliament. The German central bank, the Bundesbank, has been perhaps the most vocal advocate of a strict application of the convergence criteria. Both Hans Tietmeyer, the current president of the bank, and his predecessor, Helmut Schlesinger, have made statements on several occasions favoring a strict interpretation of the Maastricht criteria while claiming that the criteria are them selves not strict enough. For example, the 14 6 4 .6 6 0 .0 4 REVIEW JULY/AUGUST states (Protocol on the Statue o f the European System o f Central Banks and the European Central B ank, Article 2 1 ), and declares that neither the central bank nor other countries shall be liable for or assume the financial commitments of any member states (Treaty on European Union, Article 104b. 1), there are those who believe there would be pressure on the central bank to bail out countries experiencing financial difficulties.14 Bundesbank has favored an absolute lim it on inflation rather than a relative one, the latter based on the behavior of other countries. The reason for this is to ensure not simply convergence in inflation rates, but also a commitment to price stability. The Bundesbank has also attacked the deficit cri terion as setting too high a ceiling. Specifically, Mr. Tietmeyer has stated that the ceiling for the deficit ratio is at least double what it should be. He also has emphasized that the deficit criterion should be met throughout the business cycle (Financial Times, November 5, 1 9 9 4 ).12 This statement contrasts with a study prepared for the European Parliament that suggests that “It would be keeping with the spirit of the Treaty, if 3 percent were taken as the ‘full employment’ deficit during periods of econom ic expansion” (Collignon and others, 1994, p. 76). As noted above, the emphasis on a strict interpretation of the convergence criteria is based on the belief that adherence to them is necessary for a well-functioning monetary union. The proponents o f strict criteria argue that for EMU to succeed, the member states must show a prior commitment to price stability and follow sound government budgetary policies. Specifically, the empha sis on a strict interpretation of the deficit criterion is based on the idea that “a sound budget position is an indispensable precon dition for a successful anti-inflationary monetary policy.”1 There is a concern that 3 within a monetary union, expansionary national fiscal policies (as evidenced by bud get deficits in excess o f 3 percent of GDP) could conflict with the monetary policy of the supranational central bank. Such a con flict would not only create difficulties for the central bank in its effort to maintain price stability, but also could cause tension among the participants in the monetary union. Would the participants of a monetary union be willing to accept a recession brought about by the anti-inflationary polices of the central bank in an effort to combat the fiscal laxity o f other members? Furthermore, although the Maastricht Treaty prohibits the central bank from extending credit to, or directly purchasing the debt of, member 19 CONCLUSION Despite the many setbacks that have occurred since the December 1991 conclusion of the Maastricht Treaty, most of the countries of the European Union remain committed to monetary union. This commitm ent, how ever, has not been enough to produce the econom ic convergence prescribed by the treaty. Many countries have made progress in reducing their inflation rates, and the divergence in long-term nominal interest rates is declining. On the fiscal side, however, the number of countries meeting the conver gence criteria has declined. The recent recession in Europe resulted in a deterioration in the fiscal balances o f most countries. In addition, the 1992-93 exchange rate crises resulted in a reduction in the membership of the ERM. Thus, the European Union is further away from a fulfillment of the convergence criteria today than it was in the year prior to the negotiation of the Maastricht Treaty. By the end of 1996, the member states of the European Union must decide if a majority of countries are ready to proceed to EMU in 1997. As detailed above, it is implausible that a majority of countries will have fulfilled the convergence criteria by the end of 1996. EMU will m ost certainly be delayed beyond its earliest possible starting date. The M aastricht Treaty states that the final stage of EMU must begin by January 1, 1999, with the membership decided by July 1998. Even by this date, few countries are likely to satisfy the convergence criteria. Given the lack o f progress in meeting the convergence criteria, the European Union faces two options if it is to continue to pursue EMU: Relax the criteria or relax the FEDERAL RESERVE B A N K OF ST. L OU IS 15 12 Others hove claimed that even a government that has a balanced budget during upturns could have o budget deficit exceeding the Maastricht limit during a recession. See, for example, Eichengreen (1 9 9 2 ) and Kenen (199 2 ). Eichengreen orgues that it may even be optimal for disciplined gov ernments to occasionally hove deficits exceeding 3 percent of GDP (p. 50). 13 Tietmeyer (September 9, 1994). M Support for this view is given by Frationni, von Hagen and Waller (1 9 9 2 ) and (raig (199 4 ). REVIEW JULY/AUGUST Craig, R. Sean. "W h o Will Join E M U ? Impact of the Maastricht Convergence Criteria on Economic Policy Choice and Performance," Board of Governors of the Federal Reserve System International Finance Discussion Papers No. 4 8 0 (September 1 9 9 4 ). timetable for monetary union. W hich option it chooses will likely not be decided until the July 1998 deadline for determining the mem bership of EMU. The choice taken by the EU will undoubtedly be influenced by the two countries without whose participation EMU will not occur: France and Germany. Germany has strongly opposed a relax ation of the convergence criteria.15 If it maintains this position, few countries are likely to meet the membership requirements for EMU by the end of the decade. More importantly, two countries considered among the core group of EU countries — Belgium and the Netherlands — are not expected to meet the criteria.16 W ithout the participation of the core group, monetary union may not be feasible. Thus, it is likely that EMU, like its avian namesake, will remain grounded. De Grouwe, Paul. "Towards European Monetary Union Without the E M S ," Economic Policy (April 1 9 9 4 ), pp. 147-85. _ _ _ _ _ _ _. I/ie Economics of Monetary Integration. Oxford University Press, 19 9 2 . The Economist. "The Unpopularity of Two-Speed Europe" (September 1 4 , 1 9 9 1 ) . Eichengreen, Barry. "Should the Maastricht Treaty Be S a v e d ? " Princeton Studies in International Finance (December 1 9 9 2 ). Fratianni, Michele, Jurgen von Hagen and Christopher Waller. "The Maastricht W ay to E M U ,"Essays in International Finance (June 1 9 9 2 ). Gawith, Philip. "Tietmeyer Sets Out Tough Line on EMU Convergence Criteria," Financial Times (November 5 , 1 9 9 4 ) . REFERENCES Marsh, David. "E m u Strain Begins to S h o w ." Financial Times (January 1 7 , 1 9 9 5 ) . Barber, Lionel. "Sa nter steers middle course for EU ," Financial Times (January 18, 1 9 9 5 ). Bini-Smaghi, Lorenzo, Tommaso Podoo-Schioppo ond Francesco Papadia. "T he Transition to EMU in the Maastricht Treaty," Essays in International Finance (Number 19 4 , November 1 9 9 4 ), Princeton University. Collignon, Stefan with Peter Bofinger, Christopher Johnson and Bertrand de Maigret. Europe's Monetary Future. Printer Publishers, 19 9 4 . Committee for the Study of Economic and Monetary Union. Report on Economic ond Monetary Union in the European Community. Office for Official Publications of the European Communities, 19 8 9 . negotiations on the Maastricht Treaty, Germany resisted setting a fixed date for the commencement of monetary union. It believed that fixing a dote would result in a loose application of the convergence crite ria (The Economist ; September 14, 1991). 16 The other members of this core group ore France, Germany and Luxembourg. Owen, David. "M a jor S a y s Timetable for Monetary Union Unrealish'c," Financial Times (August 3, 1 9 9 3 ). Portes, Richard. 1 9 9 3 . "E M S and EM U After the Fall," World The Economy (No, 1, 1 9 9 3 ) pp. 1-16. Tietmeyer, Hans. Speech at the Association for the M onetary Union of Europe in Frankfurt/Main, September 9 , 1 9 9 4 . Printed in Bank for International Settlements, BIS Review (October 1 4 , 1 9 9 4 ) . Buiter, Willem, Giancarlo Corsetti ond Nouriel Roubini. "Excessive Deficits: Sense and Nonsense in the Treaty of Maastricht," Economic Policy (April 1 9 9 3 ), pp. 57 -100 . 15 It is interesting to note that in the 1995 Council of the European Communities, Commission of the European Communities. Protocol on the Convergence Criteria Referred to in Arhcle 109j of the Treaty Establishing the European Community. Office for Official Publications of the European Communities, 19 9 2 . _ _ _ _ _ _ _. Protocol on the Excessive Deficit Procedure. Office for Official Publications of the European Communih'es, 1 9 9 2 . _ _ _ _ _ _ _. Protocol on the Statue of the European System of Central Banks and of the European Central Bank. Office for Official Publications of the European Communities, 1 9 9 2 . _ _ _ _ _ _ _. Treaty on European Union. Office for Official Publications of the European Communities, 19 9 2 . 16 REVIEW JULY/AUGUST 1995 Donald S. Allen is an economist at the Federal Reserve Bank of St. Louis. Thom as A. Pollmann provided research assistance. ■ Changes in Inventory M anagem ent and the Business Cycle Since the 1970s, many firms have made notable changes in their inventory manage ment methods. In particular, large movements in interest rates in the early 1980s and increased global trade have combined to motivate firms to reduce inventory levels relative to sales as part of larger downsizing efforts. More efficient inventory management has been realized by implementing “just-intim e” (JIT ) management techniques and the use of bar codes. W ill these innovations in inventory management decrease the effect of inventory movements on the business cycle? This article investigates the extent of the changes in inventory management and makes some observations regarding inventory movement and the business cycle. There is evidence to suggest that the use of these innovative inventory control methods is on the rise, but the net effect on the business cycle remains ambiguous. In the first two sections, I review the role of inventory investment in postwar recessions and the motivations for holding inventory. Next, I document some of the innovations in inventory management that firms have adopted over the last 10-15 years. Finally, I discuss the potential impact of these changes on the business cycle. Donald S. Allen “I rem ember one day in the summer o f 1975 when a CBO [Congressional Budget Office] staffer returned from a congressional hearing with som e am azing news. Alan Greenspan, then President G erald Ford’ ch ief econom ic s adviser, had ju st testified that the recession was mostly an inventory correction. We all snickered at the idea that w hat was, up to then, the deep est recession since the Great Depression could have been ‘on ly’ an inventory cycle. W hen I subsequently studied the data m ore carefully, however, I learned that Greenspan had been right. L ike most o f the recessions before and since, the 1973-5 contraction was dom inated by changes in inventory investment. ” Alan S. Blinder, introduction to Inventory Theory and Consum er B ehavior (1990) THE ROLE OF INVENTORY IN POSTWAR RECESSIONS The stocks of materials and supplies, partially completed goods and finished goods in the possession of a firm are income-producing assets. These stocks are held temporarily before being sold. As inventories are increased or decreased between the begin ning and the end of a period, they add to or subtract from the investment component of GDP Unlike fixed investment, which is assumed to be the result of specific plans by firms, inventory stocks fluctuate as a result of both active decisions by firms and errors in forecasted demand. This dual effect tends to make inventory investment especially volatile around contractions, usually going he change in business inventories is usually less than 1 percent of total Gross Domestic Product (GDP), yet during cycli cal contractions this component contributes disproportionately to the change in GDP As a result, most cyclical contractions have been referred to as inventory cycles. These inventory cycles are characterized by an unanticipated drop in demand resulting in unplanned increases in inventories. Firms respond by cutting production to reduce inventory. This cut in production can exacerbate the downturn by reducing demand further. T 17 REVIEW .Y/AUGUST 1995 below the desired level, causing increased production to replenish inventory The U .S . In v e n to ry In vestm ent M o ve m e n ts in degree of undesired accum ulation and decu P ostw a r Recessions mulation is a function of the accuracy of Change in firms’ demand projections. This inventory inventory cycle phenomenon has been of varying inter Investment as est to economists, and research in this area Change in a Percentage Recession Period Change in Inventory of Change in has ebbed and flowed like the business cycle. Peak to Trough1 Real GDP2 Investment2 Real GDP Metzler (1 9 4 1 ) showed analytically how inventory cycles could be generated when 1948:4 - 1949:4 -1 4 . 5 -2 8 .3 195.2 decisions on production levels are based on expected levels of sales, and income and 1 9 5 3 :2 - 1954:2 -3 6 .9 -2 0 .0 54.2 demand are determined by production levels. 1 9 5 7 :3 - 1958:1 -6 1 .1 - 2 1 .1 34.5 Blinder and M accini (1 9 9 1 ) provide a good 1960:1 - 1960:4 - 1 5 .8 -4 5 .5 288.0 survey and bibliography of research in inven tory cycles since Metzler’s work. 1 9 6 9 :3 - 1970:4 -3 2 .2 287.5 -1 1 . 2 The role of inventory investment in 1 9 7 3 :4 - 1975:1 -1 3 5 .1 -8 4 .7 62.7 business cycle contractions has been well 1980:1 - 1980:2 -9 8 .2 - 1 0 .7 10.9 documented. Coincident declines in GDP 1 9 8 1 :3 - 1982:3 -1 1 0 .1 -3 5 .0 31.8 and inventory investment are empirical regu larities of postwar business cycles. Blinder 1 9 9 0 :2 - 1991:1 - 7 5 .1 -4 4 .5 59.3 and M accini (1 9 9 1 ) show that the average Mean 113.8 movement in inventory investment during recessionary periods in the postwar era 1 Peaks and troughs correspond to peaks and troughs of real GDP and do not always coincide with official account for 87 percent of Gross National NBER recession dates. . 2 Billions of 1 9 8 7 dollars. Product (GNP) movement from peak to trough. Computed another way, the relative movement is even greater. Table 1 shows peak-to-trough movement in inventory from positive at the beginning as a result of investment compared to the peak-to-trough unintended accumulation, to negative due to change in GDP in all postwar recessions.' The average percentage change in inventory deliberate reduction. Inventory investment averages less than investment to change in GDP is 113.8 per 1 percent of GDP, but changes in inventory cent. Admittedly, this method computes the investment can account for a substantial difference between the highest quarterly increase in business inventory and the highest portion of the change. In 1994, for example, inventory investment was $47.8 billion (in quarterly decrease in business inventory on 1987 dollars) or 0.9 percent of GDP. This an annualized basis, capturing the widest swing. However, it is evident that inventory level of inventory investment reflected an investment has been a significant contributor increase of $ 32.5 billion over 1993 or 1 The National Bureau of Economic 15.5 percent of the $209.5 billion increase to changes in GDP during contractions. Research (NBER) typically has iden in GDP in 1994. Figure 1 compares the change in GDP and tified a recession as a period with The typical inventory cycle begins with the change in business inventories since two consecutive quarters of decline an unexpected reduction in demand which 1948. Recessions are shown by shaded bars. in GDP. The peak of the cycle is the leaves firms with inventory above their Inventory level movement by itself is not quarter prior to the first quarter of desired levels. Production is reduced to the complete story. It is necessary to know decline. The hough is the last quar lower inventory levels, which can result in whether movements are active responses to ter of negative growth. The peoklayoffs and further reduction in demand. changes in the level of demand or reflect to-trough movement in inventory As inventory falls back to desired levels and errors in forecasting. The ratio of inventory investment in Table 1 is the differ to sales, defined as total stocks divided by demand resumes, production may be insuffi ence between the maximum and cient to meet demand and maintain inventory monthly sales, gives some indication o f the minimum inventory investment dur ing the recession period. nature of these movements. If we assume levels. The result is that inventory can fall Ta b le 1 18 R IE IVW JULY/AUGUST that firms plan to maintain a relatively constant level of inventory to sales, major deviations in this ratio can give clues to whether movements are planned or unplanned. If inventory accumulation is accompanied by an increasing inventory-to-sales ratio, then the accumulation may be inadvertent because inventory is rising faster than sales. If inven tory accumulation is accompanied by a con stant inventory-to-sales ratio, then the accu mulation may have been planned in response to increasing sales. An increase in inventory can also be accompanied by a decrease in the inventory-to-sales ratio, indicating that sales are increasing faster than inventories. The total business inventory-to-sales ratio in the postwar period is shown in Figure 2, with recessions indicated by shaded bars. It is also evident from this figure that the ratio peaks around the contractions, mak ing it a relatively reliable coincident indica tor. Although it cannot be claimed that inventory changes cause the business cycle, any imbalance which occurs between expect ed and actual sales shows up in inventory, and correcting this imbalance can exacerbate the cycle. Even if we do not consider inven tory investment to be a causative force but simply a barometer of forecast accuracy, most recessions appear to be marked by an inventory correction. 1995 F ig u r e 1 C han ge in G D P C o m p a re d to the C han ge in Business In ve n to rie s 1948 52 56 60 64 68 72 F ig u r e 2 Total Business In v e n to ry -to -S a le s R atio 1.8 1.7 1.6 1.5 1.4 1.3 1948 52 56 60 64 68 72 76 SOURCE: U.S. Department of Commerce, Bureau of Economic Analysis. W H Y HOLD INVENTORY IN THE FIRST PLACE? Inventory stocks represent a major utilization of resources. At the end of 1994, manufacturing and trade inventories totaled $832 billion (1 9 8 7 dollars) or 12.4 percent of annual sales. At the current prime rate, the opportunity cost of holding the 1994 level of inventory stocks amounts to more than $70 billion. This financing cost compares to the 1994 annual increase in GDP of roughly $200 billion. The capital tied up in financing inventory could also be converted into fixed invest ment in more productive capital equipment. But rational firms are motivated to hold inventory as long as the expected cost of holding it is less than the expected penalty 92 76 SOURCE: U.S. Department of Commerce, Bureau of Economic Analysis. (lost revenue or market share) for running out of stock. In other words, the optimal level of inventory in the face of uncertain sales and random supply interruption is not always zero, and there is a lim it to the savings which can be realized by lowering inventory. The motivations for holding inventories are diverse and firm-specific. Some firms minimize their delivery costs, some smooth production in the face of uncertain demand, and others stockpile against potential interruptions or anticipated price increases by suppliers. Most retailers are forced to hold inventory to accommodate the FEDERAL RESERVE B A N K OF ST. L OU IS 19 I 1995 REVIEW JULY/AUGUST 1995 INVENTORY MODELS Production Smoothing (S,s) Rule The accompanying figure on page 21 illus trates the production smoothing motivation when increasing marginal costs exist. If Q1 and Q 2 represent the demand in periods 1 and 2 , respec tively, then point A represents the average cost if Q1 is produced in period 1 and Q2 is produced in period 2. Point B represents the average cost if (Ql+Q2)/2 is produced in both periods, with the excess produced in period 1 carried over to peri od 2. The trade-off is between the cost of storage for one period versus the saving from smoothing.' The difference between A and B must be greater than the cost of holding inventory to justify smoothing. Note also that if mean demand is expected to decrease below current production for an extended period (that is, Q2 is current demand and Q1 is next period’s expected demand), then it becomes optimal to reduce production and serve part of current demand from inventory. Thus, production smoothing motivation can lead to level changes if forecast sales change direction. If costs are linear, as in the case when marginal costs are constant, and there is a significant fixed cost o f purchasing in each period, then it can be shown that “lumpy” adjustment is preferred to smoothing. An economic batch run, or a purchase which minimizes the total expected cost including the cost of storage of excess inventory, and the cost of lost sales can be determined. The inventory management technique used under these circumstances is referred to as (S,s) and entails determining maximum (S) and minimum (s) levels o f inventory.2 W hen inventories fall below (s), purchases are made to bring inventory up to (S), as long as inventories are between (S) and (s), nothing is done. The (S,s) parameters will define the upper and lower bound of inventory movement. It can be shown that the (S,s) margin is more sensitive to the mark-up of price over marginal cost than to interest rates. 1 Holt, Modigliani, Muth and Simon (1 9 6 0 ) provide the details of the production smoothing model. 2 Scarf (1 9 6 0 ) proves the optimality of the (S,s) rule under specific conditions. INVENTORY INNOVATIONS wide range of preferences and sizes of consumers. Generally, inventories are a hedge against uncertainty or a means of minimizing production costs. There are two competing models for inventory decisions, depending on the assumption about production costs. W hen firms operate in a region of increasing mar ginal costs, it becomes more economical to smooth production than to adjust to changing sales. W hen marginal costs are constant, but there are fixed costs associated with delivery or production, batch runs or bunching spread these fixed costs over larger quantities. (See the shaded insert above for discussion.) Wholesalers, retailers and manufacturing purchasers o f raw materials and supplies are more likely to face non-negligible delivery costs and therefore more likely to use batch purchasing. Is JIT Changing the Face of Inventory in America? As businesses focused on streamlining operations in the 1980s, one of the targets has been inventory stocks. Over the last 15 years, there seems to have been major shifts in the methods used to manage inven tory. In particular, many U.S. companies have studied and adopted the Japanese kanban (or JIT ) method of inventory man agement. The objective of the JIT system is to minimize the stock of parts and compo nents by having them delivered ju st in time for production, and to limit the inventory of finished goods by producing them ju st in time to fill demand. The monthly National Association of Purchasing Management survey indicates that as much as 26 percent 20 REVIEW JULY/AUGUST 1995 Just-in-Time Inventory foster the use of JIT. Intuitively, deregulation, Just-in-tim e (JIT) inventory control which reduces the econom ic lot-size of shipment, attempts to match production as closely to sales allows more continuous streams of shipment. as possible and thereby minimize the costs of holding inventory. This method, called kanban in Japan, is characteristic of Japanese industry in general and the auto industry in particular. JIT can be optimal when convex costs of production C o n v e x P r o d u c t io n C o sts exist but storage costs exceed savings from smoothing or when linear costs exist but the low mark-up, low variance of sales, low fixed costs of delivery or high costs of storage result in low values of (S,s). If firms can meet demand without holding invento ries, then inventories become superfluous. JIT can exist only in an atmosphere in which suppliers are reliable enough to minimize the risk o f stock-outs. Larson (1991) argues that deregulation of the transportation industry has resulted in innovations which of the respondents reported purchasing mate rials “hand to mouth” in January 1995, com pared to as little as 4 percent in February 1970. This suggests that the JIT philosophy has made major inroads into U.S. manufac turing. Bechter and Stanley (1992) find empirical evidence of improved inventory control along with faster speeds of adjust m ent to desired inventory levels. Prima facie evidence of the success in reducing manufacturing inventory is also seen in the consistent decline in the aggre gate inventory-to-sales ratio (shown in Figure 2), which has dropped from a peak o f approximately 1.7 during the 1990 reces sion to 1.44 in December 1994 — the lowest in about 20 years. The manufacturing sector has been reducing inventory at all stages of production. Figure 3 shows the manufac turing sector inventory-to-sales ratios by stage of processing for 1970 to 1994. The work-in-process and materials and supplies are at a low point for the last two decades, after a steady decline since the early ‘80s. Some of this decline may be attributable to factors other than JIT. For instance, a closer look shows that materials and supplies increased rapidly relative to sales during the 1973-75 recession and did not return to ear lier levels until recently. This could indicate an end to a post-oil-embargo tendency to stockpile, motivated by inflation expecta tions and sensitivity to interruptions. Some industries have been more success ful than others in lowering inventory levels relative to sales. Table 2 shows the summary statistics for the inventory-to-sales ratio by stage of processing for four manufacturing industries which have experienced significant declines in ratio. The December 1994 ratio is F EDERAL RESERVE B A N K OF ST. I O U I S 21 REVIEW .Y/AUGUST the retail and wholesale levels. The cost of financing high levels of inventory is a m ajor cost of doing business. In the early years of the industry, finance companies took on the dual role of providing credit to wholesalers and buying consumer loans initiated by deal ers. It seems appropriate, therefore, that the push to reduce inventory levels should take place in the auto industry. Minimizing inventory reduces financing needs and thus increases the competitive edge. The downside is greater vulnerability to interruptions such as strikes or to unanticipated surges in demand. U.S. automobile manufacturers appear to have embraced JIT and currently hold less than two weeks worth of sales in inventory, down from a high of 1.3 months. Figure 4 shows the changes in inventory-to-sales ratios in the motor vehicle industry by stage of processing for the period 197094. It is apparent that there has been much success in reducing inventory levels over the last 10 years. Figure 4 also reveals that the reduction occurred primarily at the work-in-process, and materials and supplies stages of produc tion with very little change in the level of finished goods relative to sales. The burden of reduced inventory has been placed on the intermediate input stage of production. As an example of the downside of lower inventory holdings, however, General Motors in 1994 experienced the shutdown of several assembly lines because of an inter ruption at a drivetrain component plant. If they had held higher levels of inventory, they would have been able to reduce the scale of the shutdown. F ig u r e 3 M a n u fa c tu rin g In v e n to ry -to -S a le s Ratios b y S ta g e of Processing SOURCE: U.S. Department of Commerce, Bureau of Economic Analysis. provided for comparison. Motor vehicle mate rials and supplies, and work-in-process inventory stocks have declined from peak ratios of over 70 percent of monthly sales each to 19 percent and 14 percent, respectively, in December 1994. In all four industries and for all stages of processing, December ratios are well below the mean for the entire period. The use of JIT tends to shift the burden of responding to uncertainty to the suppliers and to require speedy delivery methods.2 Some analysts believe that significant changes in the transportation industry, fostered in part by deregulation and increased competition, contributed to the viability of JIT. In partic ular, if a manufacturer wishes to maintain a continuous flow of materials, deliveries must take place more often in smaller batches. The deregulation in the trucking industry, which allowed competitive pricing for lessthan-truckload deliveries, and increased competition in air freight help reduce the cost of smaller, more frequent deliveries. Bar Coding The computer industry revolution and proliferation of bar coding has streamlined the inventory process in all sectors of the economy. Many retailers now use automatic scanning computer registers to record sales and track inventory immediately. These innovations have had the spillover effect of providing almost instant marketing informa tion regarding the rate of sale or use of prod ucts. The increased use of bar code scanning and more sophisticated electronic systems JIT in the Auto Industry 2 The recent eorthquoke in Kobe, Japan, emphasized the potential disadvantage which this system produces, when many Japanese manufacturers, who were otherwise unaffected, hod to shut down because of interruptions to suppliers and transportation. 1995 The evolution of the structure of the U.S. automobile industry is relatively unique and was motivated primarily by the need to smooth production, combined with a limited ability to hold inventory (see Olney, 1989). The relationship between manufacturers, wholesalers and retailers ensures that the storage of finished goods occurs primarily at NK OF ST . L OU I S 22 REVIEW T a b le 2 In v e n to ry -to -S a le s R atio (J a n u a ry 1 9 7 0 - D ecem ber 1 9 9 4 ) Stage of Processing Mean Maximum Minimum Dec. 1994 Finished Goods Motor vehicles Primary metals Electrical Non-Electrical 0.147 0.635 0.564 0.664 0.261 1.032 0.709 0.972 0.099 0.372 0.481 0.442 0.115 0.545 0.511 0.452 Work-in-Process Motor vehicles Primary metals Electrical Non-Electrical 0.298 0.862 0.960 0.878 0.730 1.338 1.174 1.123 0.137 0.564 0.676 0.614 0.139 0.677 0.689 0.614 Materials & Supplies Motor vehicles Primary metals Electrical Non-Electrical 0.366 0.863 0.679 0.647 0.762 1.365 0.893 0.185 0.567 0.541 0.437 0.185 0.605 0.554 0.514 0.886 1990:4) support the notion that inventory management methods changed significantly beginning in the ’80s. Similar work by Bechter and Stanley (1 9 9 3 ) detects changes in the speed of adjustment and desired inventory-to-sales ratio after 1981 in a buffer-stock model. The evidence supports the assertion that inventory innovations have impacted not only the quantity but also the quality of inventories held by allowing firms to more closely match patterns of use. Manufac turing has been more successful in reducing the quantity of inventory relative to sales, but the innovations in the wholesale and retail sectors should also lim it the accumulation of unplanned inventory through more direct feedback of marketing information. As a result, the innovations in all three sectors should tend to lim it the error portion of inventory accumulation. over the last 10 years has led to more efficient retail (and wholesale) inventory manage ment. This increased efficiency has not necessarily manifested itself as lower inventory levels, but allows more precise selection of stock items. In the retail sector, the inventory-to-sales ratio has actually increased slightly in contrast to the aggregate. The reasons for this increase are not obvious, but retailers must keep visible inventory on hand to stim ulate sales, and therefore have less flexibility in inventory levels. In addition, an increase in the total number of stores3 may have also contributed to the increase in aggregate retail inventory. There have been efforts at limiting inven tories at the retail level. “Quick Response” is the retail equivalent to JIT. Some retailers try to limit inventory by streamlining cus tomer orders. The effectiveness of these efforts has been limited and so far appears to have had little impact on the level of aggregate retail inventory relative to sales. Little (1992) uses quarterly manufactur ing and trade data from 1968 through 1990 to test for structural changes in inventory management. Results of regressions o f the data divided into two sub periods (1968:1 to 1982:3 and 1982:4 to IMPACT OF INNOVATIONS ON THE BUSINESS CYCLE Inventory influences business cycle con tractions primarily through unintended increases.4 How do the structural shifts in inventory management affect unintended 23 3 The Economist (March 4 ,1 9 9 5 ) reported in its retail survey that 1 9 9 3 total shopping center space in the United States was 18.5 sguare feet per head, compared with 13.1 square feet per head in 1980, according to the Schroder Reol Estate Associates. 4 Some analysts suggest that higherthan-average growth during the recovery part of the cycle reflects planned inventory investment in anticipation of increased demand. rebounded was offset by the continuing effort to reduce inventory-to-sales ratios. M o to r Vehicles In v e n to ry -to -S a le s R atio Bechter and Stanley (1 9 9 3 ) use estimated b y S tag e of Processing parameters from their buffer-stock model to 0.8 t * simulate inventory investment and conclude that the new parameters lead to larger inven tory swings for a one-time shock in sales. 0.6 Filardo (1 9 9 5 ) uses an atheoretical vector M a te ria ls and autoregression (VAR) method and a model0.4 based method to test empirically whether the changes in inventory management have muted the business cycle. He concludes 0.2 there is no evidence of a reduced role for inventory in the business cycle. As Little 0 . ■ .............................................................. (1 992) suggests, however, the innovations 1970 72 74 76 78 80 82 84 86 88 90 92 1994 are still being implemented and may not SOURCE: U.S. Department of Commerce, Bureau of Economic Analysis. have saturated the market. In this case, there is an insufficient sample size to evaluate the business cycle impact empirically. F ig u r e 5 It is difficult to separate the effect of those firms using JIT from those which do N o m in a l ln v e n to ry-to > S a le s Ratios fo r Ja p a n a n d the U n ite d States not. One approach is to see if the industries that have converted to JIT now contribute less inventory investment during the reces sion. Primary metals, electrical machinery, non-electrical machinery and m otor vehicles have shown significant decline in their inventory-to-sales ratios in the last 10-15 years. I looked at the 1980, 1982 and 1990 recessions to determine the contribution of these industries during the quarter with the biggest reduction in inventory. Together, the four industries contributed a net 22 percent to the third quarter 1980 change in business SOURCES: U.S. Department of Commerce (Bureau of Economic Analysis) and The Bank of Japan. inventory, a net 29 percent to the fourth quarter 1982 change in business inventory, but only net 1.6 percent to the fourth quarter 1990 change in business inventory. The inventory build-up? Morgan (1 9 9 1 ) sug remaining manufacturing industries con gests that a move to JIT produces a faster reaction to sales shocks and therefore will tributed 33 percent, 19 percent and 36 per cent to the change in business inventories not result in the levels of unplanned accu mulation previously observed. He also during these periods. These four industries argues that as the use of JIT increases, the that have reduced their inventory-to-sales ratios significantly over the past two decades impact will be to lessen the inventory swings contributed less to inventory swings in the during recessions. Others have tried to 1990 recession than in 1980 or 1982. directly assess the impact on the business Despite the reduction in contribution by cycle. Little (1 9 9 2 ), for example, focuses on these industries, the change in business the transitory nature of the changes and inventory contributed a higher proportion to suggests that the ongoing effort to reduce the change in GDP during the 1990-91 inventories were a drag on the recovery por downturn than in 1980 or 1981-82, but the tion of the 1990-91 recession. The expected magnitude of the decline in GDP was less in inventory accumulation after demand F ig u r e 4 1990-91 than in 1980 or 1981-82. On the surface, it appears that JIT may help reduce the magnitude of the inventory swing. Another way to test the impact of JIT on business cycles is to compare the Japanese with the U.S. experience. First, we can look for evidence that Japan does maintain lower inventory levels. Figure 5 shows the inventory-to-sales ratio for the Japanese and U.S. manufacturing sectors. The Japanese ratio is lower than the United States during the 1980s, but both ratios have converged as the United States’ decreased and Japan’s increased somewhat. Assuming that the lower inventoryto-sales ratio in Japan confirms the higher usage of JIT there, how does Japan’s business cycle experience compare with the United States’? Unfortunately, an exact comparison is not possible because Japan has not recorded many periods of declining output. Using dates from Japan’s Research Bureau Econom ic Planning Agency,5 Japan’s business cycles have had longer contractionary periods in the postwar era, averaging 16 months, compared with 11 months for the United States. Japan recorded 10 business cycles after World War II, compared to the United States’ nine. The average duration of Japan’s business cycles (50 months) and the expan sion periods (33 m onths) were shorter than the United States’ (63 months and 52 months, respectively). Table 3 shows the changes in Japanese business inventory compared to changes in GDP during its last six contractions. Three of these six contractions had countercyclical inventory movement. Similar data for the United States (Table 1) shows unambiguous procyclical movement in business inventory. The data suggest that inventory changes may play a lesser role in GDP fluctuations in Japan than in the United States. How much of this is attributable to inventory management methods and how much is due to the differ ence in business cycle definition is uncertain. Even if the use of JIT inventory manage ment methods can dampen business cycles, this method is most applicable at the manu facturing level. The contribution of manu facturing, wholesale and retail inventories to Ta b le 3 Changes in Jap a n e se In v e n to ry In vestm ent D u rin g Business Cycle Troughs Change in Real GDP2 Recession Period Peak to Trough Change in Inventory Investment2 Change in Inventory Investment as a Percentage of Change in Real GDP -5 2 7 .0 1 9 7 0 :3 - 1970:4 -1 5 8 .5 1 9 7 3 :4 - 1974:1 -5 2 9 6 .8 1 9 7 7 :2 - 1977:3 1417.0 1980:1 - 1980:2 -4 5 3 .0 4.3 1 9 8 5 :4 - 1986:1 -3 3 1 9 .0 952.2 1992:1 - 1993:4 -5 0 9 7 .0 332.5 7304.7 -1 3 7 .9 -4 9 9 .5 -3 5 .3 -0 .9 -2 8 .7 -2 8 0 7 .4 55.1 Mean 30.8 1 Peaks and trough correspond to peak ond trough (or minimum growth) of real GDP during the contractions listed by the Research Bureau of the Economic Planning Agency of Japan, but do not always coincide with the peak and trough of the period. 2 Billions of 1 9 8 5 Yen (SAAR). total trade inventories has been changing over the last two-and-a-half decades. More recently, manufacturing’s share has declined from 56 .8 percent to 4 3 .8 percent. Retail inventories have increased from a share of 24.3 percent to 31 percent. W holesale inventories’ share of the total has increased from 18.9 percent to 2 5.2 percent. The increased retail inventory-to-sales ratio and a greater retail share o f the aggregate inventory may offset the gains in dampening the cycle from JIT at the manufacturing level. CONCLUSION The data support anecdotal evidence that inventory management methods in the United States have changed significantly over the past decade or two. The result of these changes is evident in the reduced business inventory-to-sales ratio, driven almost entirely by lower inventories of work-in-process, and materials and supplies rather than finished goods. The impact 25 5 The Japanese agency uses the Lucas (1 9 7 7 ) definition, which loosely defines the business cycle in terms of deviation from trend growth. For most of the contrac tionary periods listed, Japan's GDP grew less than trend but did not experience a decline. REVIEW JULY/AUGUST 19 Larson, Paul D. "Transportation Deregulation, JIT, and Inventory Levels," The Logistics and Transportation Review (June 1 9 9 1 ), pp. 9 9 -1 1 2 . of these changes in inventory management techniques on business cycles is ambiguous. All other things being equal, inventory management innovations should reduce the probability o f unintended accumulation. But as long as firms overestimate or underestimate future demand, inventory cycles will persist. And if cutbacks in production are required to reduce inventory, then the resulting reduction in income could result in lower demand and further inventory buildup. Inventory management innovations are not a panacea for taming business cycles, but in the long run these innovations can contribute to a faster response of production to changes in demand, which in turn can reduce the boom -bust cycle in the econom y Little, Jane Sneddon. "C ha nges in Inventory Management: Implications for the U.S. Recovery," Federal Reserve Bank of Boston England New Economic Review (November/December 1 9 9 2 ), pp. 37-65. Lucas, Robert E., Jr. "Understanding Business Cycles," in Karl Brunner and Allan H. Meltzer, eds.,Stabilization of the Domestic and International Economy. Carnegie Rochester Conference Series, vol. 5. North-Holland, 19 7 7 . Metzler, Lloyd A. "The Nature and Stability of Inventory Cycles," Review of Economic Statistics (February 1 9 4 1 ), pp. 113-29 . Morgan, Donald P. "W ill Just-In-Time Inventory Techniques Dampen R e ce ssion s?" Federal Reserve Bank of Konsos City Economic Review (March/April 1 9 9 1 ), pp. 21-33. Olney, Martha L "Credit as a Production-Smoothing Device: The Case of Automobiles, 1 9 1 3 -1 9 3 8 , The Journal of Economic History (June " 1 9 8 9 ), pp. 3 7 7 -9 1 . "Retailing" Survey, The Economist (Morch 4, 1 9 9 5 ), pp. 3-18. REFERENCES Scarf, Herbert E. "T he Optimality of (S,s) Policies in the Dynamic Inventory Problem," in Kenneth. J. Arrow, Samuel Karlin and Patrick Suppes, eds., Mathematical Methods in the Social Sciences, 1 9 5 9 . Stanford University Press, 1 9 6 0 , pp. 1 9 6 -2 0 2 . Bechter, Dan M., and Stephen Stanley. "Econom ic Stability in the 1 9 9 0 s: The Implications of Improved Inventory Control," Business Economics (January 1 9 9 3 ), pp. 35-8. _ _ _ _ _ _. "Evidence of Improved Inventory Control," Federal Reserve Bank of Richmond Economic Review (January/February 1 9 9 2 ), pp. 3-12. Blinder, Alan S. Inventory Theory and Consumer Behavior. University of Michigan Press, 19 9 0 . _ _ _ _ _ _and Louis J. Maccini. "T he Resurgence of Inventory Research: W hat Hove We Le a rne d ?" Journal of Economic Surveys (No. 4 , 1 9 9 1 ) , pp. 2 9 1 -3 2 8 . Filardo, Andrew J. "Recent Evidence on the Muted Inventory Cycle," Federal Reserve Bank of Kansas City Economic Review (second quar ter 1 9 9 5 ), pp. 27 -43. Holt, Charles C., Franco Modigliani, John F. Muth and Herbert A. Simon. Planninq Production. Inventories, and Work Force. Prentice-Hall, Inc., 1960. 26 REVIEW JULY/AUGUST 1995 Alvin L. M a rly is a professor of economics and finance at the Center for Business and Government, Baruch College, City University of New York. Daniel L. Thornton is an assistant vice president at the Federal Reserve Bank of St. Louis. Jonathan Ahlbrecht provided research assistance. The authors would like to thank, without implication, Phillip Cagan, David Laidler and Allan Meltzer for comments on an earlier draft. Is There a Case for "M o d e ra te " Inflation? A lvin L. M arty and Daniel L. Thornton he proposition that inflation is a mone tary phenomenon is more widely embraced now than it was three decades ago. Moreover, it is more widely accepted that inflation is subject to long-run control by the central bank. In recent years, the cen tral banks of the United Kingdom, New Zealand and Canada have placed increased emphasis on reducing their long-run infla tion rates. In the United States, former Rep. Steven Neal, D.-N. Carolina (House Joint Resolution 55. January 5, 1993), and Sen. Connie Mack, R.-Florida, have pro posed making stable prices the primary objective of the Federal Reserve. Nevertheless, considerable opposition remains to making price stability the overrid ing objective o f U.S. monetary policy. Some argue that the benefits of price stability do not warrant the cost of achieving it. For example, although extolling the virtues of price stability, Howitt (1990) is uncertain whether the benefits are worth the costs in terms of lost output (temporary, and perhaps permanent, due to hysteresis effects). Although we are skeptical whether the empirical and theoretical analyses to date have correctly identified all of the benefits of price stability, this article addresses an issue that is logically prior to this one. Specifically, it addresses the question: If the inflation rate were zero, could society benefit from a high er rate of inflation? In other words, is mod erate inflation preferable to price stability? Several arguments have been advanced that the economy benefits from moderate infla tion. Recently, DeLong and Summers (1992) T and Summers (1 991) have suggested several rationales for why a central bank would choose moderate inflation over price stability as its long-run policy goal. This article addresses four reasons that have been suggested to prefer moderate to zero inflation: 1. Moderate inflation enhances the stability of the economy. 2. Moderate inflation results in a higher steady-state level of output per person. 3. Moderate inflation increases the efficiency of inter-industry labor market adjustments. 4. Inflation enhances the efficacy of counter cyclical monetary policy by allowing the real rate of interest to be negative, thereby stimulating effective demand in periods of recession. The first two of these arguments are wellknown to economists, but have received scant attention in public debates. Moreover, they are framed within specific, although quite different, theoretical models, so it is possible to provide a rather definitive evalua tion of their merit. The remaining two argu ments have received considerable attention, and may play a role in any public policy debate regarding the desirability of making price stability the primary monetary policy objective. The conceptual frameworks for these arguments are not well-specified, how ever, so we try to shore up their analytical footing by proposing specific interpretations. Before proceeding, several issues should be clarified. First, the hypotheses that there are econom ic benefits from moderate infla tion considered here implicitly argue against a so-called Friedman Rule (Friedman, 1969), that is, the “optimal” rate of money growth is one that generates steady-state deflation. Nevertheless, this article is not specifically about the Friedman Rule. Analyses of a Friedman Rule generally have been carried out in well-specified model economies. Second, although the last two arguments for moderate inflation lack explicit theoreti- REVIEW JULY/AUGUST cal foundations, this has not prevented them from achieving an intellectual status among some economists and policymakers. The lack of theoretical foundations forces our analysis to range from the fairly technical to the somewhat conjectural, so that we may not provide a definitive evaluation o f these arguments. In this case, we are content to present an analysis of these arguments. Third, the arguments for moderate infla tion analyzed here are based on the assump tion of a fully anticipated, steady-state infla tion. Although such inflations do not char acterize real-world economies, we make this assumption until it is relaxed when we dis cuss the reasons why price stability is prefer able to moderate inflation. Fourth, the phrase “moderate inflation” is not well-defined. Some might consider moderate inflation to be 2 to 3 percent. For others, any rate under 5 percent could be moderate. Still others may deem anything less than double-digit inflation moderate. We suggest 5 percent as the break point for moderate inflation in the United States. Finally, although we used the phrases price stability and zero inflation interchange ably, we are aware that price stability is dif ferent and more stringent than zero inflation. Price stability implies that jum ps in the price level are reversed; zero inflation need not. 1995 holdings of real balances, prices rise still fur ther, fueling expectations of further inflation. This reduces the quantity of real balances demanded still further, giving rise to a fur ther increase in prices and so on. The ques tion is: Under what conditions will this sequence converge? The answer is: Self-generating inflation cannot occur if, as the price level rises, its rate of change declines. Holding the growth of the money supply constant, this condition is illustrated in Figure 1, w hich shows two plots o f the change in the log of the price level, against the price level itself, P. In one case, the slope of the curve rises with P. In this instance, the sequence will not converge and the ultimate solution is the trivial one; the demand for real money balances approaches zero. In the other case, the slope of the curve decreases as P increases, so that the sequence converges to a steady rate of inflation. The argument that inflation enhances econom ic stability is an argument about the demand for money. To see this argument, assume that the demand for real money bal ances is solely a function of the nominal interest rate, whereby the nominal interest rate equals the constant real interest rate plus the actual rate of inflation (which is fully anticipated). Cagan (1 9 5 6 ) showed that the rate of inflation decreases as the price level rises if a/3 < 1. The parameter a is the semi-log slope of the demand for real money balances with respect to the nominal interest rate, that is, percent change in the demand for real money balances per percent age point change in the money interest rate (d In(M /P)/di). The parameter (3 is the rate at which individuals revise their expectations of inflation under adaptive expectations, dE/dt = (3(/rr - E), where rr and E are the actu al and expected rates of inflation, respectively. If a/3 < 1 and the expected rate of infla tion is initially greater than the actual rate, expected inflation falls until a stable steady state is reached at which the expected rate of inflation is equal to the actual rate. If, how ever, a [3 is greater than unity and actual inflation is initially greater than expected inflation, both expected and actual inflation grow without lim it with real balances falling. THE CASE FOR MODERATE INFLATION Is Stability of the Economy Enhanced b y Moderate Inflation? The first argument for moderate infla tion is that certain stability conditions are sturdier at a high-money (nominal) rate of interest, making the economy less vulnerable to various shocks. Understanding this argu ment requires an understanding of the notion of stability upon which it rests. To illustrate, suppose that the real rate of inter est suddenly rises, say, because of an increase in expected future profits. Given the under lying rate of inflation, this raises the money rate of interest, reducing the quantity of real money balances that individuals desire to hold. As individuals attempt to reduce their 28 REVIEW JULY/AUGUST 1995 tion.2 This suggests that the underprediction was due to a being higher at low rates of inflation. That a is inversely related to the steadystate inflation rate is plausible, but this does not imply that stability is more likely at higher rates of inflation. For example, it is plausible that individuals revise their expec tations of inflation more rapidly at higher rates of inflation, that is, that f3 is positively related to the inflation rate. Indeed, Bruno (1 9 8 9 ) provides some empirical support for a positive association between [3 and the rate of inflation. Consequently, it is not necessari ly the case that stability is greater at high inflation rates. An inflation-induced fall in a might be ju st offset, or perhaps more than offset, by an inflation-induced rise in / 3. Using adaptive expectations, no general con clusion can be reached about the stability conditions and the steady-state inflation rate. F ig u r e 1 Log of th e Price L eve l V s. A ctu a l Price Level A log P initially less than expected inflation, both actual and expected inflation fall without limit. The steady state at which the actual and expected rate are equal is unstable. Stability Through Moderate Inflation Getting on the W rong Side of the Laffer Curve If a and / are constants, the stability 3 conditions will be invariant to the steadystate inflation rate. Consequently, the sug gestion that the stability conditions are stur dier at non-zero rates of inflation is an argu ment that either a or / is inversely related to 3 the steady-state inflation rate. Specifically, it was argued that a should be smaller at high er rates of money interest. That this may be so comes from noting that the elasticity of the demand for real money balances with respect to the money interest rates, em is , equal to a i, where i is the money interest rate. Thus,’ if e m is constant, a will decline as ’ the rate of inflation and, hence, the money interest rate rises.1 All other things the same, the stability condition is more likely to be satisfied at higher rather than lower rates of money interest if the interest elasticity demand for real money balances is constant. The widely used Cagan (1956) moneydemand function assumes that a is indepen dent of the nominal interest rate. Cagan’s function significantly underpredicted real money balances during periods when prices were or had been relatively constant, but performed well during periods of high infla Bruno and Fischer (1 9 9 0 ) have revisited Cagan’s stability condition in the context of financing a given budget deficit solely through seigniorage from money creation. Although assuming that the deficit is financed solely through money creation is not realistic in developed economies like the United States, where other forms of taxation are available, the Bruno and Fischer assump tion is a useful theoretical device which allows stability conditions to be anchored by two equilibria on either side of the Laffer curve. The Laffer curve is the recognition that tax receipts do not increase continuous ly with the tax rate. Beyond some point, receipts decline as a further increase in the tax rate results in a significant erosion of the tax base. Consequently, except at that unique tax rate where tax revenue is m axi mized, there are two alternative tax rates and tax bases that generate the same tax revenue: a low tax rate and a high tax base, or a high tax rate and a low tax base. Bruno and Fischer demonstrate that if a/3 < 1, a stable equilibrium is at the socially desirable low tax rate-high tax base point. If a)3 > 1, an equilibrium is at the socially undesirable 29 1 Note that em is constant if the m money demand function is in dou ble log form: M /P=ik o ,s e=ai=k. m 2 See Bailey (1 9 5 6 ) and Friedman and Schwartz (1 9 6 3 ) for a discus sion of the issue of the empirical validity of Cagan's moneyilemand equation. REVIEW point on the Laffer curve. Consequently, the argument is not whether the system is stable or explosive, but whether equilibrium is achieved at a socially desirable point on the Laffer curve. money balances and physical capital are sub stitutes. A higher anticipated rate o f inflation induces individuals to economize on their holdings o f money balances, freeing up sav ings for capital accumulation. This leads to a higher steady-state capital/labor ratio, resulting in higher consumption per person so long as the steady state is not pushed beyond the point where consumption per person is maximized (the so-called Golden Rule point). The Tobin effect, that higher inflation induces higher levels o f capital, output and consumption per worker, is open to a number of objections. For one thing, it is dependent on Tobin’s assumption that savings are a con stant proportion of income. If the savings rate is directly reduced by higher inflation, the Tobin effect can be reversed— even in the framework of Tobin’s model (Dornbusch and Frenkel, 1973). Moreover, the Tobin effect is model-specific. The effect is absent in Ramsey-type models, in which the marginal product of capital is tied to the representa tive agent’s rate of time preference. In such models, the marginal product of capital defines a unique steady-state capital\labor ratio w hich is independent of the level of real money balances. It is now generally rec ognized that the results of both Tobin- and Ramsey-type models are sensitive to small changes in assumptions. Moreover, Orphanides and Solow (1 990) show that different models or small changes in assump tions in a particular model deliver disparate conclusions about the Tobin effect. Consequently, it is impossible as a matter of pure theory to make a compelling case that inflation increases real output. A crucial reason for the fragility of these results is that, by their very nature, these money-growth models are the wrong vehi cles for analyzing the role of money in the economy and, hence, the effect of inflation on the economy. A striking example of this is provided by Tobin’s model, which predicts that the highest level of output per person occurs in a barter economy, in which hold ings of real money balances are nil. This result stems from not taking money serious ly. Real money balances reduce transaction costs. They do this by overcoming the dou- What If Expectations Are Rational? From the condition that E = t t , it is easy to see that rational expectations are the lim iting case of adaptive expectations. Adaptive expectations approach rational expectations as / — cc. Consequently, if expectations are 3 » rational, the condition a(3 < 1 cannot be sat isfied for any value of a . In Cagan’s world, the system explodes. In the Bruno and Fischer world of a fixed real deficit, a stable equilibrium (if it exists) is achieved at a high inflation rate on the wrong side of the Laffer curve. Under rational expectations, any affect on a is completely overwhelmed by /3, which is infinite. It appears that nothing remains of the argument for stability through inflation. In the case of adaptive expectations, any possi ble reduction in a due to inflation may be offset by an increase in /3. If expectations are rational, an infinite / swamps any effect of 3 inflation on a . Indeed, the stable equilibri um is at the socially undesirable side of the Laffer curve, that is, at a high rate of infla tion (tax rate) and a low level of real cash balances (tax base). In particular, no argu ment can be made that moderate inflation produces stability on the socially desirable side of the Laffer curve. Does Moderate Inflation Lead to a Higher Level of Output? The second argument for moderate infla tion, that it leads to a higher level of steadystate output and consumption, was first for mulated by Tobin (1965). The essence of Tobin’s model is that in a growing economy, non-interest bearing real money balances augment disposable income. Given that the propensity to save out of disposable income is less than unity, an increase in real bal ances, all of which must be saved, gives rise to smaller saving in the form of physical cap ital. In Tobin’s portfolio-balance model, real 30 REVIEW we must deal with how inflation interacts with real-world institutions. It has been shown that the interaction of inflation with a less-than-fully indexed tax system works to discourage capital accumulation (Feldstein, 1976, 1979; and Tatom, 1976). The bottom line is that even within the framework of theoretical-growth models, the Tobin effect is subject to small changes in assumptions. W hen real-world institutions are included in the analysis, the weight of evidence is that inflation discourages capital accumulation. The Tobin effect is reversed.3 W hen capital is defined more realistically to include human capital, the effect of inflation is to continually reduce the levels of output per person below what they would have been under stable prices. ble coincidence of wants associated with barter and by conveying information (for example, Brunner and Meltzer, 1971). Compared to a barter economy, the reduction in total transaction costs permits society to devote more of its scarce resources to produc tion, raising output and the consumption of goods and leisure. By reducing marginal transaction costs, money also results in a higher level of trade and correspondingly higher levels of output. Although the devel opment of Solow-type growth models was an important first step in the analysis of growing economies, it is not surprising that these onecommodity models fail to capture the impor tant role that money plays in real-world economies. Thus far, we have contrasted money and non-monetary barter economies. In princi ple, similar effects occur when individuals are induced by a rise in anticipated inflation to reduce their holdings o f real balances. If inflation induces individuals to hold fewer real balances, even if one were to accept Tobin’s argument that inflation increases out put per person, any increase would be at the expense of a loss to society of the services of real balances. In fact, if money enters the production function, the Tobin effect may well be reversed; inflation then reduces out put per person, as in Stockman (1981). Inflation may not only reduce the steadystate level of per capita output, it may reduce the growth rate of output itself. For if capital is appropriately defined more broadly to include human capital, as is done in recent endogenous-growth models, inflation reduces investment in human capital, as well as in physical capital. Reduced investment in human capital lowers the growth of efficien cy per person, which reduces the growth rate itself (for example, Lucas, 1988; King and Rebelo, 1990; and Dotsey and Ireland, 1993). A small but permanent reduction in the growth rate due to inflation has an adverse effect on output levels. This continual effect on output levels is more significant than any effect o f inflation on the one-time altering of the level of output per capita explored in ear lier exogenous-growth models. Once we leave purely theoretical-growth models, and look at real-world economies, Does Moderate Inflation Enhance Relative Real-Wage Adjustments? The third argument for moderate infla tion (Tobin, 1972; Schultze, 1985; Lucas, 1989; DeLong and Summers, 1992; and Summers, 1991) asserts that declines in the price of commodities and in the real wage of workers specialized to a particular industry can be made with less friction in a world with moderate inflation than in a world of stable prices. It is argued that under moder ate inflation, the decline in a product’s price and in the real wage rate of workers can be accomplished through a rise in prices and money wages elsewhere. The belief that inter-industry adjust ments are smoother under a regime of mod erate inflation rests on the view that laborers prefer a rise in the prices of wage goods to an absolute reduction in money wages. But why should this be the case? W orkers expe rience an identical decline in real wages in both cases. One answer depends on the existence of a money illusion: A decline in real wages brought about by a rise in the prices of wage goods is incorrectly perceived as smaller than the same decline in the real wage that occurs through a reduction in money wages. The persistence of money illusion in a steady state of anticipated moderate inflation is dif ficult to rationalize. Moreover, recent evi- NK OF ST. L OU I S 31 3 After o survey of theoretical mod els, Blanchard and Fischer (1 9 8 9 ) conclude: "Calculations suggest, however, that the effects of changes in the inflation rote on cap ital accumulah'on in models of the type developed in this chapter are very small. If inflation has system atic effects on copital accumulation (and there is empirically a negative association), it is probobly for rea sons not included so for. One likely reason is that the tax system is not neutral with respect to inflation." REVIEW dence (McLaughlin, 1994; and Lebow, Stockton and Wascher, 1993) suggests there is no dearth of nominal wage cuts, even dur ing periods of moderate inflation. Furthermore, firms in a declining indus try may adjust workers’ compensation with out cutting wages. Compensation includes benefits and perks which can be adjusted rel atively easily relative to wages in cases in w hich workers have an irrational fear of nominal wage cuts. In any event, we believe that for the resistance to nominal wage cuts to be widespread, it must be motivated by considerations deeper than a pure money illusion. One possible rationale for such resis tance is that workers feel they have some control over money wages but no control over the general price level. Consequently, the same reduction in the real wage rate due to reduction in money wages brings into play factors that workers believe they can negotiate, in contrast to an increase in the prices of wage goods, which they are powerless to affect. The second possible motivation would interact with the first. Workers may have less knowledge of demand than do employ ers. Consequently, when the industry demand declines, workers may be concerned that the employer is misrepresenting the true state of nature to force an unnecessary reduction in money wages. In this case, a fall in the real wage rate due to a rise in the prices of wage goods elsewhere avoids trig gering a signal-extraction problem. We have endeavored to make the best possible case for moderate inflation as a device for smoothing inter-industry wage adjustments, but in doing so, we have ignored the existence of a cushion on money wage declines even in a regime of stable prices. If we were to introduce technical progress, even under price stability, the aver age level of money wages would rise at a rate equal to the average increase in output per person. This provides a cushion mitigating the need for an absolute decline in the money wage. Finally, we suggest the hypothesis that workers’ resistance to nominal wage cuts is not independent of the inflation regime in which they live. Under stable prices, such cuts may become more frequent and workers will become more accustomed to and less distrustful of money wage cuts. Accordingly, any “lubricant” that moderate inflation may provide to ease labor market frictions will become increasingly unnecessary in a zeroinflation regime. Although the claim that moderate infla tion facilitates inter-industry wage adjust ments cannot be definitively rejected, it does not rest on compelling theoretical or empiri cal foundations. In any event, monetary pol icy is an inappropriate and ineffective instru ment for dealing with labor market prob lems, such as market frictions or the sub optimality of the natural unemployment rate. The latter may be due to taxes on wages which make the after-tax real wage smaller than the before-tax marginal product of labor. The socially optimal amount of employment equates the disutility of labor to the before-tax real wage so that after-tax employment is sub-optimal. Moreover, high unemployment compensation increases time spent in “search unemployment.” The drift to higher unemployment in Europe and Canada is unlikely to reflect a movement up a short-run Phillips curve produced by unan ticipated deflation, but rather is due to an upward drift in the natural rate of unemploy ment. In this case, appropriate policies to reduce unemployment are reforms in taxes and unemployment compensation, not mon etary policy. Moderate Inflation Enhances the Countercyclical Efficacy of Monetary Policy A fourth argument for moderate steadystate inflation is that it enhances the counter cyclical efficacy of monetary policy by enabling the Federal Reserve to make the real rate of interest negative. The argument that the efficacy of monetary policy is enhanced by a moderate rate of steady-state inflation stems from the recognition that the money rate can never be negative, so that in a non-inflationary environment, in which the real and money rates are equal, the best that monetary policy can do is to drive both the 32 REVIEW JULY/AUGUST real and nominal rates of interest to zero.4 DeLong and Summers (1992) and Summers (1991) argue that in a zero inflation regime, monetary policy will be unable to produce a sufficiently large reduction in the real interest rate to restore full employ ment in times of large adverse shocks to aggregate demand. Is this an important argument for mod erate inflation? There are several reasons to think not. First, the argument is based on the belief that the monetary authority can exert considerable influence over real inter est rates through the so-called liquidity effect, and that monetary policy works pri marily, if not solely, through its ability to influence the real interest rate. According to this view, an expansionary monetary policy drives real interest rates down, inducing an increase in spending. But the extent and duration of the effect of monetary policy on short-term real interest rates is controversial, theoretically and empirically. The exchange between Ohanian and Stockman (1995) and Hoover (19 9 5 ) highlights the difficulties with theoretical models of the liquidity effect. The empirical evidence on the liquid ity effect is mixed. W ork by Reichenstein (1 9 8 7 ), Thornton (1 9 8 8 ), Gordon and Leeper (1994) and Pagan and Robertson (19 9 5 ) suggests the liquidity effect is rela tively weak and short-lived, although research by Christiano and Eichenbaum (1991, 1992), Cook and Hahn (19 8 9 ) and Romer and Romer (1990) suggest a more sufficiently at very low but positive interest rates, the efficacy of monetary policy is not impaired by a zero lower bound on the real interest rate. Bailey’s argument suggests that credit demand becom es very large (essential ly infinite) at very low real interest rates, so that the real longer-term interest rates do not have to be negative to significantly increase investment. Finally, despite the empirical evidence to the contrary, there appears to be a fairly widespread belief that the Federal Reserve exerts considerable influence over real short term interest rates, but much less influence over longer-term interest rates (see, for example, Goodfriend, 1993; and Greenspan, 1993). If monetary policy cannot make the long-term rate negative, it is natural to ask: Is there any gain from the possibility that the Federal Reserve may be able to make short term interest rates negative for temporary periods? In markets in which there are few impediments to the flow of funds between the long and short end of the market, consis tency of expectations requires that the cur rent long-term interest rate be equal to the expected average of future short-term rates plus a risk premium. The risk premium is affected by a num ber o f things, including uncertainty about future short-term interest rates. If the market believes that the policy does not signal an increase in policymakers’ desired steady-state inflation rates, people know that today’s policy must give rise to reversals later. W hether the difference in the magnitude of the decline and subsequent rise in short-term interest rates in the zero and moderate inflation regimes will result in sig nificantly different paths for real long-term interest rates under the two regimes is impossible to determine, a priori. Indeed, it is as easy to conjecture scenarios in which there would be no difference in the response of long-term real interest rates under the two steady-state inflation regimes as it is to con jecture scenarios in w hich there would be a significant difference.5 Given that it is unlikely that moderate inflation will enable the Federal Reserve to have a significantly larger effect on long-term real interest rates, and that very low or zero significant effect of m onetary policy on real short-term interest rates. Second, it is difficult to argue that suffi cient investment opportunities will not exist unless the real rate is negative. The issue is whether the economic outlook can become sufficiently pessimistic that the expected real return on longer-term investments is nega tive. That DeLong and Summers (1992) and Summers (1991) have raised it again sug gests that this old debate is far from settled. Bailey (1971) argues that there will always be some investments that yield a small non negative return, even if a depressed economy were not expected to return to its steadystate growth path for a period of 10 to 20 years. If investment opportunities increase 1995 33 4 No one would willingly trade a dol lar for, say, 9 5 cents a year from now, so long as the same dollar could be held for a year at zero carrying cost. 5 Indeed, Fuhrer and Madigan (1 9 9 3 ) simulate the effect of more aggressive policies that result in negative short-term interest rotes and find very small changes in long-term rates. REVIEW JULY/AUeUST real interest rates are likely to be sufficient for the Fed to offset adverse aggregate de mand shocks, the argument that moderate inflation enhances the efficacy of monetary policy seems doubtful. If some role for infla tion uncertainty is factored in, the idea that moderate inflation enhances the efficacy of monetary policy becomes even more tenuous. 1995 financial service industries expand relative to other employment of resources such as industrial output, and households sacrifice leisure to reduce their real balances when the inflation tax rises. These effects call into question the notion that, by penalizing the consumption of priced commodities, infla tion reduces work effort and increases leisure. Although it is difficult to quantify the degree to which inflation impairs the ability of the price system to signal correct informa tion, there is no doubt that the price system allocates resources most efficiently in the absence of inflation. Moreover, zero inflation is preferable to moderate inflation because inflation, even moderate inflation, distorts accounting, legal contracts and the tax system. Inflation also distorts the true cost of inventories, the depreciation o f plant and equipment, as well as the time profile of real mortgage pay ments, and other fixed-dollar denominated contracts. O f course, this analysis assumes that taxes and private contracts are not indexed against inflation. Why, it may be asked, don’t the authorities index taxes against changes in the price level so that real pay ments are unaffected? Why, in turn, doesn’t the private sector index wages and finan cial contracts to nullify the impact of price changes? In fact, the tax code is now partly indexed against inflation. Indexation, how ever, is often taken as a signal that the authorities are giving up the battle against inflation. This was the basis for the outspo ken opposition to indexation by former Federal Reserve Chairman Arthur Burns and why the Bank of Canada has opposed index ation. It is easy to construct examples in which inflation-mitigating schemes, such as indexation by reducing the marginal costs of inflation, lead to an increase in the aggregate inflation rate. Moreover, foregoing indexa tion may be a help in developing a reputa tion for credibly pursuing anti-inflation poli cies (see, for example, Fischer and Summers, 1989). For these reasons, it is not clear that indexation of tax codes is desirable. In the private sector, indexation is unlikely to occur. At the heart of the diffi- W h y Zero Inflation Is Preferable Although many estimate the output loss es of moderate inflation to be modest, this issue is far from settled. In addition to the usual problems of measuring the permanent output losses, Dotsey and Ireland (1993) have shown that in a general-equilibrium analysis, the usual effects of inflation (the inefficient economizing on real money bal ances, substituting market activity for leisure, and redirecting resources from goods production to financial activities) compound to produce a significant output loss. Dotsey and Ireland’s result stems in part from the fact that inflation lowers real output growth. Although the effects on output growth appear small, compounded over time, they are significant. Another compelling reason to prefer zero inflation is that higher inflation is asso ciated with increased variability of both inflation and relative prices. The increased variability of inflation and consequent infla tion uncertainty shorten contract lengths, thereby increasing contract costs. The greater variability also contaminates price signals, so the price system conveys less information. As the variability of inflation (associated with higher inflation) increases, it becomes more difficult to determine whether a particular commodity price change reflects a movement in the general price level, or a real shift in supply or demand resulting from taste and productivity shocks. In addition, inflation and the higher variability of the general price level cause a reallocation of resources from the production of goods to financial services for the sole purpose of hedging against inflation. Even if there is no reduction in conventional mea sures of output, inflation produces a distor tion o f output. The banking system and 34 r e v i e w JULY/AUGUST This asymmetric behavior extends to supply shocks as well.8 Adverse shocks will be accommodated; favorable ones will be ignored. Although the price level depends on many factors, including the relative inci dence of positive and negative shocks, con cern for transitional unemployment leads a central bank to pursue policies that will cause the price level to be higher than it would be otherwise. This asymmetric behavior creates an inflationary bias with the potential for accel erating inflation. As inflation increases, the monetary authority may be forced to tolerate transitional unemployment to bring the inflation down. Indeed, this appears to be what happened in the United States in the late 1970s and on a smaller scale in the late 1980s. The best way to avoid such disrup tions is to commit to a policy of stable prices. culty is a coordination problem. To be suc cessful, indexation must be implemented by a large number of diverse firms facing different information. For example, are price changes due to nominal or real variables? It is wellknown that indexing money wages to changes in prices due to real shocks is undesirable.6 Are price changes permanent or transitory? Are the price changes industry-specific or global? It is unlikely that individuals will agree on the cause of a price change and then coordinate their actions. Given these obstacles, indexation in the private sector is difficult and, hence, fairly rare. Moreover, in instances in which private indexation is fair ly widespread, as in Israel, it reduces the resolve to fight escalating inflation. Finally, it is impossible to fully index real cash balances against inflation, as previ ously discussed, because inflation leads peo ple to hold fewer cash balances and results in a loss of their services.7 For these reasons, indexation is a frail reed on which to rest hopes of mitigating inflation’s effects. SUM M ARY AND CONCLUSIONS We have reviewed several arguments in favor of moderate inflation and we find them to be lacking theoretically and, in some instances, empirically. The first argument, that moderate inflation enhances economic stability, is subject to compelling objections. If expectations are adaptive, any decline in the semi-elasticity of money demand associ ated with a higher inflation may well be off set by a more rapid revision of inflationary expectations. If expectations are rational, this must be the case. The argument that inflation leads to a higher level of output is based on theoretical models that are not robust to small specifica tion changes. W hen real-world institutions are taken into account, the weight of the evi dence is that inflation discourages capital accumulation. W hen capital is defined to include human capital, inflation may reduce not only the level o f output per capita but its rate of growth as well. Also suspect is the proposition that moderate inflation increases the efficacy of monetary policy by allowing the central bank to make the real rate of interest nega tive. Sufficient investment opportunities are likely to exist at very low but positive real Price Stability as the Objective of Monetary Policy Reducing an established moderate infla tion trend may disrupt econom ic activity, producing temporary output and employ ment losses. Given an established moderate inflation rate, Howitt (1990, p. 104) argues that despite the desirability of zero inflation, the cost of achieving it probably outweighs the benefits. This argument against moving to price stability ignores the inflationary bias (and resulting uncertainty) that characterizes policy regimes motivated by concerns for transitional output and employment losses. In the absence of a commitment to sta ble prices, a central bank concerned about transitional unemployment is likely to respond asymmetrically to shocks— tempo rary or permanent. This asymmetric behav ior has clear implications for the price level in the case of demand shocks. A monetary authority concerned with transitional unem ployment will be less willing to offset a demand shock that raises prices and employ ment than to offset an adverse demand shock that lowers prices and employment. 1995 35 6 The authorities in Israel indexed money wages to a price index which included imported goods. In fact, imported goods should be excluded since changes in import prices reflect changes in a real vari able, the terms of trade. Later on, this mistake was rectified and terms of trade effects were exclud ed from the price index. 7 In principle, to maximize the ser vices of real bolances, it would be desirable to have prices fall at the real rate of interest and set the money rate of interest to zero. People would then be induced to hold the satiety quantity of real bal ances. It would take us too far afield, however, to discuss the mer its of deflation at the real rote in comparison with stable prices. Therefore, we confine our attention to a comparison of zero with mod erate inflation. 8 In the case of the oil shock in the 1 9 7 0 s (an adverse supply shock which tended to raise inflation and reduce output), a number of econo mists advocated a quantum increase in the stock of money to offset a potential increase in unem ployment. On the other hand, how many voices were raised in favor of a reduction in the money stock when OPEC collapsed? HEVItN interest rates. Consequently, negative real rates are not required to make monetary pol icy effective. Also, even if positive inflation enabled the Fed to make real short-term interest rates negative, such actions may not lower long-term interest rates. The proposition that moderate inflation eases inter-industry wage adjustments is weak too. One argument rests on the exis tence of a money illusion; we see no econom ic rationale for money illusion in the steady state. If the asserted resistance to nominal wage cuts is based on a deeper motivation, we suggest that it should disappear entirely as the regime of zero inflation persists. Moreover, the evidence suggests that nominal wage cuts are frequent even during periods of moderate inflation. Hence, the conjecture that workers resist nominal wage cuts lacks both theoretical and empirical justification. Finally, we argue that a policy of living with inflation cannot be rationalized on the grounds that there are transitory output costs associated with reducing inflation. A policy motivated by concern for transitional unemployment is likely to have inflationary bias that will erode a commitment to any price objective. _ _ _ _a n d_ _ _ "Identification and the Liquidity Effect of a M onetary Policy Sho ck ," National Bureau of Economic Research Working Paper No. 3 9 2 0 (1 9 9 1 ). REFERENCES _ _ _ _ _and Anna Jacobson Schwartz. A Monetary History of the United States 1867-1960. Princeton University Press, 19 6 3 . Cook, Timothy, and Thomas Hahn. "T he Effect of Changes in the Federal Funds Rate Target on Market Interest Rates in the 1 9 7 0 s , " Journal of Monetary Economics (November 1 9 8 9 ), pp. 33 1 -5 1 . DeLong, J. Bradford, and Lawrence H. Summers. "Macroeconomic Policy and Long-Run Growth," Federal Reserve Bank of Kansas City Economic Review (fourth quarter 1 9 9 2 ), pp. 5-29. Dornbusch, Rudiger, and Jacob A. Frenkel. "Inflation and Growth: Alternative Approaches,"Journal of Money, Credit and Bonking (February 1 9 7 3 ), pp. 14 1-56 . Dotsey, Michael, and Peter Ireland. "O n the Welfare Cost of Inflation in General Equilibrium," unpublished manuscript (August 1 9 9 3 ), Federal Reserve Bank of Richmond. Feldstein, Martin S. "T he Welfare Cost of Permanent Inflation and Optimal Short-Run Economic Policy,"Journal of Political Economy (August 1 9 7 9 ), pp. 7 4 9-68 . _ _ _ _ _. "Inflation, Income Taxes, and the Rate of Interest: A Theoretical Analysis/'The American Economic Review (December 1 9 7 6 ), pp. 8 0 9-20 . Fischer, Stanley, and Lawrence H. Summers. "Should Nations Learn to Live with Inflation?" The American Economic Review (M a y 1 9 8 9 ), pp. 3 8 2-87 . Friedman, Milton. "The Optimum Quantity of M o ne y," The Optimum in Quantity of Money and Other Essays. Aldine Publishing Co., 19 6 9 . Bailey, Martin J. National Income and the Price Level: A Study in Macroeconomic Theory. McGrow Hill, 1 9 7 1 . Fuhrer, Jeff, and Brian Madigan. " A Monetary Policy W hen Interest Rates are Bounded at Zero," unpublished manuscript (August 2 5 , 1 9 9 3 ), Board of Governors of the Federal Reserve System. _ _ _ _ _ . "The Welfare Cost of Inflationary Finance," Journal of Political Economy (April 1 9 5 6 ), pp. 93 -1 1 0 . Goodfriend, Marvin. "Interest Rate Policy and the Inflation Scare Problem: 1 9 7 9 -1 9 9 2 , " Federal Reserve Bank of Richmond Economic Quarterly (winter 1 9 9 3 ), pp. 1-24. Blanchard, Olivier J., and Stanley Fischer, lectures on Macroeconomics. M IT Press, 19 8 9 . Brunner, Karl, and Allan H. Meltzer. "The Uses of Money: M oney in the Theory of an Exchange Economy,"The American Economic Review (December 1 9 7 1 ), pp. 78 4 -8 0 5 . Bruno, Michael. "Econometrics and the Design of Economic Reform," Econometrica (March 1 9 8 9 ), pp. 27 5 -3 0 6 . Gordon, David B., and Eric M . Leeper. "The Dynamic Impacts of M onetary Policy: An Exercise in Tentative Identification," Journal of Polihcal Economy (December 1 9 9 4 ), pp. 1 2 2 8 -4 7 . Greenspan, Alan. "Statem ent before the Subcommittee on Economic Growth and Credit Formation of the Committee on Banking, Finance and Urban Affairs of the United States House of Representatives," federal Reserve Bulletin (September 1 9 9 3 ), pp. 84 9 -5 5 . _ _ _ _ _and Stanley Fischer. "Seigniorage, Operating Rules, and the High Inflation Trap," Quarterly Journal of Economics (M a y 1 9 9 0 ), pp. 3 5 3-74 . Hoover, Kevin D. "Com m entary," this Review (M ay/June 1 9 9 5 ), pp. 26-32. Cagan, Philip. "The Monetary Dynamics of Hyper Inflation," in Milton Friedman, ed., Studies in the Quantity Theory of Money. University of Chicago Press, 19 5 6 . House Joint Resolution 55. "Directing the Federal Reserve Open Market Committee of the Federal Reserve System to Adopt and Pursue Monetary Policies Leading To, and Then Maintaining Zero Inflation," 103rd Congress, 1st Session (January 5, 1 9 9 3 ). Christiano, Lawrence J., and Martin Eichenbaum. "Liquidity Effects and the Monetary Transmission M echanism ," American Economic The Review (M a y 1 9 9 2 ), pp. 3 4 6 -5 3 . F E D E R A L RE S E RV E B A N K OF ST. L O U I S 36 HtVIEW Y/AUGUST Howitt, Peter. "Zero Inflation as a Long-Term Target for Monetary Policy," in Richard G. Lipsey, ed., ZeroInflation: The Goal of Price Stability. C. D. Howe Institute, 19 9 0 . Reichenstein, William. "T he Impact of M oney on Short-Term Interest Rates," Economic Inquiry (January 1 9 8 Z ), pp. 6Z-82. Romer, Christina D., and David H. Romer. "N e w Evidence on the Monetary Transmission M echanism ," Brookings Papers on Economic Activity (1 9 9 0 :1 ), pp. 149-98 . King, Robert G., and Sergio Rebelo. "Public Policy and Economic Growth: Developing Neoclassical Implications," Journal of Political Economy (October 1 9 9 0 , Part 2 ), pp. S I 26-50. Schultze, Charles L. "Microeconomic Efficiency and Nominal Wage Stickiness," The American Economic Review (March 1 9 8 5 ), pp. 1-15. Lebow, David E., David J. Stockton and William L. Wascher. "Inflation, Nominal W age Rigidity, and the Efficiency of Labor M arkets," unpub lished manuscript (August 1 9 9 3 ), Board of Governors of the Federal Reserve S y s te m . Stockman, Alan C. "Anticipated Inflation and the Capital Stock in a Cash-in-Advance Economy,"Journal of Monetary Economics (November 1 9 8 1 ), pp. 38 7 -9 3 . Lucas, Robert E., Jr. "O n the Mechanics of Economic Development," Journal of Monetary Economics (July 1 9 8 8 ), pp. 3-42. Summers, Lawrence. "H o w Should Long-Term Monetary Policy Be D eterm ined?" Panel Discussion, Journal of Money, Credit and Banking (August 1 9 9 1 , Part 2), pp. 62 5 -3 1 . Lucas, Robert F. "The Bank of Canada and Zero Inflation: A New Cross of G o ld ?" Canadian Public Policy (March 1 9 8 9 ), pp. 84 -93. Tatom, John A. "The Welfare Cost of Inflation," this Review (November 1 9 7 6 ), pp. 9-22. McLaughlin, Kenneth J. "Rigid W a g e s ?Journal of Monetary " Economics (December 1 9 9 4 ), pp. 3 8 3 -4 1 4 . Ohanian, Lee E., and Alan C. Stockman. "Theoretical Issues of Liquidity Effects," this Review (M ay/June 1 9 9 5 ), pp. 3-25. Thornton, Daniel L. "The Effect of M onetary Policy on Short-Term Interest Rates," this Review (M ay/June 1 9 8 8 ), pp. 53-72. Tobin, James. "Inflation and Unemployment," American Economic The Review (March 1 9 7 2 ), pp. 1-18. Orphanides, Athanasios, and Robert M . Solow. "M oney, Inflation and Growth," in Benjamin M. Friedman and Frank H. Hahn, eds., Handbook of Monetary Economics, vol. I. North-Holland, 19 9 0 . _ _ _ _ _ _ _ _ _ _ _ _ _ . "M o n e y and Economic Growth," Econometrica (October 1 9 6 5 ), pp. 6Z 1 -84. Pagan, Adrian R., and John C. Robertson. "Resolving the Liquidity Effect," this Review (M ay/June 1 9 9 5 ), pp. 33 -54. 1 9 9 5 UIS 37 KEVIN) JULY/AUGUST 1995 David C. Wheelock is a senior economist at the Federal Reserve Bank of St. Louis. Paul W. Wilson is an associate professor of economics at the University of Texas at Austin. Heidi L. Beyer provided research assistance. Evaluating the Efficiency of Commercial Banks: Does O ur V ie w of W hat Banks Do M atter? estimation techniques have been proposed, each with advantages and disadvantages. The problem is complicated by the myriad of different services that commercial banks perform. Researchers deal with complex issues in measuring bank production: Is a deposit an input to the production process, or an output? Should outputs be measured in terms of the number of a bank’s accounts, the number of transactions it processes or the dollar amounts of its loans or deposits? Perhaps not surprisingly, estimates of com mercial bank inefficiency vary considerably across studies that use different techniques, conceptions of bank production and data samples. This article investigates the sensitivity of efficiency measures to broadly different conceptions of how banks operate. We use a single-estimation technique and a common pool of banks to compare efficiency measures based on alternative views of bank production. We find substantial differences in mean efficiency across models and low, though statistically significant, correspondence in the rankings of banks by efficiency scores across models. First, we discuss why measuring commercial bank efficiency is useful, some alternative measures of efficiency and techniques for estimating efficiency. A description of the approach we take, our data and our results follow. David C. W heelock and Paul W . Wilson n the past 15 years, the banking industry has faced growing competition from other financial service firms and financial mar kets and, at the same time, has undergone substantial deregulation and change. Proponents of further deregulation, such as the removal of barriers to the commingling of commercial and investment banking, argue that such changes would enhance the efficiency and viability of American banks. The impact of competitive and regulatory changes on banks can be judged by gross measures of performance, such as profitability and failure rates. Econom ists are also inter ested in how such changes affect the efficiency with which banks transform resources into various financial services. Inefficiency implies that resources are wasted, that is, that firms are producing less than the feasible level of output from the resources employed, or are using relatively costly combinations of resources to produce a particular mix of products or services. Thus, a goal of policymakers, as well as stockholders and managers, is to devise policies that improve the efficiency o f commercial banks. Unfortunately, economists do not agree upon the appropriate methodology for mea suring the efficiency of banks. Several I W H Y DO W E CARE AB O UT THE EFFICIENCY OF COMMERCIAL BANKS? The performance of firms is often described in terms o f their efficiency. The measured efficiency of a production unit (a firm or plant) is generically interpreted as the difference between its observed input and output levels and the corresponding optimal values. An output-oriented measure of efficiency compares observed output with the maximum output possible for given input levels. Alternatively, an input-oriented 39 REVIEW JULY/AUGUST efficiency measure compares the observed level of inputs with the minimum input that could produce the observed level of output. These are measures of technical efficiency, and as such ignore the behavioral goals of the firm. Measures of allocative efficiency com pare the observed mix of inputs or outputs with the optimal m ix that would minimize cost, maximize profit or obtain any other behavioral goal. Allocative efficiency can be combined with technical efficiency to measure overall efficiency. In addition, measures of technical efficiency can be used to construct measures of scale efficiency, which involve comparison of observed and optimal scale, or size, of the firm. One can also measure scope efficiency, which involves comparison of the cost of producing the observed m ix of outputs in a single firm with the costs that would prevail if each output was produced in a separate firm. Researchers have found that banks suffer more from technical inefficiency than from scale or scope inefficiency (for example, Berger and Humphrey, 1991). The efficiency of commercial banks is important for at least two reasons. First, efficiency measures are indicators of success, by which the performance of individual banks, and the industry as a whole, can be gauged. Banks face growing competition, both from other banks and from firms and markets out side the industry (see W heelock, 1993), and presumably banks will be more successful in maintaining their business if they operate efficiently. Berger and Humphrey (1992) find that during the 1980s high-cost banks experienced higher rates of failure than more efficient banks. Similarly, in a study of bank failures during the 1920s, W heelock and W ilson (1995) find that the less technically efficient a bank was, the greater its likeli hood of failure. A second reason to investigate the effi ciency of commercial banks is the potential impact of government policies on efficiency. One might gauge the impact of a regulatory change by measuring its effect on commer cial bank efficiency, or examine efficiency among banks in different states to measure the effect of differences in branching restric 1995 tions or other regulations. Recent proposals to end the Glass-Steagall separation of com mercial and investment banking stem in part from a view that broader powers could enhance the efficiency of banks and other financial institutions. Obviously, this change could enhance the scope efficiency of banks if there are complementarities in the produc tion of commercial and investment banking services. Conceivably, such change could also improve scale or overall efficiency. Improved efficiency is also one argument made in support of interstate branching and, indeed, Grabowski, Rangan and Rezvanian (1 9 9 3 ) find that branch banking organiza tions are more efficient than multiple-office bank holding companies. Other studies have considered whether bank mergers enhance efficiency. Using dif ferent approaches, Rhodes (1 9 9 3 ) finds that mergers have not generally improved effi ciency, though Fixler and Zieschang (1993) conclude the opposite. Shaffer (1 9 9 3 ), on the other hand, evaluates potential mergers and concludes that they could significantly reduce inefficiency for many banks of less than $10 billion of assets. The im pact of ownership or manage ment structure on efficiency has also been studied. Pi and Timme (1 9 9 3 ), for example, find that banks whose chief executive officer also serves as board chairman are less effi cient than other banks, and Mester (1993) shows that mutual savings and loan associa tions are more efficient than stock S&Ls. MEASURING COMMERCIAL BANK EFFICIENCY The efficiency of commercial banks has been studied using a variety of techniques and samples, and, as noted above, has been used to address numerous policy issues. Recent studies typically use techniques that accommodate the multiple outputs of banks and measure the efficiency of individual banks relative to a standard set by peer institutions. Readers interested in a survey of this research can refer to Berger, Hunter and Timme (1993). To date, no technique for measuring efficiency has been generally accepted and different methodologies appear to generate 40 JULY/AUGUST considerable differences in measured effi ciency, even when common bank samples are used. Variants of four techniques are common in the literature. The “stochastic cost frontier” approach is an econometric methodology in which deviations of a firm’s actual cost from predicted cost are presumed to be due to random error and inefficiency, each of which is assumed to have a particular statistical distribution (usually the normal distribution for the random error and a half normal for inefficiency). The “thick frontier” approach is a variant in which deviations from predicted cost within the lowest average cost quartile of banks are assumed due to random error, and the differences between the predicted costs of banks in the highest and lowest quartiles are assumed to be due to inefficiency. The “distribution-free” approach is applicable when data for more than one year are available. It assumes that inefficiency is stable over time, while random errors average out over time. That is, a bank’s inefficiency for a span of years is taken to be the mean of its measured inefficiency across all years within the period. Finally, “Data Envelopment Analysis” (DEA) is a non-parametric methodology in which linear pro gramming is used to measure the distance of individual banks from the efficient, or “best-practice,” frontier. All deviations from the efficient frontier are assumed to be due to inefficiency. Researchers have found that estimates of inefficiency are sensitive to the choice of technique. Ferrier and Lovell (1990), for example, apply the stochastic cost frontier and DEA techniques to a common sample of banks and arrive at different estimates of inefficiency. Berger (19 9 3 ) finds substantial differences in measured efficiency from two variants of the distribution-free approach. A second reason why different studies of commercial bank efficiency often reach seemingly contradictory findings might stem from differences in how a banking firm is modeled. Regardless of which of the four measurement techniques is used, the researcher must specify a list of inputs and outputs. The question, “W hat do banks produce?” is not simple to answer. Banks provide a variety of services, from loans 1995 and deposit accounts to trust services, safe deposit box rentals, mutual fund sales and foreign exchange transactions. Moreover, changes in regulation, technology and cus tomer demands have caused the types of services that banks perform to change over time. For example, banks now provide a variety of securities-related services, such as underwriting and mutual fund sales, which regulators forbid a few years ago. To tractably measure efficiency, researchers are forced to begin with simplified models of the banking firm. Unreliable estimates of efficiency can stem from the use of models that omit key features of bank production. Some researchers view banks as producers of loans and deposit accounts, and measure output by either the number of transactions or accounts serviced. This view is referred to as the “production” approach. Others argue that a bank’s output should be measured in terms of the dollar volume of loans or deposits it provides, a view known as the “intermedi ation” approach. Most studies of inefficiency use the intermediation approach, in part because the necessary data are more readily obtained. We are aware of only one recent study taking the production approach (Ferrier and Lovell, 1990), though in the 1970s and early 1980s such studies were more common (see Gilbert, 1984). The production approach focuses on operating costs and ignores interest expense. The intermediation approach, on the other hand, includes both operating and interest expenses, and hence may be of more interest for studying the viability of banks (see Berger, Hanweck and Humphrey, 1987; or Ferrier and Lovell, 1990). For analysis of the operating efficiency of banks, however, the production approach may be of interest. Among those who use the intermedia tion approach are researchers who hold the view that banks produce various loans and other investments from deposits, other fund ing sources, labor and materials. This “asset” approach has been criticized because it ignores the fact that banks expend consider able resources supplying transactions and savings deposits (Berger and Humphrey, 1992). Some researchers apply empirical criteria to determine what services to consider as bank outputs and what to consider as inputs. F E DE RA L RESERVE B A N K OF ST . L OU I S REVIEW JULY/AUGUST Berger and Humphrey (1 9 9 2 ), for example, classify activities for which banks create high added value, such as loans, demand deposits and time and savings deposits as important outputs, with labor, physical capital and pur chased funds classified as inputs. Alternatively, Aly, Grabowski, Pasurka and Rangan (1 9 9 0 ), Hancock (1991) and Fixler and Zieschang (1993) adopt a “user-cost” framework, whereby a bank asset is classified as an output if the financial return on the asset exceeds the opportunity cost of the investment, and a lia bility is classified as an output if the financial cost of the liability is less than its opportunity cost. Even though their details differ, the two approaches empirically tend to suggest simi lar classifications of inputs and outputs. The main exception is classification of demand deposits as an output in most user-cost stud ies, and as both an input and an output when the value-added approach is used (see Berger and Humphrey, 1992, for more detail). Table 1 summarizes six recent studies of commercial bank production efficiency. Although representative, this list is far from exhaustive. These studies employ a variety of estimation techniques and include a variety of different inputs and outputs in modeling the banking firm. The studies typically report inefficiency measures by bank-size grouping and for more than one type of inefficiency, though for brevity we report ju st the mean overall inefficiency. The reported percent ages indicate the extent to which the average bank overused inputs to produce a given level of output. Thus, the 35 percent ineffi ciency found by Aly and others (1 9 9 0 ) indi cates that the average bank could have pro duced the same level of output with ju st 65 percent of the input levels actually used. Measured inefficiency clearly varies with estimation technique, model specification and, probably, the sample of banks used by the researcher. In the remainder of this article, we investigate the extent to which measures of efficiency and the rankings of individual banks depend on whether the intermediation approach or production approach is employed. Because we are interested in the impact of the approach taken on measured efficiency, we use a single technique— DEA— applied 1995 to a common pool of banks. Our findings might, of course, be different if we used another technique or sample, but the purpose of this article is to investigate how sensitive efficiency measures are to the model of bank production employed. M ETHO D OLO G Y We trace our measures of efficiency to the work of Debreu (1951) and Farrell (1957). Boles (1966) was one of the first to use linear programming methods to measure efficiency in production using their ideas. Other exten sions have collectively come to be named Data Envelopment Analysis (DEA), a term coined by Chames, Cooper and Rhodes (1978). Lovell (1 993) summarizes this literature. Details about the efficiency measures used in this article are contained in the shad ed insert on page 6 . The essential ideas, however, are illustrated in Figure 1, which considers the case of a sample of firms pro ducing a single output from two inputs, x l and x 2. Suppose firms A, B and C each pro duce a given level of output; A and B lie on the production frontier XX ', while C lies in the interior of the production set. The fron tier XX' is the set of all combinations of inputs which can produce the same level of output, and where the reduction of at least one input necessarily causes output to fall. Hence, firms A and B are regarded as effi cient, whereas firm C is regarded as ineffi cient. Inefficient firms such as C may lie in the interior of the production set due to imperfect information, managerial incompe tence or perhaps other reasons. For firm C, input weak technical efficiency (iW E) is defined as the ratio of distances OC'/OC in Figure 1. By reducing the input quantities used by firm C by this amount, the firm could move to point C' and would be consid ered efficient in the IW E sense. Next, we define input overall efficiency (IO E). In terms of Figure 1, the isocost line is given by PP'. For firm C, the IOE score is given by the ratio of distances OC"/OC. Although the point C " lies outside the pro duction set boundary, and hence is not feasi ble, input costs at C " are the same as at B, which is a feasible point. Hence, if firm C F E D E R A L RESERVE B A N K OF ST. L OU I S 42 REVIEW JULY/AUGUST 1995 T a b le 1 Selected Studies of Com m ercial B a n k Production Inefficiency Study, Technique, Approach Inputs Outputs Sample Aly and others (1990); DEA; intermediation labor, physical capital, loanable fu nd s1 real estate loans, com random , 3 2 2 m ercial loans, consum er banks, 1 9 8 6 data loans, all other loans, d em and deposits overall inefficiency: 3 5 % Berger and Humphrey (1991); thick frontier; intermediation labor, physical capital, purchased fu nd s7 d em a nd deposits, all banks, 1 9 8 4 retail time and sa vin gs data deposits, real estate loans, com mercial loans, installm ent loans total inefficiency3: 2 4 % (branching states), 1 9 % (unit b a n kin g states) Elyasiani and Mehdian (1990); DEA intermediation labor, physical capital, d em a nd deposits, tim e and savings deposits securities held, real 191 b a n k s with estate loans, commercial assets over S 3 0 0 loans, all other loans million, 1 9 8 0 data technical inefficiency: 1 0 % Ferrier and Lovell (1990) DEA and stochastic cost frontier; production labor, occupancy costs, expenditure on material 5 7 5 banks, 1 9 8 4 num ber of: dem and data deposit accounts, time and sa v in g s deposit accounts, real estate loans, installm ent loans, com mercial loans overall inefficiency: 2 1 % (DEA), 2 6 % (stochastic cost frontier) Hunter and Timme (1995) distribution free; intermediation labor, physical capital/ purchased funds, transactions accounts/ non-transactions accounts under S I 0 0 ,0 0 0 4 com mercial and security loans, consum er loans, all other loans, n on interest income overall inefficiency: 3 0 - 5 4 % (depend in g on m odel); 2 3 3 6 % (om itting 1% extrem e values) Kaparakis and others (1994); stochastic cost frontier; intermediation labor, physical capital, interest bearing deposits under $ 1 0 0 ,0 0 0 , non-interest bearing deposits/ purchased funds loans to individuals, real 5 ,5 4 8 b a n k s with estate loans, commercial assets over S 5 0 loans, other5 million, 1 9 8 6 data 3 1 7 b a n ks with assets over $1 bil lion, 1 9 8 5 -9 0 data Results overall inefficiency: 1 0 % 1 time and savings deposits, notes and debentures and other borrowed funds. 2 federal funds purchased, time deposits over 51 0 0 ,0 0 0 , foreign deposits and other borrowed funds. 3 includes inefficiencies due to excessive deposit interest paid and purchased fund interest paid. 1 input treated as "quasi-fixed ," that is, not variable in the short run. 5 fed funds sold, securities held, securities and other assets in trading accounts. were to become efficient in the IOE sense, its input mix would have to be altered; the IOE score, however, can be obtained by considering the hypothetical proportionate reductions of inputs represented by point C". In terms of Figure 1, allocative efficiency for firm C is given by the ratio of distances OC'VOC'. Allocative inefficiency arises from using a combination of inputs that does not minimize total cost, as opposed to technical inefficiency, which is a proportionate overuse of all inputs. Finally, we can determine scale efficiency by comparing IW E computed under the assumption that the firm is operating at constant returns-to-scale with IW E obtained previously. A score of unity implies that the firm is operating under constant returns. W hile a score other than 1 does not translate 43 REVIEW JULY/AUGUST 1995 A MATHEMATICAL DESCRIPTION OF EFFICIENCY MEASUREMENT We use measures of efficiency discussed by Fare, Grosskopf and Lovell (1 9 8 5 ). First, we compute the input weak technical efficiency (IW E) score for the ith firm in a sample by solving the linear programming problem: where X, Y, x, and y, are defined as in equation 1, p( is a (1 X m) vector of input prices, and x * is an (m X 1) vector of effi cient inputs to be computed. The IOE score may be defined as (3) min Wj subject to (1) X qi < W ix i Oi = p ix ; / p ix i. The constraints in equation 2 are similar to those in 1. The same reference technology is defined by the constraints in 2 , but instead of proportionately reducing inputs until the ith firm lies on the refer ence technology inputs are further reduced proportionately until the firm lies on the isocost plane tangent to the production set boundary. An allocative efficiency score, Af, may be defined by dividing the IOE score by the IW E score: Yq i > y t Iq, = i q,e9l?, where n firms produce s outputs using m inputs, q t is a (N X 1) vector of weights to be computed for the ith firm, 0 < W, £ 1 is a scalar, X; is a (m X 1) vector of inputs for the feth firm, y { is a (s X 1) vector of out puts for the fcth firm, X = [x, ,...,x-V ] is a (m X N) matrix of observed inputs, Y = [y!,...,yN is a (s X N) matrix of observed ] outputs and I is a ( 1 X N) vector of ones. The minimand V , in equation 1 mea V sures the input weak efficiency of the ith firm. The inequality constraints in equa tion 1 define a reference technology with strong disposability of inputs; constraining the weights in q to sum to unity allows the reference technology to exhibit variable returns to scale. For the ith firm, W, gives the proportion by which inputs can be reduced to move the firm from the interior of the production set onto the piecewise-linear boundary of the production set corresponding to the reference technology in 1. Next, we compute input overall effi ciency (IOE) score O, for the ith firm by first solving the linear program: (4) Ai = O i / W t. The efficiency scores obtained from 1 measure technical efficiency as the dis tance to the relevant isoquant, but do not consider where the firm is situated along the variable-returns production frontier. To measure scale efficiency, equation 1 must be recomputed for each firm, first assuming constant returns to scale by removing the restriction C qt = 1, and then assuming non-increasing returns-to-scale by imposing the restriction Cq, < 1 . In the case of IW E, this produces efficiency scores W ™5 and W.V S, respectively, for ,R the ith firm. The scale efficiency score corresponding to 1 is then defined as (5) min ptx' S, as W,C 5/W,. R X* subject to Clearly, 0 < S( ^ 1 since WtC S£ WiN S < W,. R K If Sj= 1, then the ith firm is scale-efficient, that is, the firm is operating at the point of constant returns on the production frontier. If Sj < 1, then the firm is scale-inefficient due to either decreasing returns if W;M RS=W,, or increasing returns if WiN S< W,. K Xq t < x ‘ Yqi > y i (2) lq,=l qt e X x; e X FEDERA L RESERVE B A N K OF ST. L OU I S 44 REVIEW Y/AUGUST easily into a specific percentage deviation from constant returns, the scores are useful for ranking firms by the extent of their inefficiency. Each of the efficiency scores described above measures efficiency in an input orien tation; efficiency is measured by holding output fixed and determining the maximum feasible reduction in inputs. Efficiency can also be measured by holding inputs fixed and determining the maximum feasible expansion of outputs. Since the efficiency measures we use do not imply underlying assumptions regarding the behavior of firms, the choice between input and output orien tations is somewhat arbitrary; one might compute both types of efficiency measures to get more information than can be obtained from either the input or output orientations alone. Note that both IWE and IOE are radial measures of efficiency, that is, in each case efficiency is measured along a ray emanating from the origin and passing through the firm in input-output space. Consequently, the efficiency scores are independent of the units of measurement used for both inputs and outputs, which is advantageous since units of measurement may always be defined arbitrarily. Fare and others (1985) observe that some DEA formulations do not share this property. F ig u r e 1 M e a s u rin g Technical, A llo c a tiv e a n d O v e ra ll Efficiency age total assets at the end of 1993 for FCA program banks was $1 6 3 .6 million, with a range from $8 .0 million to $ 2 ,6 0 2 .8 million, average total assets were $ 3 0 0 .7 million, with a range from $1.0 m illion to $1 0 8 ,2 2 3 .0 million, for all U.S. commercial banks (as reported in the Federal Deposit Insurance Corporation Reports of Condition, that is, the “Call Reports”). The average return on assets of 1.15 percent for the banks in our sample, however, was approximately the same as the average for all banks ( 1.12 percent). Nevertheless, because our sample of banks is not random, the efficiency measures calculat ed here should not be interpreted as reflecting the efficiency of commercial banks in general. For the production approach to modeling bank activities, we construct variables using definitions from Ferrier and Lovell (1 9 9 0 ): EMPIRICAL IMPLEMENTATION For our empirical analysis of commercial bank efficiency, we use a sample of banks participating in the Federal Reserve System’s Functional Cost Analysis (FCA) program for 1993. Participants in this program supply information about their operations and costs which are not generally available for banks, and which are necessary to measure efficien cy using the production approach. After eliminating observations with missing values and observations for depository institutions other than commercial banks, data for 269 banks remain. Because participation in the FCA pro gram is voluntary, the banks in our sample may not be representative of the industry as a whole. For example, whereas the aver 1995 Outputs: y l = number y 2 = number y 3 = number y4 = number y 5 = number of demand deposit accounts of time deposit accounts of real estate loans of installment loans of commercial loans Inputs: X! = number of employees x2 = occupancy costs and expenditure on furniture and equipment x3 = expenditure on materials F EDERAL RESERVE B A N K OF ST. L OU I S 45 AUGUST 1995 T a b le 2 D escrip tive Statistics Variable Mean Standard Deviation Minimum Maximum y , 7 9 0 2 .9 1 1 3 0 9 1 .7 6 4 6 9 .0 0 1 7 3 3 6 2 .0 0 Yl 7 6 1 8 .4 3 1 0 6 8 0 .6 4 4 1 3 .0 0 1 0 6 8 2 1 .0 0 Yi 9 9 8 .2 3 1 5 0 3 .9 9 0.00 1 3 4 5 6 .0 0 Y * 3 1 3 4 .0 1 6 9 1 8 .6 3 4 4 .0 0 87 79 4.00 8 9 9 .6 3 1 7 4 4 .3 7 0.00 2 3 9 9 8 .0 0 4 0 7 6 4 3 1 6 .6 0 6 0 0 2 0 7 0 0 .0 9 1 3 8 7 9 6 3 .0 0 65 35 190 00.0 0 Y 9 9 2 7 2 7 7 5 .8 9 1 4 4 1 4 1 8 9 0 .7 9 5 1 0 3 0 0 0 .0 0 1 6 1 6 6 9 1 0 0 0 .0 0 Y 4 2 0 3 8 3 3 1 .6 1 5 4 5 1 2 3 7 6 .3 8 0.00 37 74 4 9 0 0 0 .0 0 y 1 6 2 5 4 8 3 2 .7 7 4 3 8 3 8 0 4 2 .2 4 1 2 7 8 9 0 .0 0 6 1 7 1 3 6 0 0 0 .0 0 Y s 2 8 5 5 3 7 7 5 .5 2 6 4 1 9 7 1 9 4 .1 7 0.00 8 9 5 4 7 1 0 0 0 .0 0 *1 88.02 1 4 4 .0 9 3.8 9 1 7 3 0 .0 7 x, 7 7 63 04.50 1 2 6 0 9 2 0 .3 7 7 9 0 .0 0 1 3 0 9 0 8 3 4 .0 0 h 3 3 0 1 0 6 .6 1 5 7 69 12.53 1 1 8 5 9 .0 0 6 9 4 8 5 5 2 .0 0 »i 30 82 7.13 5 9 0 5 .6 5 1 9 2 2 2 .3 8 6 7 83 2.63 Ys Y\ 0 .0 0 5 4 0.0001 0 .0 270 0 .0 0 2 3 *2 0 .0 0 2 5 0 .0 0 0 7 0.0010 0 .0 0 5 7 30 82 7.13 5 9 0 5 .6 5 1 9 2 2 2 .3 8 67 83 2.63 rz 5 2 .0 6 4 8 .4 3 0 .7 5 7 1 5 .1 7 r, 21.88 1 1 .69 2 .3 6 1 3 1 .8 6 1 7 0 4 5 .2 0 3 7 1 4 0 .7 1 1 6 6 .0 0 4 3 1 2 2 7 .0 0 5 4 4 1 7 .5 3 9 6 1 4 4 .1 6 2 1 3 .0 0 1 2 7 9 9 6 2 .0 0 1 6 7 5 5 .4 0 2 9 2 8 2 .6 0 9 3 .0 0 2 7 0 4 8 0 .0 0 5 9 8 0 .0 3 1 2 0 4 8 .5 1 100.00 1 3 7 3 0 0 .0 0 1 1 0 4 9 0 .9 8 1 5 8 7 6 5 .4 7 5 1 3 8 .0 0 1 8 9 4 4 7 7 .0 0 4 1 0 2 4 9 .0 0 w ; 1 8 3 2 3 .8 6 43 31 2.12 200.00 "3 89.32 150.08 4 .0 0 1 8 5 6 .0 0 « 4 3008.39 5 2 5 2 .8 5 1 3 .0 0 4 7 5 1 1 .0 0 0 .0 3 4 7 P2 0 .0 0 4 8 0 .0 143 0 .0 3 9 2 P i 0.0110 0 .0 0 6 6 0 .0 4 8 9 0.1000 P3 31.11 6 .1 8 2 0 .3 8 77.91 P4 0 .3 4 0.31 0 .0 8 2 .8 3 Input prices: Outputs: V! = loans to individuals for household, family, and other personal expenses v2 = real estate loans v3 = commercial and industrial loans v4 = federal funds sold, securities purchased under agreements to resell, plus total securities held in trading accounts Wi = total expenditure on salaries and fringe benefits/Xj w2 = x 2/level of deposits w , = x 3/level of deposits For the intermediation approach to bank production, we construct variables using definitions from Kaparakis, Miller and Noulas (1994): F E D E R A L RE S E RV E B A i : K OF ST. L O U I S 46 REVIEW JULY/AUGUST RESULTS OF EFFICIENCY MEASUREMENT Inputs: uj = interest-bearing deposits except certificates of deposit greater than $ 100,000 u2 = purchased funds (certificates of deposit greater than $ 100,0 0 0 , federal funds purchased, and securities sold plus demand notes) and other borrowed money u3 = number of employees u4 = premises and fixed assets We compute the various efficiency mea sures for the five models summarized below: Model Inputs Outputs Input Prices 1 x ,-x 3 Y\-Ys w-w, 2 X i-X j y~Ys w\- 3 x ,-x 3 Y’ -Ys HV - w 2 4 x ,-x 3 Y'-Yi w\- Wi 5 Input prices: p 1 = average interest cost per dollar of p 2 = average interest cost per dollar of u2 p3 = average annual wage per employee p4 = average cost of premises and fixed assets. u-1/4 A -ft Models 1 and 2 correspond to the Ferrier and Lovell (1 9 9 0 ) specification, with alterna tive price definitions. Models 3 and 4 provide a bridge to the intermediation approach by replacing the number of accounts and loans in the output variables with dollar amounts. Model 5 is the Kaparakis and others (1994) specification. Table 3 presents the mean scores for each type of efficiency described in the preceding section. Note that since the same inputs and outputs are used in models 1 and 2, and models 3 and 4, the technical and scale efficiency scores are the same for these models. For each efficiency measure, Table 3 also shows the standard deviation of the scores across the 269 banks in the sample, the number of banks having an efficiency score of unity (labeled “Number Efficient”), that is, the number of banks operating on the efficient frontier, as well as 90 and 95 percent confidence intervals for the mean. Given the large number of banks with efficiency scores of unity, particularly in the case of technical efficiency, and since all of the effi ciency scores are defined to lie between zero and 1, the underlying distributions of the individual efficiency scores are clearly non-normal. Results from Atkinson and W ilson (1 9 9 5 ), however, suggest that our sample size of 269 is easily large enough for us to rely on the asymptotic normality of the sample means implied by the central limit theorem, and thus to compute confidence intervals based on a normal distribution. In several cases, the confidence intervals for the means reported in Table 3 overlap. In addition, Kaparakis and others also define quasi-fixed input, non-interest bearing deposits, for which there is no corre sponding price. Other studies adopting the intermediation approach have ignored this item, as we do in the results reported below. Including non-interest bearing deposits as a fifth input when measuring technical or scale efficiency seems to have little effect on the results. To form a specification midway between the production approach represented by the Ferrier and Lovell (1990) specification and the intermediation approach represented by the Kaparakis and others (1994) specifica tion, we define y'v ...,y'5 as the dollar amount of each account or loan corresponding to y x,...,y5, respectively. Because outputs are now measured in dollar amounts, this model is best classified as representing the interm e diation approach, even though the choice of variables is based on Ferrier and Lovell (1 990). In addition, we define an alternative price system, Wj, w2, w', for the Ferrier and Lovell specification, where = w1; and w and W are computed similarly to w2 '2 3 and w3 except that level of deposits is replaced by the num ber of time and demand deposits. This seems to us to make the mapping of inputs and outputs under the production approach more consistent. We report sum mary statistics for each of the variables in Table 2. 1 9 9 5 F EDERAL RESERVE B A N K OF ST. L OU I S 47 REVIEW JULY/AUGUST 1995 T a b le 3 Efficiency Scores ( 2 6 9 o b se rva tio n s) M odel M e a n Stan d ard E rro r Technical e ffic ie n c y ( W k): Num ber E ffic ie n t 9 0 % C o n fid e n c e In t e r v a l 9 5 % C o n fid e n c e In t e r v a l 1,2 0.6348 0.0141 42 0.6115 0.6581 0.6071 3,4 0.7675 0.0103 52 0.7505 0.7846 0.7472 0.0107 75 0.7911 5 0.8088 0.8265 0.7877 0.6626 0.7879 0.8299 S c a le e ffic ie n c y ( S k): 1,2 0.8833 0.0067 32 0.8723 3,4 0.9452 0.0047 49 0.9374 5 0.9414 0.8943 0.9530 0.8701 0.8964 0.9359 0.9545 0.9527 7 0.9319 0.9509 0.9301 0.7562 0.7833 0.7536 0.4814 0.5170 0.4780 0.5204 7 0.7812 0.0057 0.8036 0.7790 0.8057 A llo c a t iv e e ffic ie n c y ( A k): 1 0.7698 0.0082 7 4 2 0.4992 0.0108 3 0.7924 0.0068 4 0.6340 0.7838 7 0.6162 0.0108 5 0.0080 13 0.7706 0.6518 0.7970 0.6128 0.7680 0.7860 0.6552 0.7996 O v e r a ll e ffic ie n c y ( 0 k): 1 0.4835 0.3356 7 0.4644 0.0116 4 0.5026 0.4607 0.5063 0.3607 0.3146 0.3567 0.3105 3 0.6053 0.0095 7 0.5897 0.6209 0.5867 4 0.4928 0.0121 7 0.4727 0.5128 0.4689 0.6144 0.6496 0.6110 2 5 0.6320 0.0128 0.0107 13 We test for significant differences among the means of each efficiency measure across different models. At the 0.05 significance level, we are unable to reject the null hypoth esis of equivalent means in the following cases: (1) scale efficiency for models 3, 4 and 5; (2) allocative efficiency for models 1 and 5, and models 3 and 5 (we do reject the null hypothesis when comparing allocative efficiency among models 1 and 3 ); and (3) overall efficiency for models 1 and 4. In all other instances, we reject the null hypothesis of no difference. Even the seemingly innocuous modification of redefining the input prices between models 1 and 2 , and between 3 and 4 has a large effect on mean allocative and overall efficiency. Note also that for the m ost extreme comparison, models 1 (the production view) and 5 0.6239 0.5166 0.6529 (the intermediation view), we reject the null hypothesis of equal levels of technical and overall efficiency. This suggests that, at least for this sample of banks, average efficiency does depend on the view of bank production assumed by the researcher. We find that average technical and overall efficiency is higher under the intermediation approach (model 5) than under the production approach (model 1). Our finding for overall inefficiency of 37 percent using model 5 is similar to what Aly and others (1 9 9 0 ) found for their sample, though substantially greater than what Kaparakis and others (1994) found for theirs (see Table 1). It is possible to determine whether a particular bank lies on the increasing (IRS), constant (CRS) or decreasing (DRS) returns portion of the technology. Table 4 shows 48 R E V I E W JULY/AUGUST the results of this analysis, considering only banks that were found to be technically effi cient.1 Thus, for example, 16 banks, or 38.1 percent of all technically efficient banks, operated on the constant-returns portion of the technology under models 1 and 2 . We test the null hypothesis of no associ ation among the rows and columns of the matrix represented by Table 4 using Pearson’s chi-square test, the likelihood ratio chi-square statistic, and Fisher’s exact test.2 For the entire matrix, all three tests reject the null hypothesis of no differences in the propor tions in each row and column. However, when we perform pairwise tests by deleting individual rows from Table 4, we fail to reject the null hypothesis of no difference for models 3, 4 and 5, and for models 1, 2 and 3, 4. Each of the three tests fail to reject at the 90 percent level. In the case of models 1, 2 and 5, we reject the null hypothesis of no difference in the proportions at greater than 99 percent. Thus, while we find evidence of similarity in terms of returns-to-scale when comparing models 3, 4 with either models 1, 2 or 5, models 1, 2 and 5 appear different in terms of returns-to-scale. More banks appear to be operating under constant returns-to-scale when the intermediation approach is taken (model 5) than when the production approach is used (model 1). Since returns-to-scale at a given location on the production frontier depend upon the shape of the variablereturns technology, these results indicate that the technology implicitly estimated by models 3, 4 is similar to the technologies implied by models 1, 2 and 5, which in turn are significantly different. This is consistent with our view of models 3, 4 as a bridge between the production approach represent ed by models 1, 2 and the intermediation approach represented by model 5. The result also suggests that differences between the two approaches might be due not only to use of number of accounts and loans versus dollar amounts, but also to the treatment of time deposits as an output or an input. In addition to comparing mean efficiency scores, we use the W ilcoxon matched-pairs signed-ranks test, a sign test for equality of medians and Kendall’s T-statistic to further 1995 T a b le 4 Returns to Scale Model 1,2 CRS DRS 16 24 (3 8 .1 % ) 3,4 (4 .8 % ) 20 (5 9 .6 % ) 1 (3 8 .5 % ) 48 (6 4 .0 % ) 2 (5 7 .1 % ) 31 5 IRS (1 .9 % ) 21 (2 8 .0 % ) 6 (8 .0 % ) examine the similarity of efficiency scores across the five models. We report the results of these tests in Table 5 .3 The W ilcoxon matched-pairs signedranks test analyzes the equality of distribu tions without making assumptions regarding the form the distributions might take. The values shown in parentheses in the second column of Table 5 give the two-tailed normal probabilities associated with the test statistic. Hence, we fail to reject the null hypothesis of identical distributions when comparing scale efficiency scores from models 3, 4 and 5, when comparing allocative efficiency scores from models 1 and 5, and from mod els 3 and 5, and when comparing overall effi ciency scores from models 1 and 4. In all other cases, we reject the null hypothesis. It appears that, for the most part, the distrib utions of the various efficiency scores do vary across models. The sign test for equivalence of medians yields a two-tailed binomial probability, which we also report in Table 5. In only two instances do we not reject the null hypothesis of equal medians: when comparing scale efficiency scores from models 3, 4 and 5, and when comparing allocative efficiency scores from models 3 and 5. These results are consistent with our finding that, in most cases, average efficiency varies across models. Finally, rather than comparing the distributions of efficiency scores from different models, we use the scores to rank banks in terms of their estimated efficiency Kendall’s X-statistic measures the correlation among the ranks of banks from two models and provides a statistical test of the null hypothe sis of no association between two sets of F EDERAL RESERVE B A N K OF ST. L OU I S 49 1 Since we ate using an input orientation, we could also examine whether inefficient banks would lie on the increasing-, constantor decreasing-returns portion of the technology if inputs were pro portionately contracted to move the bank to the frontier. However, since the path a bank might actually toke to reoch the frontier if the sources of inefficiency were removed depends upon behavioral goals, we ignore technically ineffi cient banks here. 1 Details on these computations moy be found in the Stala Reference Manual: Release 3 .1, Stato Corporation (199 3 ). 3 See Snedecor ond Cochran (1 9 8 9 ) and Kendall and Gibbons (1 9 9 0 ) for o discussion of these tests. Computational details are given in the Stata Reference Manual: Release 3.1, Stata Corporation (19 9 3 ). REVIEW JULY/AUGUST 1995 T a b le 5 T a b le 5 (c o n t.) Statistical C om parison of Efficiency Scores Statistical C om parison of Efficiency Scores (p ro b a b ility v a lu e s in parentheses) (p ro b a b ility va lu e s in parentheses) M odel W i lc o x o n S ig n e d - R a n k s S ig n T e st K e n d a ll's r W i lc o x o n S ig n e d - R a n k s M odel S ig n T e st K e n d a ll's r O v e r a ll e ffic ie n c y ( 0 t ): T ech nical e ffic ie n c y ( I V J : (2 ,2 )/(3 ,4 ) -8 .2 4 (0.0001) 0.0000 0.2513 (0.0000) (1,2) 12.31 (0.0001) 0.0000 0.5874 (0.0 0 0 0 ) (1 ,2 1 /5 -8 .9 0 (0.0001) 0.0000 0.0909 (0.0262) (1,3) -9 .0 8 (0.0001) 0.0000 0.2518 (0.0 0 0 0 ) (3 ,4 1 /5 - 3 .2 5 (0.0012) 0.0034 0.1081 (0.0082) (1,4) - 1 .0 6 (0.2903) 0.0327 0.3072 (0.0000) (1,5) -9 . 1 7 (0.0 0 0 1 ) 0.0000 0.1329 (0.0012) (2,3) -1 2 .7 4 (0.0001) 0.0000 0.1704 (0.0 0 0 0 ) (2,4) - 1 1 .6 2 (0.0001) 0.0000 0.5000 (0.0000) (2,5) -1 2 .3 6 (0.0001) 0.0000 0.0405 (0.3222) (3,4) 9.76 (0.0001) 0.0000 0.4128 (0.0000) (3,5) - 2 .6 7 (0 .0 0 7 6 ) 0.0015 0.1034 (0.0115) (4,5) -9 .0 1 (0.0001) 0.0000 0.1831 (0.0000) S c a le e ffic ie n c y ( S 4): (1 ,2 )/(3 ,4 ) -9 .4 6 (0.0001) 0.0000 0.2333 (0.0000) (1 ,2 )/ 5 - 7 .5 8 (0.0001) 0.0000 0.1114 (0.0065) !3 ,4 )/5 -0 .3 2 (0.7482) 0.4644 0.1805 (0.0000) A llo c a t iv e e ffic ie n c y ( 4 t ): (1,2) 12.88 (0.0001) 0.0000 0.0068 (0.8680) (1,3) - 3 .5 8 (0.0003) 0.0003 0.4179 (0.0000) (1,4) 8.40 (0.0001) 0.0000 -0.0964 (0.0185) (1,5) - 1 .4 5 (0.1481) 0.0327 0.0934 (0.0225) (2,3) -1 3 .3 9 (0.0001) 0.0000 0.0267 (0.5135) (2,4) -1 2 .2 8 (0.0001) 0.0000 0.5661 (0.0000) -1 3 .0 3 (0.0001) 0.0000 0.0342 (0.4036) 0.0000 -0.0047 (0.9092) (2,5) (3,4) 10.23 (0.0001) (3,5) 0.32 (0.7515) 0.6258 -1 0 .4 9 (0.0001) 0.0000 x 4N + 10 VAR(t ) = ---------------, 9 N (N —1) where N gives the number of observations. By definition, the statistic lies between -1 and + 1, taking a value of +1 if rankings are in complete agreement, or -1 if the ranks are completely reversed. Kendall and Gibbons (1 990) suggest that the T-statistic may also be viewed as a measure of concordance. Any two pairs of ranks (u;,V;) and (uj, Vj), i, j = 1,...,N, i # j, are defined as concordant if v, < v( when uf< uf or vi > vj when ul > u J. Similarly, they are defined as discordant if V;<Vj when u ,> u ( or vl > v j when Uj > U j . The total number of pairs is 0.0627 (0.1255) (4,5) rankings. The statistic is approximately normally distributed, with zero expected value and with variance 0.1648 (0.0001) F EDERAL RESERVE B A N K OF ST. L OUI S 50 HEVIEW JULY/AUGUST N (N -1)/2, and t can be shown to be equiva lent to the proportion of concordant pairs minus the proportion of discordant pairs. The last column of Table 5 gives the r-statistic for the various pairs of models for each measure of efficiency, along with signifi cance levels as shown in parentheses. We fail to reject the null hypothesis of no associ ation among ranks in only five instances when comparing allocative efficiency scores. (In particular, we fail to reject the null hypothesis for the following pairs of models: 1, 2; 2, 3; 2, 5; 3, 4; and 3, 5, and in only one case when comparing overall efficiency scores— for models 2 and 5). In all other cases, we reject the null hypothesis of no correlation. Note, however, that when we reject the null hypothesis of no association, the t-statistic is usually rather small in absolute terms; the largest value of the statistic shown in Table 5 is 0.5 8 7 4 (in the case of overall efficiency for models 1 and 2). As is typical in classical hypothesis testing, rejection of the null hypothesis does not necessarily imply acceptance of an alternative hypothe sis. That is, our statistical test may reject the hypothesis that the rankings are not associated, but that does not necessarily imply that the rankings are associated-the test is simply not that powerful. Figure 2 plots the rankings of overall efficiency scores for model 1 against those for model 5. Note that the value of Kendall’s T-statistic for this comparison is significantly different from zero at the 0.0012 level, indicating that the two sets of rankings are not discordant. The low value of the test statistic (0 .1 3 2 9 ), however, suggests that nei ther are they concordant. Our results based on the W ilcoxon matched-pairs signed-ranks test and the sign test for equivalence of medians are consistent with our observations on the differences of mean efficiency scores across the models discussed earlier. Taken together, our results indicate that different model specifications are likely to produce different measures of the level of inefficiency among a sample of banks, but not necessarily dissimi lar rankings of individual banks in terms of measured efficiency. For our data, the rankings are similar enough to reject the 1995 F ig u r e 2 R a n k in g s of O v e ra ll Efficiency Scores M odel 1 M odel S null hypothesis of no association, but in many cases are far from being in complete concordance. Concordance is relatively high for the technical and scale efficiency mea sures, which do not rely on price data. The introduction of price data to measure alloca tive and overall efficiency might also intro duce more sources of noise or error. CONCLUSION Like other studies of commercial bank efficiency, we find considerable inefficiency among banks in our sample. Other studies have found substantial variation in efficiency measures in applying different estimation techniques to a common pool o f banks. We find that measured efficiency also depends on the researcher’s conception of what banks do. In this article, we measure various types of production efficiency under two very differ ent views of banking. We find that, on aver age, technical and overall efficiency is higher under the intermediation view of the bank ing firm than under the production view. Mean allocative efficiency is, however, similar N K OF S T . L O U I S 51 RVW EIE JULY/AUGUST 1995 Fore, Rolf, Shaw na Grosskopf and C. A. Knox Lovell.The Measurement of Efficiency of Production. Kluwer-Nijoff Publishing, 1 9 8 5 . under the extreme versions of each approach. Under the intermediation view, we also find somewhat less scale inefficiency and more banks operating on the constant-returns por tion of the efficient frontier. Despite the dif ferences in mean measured efficiency across the different conceptions of how banks oper ate, however, we find some similarity in the rankings of efficiency scores of individual banks. Further research will, of course, be necessary to determine how sensitive these findings are to the particular dataset and esti mation techniques employed in this article. Farrell, M . J. "The Measurement of Productive Efficiency," Journal of the Royal Statistical Society, Series A (1 9 5 7 ), pp. 25 3 -8 1 . Ferrier, Gary D., and C. A. Knox Lovell. "M easuring Cost Efficiency in Banking: Econometric and Linear Programming Evidence," Journal of Econometrics (October/November 1 9 9 0 ), pp. 2 2 9 -4 5 . Fixler, Dennis J., and Kimberly D. Zieschang. "A n Index Number Approach to Measuring Bank Efficiency: An Application to M ergers," Journal of Banking and Finance (April 1 9 9 3 ), pp. 4 3 7 -5 0 . Gilbert, R. Alton. "B a n k Market Structure and Competition: A Survey," Journal of Money, Credit and Banking (November 1 9 8 4 ), pp. 61 7 -4 5 . Grabowski, Richard, Nanda Rangan and Rasoul Rezvanian. "Organizational Forms in Banking: An Empirical Investigation of Cost Efficiency," Journal of Bonking and Finance (April 1 9 9 3 ), pp. 531-38 . REFERENCES Aly, Hassan Y., Richard Grabowski, Carl Pasurka and Nanda Rangan. "Technical, Scale, and Allocative Efficiencies in U.S. Banking: An Empirical Investigation," Review of Economics and Statistics (M a y 1 9 9 0 ), pp. 2 1 1-18 . Hancock, Diana. A Theory of Production for the Financial Firm. Kluwer Academic Publishers, 1 9 9 1 . Atkinson, Scott, and Paul W. Wilson. "Comparing M ean Efficiency and Productivity Scores Across Small Samples: A Bootstrap Methodology," Journal of Productivity Analysis (forthcoming). Hunter, William C., and Stephen G. Timme. "Core Deposits and Physical Capital: A Reexamination of Bank Scale Economies and Efficiency with Quasi-Fixed Inputs,"Journal of Money, Credit and Banking (February 1 9 9 5 ), pp. 16 5-85 . Kaparakis, Emmanuel I., Stephen M . Miller and Athanasios G. Noulas. "Short-run Cost Inefficiency of Commercial Banks: A Flexible Stochastic Frontier Approach,"Journal of Money; Credit ond Banking (November 1 9 9 4 ), pp. 8 7 5 -9 3 . Berger, Allen N. "Distribution-Free Estimates of Efficiency in the U.S. Banking Industry and Tests of the Standard Distributional Assum ptions,"Journal of Productivity Analysis (1 9 9 3 ), pp. 2 6 1 -9 2 . _ _ _ _ _ _ _, G. A. Hanweck and D. B. Humphrey. "Competitive Viability in Banking: Scale, Scope, and Product M ix Economies," Journal of Monetary Economics ( 1 9 8 7 ), pp. 5 0 1-20 . Kendall, M., and J. D. Gibbons. Rank Correlation Methods. Oxford University Press, 19 9 0 . _ _ _ _ _ _ _and David B. Humphrey. "M easurem ent and Efficiency Issues in Commercial Banking," in Zvi Griliches, ed., Output Measurement in the Services Sector. University of Chicago Press, Lovell, C. A. Knox. "Production Frontiers and Productive Efficiency," in Hal Fried, C. A. Knox Lovell and Shelton S. Schmidt, eds., The Measurement of Productive Efficiency: Techniques and Applications. Oxford University Press, 1 9 9 3 , pp. 3-67. 1 9 9 2 , pp. 2 4 5 -7 9 . _ _ _ _ _ _ _o n d_ _ _ _ _ _ _ . "The Dominance of Inefficiencies Over _ Scale and Product M ix Economies in Banking," Journal of Monetary Economics (August 1 9 9 1 ), pp. 117-48. Mester, Loretta J. "Efficiency in the Savings and Loan Industry," Journal of Banking and Finance (April 1 9 9 3 ), pp. 2 6 7-86 . Pi, Lynn, and Stephen G. Timme. "Corporate Control and Bank Efficiency," Journal of Banking and Finance (April 1 9 9 3 ), pp. 515-30. _ _ _ _ _ _ _, William C. Hunter and Stephen G. Timme. "The Efficiency of Financial Institutions: A Review and Preview of Research, Past, Present and Future," Journal of Bonking and Finance (April 1 9 9 3 ), pp. 22 1 -4 9 . Rhoades, Stephen A. "Efficiency Effects of Horizontal (In-Market) Bank M ergers," Journal of Banking and Finance (April 1 9 9 3 ), pp. 4 1 1 -2 2 . Boles, J. "Efficiency Squared: Efficient Computations of Efficiency Indexes," Western Farm Economics Association Proceedings (1 9 6 6 ), pp. 137-42 . Shaffer, Sherrill. "C a n M egam ergers Improve Bank Efficiency?" Journal of Bonking and Finance (April 1 9 9 3 ), pp. 4 2 3 -3 6 . Snedecor, G. W., and W. G. Cochran. Statistical Methods. Iowa State University Press, 1 9 8 9 . Charnes, A., W. W. Cooper and E. Rhodes. "M easuring the Efficiency of Decision M aking Units,"European Journal of Operational Research (1 9 7 8 ), pp. 4 2 9-44 . Stata Reference Manual: Release 3 . 1. State Corporation, 19 9 3 . Debreu, Gerard. "The Coefficient of Resource Utilization," Econometrica ( 1 9 5 1 ), pp. 2 7 3-92 . Elyasiani, Elyas, and Seyed M. Mehdian. " A Nonparametric Approach to Measurem ent of Efficiency and Technological Change: The Case of Large U.S. Commercial Banks,"Journal of Financial Services Research (July 1 9 9 0 ), pp. 157-68 . 52 Wheelock, David C. " I s the Banking Industry in Decline? Recent Trends and Future Prospects from a Historical Perspective," this Review (September/October 1 9 9 3 ), pp. 3-22. _ _ _ _ _ _ _and Paul W. Wilson. "Explaining Bank Failures: Deposit Insurance, Regulation and Efficiency,"Review of Economics and Statistics (forthcoming). A ll nonproprietary and nonconfidential data and programs for the articles published in R e vie w are available to our readers and can be obtained from three sources: 1 . FRED (F e d e ra l R e se rve Econom ic Federal Reserve Bank of St. Louis, Post Office Box 442, St. Louis, MO 63166-0442. You can request data and programs on either disk or hard copy. Please include the author, title, issue date and page numbers with your request. 2 . Research D e p a rtm e n t, D a ta ), an electronic b u lle tin b o a rd You can access FRED by dialing (314) 621-1824 through a modem-equipped personal computer. For a free brochure on FRED, please call (314) 444-8809. service. 3 . In te r-u n iv e rs ity C onsortium lo r Political a n d Social Research (IC P S R ). Member institutions can request data through the CDNet Order facility. Nonmembers should write to: ICPSR, Institute for Social Research, PO. Box 1248 Ann Arbor, Michigan 48106, or call (313) 763-5010. Federal Reserve Bank of St. Louis Post Office Box 4 42 St. Louis, Missouri 6 3 1 6 6 -0 4 4 2 recycled paper