The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Federal Reserve Bank of St. Louis I S E P T E M B E R / OCTOBER 1 9 9 5 • V O L U M E 77, N UM BER 5 Performance Contracts for Central Bankers Capacity Utilization and Prices W ithin Industries Deflation and Real Economic Activity Under the Gold Standard P resident Th om a s C. M e lz e r D irecto r o f R esearch W illia m G . D e w a ld A sso cia te D irecto r o f R esearch Cletus C. C o u g h lin R esearch C o o r d in a to r and Review E d itor W illia m T. G a v in B a n k in g R. A lto n G ilb e rt D a v id C. W h e e lo c k In tern a tio n a l C h risto p h er J . N e e ly M icha el R. P a k k o P atricia S. P o lla rd M a cro econ om ics D o n a ld S. A lle n R ichard G . A n d e rso n Ja m e s B. B u lla rd M icha el J . D u e k e r Joseph A . Ritter D a n ie l L. Th o rn to n P eter S. Yoo R egion al M ich e lle C la rk N e e ly K e v in L. K liesen A d a m M . Z a re ts k y D irecto r o f E d ito r ia l S erv ices D a n ie l P. B rennan M an agin g E d ito r Charles B . H en derson G r a p h ic D esigners B ria n D . E bert Inocencio P. Boc R eview is p u blished six tim es per year by the Research D ep artm en t o f th e F e d e ral R eserve B a n k o f St. L ou is. S in g le -co p y s u b sc rip tio n s are availab le free o f ch arge. S en d re q u e sts to: F e d e ra l R e serv e B a n k o f S t. L o u is, P u b lic A ffa irs D e p a rtm e n t, P.O. B o x 4 4 2 , S t. L o u is, M issou ri 6 3 1 6 6 - 0 4 4 2 . T h e view s expressed are those of the individual auth ors and do n o t n e cessa rily re flect o fficia l p o sitio n s o f the Fed eral Reserve B an k o f St. L ou is, the Federal Reserve S y ste m , o r th e B oard o f G o v e rn o rs. A rtic le s m ay be reprinted or reproduced if the sou rce is credited. Please provide the Public Affairs D epartm ent w ith a copy o f the reprinted m aterials. results for specific industries provide insight about the robustness of the relationship between the capacity utilization rate and price changes. V olum e 7 7 , N u m b e r 5 27 D e fla tio n a n d Real Economic A c tiv ity U n d e r the G o ld S ta n d a rd 3 P erform an ce Contracts fo r C entral B ank ers C h ristop h e r J . W a lle r Discretionary monetary policy actions aimed at expanding a nation’s economy may simply produce an excessive rate of inflation without a corresponding increase in output or employment levels. Various proposals have been advanced to eliminate this inflationary bias, such as building a reputation for price stability and making the central bank independent of political pressure. Christopher J . Waller examines a new proposal: performance contracts, which would provide the proper finan cial incentives for the central banker to pursue price stability. But what is a performance contract and how does it work? More important, how difficult would it be to actually use one? 15 C apa city U tiliz a tio n an d Prices W ith in Industries P e te r S. Yoo The capacity utilization rate is a com monly used indicator of future price changes. Studies often find that the total industry capacity utilization rate and inflation are correlated. Peter S. Yoo examines capacity utilization rate data and price data for 23 industrial sectors to see if the two variables show signifi cant correlations within individual industries. He finds that the regression C h risto ph er I . N e e ly a n d G e o ffre y E. W o o d Targeting the price level, rather than the inflation rate, permits the future price level to be known and long-run plans to be made more easily. Despite these advan tages, countries have adopted inflation targets because price level targets require policymakers to reduce the price level to a pre-announced value after an infla tionary shock. Critics claim making such a commitment is undesirable, because deflation— a fall in the general price level— can have harmful effects. Christopher J . Neely and Geoffrey E. Wood examine the facts surrounding the temporary periods of deflation that occurred under the gold standard from f 8 7 0 t o l 9 1 3 . Although they caution against drawing conclusions from 100-year-old data generated under a much different monetary regime, they think that another look at this period is warranted, because much of the modern fear about falling prices is derived from the experiences of the era. REVIEW SEPTEMBER/ OCTOBER 1995 Christopher J. W a lle r is an associate professor o f econom ics at In d ia n a U n ive rsity and se rve d as o visiting scholar at the Federal R eserve B ank o f St. Louis w h e n this article w as b e gu n . Stephen M . Stohs p ro vid e d research assistance. Th e a u th o r tha n ks Jo e Ritter, Chris N eely, Bill G a vin , Carl W alsh a n d Steve Stohs fo r the ir com m ents on e a rlie r drafts. Performance Contracts for Central Bankers Christopher J. W aller ince the end of World W ar II, econom ies around the world have been plagued by historically high and persistent inflation. This raises a question: If inflation is socially undesirable, w hy do policym ak ers produce it? O ne explanation is that discretionary m onetary policy may lead to an inflationary bias. T his explanation is based on the “tim e-inconsistency” prob lem , first outlined by Kydland and Prescott (1 9 7 7 ) and illum inated by Barro and G ordon (1 9 8 3 a ). T he typical version o f this explanation assum es that society wants the m onetary authority to follow a low inflation policy, w hich it prom ises to do. O nce private agents com m it them selves to nom inal wage contracts based on a low expected inflation rate, however, the m onetary authority is assum ed to have an incentive to create “surprise” inflation and inflate away the real value o f the contract ed nom inal wage. As a result, firm s hire m ore labor and produce m ore output. But, because private agents are aware o f this incentive, they do n o t believe that the central bank w ill carry through w ith its prom ise to m aintain inflation at a low level. H ence, w orkers set their nom inal wages high enough so that the extra infla tion created by the central bank leaves real wages at their desired levels. Consequently, no additional output or em ploym ent is created bu t society suffers from an inflation bias. F o r the past decade, researchers have investigated an array o f methods with w hich to reduce this inflation bias. A lthough m ost m ethods prom ise to lower S the inflation bias, they usually do so at the co st o f creating greater output variability. However, a recen t proposal by W alsh (1 9 9 5 a ) and Persson and Tabellini (1 9 9 3 )— the adoption o f perform ance con tracts for central bankers— has created a stir am ong econom ists w orking in this area. T h e purpose o f this article is to sur vey the w ork on perform ance contracts and com pare it to earlier proposals for m it igating the inflation bias. T he rem ainder o f the paper proceeds as follows: T he sec ond section contains a m odel describing the basic tim e-inconsistency problem and reviews previous suggestions for elim inat ing the inflationary bias. Follow ing that is a discussion o f the nature o f perform ance contracts and how they w ork. T h e fourth section probes the principal-agent nature o f central banking and its relationship to central bank independence. In the final section, I offer concluding com m ents. THE TIME-INCONSISTENCY PROBLEM A Model of Discretionary Monetary Policy A general description o f how monetary policy is determined would go something like this: Society (the principal) delegates the power to create money to the central bank (the agent). Society instructs the central bank to use its money creation powers to “do good.” W hat is meant by doing good is often not well-defined; nevertheless, it can be interpreted to mean that the central bank should produce a policy that improves the well-being o f society. The central bank then enacts policy according to some objective function. Presumably, its objective is to m axim ize social welfare, but it could also be to m axim ize something other than society’s welfare. Finally, after policy is enacted, the monetary authority may be asked to account for its actions. RfVIIW SEPTEMBER/OCTOBER To illustrate the nature o f the tim eincon sistency problem , consider the follow ing version o f the Barro and G ordon model: (1 ) it now for com parison later. E qu ation 3 looks very m uch like society’s u tility fu n c tio n except that the central b an k is allow ed to have a potentially different ou t put target, _y" + k M, than society’s. If k M= k\ then the cen tral ban k’s objective is id enti cal to society’s. If k M+ k\ then the central ban k uses policy to pursue an agenda that is different than that o f society as a w hole. T h e reason the central b an k has a different agenda is im portant and is a crucial part of the perform an ce-con tract debate, as dis cussed later in this article. Finally, for ease o f analysis, the m onetary authority is assum ed to con trol the inflation rate directly and thus chooses tt to m axim ize equation 3 given equation 1. Consider the case in w hich the central b an k has only society’s interests at heart, that is, k M= k 5 = k. Since society w ants inflation to be zero (o n average), suppose the central bank can pre-com m it to a policy w hereby it w ill n ot create system atic infla tion. T his im plies that expected inflation is zero. Substituting equation 1 into 3 and m axim izing su b ject to the con strain t k M= ks = 0 yields w hat is called the so cia lly op tim al or “p re-com m itm en t” solution for inflation and output: y = y " + ( K - K c) + u (2 ) U s = - ( y - y ' ' - k s ) 2- h jr (3 ) UM= w - ( y - y n- k M) 2- b K 2 , 1 9 95 where y is real output, y n is the trend level of output, 7r is the inflation rate, ttc is the expected inflation rate and u is a m ean zero, serially uncorrelated real output shock. Equation 1 describes how output is influenced by inflation and inflation expectations. W orkers are assum ed to sign nom inal wage contracts prior to the setting o f m onetary policy and the contracted wage is based on the expected rate o f infla tion. An inflation surprise reduces the real value o f the contracted nom inal wage, thereby inducing firms to hire m ore labor and produce m ore output. E quation 2 is society’s utility function and shows that society suffers from output and inflation fluctuations about their tar geted levels. Society’s target output level is y" + k s, w here y" is the natural or trend level o f output and k s is a positive co n stant. T he param eter k s is assum ed to reflect society’s b elief that distortions in the econom y m ake trend output undesir ably low. Society’s preferred inflation rate is assum ed to be zero. T h e param eter b m easures the relative w eight society places on losses arising from inflation. The w eight on losses arising from output has been set equal to 1 for n otational ease. Equation 3 is assumed to be the central ban ker’s objective function. T he param e ter w is the salary or budget the central banker receives for doing the jo b . This term is irrelevant in the standard Barro and G ordon m odel and is usually ignored. But this term plays a key role in the perform ance-con tract literature, so I w ill include From equation 4 , the central ban ker par tially offsets the output sh o ck by allow ing inflation to vary m ore. E xpected inflation is zero, and expected output is y n. In this w orld, p re-com m itm ent refers to the idea that the central bank can com m it itself to m aking the inflation rate zero on av erag e, b u t w ill vary the period-by-period inflation rate to stabilize output in a way that m axi m izes social welfare. T he central bank m akes no attem pt to expand output above the trend level even though it has a desire to do so. In short, even though k > 0, pre-com m itm ent m eans the central ban k is F E D E R A L RESERVE B A N K OF ST . L O U I S 4 able to credibly prom ise to act as if k = 0. Now suppose that the central bank cannot com m it itself to acting as if k = 0. Now the central banker chooses v , taking 7t‘ as given, to m axim ize its objective fu n c tion. M axim izing equation 3 yields the follow ing expression: ( 6) K= be in the pre-com m itm ent case. Thus, even though the central banker does w hat society w ants him to do, the use o f discre tionary policy m akes society w orse off in equilibrium . There are three points to note about equations 8 and 9. F irst, the inflation bias is a constant— it is n ot a random variable n or does it vary over time. Second, the bias does n o t depend on the output shock. Third, the stabilization response to the output sh o ck u is the same in both the socially optim al solu tion and the discre tionary solution. These features all com e into play w hen discussing the optim al design o f perform ance contracts. (K e+ k M- u ) Rational expectations im plies that ttc m ust be set consistent w ith equation 6. T his im plies that (7) k^_ n RESOLVING THE TIM E INCONSISTENCY PROBLEM b w hich yields the follow ing solutions for the d iscretion ary equilibrium : (8) (9) n - kM 1 b 1+ b y=y + Since the pu blication o f the Barro and G ordon (1 9 8 3 a ) paper, research has focused on ways o f elim inating this infla tionary bias. There have been two distinct d irections o f research: the reputationbuilding approach and the institutionaldesign approach. ■u . 1+ b Reputation Building T he reputation-building approach focuses on the use o f “p u n ish m en t” strate gies by private agents to deter the central ban k from generating the inflation bias. In these m odels, w orkers believe the central ban k w ill follow a low inflation policy as long as it has n o t tried to surprise w orkers in the past. O therw ise, they “p u n ish ” the central ban k by expecting a high inflation rate, w hich the central ban k validates to avoid creating a recession. By using this type o f m echanism , the private sector is able to persuade the central banker to develop a reputation for enacting the announced policy. Barro and G ordon (1 9 8 3 b ) show ed that reputation building would generate a low er inflation bias but would n ot elim inate it. Barro and G ordon’s early m odel o f rep utation was done under the assum ption of perfect inform ation. Subsequent research exam ined how robust the reputation- T he only difference betw een these expres sions and those from the pre-com m itm ent solution is that there is now an inflationary bias, given by kM/b > 0; output is the same. W hy does the inflation bias arise? Because the target level o f output is higher than the trend value. O nce wage contracts are signed, the central bank can increase output above trend by creating an inflation surprise. T he central bank does this not out o f self-interest bu t because society w ants it to. Even though society as a w hole desires this, however, individual agents have no incentive to allow their wages to be inflated away. C onsequently they set expectations and nom inal wage demands accordingly In equilibrium , the econom y suffers from excessive inflation w ith no additional gains in output. It can be show n that the loss from the discre tionary equilibrium is higher than it would 5 BfVIEN SEPTEMBER/OCTOBER building approach was to inform ation im perfections. C anzoneri (1 9 8 5 ) showed that the econom y w ould suffer inflation “cy cles” due to occasional breakdow ns in credibility if private agents were unable to separate exogenous inflation shocks from system atic policy actions. Backus and Driffill (1 9 8 5 ), Barro (1 9 8 6 ) and Rogoff (1 9 8 7 ) show ed that if private agents are unsure o f the central banker’s type— infla tion haw k or dove— then a recession w ill frequently o ccu r early in a central ban ker’s term. T his is because private agents’ expectations o f inflation are an average of the haw k’s and the dove’s equilibrium inflation rates. If the central banker is a hawk, inflation is set low er than expected and a recession occurs. If the central banker is a dove, he may act like a haw k and create a recession to build a reputation as a hawk. T he reason is that if the dove inflates immediately, he reveals h im self as a dove and inflation expectations w ill be higher for the rem ainder o f his term in office. By acting like a haw k, he manages to keep inflation expectations low. The dove, however, eventually chooses to cre ate an inflation surprise and expand ou t put for a short period o f time. Thu s, w hile inflation is low er on average, output and inflation are m ore variable. Although reputation m odels are able to generate low er equilibrium inflation rates, albeit at som e co st o f greater output variability, they have several unappealing aspects. F irst, there are an infinite num ber of p unishm ent strategies that could be adopted, and it is not obvious w hich is the correct one to use. F o r exam ple, how long should the punishm ent last? Second, the m ultiplicity o f strategies suggests that private agents w ould have to coordinate their actions to send a clear sig nal to the central bank as to how they would behave in the event that they are surprised. B ut how is su ch coordination to be achieved? Large, national trade u nions may be sufficient for coordinating actions in som e countries, bu t this is n o t a feasible solution in the relatively atom istic labor m arkets that characterize the U.S. economy. 1995 Third, the reputation approach tends to focus on the personality and reputation o f individual central bankers. Because individuals do n o t serve as the central banker for long periods in the real w orld, this approach suggests that there w ill be considerable uncertainty and variability o f policy as central bankers turn over. Thu s, we should focus on ways o f developing the institutional reputation o f the cen tral bank instead o f the reputation o f individual cen tral bankers. Finally, the reliance on the private sec tor to enforce the appropriate path o f m o n etary policy is a b it unpleasant from a pub lic policy perspective. Th e rep u tation building approach does n o t try to change the central ban k’s objective fun ction directly; rather, it alters the central ban k’s behavior by m aking the policy choice dynam ic, that is, by m aking today’s policy actions have future consequences. But if the in stitutional structure o f the central ban k provides it w ith the wrong policy incentives, then it w ould seem prudent to change the in stitu tion rather than rely on private agents to solve the problem . To illustrate this point, consider the response to airline h ijackings. O ne way o f dealing w ith h ijackers is to arm the pas sengers and let them enforce peace on the airplane. T his is akin to w hat the reputa tion approach does for the inflation bias. A better idea is to change the environm ent for boarding a plane so that the likelihood o f a h ijackin g is reduced— h ence, the use o f m etal detectors. As a result o f these problem s w ith the reputation -based approach, researchers began to investigate in stitutional reform s for the central b an k that would m itigate the inflationary bias. Institutional Design T h e in stitu tio n al-d esign approach focuses on using legislative m eans to restrain the central bank from engaging in high-inflation policies. T h e in ten t is to m anipulate the central ban k’s objective function directly through legislative action. Som e w ork in this area has 6 BIVIIW Septem ber / o c t o b ir i 99 5 punishm ent). Consequently, the central ban ker’s self-interest plays a large bu t hidden role in these types o f m odels. focused on legislation that restricts the d ay -to -d ay operating procedures o f the central bank; other research show s how the appointm ent process for central bankers can be used to elicit better infla tion perform ance. Advocates o f the latter line o f research recom m end m aking the central bank independent from elected leaders as a m eans o f reducing the infla tionary bias. Conservative Central Bankers The appointm ent and reappointm ent o f a central banker w ho sets policy accord ing to his own self-interest plays a large role in other institutional schem es for dealing w ith the inflation bias. Thom pson (1 9 8 1 ) and Rogoff (1 9 8 5 ) proposed appointing a “conservative” central banker who dislikes inflation m ore than everyone else in society. A conservative central banker generates a low er inflationary bias b u t does so by n ot stabilizing the econom y in a socially optim al fashion.2 To illustrate this point, suppose that society appoints a central ban ker w ho puts m ore w eight on inflation than it does. T h e central banker w ould then have a larger value o f the para m eter b in equation 3 to use in setting pol icy. From equations 8 and 9 , however, we see that a larger value o f b reduces the inflation bias bu t m akes output m ore variable. F or the conservative central banker’s policies to be credible, society m ust believe that he cannot be rem oved ex post by the current governm ent. Thus, the central banker m ust have som e degree of independence to pursue policies that are n ot desired by the current adm inistration (and, implicitly, the electorate). Subsequent research by Flood and Izard (1 9 8 9 ) and Lohm ann (1 9 9 2 ) show ed that com plete independence was n ot socially optim al— for certain bad states o f the w orld, society benefits from firing the conservative central banker and stabilizing output. Targeting Regimes Legislative restrictions on the central bank often take the form o f im pos ing m onetary targeting or adopting simple rules (w hich are actually targeting regimes w ith a horizon o f one period). T he adop tion o f Friedm anesque k-percent rules has been studied by Alesina (1 9 8 8 ) and Lohm ann (1 9 9 2 ). They show that these rules elim inate n ot only the inflationary bias, bu t also stabilization o f output by the m onetary authority. Hence, there is a trade-off betw een reducing inflation and stabilizing output. Sim ple rules dom inate d iscretion w hen output shocks are small and relatively rare.1 M ulti-period targeting horizons have been exam ined by C anzoneri (1 9 8 5 ) and G arfinkel and O h (1 9 9 3 ). In these m odels, the central ban k m ust follow policies so that the targeted inflation rate occurs on a v era g e over som e time interval. In this environm ent, the central bank creates an inflation bias early in the targeting horizon, but it is sm aller than it would have been in the absence o f target ing. However, it produces sub-optim ally low inflation (or even deflation) at the end o f the targeting horizon to h it the targeted inflation or m oney growth rate. Stabilization is also sacrificed in the nam e o f inflation, since shocks early in the period are n ot stabilized in an optim al fashion because those actions m ust be reversed later in the targeting period. An im plicit assum ption in these tar geting m odels is that the central banker’s w orst penalty for m issing the target is dism issal (shooting him is not a realistic A NEW INSTITUTIONAL DESIGN: PERFORMANCE CONTRACTS A con sisten t them e o f b o th the reputa tion-building and in stitu tio n al-d esign m odels is that the inflation bias can be reduced or elim inated, b u t usually at the cost o f having the central ban k reduce its F E DE RA L RESERVE B A N K OF ST . L O U I S 7 1 Recently, Haubrich and Ritter (1 9 9 5 ) have argued that this comparison between simple rules and discretion is biased in favor of rules, because it assumes that the choice between adopting a simple rule over discretion is a one-time decision. In fact, m onetary authority has the option of waiting before com m itting to a k-percent rule, and this option has value that is typically ignored in the Alesina and Lohm ann analyses. Thus, they argue that discretion is more likely to be preferred than is typically shown. 2 Faust (1 9 9 4 ) has shown that the appointm ent of o central banker w ho prefers a lower trend inflation rate than the m edian voter con im prove social welfare if the m ajority of voters are net nom inal debt holders. Stabilization issues, however, are not studied in Faust's m odel. REVIEW SEPTEMBER/OCTOBER em phasis on stabilizing output. Thus, there appears to be a trade-off betw een reducing average inflation and stabilizing the real econom y.3 Debate has centered on the relative benefits and costs o f this trade off in determ ining the goals o f m onetary policy, and the types o f legislative restraints to place on the central bank. Recently, however, a new idea has sur faced in the institutional-design literature for dealing w ith the inflation bias. The idea is to offer the central banker a perfor m ance contract, w hereby the central banker’s salary or the ban k’s budget is tied directly to the perform ance o f im portant m acroeconom ic variables such as GDP and the inflation rate. By giving the central banker the proper financial incentives, these researchers have show n that the cen tral bank can be induced to generate low inflation w ithout forsaking its stabilization responsibilities. those incentives do n ot appear, at first glance, to be con sisten t w ith m axim izing society’s w ell-being. T he problem is deter m ining w hat those incentives should be. Follow ing the principal-agent litera ture, W alsh proposed offering the central bank a perform ance contract. T his co n tract ties the central banker’s personal com pensation or the size o f the ban k’s budget to the perform ance o f the economy. O nce the con tract is signed, society encourages the central banker to pursue his ow n self-interest and adopt policies that increase his incom e or the ban k’s bud get. T he trick is to structure the con tract in such a way that by trying to increase his own resources, the central banker m axi mizes social welfare in the process. This approach is a radically different way to deal w ith policym akers. Under this in stitutional design, society exploits the pursuit o f self-interest by the central banker to achieve the socially desirable outcom e. T his differs from the traditional view o f appointing a benevolent central banker and then instru cting him to do good. Under the perform ance con tract approach, society essentially says: “You can do w hat you w ant, but you will pay personally for undesirable ou tcom es.” M aking the central bank accountable for its actions is a prom inent them e o f perfor m ance contracts. Performance Contracts 3 Empirical evidence on this point is m ixed. For exam ple, some researchers have shown that greater central bank indepen dence is associated with lower average inflation rates but has no relationship with the vari ance of GDP. Other w ork has shown that countries with inde pendent central banks tend to suffer greater output losses dur ing disinflations, which sug gests that there is a trade-off between reducing inflation and stabilizing output variability. 4 Persson and Tabellini (1 9 9 3 ), working from an early draft of W alsh's paper, extended his approach to a m ore general fram ework. 1995 W alsh (1 9 9 5 a ) suggested that the m on etary policy game be viewed as a prin cipalagent p ro b lem .4 In a prin cip al-agen t m odel, one individual or group (th e prin cipal) delegates control over a policy vari able to another individual or group (the agent). Although the principal w ould like the agent to set policy so that the p rin ci pal’s welfare is m axim ized, the agent has a different objective and opts for a policy that does not give the principal its m ost desired outcom e. The solution to this problem is for the principal to offer the agent a contract that gives the agent the incentives to enact the policy desired by the principal. By viewing m onetary policy as a prin cipal-agent m odel, W alsh redirected atten tion to the source o f the problem — the central banker is confronted w ith a set of preferences that do n ot yield the outcom e that society prefers m ost. So rather than w orry about appointing conservative cen tral bankers or adopting appropriate repu tation strategies, W alsh argued that we should provide the central banker w ith the incentives to “do the right thin g”— even if Designing a Performance Contract W h at does a perform ance con tract look like? C onsider the follow ing co m pensation con tract for setting the central banker’s salary (w in equation 3 ): (1 0 ) w = s —X n , w here s denotes the central banker’s base salary or the budget o f the central bank. This contract specifies that the central banker be paid a base salary s, w hich will be reduced if any inflation occurs. The degree o f salary reduction is determ ined by the param eter X. A key feature o f this co n tract is that it is based solely on the pub licly observed inflation rate; it is n ot based F E D E R A L RESERVE B A N K OF ST . L O U I S 8 SEPTEMBER/OCTOBER 1 9 9 5 on item s that are unverifiable (such as how hard the central banker w orks). O nce the co n tract is in place, society tells the central banker to set policy in any m anner he sees fit; there is no m ention o f pursuing the public good. Therefore, given equations 1, 3 and 10, the central banker chooses 7t to m axim ize (11) Furtherm ore, output and inflation are stabilized in the socially optim al fashion. The reason this can be accom plished is that the inflation bias is con stan t and in d e pendent o f the output sh o ck u. So a sim ple linear penalty for inflation is sufficient to deter the central ban k from inflating. But the key p oin t is that elim inating the inflation bias through appropriate in cen tives does n ot require the central banker’s stabilization response to be distorted. Therefore, there is no cost for elim inating the inflation bias. By careful construction o f the central banker’s com pensation, so ci ety is able to elim inate the inflation bias and have output optim ally stabilized. This is indeed a pleasant result. T h e co n tract could take a variety o f different form s and still generate the opti mal outcom e. Every con tract, however, m ust have the feature that the central bank pays m ore attention to inflation (o r less attention to output) than society does. T his sim ply reflects R ogoff’s (1 9 8 5 ) notion o f a conservative central banker. T he only difference is that in R ogoff’s fram ew ork, society carefully selects a central banker w ho has the “righ t” personal attributes to reduce inflation, w hereas the con tract approach gives any arbitrarily chosen cen tral banker the appropriate incentives to produce low inflation. In general, the principle o f R ogoff’s idea is still relevant; the issue is how to define “conservative.” R ogoff’s definition o f a conservative central banker was som eone w ho put more w eight on inflation relative to stable out put. But w e could define a conservative central banker as som eone w ho has a low er inflation rate target or low er output target than the rest o f society. In all cases, the central ban ker cares relatively m ore about inflation than output. F or exam ple, consider the follow ing perform ance contract: UM= s - A x - ( y - y " - k M) 2- b 7 r 2. T his yields the follow ing expression for the inflation rate: (12) n= V1 + b y [feM+ 7T‘ - A - u ] . Im posing rational expectations yields the follow ing equilibrium solutions for infla tion and output: (1 3 ) k M- A 1 n —---------------------u b 1+ b (1 4 ) y=y + 1 + b G iven these expressions for w hat inflation and output w ill be w hen the central banker pursues his own self-interest, society would like to set the w eight A such that the expressions in 13 and 14 are exactly the same as those given by the pre com m itm ent solutions in equations 4 and 5. T h is result can be accom plished by setting: (1 5 ) A = k M. By setting A = kM, the reduction in salary from creating an inflation surprise ju s t off sets any benefits that would accrue from expanding output towards y" + k M. H ence, on the margin, the loss o f incom e for the central banker is ju s t equal to the utility gain from creating surprise inflation and expanding output, so he chooses n o t to create surprise inflation and no inflation bias occurs. (1 6 ) ■2kM( y - y " ) + ( k M) 2 In this exam ple, society sim ply offers the central ban ker a con tract that penalizes him if output is above the natural rate, plus adds a fixed am ount to the base salary F E D E R A L RESERVE B A N K OF ST. L O U I S 9 according to the magnitude o f k M. Substituting 16 into 3 and rearranging yields (1 7 ) com e. Because the central banker’s targeted inflation rate is less than society’s preferred rate, the central banker appears m ore con servative than the rest o f society; in contrast to Rogoff’s model, however, this type of conservative central banker does not cause stabilization to be sub-optim al. T he key p oin t o f this discussion is that offering the central ban ker a perform ance co n tract m ay be equivalent to appointing an appropriately defined conservative cen tral banker. O nce we realize this, there is no reason to believe that these central bankers w ill understabilize the economy. JJM = s - ( y - y " ) 2- b n 2 . T he con tract in 16 leads to an o b jec tive fun ction for the central ban ker that is equivalent to appointing a central banker w ith a low er output target than the rest o f so ciety since the param eter feMdisappears. W ith this contract the central banker will use discretion to produce the socially opti mal outcom e. Alternatively, Svensson (1 9 9 5 ) proposes a contract o f the form: (1 8 ) IS TIME INCONSISTENCY THE SAME AS A PRINCIPALA G ENT PROBLEM? : s + 2 b n c n - b (jz c ) 2, In the perform ance co n tract approach above, it was show n that appropriately chosen contracts can ind u ce the central banker to produce the socially optim al outcom e. T h is result was dem onstrated w ithout any reliance on the assum ption that the central ban ker’s output target was equal to society’s. W alsh conducts his analysis under the assum ption that society and the central banker have the same objective functions, that is, k M= k s. T his assum ption is com m on in the tim e-incon sistency literature, bu t is n o t con sisten t w ith the principal-agent m odel. U sually in a principal-agent problem , the agent has a different objective than the principal. A more classical d epiction o f the principalagent problem would lo o k like the follow ing utility functions: w here it ' is an arbitrary constant to be determ ined by society. Substituting 18 into 3 and rearranging yields (1 9 ) UM= s - ( y - y n- k M) 2- b ( x - 7 T ‘ ) 2. If k M= k s, this contract looks very m uch like society’s utility fu nction except that the central banker’s target inflation rate is now different from zero. Thu s, the co n tract in 18 is observationally equivalent to appointing a central banker w ith a differ ent inflation target than the rest o f so ci ety’s. A central banker w ith this con tract w ill set policy such that, in equilibrium , inflation and output are given by (2 0 ) ( 21 ) kM 1 n = — + n c --------- u b l+b l+ b ( 22) u . (2 3 ) Notice that in setting ttc = - k M/b, we obtain the socially optimal solution. Thus, by having the central banker target a desired inflation rate o f minus the inflation bias, society obtains its m ost preferred out- U s = —(y - y " ) 2— bn1 UM= w - ( y - y " - k M) 2-b7 T 2 W ith this form ulation, society has prefer ences that are consistent w ith the socially optim al solu tion given in 4 and 5. T he central bank, on the other hand, wants 10 IEMEW SEPTEMBER/OCTOBER 1 9 9 5 output to be higher than its trend value (for som e unspecified reason). Thus, the central banker uses his discretionary pow ers to create an inflation surprise, thereby expanding output. Rational agents foresee this and adjust wages so that they are not fooled. T he outcom e is an inflation bias w ith no additional output gains. Although the story is the same as the tim e-inconsistency m odel described above, there is one fundam ental difference: Society does not w ant the central bank to try to expand output above trend. The central bank does so in pursuit o f its own self-interest. This situation is w hat perfor m ance contracts were designed for: en tic ing a “m isbehaving” agent to produce the principal’s desired policy. But if the perform ance con tract gener ates the socially optim al outcom e regard less o f w hether society and the central banker have the same output targets, why is it im portant to classify the problem as a tim e-inconsistency problem rather than a principal-agent problem ? The reason is that if the policy game is described as the principal-agent problem as in equations 22 and 23 above, the credibility o f con tract enforcem ent is not an issue. The principal very clearly wants the socially optim al pol icy to be im plem ented and has every incentive to hold the central banker to the contract and not renegotiate it. But in the case in w hich the central banker is trying to give society w hat it w ants, society is in consistent— it wants higher output, w hich can only be achieved by being “fooled;” yet, society does n o t w ant to be fooled. If the central banker is m axim izing social welfare, then society should renege on the perform ance contract once private agents set their wages— it should let itself be fooled. Since it is optim al ex post to renege on the perform ance contract, then private agents w ill never believe it changes the central banker’s incentives, and we are right back where we started. The credibility o f contract enforce m ent raises an im portant point: Tim einconsistency and principal-agent relation ships are not the same thing, even though perform ance contracts appear to solve b oth types o f problem s. Thus, one needs to be careful in using solution concepts interchangeably. Enforceability o f the perform ance co n tract corresponds to M cC allum ’s (1 9 9 5 ) second fallacy o f central bank indepen dence. M cC allum argues that a perfor m ance con tract “does n o t actually over com e the m otivation for dynam ic in con sis tency; it m erely relocates it” (p. 2 1 0 ). As long as the central banker is presum ed to be m axim izing social welfare, this argu m ent is correct. But if the inflation bias is actually the result o f a “tru e” principalagent problem rather than a tim e-inconsistency problem , society can pre-com m it itself to enforcing the contract. Actually, M cC allum ’s criticism of perform ance contracts is too strong. W hile it is correct to say that a perfect com m itm ent technology or institutional design does n o t exist (for exam ple, even the U.S. C onstitution is n ot a perfect co m m itm ent to liberty because we can change it anytim e we w ant), it is possible to make the costs of reneging on prom ises more costly and thus m ake m onetary policy more credible. T he basic idea o f perfor m ance contracts, and the prem ise behind the entire institutional-design literature, is to increase the cost o f reneging on a cooperative arrangem ent. Som e in stitu tions have low reneging costs (a policy target), w hile others have very high reneg ing costs (abolishing the F ed ). By relocat ing the source o f dynam ic consistency, perform ance contracts attem pt to increase the costs o f reneging on low inflation prom ises. A TALE OF TW O PRINCIPALA G EN T PROBLEMS In equations 22 and 2 3 , the central banker has different objectives than so ci ety as a w hole in that he w ants to increase output above the current trend value. This m athem atical form corresponds to the tra ditional principal-agent problem . But why would the central bank have an objective F E D E R A L RESERVE B A N K OF ST. L OU I S 11 R EVIEW SEPTEMBER/OCTOBER electorate (why elected leaders do n o t sim ply take control o f m onetary policy then is som ew hat puzzling). According to this scenario, central bank independence is an undesirable in sti tutional structure. T h e perform ance co n tract approach can w ork only if the elected leaders have control over the central bank through the setting o f budgets and salaries, and the ability to dism iss the central banker over policy actions. F or exam ple, W alsh (forthcom ing) show s that if ad just ing the ban k’s budget and salaries is infea sible, then threatening to dism iss the cen tral banker if certain poor policy outcom es arise can replicate the equilibria supported by perform ance contracts. W alsh refers to these optim ally designed threats as “dis m issal co n tracts,” since the central banker know s exactly w hich conditions w ill lead to his dism issal and agrees to such an arrangem ent. The im plications for central bank independence in this setting are very dif ferent from w hat is generally thought to be. C entral bank independence is general ly believed to be a crucial elem ent o f good inflation perform ance, and the em pirical evidence to date is con sisten t w ith that view (see Alesina and Sum m ers, 1 9 9 3 ). Because o f this theoretical and em pirical evidence, legislation has been introduced around the w orld that aim s at increasing the independence of central banks. W hy do the im plications for central bank independence forthcom ing from the principal-agent story described above dif fer so m uch from w hat is actually happen ing in the world? A likely explanation is that this principal-agent story is n ot the correct view. that differs from w hat society wants? T he answ er to this question lies in the policy structure o f m ost dem ocracies. The general public elects a leader who either conducts policy h im self or delegates the control o f policy to som eone else. M onetary policy typically falls in the dele gation category. In the U nited States, for exam ple, voters elect the President and m em bers o f Congress w ho, in turn, dele gate the control o f m onetary policy to the Federal Reserve. Although they delegate control o f m onetary policy, the President and the Senate jo in tly determ ine who shall serve as the head o f the Federal Reserve. T hu s, there are typically three actors in any m onetary policy m odel: the voters, the elected leaders and the central banker. In the tim e-inconsistency m odel, all of these actors are assum ed to have the same objective. From a principal-agent perspec tive, however, the presum ption is that they have differing objectives. The "Rogue" Central Banker 5 Actually, this would im ply that the central banker has a differ ent inflation rate target than society's rather than a different output target. 1995 Consider the follow ing principal-agent problem. The voters and elected leaders have the same policy objective, given by equation 22, while the central banker has the objective function given in 23. In this case, the central banker is a “rogue” policy m aker who sets policy to m axim ize his . self-interest rather than society’s or the elected leaders’ and w ho, by doing so, creates an inflation bias. W hy would the central bank behave this way? C entral bankers may w ant to m axim ize their am enities such as the num ber o f staff m em bers, the lu xuriance of buildings and the size of travel budgets, all o f w hich are funded by excessive seignior age creation.5 Or if the central bank is unduly influenced by a special interest group, say the banking/financial sector, it may pursue policies that benefit these sec tors rather than society. Regardless o f the source o f the problem , perform ance co n tracts are a desirable way o f dealing w ith it. Society and the elected leaders use a perform ance contract to rein in the central banker and m ake him accountable to the Elected Leaders As Monetary Authority C onsider an alternative principalagent problem proposed by Fratianni, von Hagen and W aller (1 9 9 5 ). Suppose that voters face an agency problem w ith elected leaders. Voters w ant leaders to carry out policies con sisten t w ith their objective fun ction in equation 2 2 , bu t leaders may 12 REVIEW SEPTEMBER/OCTOBER 1 99 5 have incentives to m isuse m onetary policy for political reasons. F or exam ple, elected leaders may follow policies that benefit special interest groups or that further their short-run re-election chances. If unusually high levels of output increase an in cu m b en t’s chances o f being re-elected, he may try to create surprise inflation to expand output above trend. Furtherm ore, signifi cant partisanship in the policy process may lead to a redistribution o f resources that does n ot prom ote the public good. These are all reasons the elected leaders may have an objective function sim ilar to equation 23, if they controlled m onetary policy directly. If elected leaders have an incentive to m isuse m onetary policy, it is in society’s interest to delegate policy to a non-political agent who will enact the policies desired by the general public. T his agent would have society’s objective function as his own. T he problem is: How is this non-political agent chosen? E lections w ill n ot w ork since getting re-elected may be why policy is m isused in the first place. Th e central banker needs to be appointed, but this is typically done by the elected leaders.6 Thus, elected leaders can use appointm ent or the threat o f non-reappointm ent to pressure the central bank into im plem ent ing policies aimed at helping the incum bent leaders. If the central ban k’s budget or the central bankers’ salaries are under legislative control, then the central bankers can be pressured through bud getary cuts to pursue sub-optim al policies. In this fram ew ork, the central bank would like to do the right thing bu t its im m ediate principal— the elected lead ers— have objectives that differ from the general public. The elected leaders, not the central bank, need to be made accountable. Accordingly, society benefits by m aking the central bank as free o f p olit ical interference as possible, since inflation will be reduced and output w ill be stabi lized optimally. Thus, central bank inde pendence is crucial for good m onetary pol icy; w ithout it, the central bank is merely a veil for political leaders. Anything that makes the central bankers appointm ent and budget less susceptible to political pressure will lead to better monetary policy.7 T his view of the principal-agent nature o f m onetary policy has led academ ic econom ists to support the m ovem ent toward greater central bank independence. W h at would be the purpose o f central bank perform ance contracts in this latter version o f the principal-agent problem ? If the elected leaders are the ones who write and enforce the central ban k’s perform ance contract, then they probably will not solve the problem . Clearly, enforcem ent of the contracts would lack credibility since elected leaders have an incentive to forgive any transgressions the central bank m akes (as long as the transgressions benefit the elected leaders). There is one potential benefit of using performance contracts in this environment. Performance contracts make policy more vis ible and the goals o f the monetary authority more transparent. Presumably, this visibility would lead to better policy actions, since deviations from the socially optimal path would have to be explained publicly at speci fied intervals of time. Individuals who employ political pressure on the central bank would be brought into the public limelight and the personal costs to elected leaders from this attention, we hope, would deter them from putting pressure on the central bank. Furthermore, although it is a blunt instrument, the ballot box may provide enough credibility in the enforcement o f the contract such that better macroeconom ic performance would be achieved. CONCLUSIONS A lthough theoretically appealing, per form ance contracts may n ot be feasible in practice. In fact, political infeasibility may w ell be the reason we do n ot observe this type o f institutional arrangem ent in the real world. N evertheless, the perform ance con tract research we see today could well turn out to be the foundation for the design o f central banks in the 21 st century. But w e’ll need to try a few experim ents first to see how well they w ork in practice. New Zealand’s recent reform s o f its central F EDERAL RESERVE B A N K 13 OF ST. L OU I S 6 The interested reader should see W aller (1 9 9 2 ,1 9 9 5 ) for an exam ple of such an appoint m ent process. 1 W aller (1 9 9 2 ,1 9 9 5 ), W aller and Walsh (1 9 9 5 ) and Alesina and Gatfi (1 9 9 5 ) show how reducing the degree of political influence in the appointm ent process can lead to superior m acroeconom ic outcom es. REVIE W SEPTEMBER/OCTOBER 1995 Hutchison, M ichael. "Central Bank Credibility and Disinflation in New Zealand," Federal Reserve Bank of San Francisco Weekly Letter (February 1 9 9 5 ). bank structure seem to be very sim ilar to a perform ance contract and may well be the test case we need. Evidence to date is sparse, but the reform s appear to have played a role in reducing inflation and inflation expectations.8 Future designs o f central bank in stitu tions w ill probably reflect a com bination of independence and perform ance contracts. The result would be highly autonom ous central banks that are clearly held accou n t able to the electorate. W h at more could we ask for? Kydland, Finn, and Edward Prescott. "Rules Rather Than Discretion: The Inconsistency of Optim al Plans," Journal of Political Economy (1 9 7 7 ), pp. 4 7 3 -9 2 . Lohm ann, Susanne. "O ptim al Com m itm ent in M onetary Policy: Credibility Versus Flexibility," The American Economic Review (1 9 9 2 ), pp. 2 7 3 -8 6 . M cCallum , Bennett. "Tw o Fallacies Concerning Central Bank Independence," The American Economic Review Papers and Proceedings (1 9 9 5 ), pp. 2 0 7 -1 1 . Persson, Torsten, and Guido Tabellini. "Designing Institutions for M onetary Stability," Carnegie-Rochester Conference Series on Public Policy (1 9 9 3 ), pp. 5 5 -8 3 . REFERENCES Alesina, Alberto. "Alternative M onetary Regimes: A Review Essay," Journal of Monetary Economics (1 9 8 8 ), pp. 1 7 5 -8 4 . Rogoff, Kenneth. "Reputational Constraints on M onetary Policy," Carnegie-Rochester Conference Series on Public Policy (1 9 8 7 ), pp. 1 4 1 -8 2 . _ _ _ _ _ _ _ and Roberta Gatli. "Independent Central Banks: Low Inflation at No C o st?" American Economic Review Papers and Proceedings (1 9 9 5 ), pp. 1 9 6 -2 0 0 . _ _ _ _ _ _ _ "Th e Optim al Degree of Com m itm ent to an Interm ediate Target," Quarterly Journal of Economics (1 9 8 5 ), pp. 1 1 6 9 -9 0 . _ _ _ _ _ _ _ and Lawrence Sum m ers. "Central Bank Independence and M acroeconom ic Performance: Som e Com parative Evidence," Journal of Money, Credit, and Banking (1 9 9 3 ), pp. 1 5 1 -6 2 . Spiegel, M ark. "Rules vs. Discretion in N ew Zealand M onetary Policy," Federal Reserve Bank of San Francisco Weekly Letter ( M arch 1 9 9 5 ). Backus, David, and John Driffill. "Inflation and Reputation," The American Economic Review (1 9 8 5 ), pp. 5 3 0 -3 8 . Svensson, Lars. "Inflation Targeting Need Not Increase Output Variability: Com m ent on Rogoff and W a lsh ," w orking paper (1 9 9 5 ), Institute for International Econom ic Studies, Stockholm University. Barro, Robert J. "Reputation in a M odel of M onetary Policy with Incom plete Inform ation," Journal of Monetary Economics (1 9 8 6 ), pp. 1 -2 0 . Thom pson, Earl. "W h o Should Control the M oney Supply," The American Economic Review Papers and Proceedings (1 9 8 1 ), pp. 3 5 6 -6 1 . _ _ _ _ _ _ _ and David B. Gordon. "A Positive Theory of M onetary Policy in a Natural Rate M o d e l," Journal of Political Economy (1 9 8 3 a ), pp. 5 8 9 -6 1 0 . Waller, Christopher J. "Appointing the M edian Voter to a Policy Board," working paper (1 9 9 5 ). _ _ _ _ _ _ _ "A Bargaining M odel of Partisan Appointm ents to the Central B on k," Journal of Monetary Economics (1 9 9 2 ). _ _ _ _ _ _ _ a n d _ _ _ _ _ _ _ "Rules, Discretion and Reputation in a Model of Monetary Policy," Journal of Monetary Economics (1 9 8 3 b ), pp. 1 0 1 -2 2 . _ _ _ _ _ _ _ and Carl E. W alsh. "Central Bonk Independence, Economic Behavior, and Optim al Term Lengths," working paper (1 9 9 5 ). Canzoneri, M atthew . "M onetary Policy Gam es and the Role of Private Inform ation," The American Economic Review (1 9 8 5 ), pp. 1 0 5 6 -7 0 . Faust, Jon. "W h o m Con W e Trust to Run the Fed: Theoretical Support for the Founders' V iew s," working paper (1 9 9 4 ). W alsb, Carl E. "Is N ew Zealand's Reserve Bank Act of 1 9 8 9 an Optim al Central Bank Contract?" Journal of Money, Credit, and Banking (forthcom ing). Fratianni, M ichele, Jurgen von Hagen and Christopher J. W aller, "Central Banking as a Political Principal-Agent P roblem ," Economic Inquiry (forthcom ing). _ _ _ _ _ _ _ "O ptim al Contracts for Central B ankers," The American Economic Review (1 9 9 5 ), pp. 1 5 0 -6 7 . _ _ _ _ _ "W he n Should Central Bankers Be Fired?" w orking paper (Septem ber 1 9 9 4 ), University of California at Santa Cruz. Flood, Robert, and Peter Izard. "M onetary Policy Strategies," International Monetary Fund Staff Papers (1 9 8 9 ), pp. 6 1 2 -3 2 . Garfinkel, Michelle, and Seonghw an O h. "Strategic Discipline in M onetary Policy with Private Inform ation: Optim al Targeting H orizons," The American Economic Review (1 9 9 3 ), pp. 9 9 -1 1 7 . 8 See Hutchison (1 9 9 5 ) and Spiegel (1 9 9 5 ). Haubrich, Joseph, and Joseph Ritter. "D ynam ic Com m itm ent and Im perfect Policy Rules," Federal Reserve Bank of St. Louis Working Paper No. 95-015 (N ovem ber 1 9 9 5 ). NK OF ST . L O U I S 14 Peter S. Yoo is an econom ist ot the Federal R eserve B an k o f St. Louis. R ichard D . Ta ylo r p ro vid e d research assistance. Capacity Utilization and Prices W ithin Industries producer price index (PPI) finished goodsbased measures of inflation. Since inflation is an aggregate phenomenon, their focus is undoubtedly justified. Yet, the economic analysis that links inflation to capacity uti lization should apply to any product market, regardless of its size. Therefore, the relation ship between price and capacity use should also be evident in industry level data— per haps more so. In this paper, I use two-digit standard industrial classification (SIC) industry mea sures of capacity utilization to explore the robustness of the relationship between capacity utilization and prices. The results suggest that such measures do not have a consistently strong and simple relationship with each industry’s price data. Peter S. Yoo he strength of the econom ic expansion during the past two years has renewed fears of accelerating inflation. As these fears have grown, people have turned to various statistics to substantiate any signs of rising inflation. Commodity prices, wages, sales-to-inventory ratios, civilian unemployment rates and capacity utilization rates are some of the statistics commonly used to predict the future path of inflation. These measures embody the basic idea of supply and demand: As the demand for scarce goods increases, their prices must also increase. The staff of the Board of Governors of the Federal Reserve System measures capacity utilization as the ratio of industrial production to industrial capacity.1 Since the denominator in this ratio normalizes industrial production by a measure of the potential industrial output of the economy, the ratio provides a cyclical measure of industrial output. The Board’s measure of capacity assumes that a firm’s or an indus try’s production capacity is fixed over some moderate time horizon, usually due to the quantity of the available plant and equipment stock. W hen firms attempt to produce beyond their “normal” levels, the cost of producing the additional output becomes increasingly expensive if the firm’s production process exhibits diminishing retums-to-scale. The higher cost then translates into higher prices. Most of the empirical researchers on this subject use total industrial capacity utiliza tion and the consumer price index (CPI) or T THE RELATIONSHIP BETWEEN PRICE A N D CAPACITY UTILIZATION Econom ists typically have used two frameworks to estimate the relationship between prices and the strength of economic activity. First is the supply curve, a relation ship between prices and quantities. Shea (1 993) finds that the supply curve of several four-digit SIC industries is upward sloping: Any increase in demand is met by a combina tion of additional output and higher prices. Over some moderate time frame in which firms have finite and fixed capacity, any increased production then implies higher rates of capacity utilization, which creates a positive relationship between price changes and capacity utilization. The second and more common frame work is a forecasting relationship between capacity and inflation. Such studies include Garner (1 9 9 4 ), McElhattan (1 9 7 8 , 1985) and Finn (1 9 9 5 ). G am er and Finn estimate simple linear equations in which the current rate of inflation is a function of previous periods’ inflation and total industrial capacity utilization rates. McElhattan assumes there is a boundary point of total industrial capacity N K OF ST . L O U I S 15 1 See Federal Reserve Measures of Capacity and Utilization (1 9 7 8 ) end Shopiro (1 9 9 2 ) for discussions obout the construction of the series. REVIE W SEPTEMBER/OCTOBER ^ I do not use the term inflotion when referring to industry data because inflation is an increase in the overall price level, while an increase in an industry's price level is not. utilization, beyond which inflation increases or decreases, a concept analogous to the non accelerating inflation rate of unemployment. Therefore, she regresses changes in inflation on previous changes in inflation and on lagged capacity utilization rates. All three of these studies find a statistically significant relationship between total industrial capacity utilization rates and inflation. The accompanying figures show the relationship between price changes and capacity utilization for 23 two-digit indus tries and three aggregate groups: total, mining and manufacturing industries.2 The price changes in the figures are monthly per centage changes in each industry’s net output price level without their seasonal components. (I used regressions with 12 monthly dummy variables to remove the seasonal component from each industry’s monthly percentage price changes.) The finished goods producer price index is the price index associated with total industrial capacity utilization rates. The sample covers the period of 1987-94. The figures yield mixed signals about the relationship between capacity utilization and prices. Total industrial and manufacturing capacity utilization rates seem to track price changes from late 1990 to early 1993, but otherwise show no obvious relationship. The mining aggregate shows volatile price changes, but with little connection to changes in capacity utilization. The 23 two-digit industries show similar ambiguity. Some industries, such as paper and fabricated metals, show an extremely close relationship between capacity use and percentage price changes. The figures for these two industries indicate that capacity utilization rates and price changes moved in tandem from 1987 to 1994. Other industries, such as the leather industry, show no discernible relationships between capacity constraints and price changes. Still others, like stone, clay and glass products, show signs of positive co movements for a portion of the sample period but not for the entire sample period. 1995 rates to forecast price changes within the context of a simple linear relationship. Current price changes are functions of past price changes and capacity utilization rates in forecasting equations: 77, = / ( 7Tm , C U J , where v t is the monthly percentage change in an industry’s net output price level, the TTt i's are lagged price changes, and the cu, ’s are that industry’s current and lagged capacity utilization rate. Unlike in Shea’s study, esti mates of the above relationship cannot be interpreted as supply curves, because capacity utilization and price changes are equilibrium values determined by the intersection of the demand and supply schedules. This causes an identification problem because it is impos sible to determine whether prices increased because the demand schedule shifted out or because the supply schedule shifted in. Still, many people estimate such relationships and use capacity utilization rates as sufficient indicators of future price changes. Indeed, the media and other popular sources of busi ness news usually promote the idea that high current rates of capacity utilization indicate imminent price pressures. Most macroeconomic data series have persistence, that is, current and past values are significantly related. Therefore, a regres sion that attempts to estimate the relation ship between capacity utilization and price changes should include lagged values of price changes to account for their persistence rather than attributing it all to movements in capacity utilization. Including past price changes then allows one to estimate the marginal information contained in capacity utilization about current and future price changes. Unfortunately, determining the number of lags to include in a regression is a problem. Including too many lags can reduce the pre cision of the estimated coefficients or yield spurious significant correlations, whereas using too few lags will not capture all of the persistence in the data. The Schwarz infor mation criterion provides a way to capture the amount of persistence in price changes. It weighs the gains in explanatory power against the number of additional variables REGRESSIONS I now turn to linear regressions to examine the ability of capacity utilization F EDERAL RESERVE B A N K 16 OF ST. LOUIS REVIE W SEPTEMBER/OCTOBER included in the regression, analogous to an adjusted R2 measure. I use this criterion because Geweke and Meese (1981) found that it outperformed most others in the consistency of lag-length selection. I therefore estimate a linear equation in which current price changes are functions of: previous price changes; capacity utiliza tion rates using monthly percentage price changes; and capacity utilization rates that have had their seasonal components removed.3 The sample starts in 1986 and extends through 1994. To determine the number of lags of price changes and capacity utilization for each regression, 1 use the Schwarz informa tion criterion, allowing up to 24 lags of both price changes and capacity utilization rates. Table 1 shows the results of the search, in which an entry of zero indicates that only contemporaneous capacity utilization rates are included. Table 1 shows that most two-digit industry price changes have a simple relationship with lagged price changes and capacity utilization rates. Eleven of the 23 industries appear to be well-described by their previous month’s price change and contemporaneous capacity uti lization. Among those industries with more complex relationships, only two industries— lumber and electrical machinery— show any link between additional lags of capacity uti lization and current price changes. Moreover, none of the industries shows a noticeable relationship between current price changes and either lagged price changes or capacity changes are positively and significantly related (at the 5 percent level) to previous price changes, with a percentage-point increase in the previous month’s price change associated with 0.38 and 0.47 percentage-point increases in current prices, respectively. The same two groups also show positive and statistically significant relationships with contemporane ous capacity utilization. The estimates indicate that a percentage-point increase in capacity utilization is associated with a 0.04 percentagepoint increase in prices in the current period and ju st over a 0.06 percentage-point increase in the long run. W hile the effect is significant and has the correct sign, the size is an order of magnitude smaller than that of lagged price changes. The regression results for the two-digit industries also reveal a strong relationship between current and previous price changes. Seventeen of 23 regressions show statistically significant relationships between current and lagged price changes, with 16 of the 17 industries statistically significant at the 5 percent level and coal mining significant at 10 percent. Most of the statistically signif icant relationships between current and lagged price changes indicate a positive and sizable correlation. On average, a 1.0 percentage-point increase in the previous month’s price change is associated with a 0 .30 percentage-point increase in current prices. The coefficients of the previous peri od’s price change vary from -0 .4 0 to 0.52, and the cumulative sums for multiple lags of price changes range from 0 .10 to 0.79. The relationship between current price changes and capacity utilization, however, is not as clear. Among the forecast equations for two-digit industries that include only contemporaneous capacity utilization, seven— furniture and fixtures, paper products, printing and publishing, rubber and plastic products, primary metals, fabricated metals and mis cellaneous manufacturing— indicate statisti cally significant and positive coefficients at the 5 percent level, with one— textile mill products— at the 10 percent level. Together, these eight industries produce 26.5 percent of industrial output. The magnitudes of the coefficients are not very large, ranging from 0.01 to 0.02, noticeably smaller than the u tilizatio n b eyon d th ree m o n th s. Given the results in Table 1 , 1 estimate the simple forecasting equations for the 23 two-digit industries and three aggregated groups (mining, manufacturing and total industrial). Each industry’s equation includes the number of lags indicated by Table 1. In addition, I calculate the sum of the coeffi cients of the capacity utilization variables to measure the cumulative relationship between capacity utilization and price changes.4 Table 2 shows the regression results from estimating the above equation over the sample period of January 1986 through December 1994, with t-ratios in parentheses.5 Two of the three aggregate groups, total industrial and manufacturing, indicate that current price 1995 N K OF ST. 17 LOUIS 3 The Board of Governors does not release capacity utilization in a seasonolly unadjusted form. It does, however, release industrial produc tion seasonally unadjusted. Because the published capacity measure does not have o seasonal component, I define seasonally unadjusted capacity utilization as seasonally unadjusted industrial production divided by capacity. This measure allows me to filter the sea sonality of price changes and capac ity utilization rates in a similar man ner, so any distortions introduced by the filter will be minimized. 4 1 did not consider first-differencing the dato because none of the price change series indicate a unit root and, moreover, it seems unlikely that prices are 1(2) processes. 5 1 use Newey-W est robust standard errors when calculating the t-ratios to correct any remaining serial cor relation of the residuals and heferoskedasticity. REVIEW SEPTEMBER/OCTOBER 6 Geweke and Meese (1 9 8 1 ) found (hot although (tie Schwarz criterion was consistent in its estimation of lag-length selection, it can underes timate the lag length. They found the degree of underestimation to be very small, however. 1995 of forecasting inflation based on the relation ship between capacity constraints and prices is appealing, the evidence from two-digit industry data is weak. The simple forecasting results reported in this article have not iden tified strong, consistent relationships between prices and capacity constraints. Second, even among the industries with a statistically significant relationship, the size of the rela tionship is small. These results suggest that current price changes are the best indicators of future price changes, and that the fore casting information contained in the current period’s capacity utilization rate is smaller in magnitude than the informational content of past price changes. typical coefficient on previous price changes. These estimates indicate that a 1.0 percentage-point increase in capacity utilization is associated with a 0.01-to-0.16 percentagepoint increase in prices in the long run. The forecast equations for the two industries with lagged capacity utilization rates included in the regressions (lumber products and electri cal machinery) show very small, statistically insignificant, cumulative relationships with current price changes. O f course, it is possible that the number of lags included in these equations is not suf ficient to capture the dynamic relationship between prices and capacity utilization, espe cially if the Schwarz criterion underestimates the number of lags.6 To check the robustness of the specification, I also select a common forecasting equation for each of the industries, using three lags of price changes and contemporaneous-plus-three lags of capacity utiliza tion. The additional lags allow some latitude for possible misspecification, but do not impose a large penalty for the number of additional regressors. Table 3 shows the regression results from estimating the forecasting equation with the additional lags over the same sample period. Forecast equations for six of the two-digit industries— coal mining, printing and publish ing, chemical products, leather products, primary metals and miscellaneous manufac turing— as well as total industrial and manu facturing aggregates, show statistically signif icant coefficients (at the 10 percent level) for the added capacity utilization lags. The sums of the capacity utilization coefficients suggest that the conclusions about the relationship remain essentially unchanged. Nearly all of the sums equal the single coefficient shown in Table 2, and with the exception of stone and earth minerals, stone, clay and glass products, and primary metals, the significance of the total estimated relationship between capacity uti lization and price changes remains unaffected by the change in the forecasting equations lag structure. REFERENCES Board of Governors of the Federal Reserve System . Federal Reserve Measures of Capacity and Capacity Utilization (February 1 9 7 8 ). Finn, M a ry G. "Is 'H ig h ' Capacity Utilization Inflationary?," Federal Reserve Bank of Richmond Economic Quarterly (w inter 1 9 9 5 ) pp. 1 -1 6 . Garner, C. Alan. "Capacity Utilization and U .S . Inflation," Federal Reserve Bank of Kansas City Economic Review (fourth quarter 1 9 9 4 ) pp. 5 -2 1 . Gew eke, John and Richard M eese. "Estim ating Regression Models of Finite but Unknow n O rder," International Economic Review (February 1 9 8 1 ) pp. 5 5 -7 0 . M cElhattan, Rose. "Inflation, Supply Shocks and the Stable-lnflation Rate of Capacity Utilization, "Federal Reserve Bank of San Francisco Economic Review (w inter 1 9 8 5 ) pp. 4 5 -6 3 . _ _ _ _ _ _ _ . "Estim ating a Stable-lnflation Capacity Utilization R ate," Federal Reserve Bank of San Francisco Economic Review (fall 1 9 7 8 ) pp. 2 0 -3 0 . Shapiro, M atthew D. "Assessing the Federal Reserve's M easures of Capacity and Utilization," Brookings Papers on Economic Activity (vol. 1 ,1 9 8 9 ) pp. 1 8 1 -2 2 5 . Shea, John. "D o Supply Curves Slope U p ? " Quarterly Journal of Economics (February 1 9 9 3 ) pp. 1 -32. CONCLUSIONS Two conclusions emerge from the analysis in this article. First, although the possibility F E D E R A L RESERVE B A N K OF ST. L O U I S 18 SEPTEMBER/OCTOBER 1 9 95 C apa city U tiliz a tio n a n d N e t O u tp u t Price C u rve s fo r Selected Industries Total In d u s tria l C a p a city U tiliz a tio n a n d Finished G o o d s PPI Capacity Utilization Percent Price Changes Capacity Utilization Percent Price Changes M e ta l M in in g Total M in in g Capacity Utilization M a n u fa ctu rin g Percent Price Changes Capacity Utilization Percent Price Changes 10 1987 93 1994 O il a n d G a s E x tra ctio n Coal M in in g Capacity Utilization 88 Percent Price Changes F E D E R A L RESERVE B A N K OF ST. 19 Capacity Utilization LOUIS Percent Price Changes SEPTEMBER/OCTOBER 1 9 95 C apa city U tiliz a tio n a n d N e t O u tp u t Price C urve s fo r Selected Industries Foods S lo n e a n d Ea rth M in e ra ls Percent Price Changes Capacity Utilization 93 1.5 Capacity Utilization Percent Price Changes 8 3 .5 1 2 83- ■1 82.5 • 0.5 82- • 0 81.5 11- • -0 5 — 1 1987 88 93 1994 8 0 . 5 -f 80 i 1987 88 93 1994 A p p a re l Products Te x tile M ill Products Percent Price Changes Capacity Utilization Capacity Utilization Percent Price Changes 95 92 93 1994 F urn itu re s a n d Fixtu re s Lu m b e r Products Capacity Utilization Percent Price Changes Capacity Utilization 4 ■ -2 FEDERAL RESERVE B A N K OF ST. L OU I S 20 Percent Price Changes H[V|[Hf SEPTEMBER/OCTOBER C apa city U tiliz a tio n a n d N e t O u tp u t Price C u rve s fo r Selected Industries P a p e r Products P rin tin g a n d P u b lish in g Percent Price Changes Capacity Utilization Capacity Utilization Percent Price Changes 1 1987 88 89 90 91 93 1994 1987 Percent Price Changes Capacity Utilization Percent Price Changes Capacity Utilization Percent Price Changes 93 1994 P e tro le u m Products Chem ical Products Capacity Utilization 88 2 ■0 ■-1 R u b b e r a n d Plastics Products Capacity Utilization Percent Price Changes 21 R EVIEW SEPTEMBER/OCTOBER 1995 C apa city U tiliz a tio n a n d N e t O u tp u t Price C urve s fo r Selected Industries S lo n e , C lay a n d Glass Products Capacity Utilization P rim a ry M eta ls Percent Price Changes Percent Price Changes Capacity Utilization Percent Price Changes Tra n s p o rta tio n Equ ip m e n t Electrical M a c h in e ry Capacity Utilization Percent Price Changes N o n -E le c trica l In d u stria l M a c h in e ry F a bricate d M eta ls Capacity Utilization Capacity Utilization Percent Price Changes Capacity Utilization 22 Percent Price Changes IMlfW SEPTEMBER/OCTOBER 1 9 9 5 C apa city U tiliz a tio n a n d N e t O u tp u t Price C u rve s fo r Selected Industries Instrum ents M iscellan e o u s M a n u fa ctu rin g Percent Price Changes Capacity Utilization Percent Price Changes Capacity Utilization Ta b le 1 S ch w a rz In fo rm a tio n C rite ria fo r Selecting the N u m b e r of Lags Industry Capacity Price Utilization Changes Rates Industry Capacity Price Utilization Changes Rate Total industrial 1 0 P rinting a n d p ublishing 1 M a n u fa ctu rin g 1 0 Chem ical products 2 0 M in in g 1 0 P etroleum products 1 0 M etal m in in g Coal m in in g 1 1 0 0 R ubber a n d plastics products Leather products 2 2 0 0 O il a n d gas extraction 1 Stone a n d e a rth m inerals Foods 1 0 0 Stone, clay and glass products P rim a ry m etals 1 1 1 0 Fabricated m etals 3 0 0 Textile m ill products 3 0 0 3 0 N o n-electrical industrial m a ch in e ry Electrical m a ch ine ry 3 A p p a re l products Lu m b e r products 1 2 Tra n sp o rtatio n e qu ip m en t 1 3 0 Fu rn itu re a n d fixtu re s 1 2 0 Instrum ents 2 0 P aper products 2 0 M iscellaneous m a nu factu rin g 1 0 i B A N K OF ST. L OU I S 23 0 0 MAY/JUNK 1 9 95 Ta b le 2 R e g re s s io n S u m m a ry , V a ria b le N o . o f La gs - P rice C h a n g e s b y 2 - D i g i t In d u s trie s : 1 / 8 6 - 1 2 / 9 4 77(-) TT ,.2 TTf -3 Stffll rfs CU, CU(.j CU,.2 CU,.3 Sum Cu's R2 Aggregate Groups Total industrial M anufacturing M ining 0 .3 8 " (3 .7 4 ) 0 .4 7 ” (4 .2 4 ) 0 .1 9 (1 .6 2 ) 0 .0 4 “ (2 .1 7 ) 0.04** (2 .5 8 ) 0 .0 3 (0 .5 1 ) 0.3 9 * * (5 .3 3 ) -0 .4 0 * (1 .7 0 ) 0 .1 9 (1 .5 6 ) -0 .3 4 “ (4 .0 4 ) -0 . 0 3 (1 .0 3 ) -0 .0 1 (0 .4 0 ) 0 .0 2 (0 .3 2 ) 0.01 (1 .1 6 ) 0 .2 5 " (3 .0 5 ) 0 .0 7 (0 .7 6 ) 0 .0 5 (0 .6 0 ) 0 .4 6 “ (4 .8 3 ) -0 . 1 2 (1 .2 5 ) 0.40** (3 .9 8 ) -0 . 0 8 (0 .6 9 ) 0 .3 9 “ (4 .1 0 ) 0 .4 0 “ (3 .5 2 ) 0 .1 8 (1 .3 4 ) -0 . 0 5 (0 .6 3 ) 0 .2 6 “ (2 .7 5 ) 0.5 2 * * (5 .0 7 ) 0.11 (1 .0 4 ) 0 .2 1 “ (2 .7 2 ) -0 . 0 0 (0 .0 4 ) -0 . 0 8 (0 .7 0 ) 0 .0 7 (0 .9 7 ) 0 .1 4 (1 .6 2 ) -0 . 0 2 (0 .4 1 ) 0 .0 1 * (1 .6 7 ) 0 .0 0 (0 .6 6 ) 0 .1 0 “ (2 .1 0 ) 0 .0 2 “ (3 .1 3 ) 0 .0 6 “ (3 .3 2 ) 0 .0 1 “ (3 .4 2 ) 0 .0 4 (1 .5 5 ) 0 .1 5 (0 .9 4 ) 0.02** (2 .6 1 ) -0 . 0 0 (0 .4 2 ) 0.01 (1 .5 1 ) 0 .0 2 “ (3 .0 7 ) 0 .0 1 “ (3 .2 6 ) 0 .0 0 (0 .9 3 ) 0.01 (0 .6 6 ) 0.01 (0 .6 3 ) 0 .0 2 (1 .4 9 ) 0 .0 2 ” (3 .0 7 ) 0 .2 3 0 .3 5 0 .0 4 Mining Industries M etal m ining Coal m ining O il and gas extraction Stone and earth m inerals 0 .1 7 0 .0 9 0 .0 4 0 .1 2 Manufacturing Foods Textile m ill products Apparel products Lum ber products fu rn itu re a n d fixtures Paper products Printing and publishing Chem ical products Petroleum products Rubber and plastics products Leather products Stone, clay and glass products P rim a ry m etals Fabricated m etals N o n -e le ctrica l m a chinery Electrical m achinery Transportation equipm ent Instrum ents Miscellaneous m anufacturing 0 .2 7 “ (3 .7 0 ) 0.2 4 * * (2 .9 5 ) 0 .2 8 “ (3 .0 9 ) 0.2 5 * * (3 .4 6 ) 0.2 2 * * (2 .0 0 ) 0.2 2 * * (2 .8 0 ) 0 .1 0 (0 .6 7 ) 0 .6 1 “ (5 .0 4 ) 0 .7 4 * * (8 .8 1 ) 0.3 6 * * (3 .6 8 ) 0.5 4 * * (4 .5 4 ) 0 .2 3 * (1 .8 1 ) 0.3 7 * * (4 .2 4 ) 0 .2 8 “ (2 .9 8 ) 0.3 5 * * (3 .7 3 ) 0.32** (4 .0 8 ) 0 .1 8 (1 .6 2 ) 0 .6 2 “ (5 .7 0 ) 0.5 4 * * (5 .5 3 ) 0 .2 3 “ (2 .9 2 ) 0 .2 6 “ (3 .2 6 ) 0 .6 9 “ (6 .1 0 ) 0 .7 9 * * (1 2 .8 1 ) 0 .2 5 * (1 .7 0 ) 0 .0 6 0 .2 7 0 .1 6 0 .1 0 (1 .4 8 ) -0 . 0 2 (0 .7 2 ) 0 .3 4 0 .1 7 0 .5 0 0.11 0 .5 3 0 .1 8 0 .3 0 0 .0 9 0 .1 0 0 .3 9 0 .5 8 0 .4 4 -0 .0 4 ” (2 .0 0 ) 0.01 (0 .2 1 ) 0 .0 2 (1 .1 8 ) 0 .0 0 (0 .4 5 ) 0.01 0 .0 6 0 .0 8 0 .1 6 /-ratios in parentheses. 1 denotes significance at 10 percent. ** denotes significance at 5 percent. -0 .2 1 ” (3 .9 9 ) 24 MAY/JUNE 1 9 9 5 Ta b le 3 R e g re s s io n S u m m a r y , F ix e d N o . o f La g s - P ric e C h a n g e s b y 2 - D i g i t In d u s trie s : 1 / 8 6 - 1 2 / 9 4 TTf.j iT t.2 7t , . 3 Sum i f s cu, cuf_| cuf.2 cu,.3 Sum tu's R2 0.34** (2 .9 3 ) 0.4 6 * * (2 .6 9 ) 0 .1 3 (0 .9 7 ) -0 . 0 5 (0 .4 6 ) -0 .2 0 * (1 .8 8 ) 0 .0 7 (0 .9 0 ) -0 . 1 0 (1 .1 9 ) 0 .0 3 (0 .3 0 ) -0 . 0 5 (0 .5 2 ) 0 .1 9 (1 .4 8 ) 0.2 8 * * (2 .3 2 ) 0 .1 5 (1 .0 6 ) -0 .0 1 (0 .1 7 ) 0 .0 5 (1 .2 1 ) 0 .2 8 (1 .0 7 ) 0.1 8 * * (1 .9 7 ) 0 .0 8 (1 .1 4 ) 0 .1 3 (0 .3 8 ) -0 .1 5 * * (2 .0 6 ) 0 .0 2 (0 .2 8 ) -0 .4 1 (0 .8 3 ) 0 .0 2 (0 .4 5 ) -0 .1 2 * * (2 .6 4 ) 0 .0 0 (0 .0 1 ) 0.0 4 * * (2 .9 0 ) 0 .0 4 * * (3 .0 7 ) 0 .0 0 (0 .0 2 ) 0.21 0.47** (5 .9 3 ) -0 .4 2 * (1 .8 8 ) 0 .1 2 (0 .9 0 ) -0 .3 2 * * (3 .2 4 ) -0 .2 2 * (1 .9 3 ) -0 . 1 2 (1 .1 8 ) 0 .0 7 (0 .9 1 ) -0 . 0 3 (0 .2 9 ) -0 . 0 2 (0 .2 0 ) 0 .0 5 (0 .5 4 ) -0 . 0 9 (1 .0 0 ) 0 .0 6 (0 .7 0 ) 0.2 4 * * (2 .1 2 ) -0 . 4 9 (1 .5 5 ) 0 .0 9 (0 .6 1 ) -0 . 2 8 (1 .3 7 ) -0 . 0 4 (0 .4 8 ) -0 . 0 6 (1 .3 6 ) 0.8 6 * * (2 .4 6 ) -0 . 0 2 (1 .3 5 ) 0.01 (0 .0 6 ) 0 .0 2 (0 .7 7 ) -0 . 5 6 (0 .9 5 ) 0 .0 3 (1 .6 2 ) -0 . 0 6 (0 .6 6 ) 0.01 (0 .2 0 ) -0 . 1 6 (0 .2 7 ) -0 . 0 0 (0 .0 1 ) 0 .0 5 (0 .4 3 ) 0.0 5 * * (2 .5 8 ) -0 . 1 5 (0 .4 2 ) 0.01 (0 .3 8 ) -0 . 0 4 (1 .5 2 ) 0 .0 2 (0 .7 1 ) -0 .0 1 (0 .1 3 ) 0 .0 2 * (1 .7 8 ) 0.2 1 * * (2 .4 4 ) 0 .0 7 (0 .8 0 ) 0 .0 4 (0 .4 9 ) 0.4 8 * * (4 .0 0 ) -0 .2 2 * (1 .7 4 ) 0.3 5 * * (3 .3 6 ) -0 . 1 2 (1 .0 0 ) 0.4 2 * * (4 .9 7 ) 0.4 3 * * (3 .3 1 ) 0.11 (0 .6 7 ) -0 . 0 9 (1 .0 6 ) 0.2 0 * * (2 .3 7 ) 0.4 1 * * (4 .6 0 ) 0 .0 9 (0 .9 1 ) 0 .2 1 * * (2 .6 1 ) -0 . 0 3 (0 .2 5 ) -0 . 0 8 (0 .7 5 ) 0 .0 5 (0 .6 8 ) 0 .1 4 (1 .5 0 ) 0 .0 9 (1 .0 4 ) 0.2 7 * * (3 .5 8 ) 0.2 4 * * (2 .9 2 ) -0 . 1 3 (1 .0 9 ) 0 .1 5 (1 .5 2 ) 0.1 7 * * (2 .2 4 ) -0 . 0 3 (0 .3 5 ) 0.47** (4 .5 3 ) -0 .2 1 * * (2 .1 9 ) 0.3 2 * * (4 .4 7 ) 0.2 7 * * (3 .1 9 ) -0 . 0 0 (0 .0 3 ) 0.01 (0 .0 8 ) 0.3 9 * * (4 .6 7 ) 0 .3 1 * * (4 .3 4 ) 0 .1 3 (1 .3 3 ) -0 .2 0 * * (2 .2 8 ) 0 .1 7 (1 .5 1 ) 0.11 (1 .3 1 ) -0 . 0 6 0 .2 4 * (0 .4 0 ) (1 .8 9 ) 0.2 6 * * 0 61** (3 .0 0 ) (5 .3 3 ) 0.2 5 * * 0.5 3 * * (3 .6 0 ) (5 .1 9 ) -0 . 0 4 0 .3 1 * * (0 .4 1 ) (2 .1 6 ) 0 .0 4 -0 . 0 2 (0 .3 5 ) (0 .1 5 ) 0 .0 5 0.5 6 * * (0 .6 0 ) (3 .5 3 ) 0 .1 0 -0 . 0 5 (1 .0 0 ) (0 .2 6 ) -0 . 1 5 0 .7 4 * * (9 .4 2 ) (1 .6 1 ) 0 .1 0 0.3 2 * * (0 .8 3 ) (2 .7 4 ) 0 .0 7 0.4 9 * * (0 .5 8 ) (4 .2 1 ) 0.11 0.2 9 * * (1 .5 7 ) (2 .6 6 ) 0.25** 0.4 5 * * (2 .4 4 ) (2 .8 6 ) 0 .1 5 0.5 8 * * (1 .3 6 ) (4 .6 1 ) 0.2 4 * * 0.7 2 * * (3 .1 0 ) (5 .5 6 ) 0.2 7 * * 0.7 9 * * (3 .5 1 ) (1 0 .9 1 ) 0.20** 0.30** (2 .0 9 ) (2 .7 6 ) 0 .0 3 -0 . 2 5 (0 .4 5 ) (1 .3 1 ) 0 .0 9 0.3 1 * * (1 .3 8 ) (2 .2 0 ) -0 .1 6 0 .1 0 (1 .5 8 ) (0 .7 8 ) -0 . 0 2 (0 .2 7 ) -0 . 0 0 (0 .2 1 ) 0 .0 0 (0 .1 2 ) 0.1 0 * * (2 .2 7 ) -0 .0 1 (0 .5 4 ) 0 .0 3 (1 .2 5 ) 0 .0 3 * (1 .9 7 ) 0 .0 7 * (1 .8 0 ) 0 .2 7 (1 .2 7 ) 0 .0 0 (0 .1 5 ) 0.01 (0 .4 6 ) -0 .0 1 (0 .2 5 ) 0.0 7 * * (3 .1 0 ) 004** (2 .1 0 ) -0 .0 1 (0 .5 8 ) 0.01 (1 .0 7 ) -0 .0 1 (0 .3 4 ) 0 .0 2 (0 .8 6 ) 0.0 3 * (1 .9 2 ) -0 . 0 4 (0 .4 9 ) 0.01 (0 .8 1 ) -0 .0 1 (0 .6 7 ) 0 .1 1 * (1 .6 8 ) -0 .0 1 (0 .5 8 ) 0 .0 5 (1 .5 9 ) -0 .0 6 * * (3 .5 1 ) 0 .0 7 (1 .2 6 ) -0 . 3 0 (0 .7 6 ) -0 . 0 2 (0 .6 4 ) 0.01 (0 .4 9 ) -0 . 0 0 (0 .0 8 ) -0 . 0 0 (0 .1 8 ) -0 . 0 3 (1 .3 6 ) 0 .0 2 (1 .1 0 ) -0 .0 3 * * (2 .0 2 ) 0.01 (0 .3 5 ) -0 . 0 3 (1 .2 2 ) -0 .0 1 (0 .4 7 ) 0 .1 0 (1 .0 9 ) 0.01 (1 .1 2 ) 0.01 (0 .3 1 ) -0 .1 4 * * (2 .0 6 ) 0 .0 2 (1 .1 4 ) 0.01 (0 .2 3 ) 0 .0 0 (0 .0 5 ) -0 .0 9 * * (2 .5 8 ) 0 .3 4 (0 .5 6 ) 0 .0 3 (0 .6 0 ) 0 .0 3 (1 .3 1 ) 0 .0 0 (0 .1 0 ) -0 .0 1 (0 .3 2 ) -0 . 0 0 (0 .2 2 ) -0 .0 1 (0 .7 7 ) 0 .0 0 (0 .0 1 ) -0 . 0 2 (0 .7 7 ) 0.01 (0 .3 9 ) 0 .0 2 (0 .8 9 ) -0 . 0 2 0 .0 2 (0 .3 0 ) (0 .3 4 ) -0 .0 1 0 .0 1 * (1 .0 5 ) (1 .9 0 ) 0.01 0 .0 0 (0 .5 9 ) (0 .8 2 ) -0 . 0 9 -0 . 0 2 (1 .6 2 ) (1 .0 8 ) 0 .0 2 0.0 2 * * (1 .4 6 ) (4 .3 3 ) -0 . 0 0 0 .08** (0 .0 7 ) (3 .3 2 ) 0.0 5 * * 0.0 1 * * (3 .1 7 ) (3 .2 9 ) -0 . 0 2 0 .0 2 (0 .6 4 ) (1 .0 8 ) -0 . 2 9 0 .0 3 (0 .7 3 ) (0 .1 9 ) 0 .0 2 0.0 3 * * (3 .3 4 ) (0 .7 6 ) -0 .0 5 * * -0 .0 1 (2 .9 5 ) (1 .1 5 ) 0.01 0 .0 1 * (0 .8 8 ) (1 .7 1 ) -0 .0 4 * * 0.01 (2 .0 8 ) (1 .2 4 ) 0.01 0.0 1 * * (0 .8 0 ) (2 .6 1 ) 0 .0 0 0 .0 0 (0 .0 1 ) (0 .8 1 ) 0 .0 2 0 .0 0 (1 .0 9 ) (0 .5 2 ) 0 .0 3 0.01 (0 .9 9 ) (1 .1 8 ) 0 .0 2 0 .0 2 (0 .8 6 ) (1 .6 1 ) -0 .0 2 * 0.0 2 * * (1 .7 4 ) (3 .1 7 ) Aggregate Groups Total industrial M anufacturing M ining 0 .3 5 0 .0 6 Mining Industries M etal m ining Coal m ining O il and gas extraction Stone and earth m inerals 0.21 0 .1 5 0 .0 8 0 .1 4 Manufacturing Foods Textile m ill products Apparel products Lum ber products Furniture and fixtures Poper products Printing and publishing Chem ical products Petroleum products R ubber and plastics products Leather products Stone, clay and glass products P rim a ry m etals Fabricated m etals N o n -e le ctrica l m achinery Electrical m achinery Transportation equipm ent Instrum ents M iscellaneous m anufacturing t-ratios in parentheses. * denotes significance at 10 percent. ** denotes significance at 5 percent. F E D E R A L RE S E RV E B A N K OF S T . L O U I S 0 .0 7 0 .2 9 0 .1 7 0 :3 7 0 .2 4 0 .5 2 0 .2 5 0 .5 7 0 .1 7 0 .3 4 0 .1 7 0 .1 7 0 .4 5 0 .6 0 0 .4 5 0.11 0 .0 7 0 .1 0 0 .2 5 C hristopher J. N e e ly is a n econom ist at the Federal R eserve B an k o f St. Louis. G e o ffre y E. W o o d is a professor at City U n ive rsity Business School in London a n d w as a visiting scholar at the Federal R eserve B an k o f St. Louis w h e n this article w as b e g u n . K elly M . M o rris p ro vid e d research assistance. Deflation and Real Economic Activity Under the Gold Standard Christopher J. Neely and Geoffrey E. Wood n the past few years, several countries have announced explicit target ranges for inflation. New Zealand did this in 1990, Canada in 1991, the United Kingdom in 1992, and Sweden and Finland in 1994. Even when an inflation target is achieved, the future price level is not easy to predict because none of these countries has committed itself to reversing the consequences of shocks to the price level. Indeed, in New Zealand there is an explicit commitment not to reverse cer tain such shocks. One alternative to inflation targeting is price level targeting.1 The adoption of a constant price level target would have several advantages over an inflation target. Chief among these is that consumers and firms could write simpler contracts and make long-run plans without worrying about inflation. Price level targeting also may avoid the “time-inconsistency” problem of an inflation targeting regime in that the monetary authority would have less incen tive to inflate the economy in a one-time bid to increase output temporarily. Under a price level target, any “surprise” inflation must be reversed. Critics o f price level targeting argue that making a commitment to reverse sur prise increases in the price level is undesir able because a fall in the general price level, or deflation, can have harmful effects. One such critic, Stanley Fischer, put it this way: I “I argue for the inflation target because I fear the consequences of having to aim to deflate the economy half the time, which is what the price level approach requires.”2 Since the end of World War II, year-overyear declines in the price level have been rare in the industrialized world; during the period of the gold standard, however, both long downward trends in the price level and much shorter periods of falling price levels were common.3 Ironically, although Irving Fisher advocated a price level target precisely to avoid the protracted downward (and upward) swings in the price level observed under a gold standard, the experience of this period provokes, in part, the criticism of price level targeting today. Perhaps more important for these beliefs about deflation is the deflation ary period (not examined here) from 1929 through 1933, in which the price level fell by 20 to 30 percent. Bernanke (1 995) argues persuasively that this price decline, caused by the U.S. determination to stay on the gold standard, was a major contributor to the severity of the Great Depression. This article reexamines the facts surrounding temporary periods of deflation that occurred under the gold standard from 1870 to 1913. We first describe the behavior of price, money and output data, then perform some simple tests to determine whether output growth grew more slowly during periods of falling prices and whether knowledge of a falling price level would, in fact, have helped predict lower output growth. Although we must be cautious about drawing conclusions from 100-year-old data generated under a much different mone tary regime, another look at this experience is warranted because several countries have adopted policies that are likely to be associ ated with temporary periods o f deflation. The next section briefly reviews why deflation may affect real output. A descrip tion of our data set and an explanation of our statistical tests follow. We then report the results of our tests, before concluding with some ideas for future work. 1 A price level is o weighted overage of prices in a country. Price level targets may be either constant over time (static) or have a trend. In this paper, w e will use price level targeting to refer to a static price level target. The shaded insert on pp. 34 ond 35 distinguishes price level and inflation targets. 2 The Financial Times, June 24, 1 9 94. Note thot Fischer refers to a static price level target. A price level target with a positive trend would only require the monetary authority to "disinflote" half the time, that is, to run o rate of infla tion below the long-run trend. Disinflation is not the only poten tial drawback of price level targets. Some oppose them because they might lead to greater short-run volatility in the inflation rate. 3 Periods in which prices fall on a yeor-over-year basis ore considered periods of deflation. REVIEW IBKR / O C T O B E R 4 An excellent review of these issues con be found in McCollum (1 9 8 9 ), Chapter 9. Ohanian ond Stockman (forthcoming) consider the conse quences of monetary shocks for the economy when some, but not all, prices are sticky. That paper also sets out several explanations for price stickiness in addition to those reviewed in McCallum. 5 Barro (1 9 9 5 ) finds benefits of lower inflation in the form of higher long-run growth in a cross-country study. Here, we are concerned with short-run effects. 6 See W ynne (1 9 9 5 ) for a survey of price stickiness and Craig (1 9 9 5 ) for evidence on wage rigidity. 7 Advocates of this view might point out that real wages rose substan tially during the severe deflation of the Great Depression. 8 Various series existed before publi cation of that volum e, but they had deficiencies that were reme died (as well as some new data permitted) by Capie ond Webber. See Capie and Webber for discus sion of previous series deficiencies and how they are remedied. The crucial point is that these previous series contained a spurious trend. 9 There has recently been some dis cussion of the reliability of that out put series— see the interchange between Greasley (1 9 8 6 ,1 9 8 9 ) and feinstein (1 9 8 9 ), ond the dis cussion in Crafts, Leybourne and Mills (1 9 8 9 )— but there seems to be general agreement that whatev er its deficiencies, it is the best available. 10 W e dropped Japan from the sample because it did not have a metallic standard during the 19th centuiy and because its national banking and financial system was just forming (see Backus and Kehoe, 1 9 9 2 ). Uniquely, Japan's growth under falling prices (5 .4 percent) was substantially higher than its growth under rising prices (1 .5 percent). PRICE STICKINESS, DEFLATION AND OUTPUT 1 9 95 despite evidence to the contrary, many econ omists continue to believe that some prices are inflexible downward and that even tem porary periods of deflation might reduce output through this channel.7 Bernanke and Jam es (1 9 9 1 ) argue that deflation might alternatively affect the economy by increasing the real value of nominally denominated debt. For example, a 2 percent annual deflation would translate a nominal interest rate of 4 percent into a real interest rate of 6 percent. Increasing the real rate of interest might promote debtor insolvency and financial distress. The opposition to price level targeting from those who fear the results of deflation, either because o f downward price rigidity or the consequences of debt-deflation, makes the study of the historical association between output and deflation worthwhile. A review of the evidence would be a first step in con sidering whether a central bank should now adopt a price level target. It is now widely accepted that there is no long-term trade-off between inflation and output or employment; the existence of a short-run trade-off, on the other hand, is not generally denied. There are several explana tions for this trade-off: lags between actual and expected inflation (see Hume, 1752; Fisher, 1926; and Friedman, 1968); misperceptions about relative and general price shifts (Lucas, 1972); and staggered wage or price setting (Fischer, 1977; Taylor, 1980).4 None of these theories, however, predicts that lowering the price level is more costly than lowering infla tion. Nevertheless, prices have not fallen (by anything more than a trivial amount) in any major economy since 1945. The means by which deflation might reduce output, however, are often not explicitly stated.5 One view is that deflation interferes with the adjustment of relative prices because nominal wages or some prices do not adjust downward easily. If wages and/or prices are sticky downwards, a negative demand shock will tend to cause persistent unemployment as prices and wages are slow to fall as required to clear markets. W ith a sufficiently high inflation trend, relative prices can adjust to a negative demand shock without any actual prices having to fall. Because markets work better with a little inflation, according to this view, output will be less variable over busi ness cycle horizons and, perhaps, even higher in the long run. Critics of the theory of downward price rigidity point out that many wages and prices do, in fact, decrease, and that the extent to w hich prices are sticky depends on whether people expect inflation. An atmosphere of overall price stability will make people more willing to accept reductions in their wages or prices. There is mixed evidence from m icroeco nomic data on the idea that prices are sticky; certainly, some prices change more frequently than others. There is, however, little evidence of asymmetry in price stickiness.6 Blinder (1991) presents the results of a survey in which firms report asymmetric price rigidity. He finds greater upward rigidity. Nevertheless, THE RELATION BETWEEN PRICE A N D O UTPUT GROW TH DURING THE GOLD STANDARD ERA We use two sets of data. The first con sists of 44 annual observations on money, prices, interest rates and output in the United Kingdom from 1870 to 1913. The period 1880-1913 is generally considered the heyday of the classical international gold standard. We end the sample before the beginning of World War I in 1914. The source for the monetary series is Capie and Webber (1 985) ,8 The interest rate is a short term one from the last quarter o f each year. The output series is Feinstein’s (1 9 7 2 ) com promise estimate of GDP and, therefore, his implicit price deflator is used as the price series.9 All data are annual to conform to the necessity of using annual GDP data. The second data set consists of 44 annual output and inflation observations (1 8 7 0 to 1913) from nine of 10 industrialized countries compiled for comparison of international business cycles by Backus and Kehoe (1 9 9 2 ), from which more complete description of the data is available.10 28 REVIE W SEPTEMBER/OCTOBER 1 9 9 5 F igu re 1 Tim e Series of the Levels of the U n ite d K in g d o m D ata O u tp u t Im p licit Price D e fla to r 125 120 115 110 105 Inte rest Rates 1200 - / 1100 1000 900 - 700 600 Bill 500 1870 The Time Series I i■ '■ i ■1I 111 I 78 82 86 90 94 98 States and France returning to the gold stan dard, raising the demand for and price of gold. The nominal interest rate seems to display typical cyclical fluctuations around a stationary mean. Figures 1 and 2 display the time series of the log levels and log differences of the four United Kingdom series from 1870 to 1913. The shading in the figures represents periods in which the price level fell (not periods of recession). The monetary series, M3, and the output series generally grew over time. The price deflator series does not display the consistent rise typical of modem price indices; rather, periods of rising and declining prices seem to be nearly equally common. The long downward trend in the price level until 1896, followed by an upward swing through the end of the sample in 1913, was caused by fluctu ations in the world supply of and demand for gold. For example, the downward drift in prices until 1896 was partly due to the United I 74 Inflation and Output During the Two Subperiods Figure 3 shows the higher average rates of inflation, displaying a scatterplot of the mean rates of output growth vs. mean inflation rates for each of nine countries from the Backus and Kehoe data set for each of the two subperiods (1 8 7 0 -9 6 and 1897-1913). The figure shows that average inflation rates were uniformly lower in the first period (1 8 7 0 -9 6 ) than they were in the second period (1 8 9 7 -1 9 1 3 ). FED ER A L RESERVE B A N K OF ST. L O U I S 29 02 06 10 1913 Tim e Series of the D ifferences of the U n ite d K in g d o m D ata O u tp u t G ro w th In fla tio n 0.075 0.075 n 0.050 0.025 0.000 -0 .0 25 -0 .0 5 0 T 11I■ I 1 -0 .0 7 5 *i 1870 74 78 82 86 90 94 98 1870 Inte rest R ate Changes 74 78 82 86 90 94 98 02 06 82 86 90 94 98 02 06 10 1913 M o n e y G ro w th 0.10 0.08 0.06 0.04 0.02 0.00 - 0.02 -0 .0 4 -0 .0 6 74 78 82 86 90 94 02 06 10 1913 1870 Consistent with the idea that deflation reduces output growth, the mean levels of output growth also appear to be lower during the first period. Curiously, across countries there seems to be a negative relationship between output and price changes in the first period and a pos itive relationship in the second. Output Growth and Deflation O ver Short Horizons Examining inflation and output growth over the two long subperiods is a convenient way to examine the relationship between average inflation and average output growth over longer periods. It does not, however, get direcdy at the question of whether price declines were associated with lower output 74 78 growth over short periods. To see this, we sort the data on output growth by the rise or fall of prices. For the purpose of categoriza tion, we define a deflationary period as any year in which prices fell; we make no dis tinction between the episodes on the basis of length, severity or cause. For the United Kingdom data, five of nine deflationary episodes lasted more than one year, and three lasted more than two years. Table 1 (page 32) provides some summary statistics for data from the nine countries used by Backus and Kehoe for the period 1870-1913. The first two columns provide the unconditional means of output growth and inflation. The third column shows the percentage of the time that prices were rising during the sample period. Mean REVIEW SEPTEMBER/OCTOBER 1 9 9 5 price declines were of comparable magnitude to mean price rises, and periods of mild price rises were only slightly more common than periods of declining prices; the data show that prices rose about 46-67 percent of the time during the sample. Figure 4 is analogous to Figure 3 in that it depicts mean output growth for the nine countries from the Backus and Kehoe sample, conditioned on whether prices rose or fell. Again, the means of output growth during periods of rising prices appear gener ally higher than the means during periods of falling prices. This positive relationship between price changes and output growth is again consistent with the idea that defla tionary periods were associated with relatively hard times. F ig u re 3 M e a n O u tp u t G ro w th in the First P e rio d ( 1 8 7 0 -9 6 ) a n d the Second P e rio d (1 8 9 7 -1 9 1 3 ) Output growth 6.5 5.5 4.5 r n +Sweden 2 GeraT y 2 J-Norway 2 Italy 2 +(jK 2 + Italy 1 0.5 -----1------- 1--------1--------1--------r— — r— -2 .0 -1 .5 -1 .0 -0 .5 0 0.5 1.0 — i— 1.5 — i— 2.0 —i 2.5 Inflation Note: Sample TESTING THE RELATIONSHIP BETWEEN DEFLATION AND OUTPUT GROW TH 1= 1870-96, sample 2 = 1897-1913. F ig u re 4 M e a n O u tp u t G ro w th C o n d itio n a l on In fla tio n o r D e fla tio n The positive relationship between price changes and output growth must be inter preted with a great deal of caution. First, the positive correlation between price changes and output growth could be due to chance. In other words, how likely is it that the observed data would have been generated if the means of output growth were equal under deflation and inflation? Second, the previous section only examined the relation ship between price changes and output growth period by period; we would like to know about their relationship over time as well. Third, even if deflation is statistically associated with lower output growth, that does not mean it causes lower output growth— a third factor could be causing both. Output growth 4.5 Canada l+ +US1 + US2 Denmark 1+ Canada 2 3.5 Australia 2 2.5 - ►UK2 Norway 2 0.5 -0 .5 - 4 — i-------- 1---------1---------1---------1---------r~ - 3 - 2 - 1 0 1 2 Inflation Note: Sample 1= inflation, sample 2 = deflation. tical significance of the correlation. The second and third columns of Table 2 present results of the F-tests of the hypothesis that the mean output growth for each of the nine countries in Figure 3 was the same dur ing the second period (1 8 9 7 -1 9 1 3 ) as in the first period (1 8 7 0 -1 9 1 3 ). The third column gives the probability that we would obtain at least as extreme a result if the means were truly the same. This number, called the “p-value,” is often loosely interpreted as the strength of the evidence against the hypothe sis that the means are the same. Values less To test whether the apparent relationships between output growth and price level changes pictured in Figures 3 and 4 could be coinci dence, we can determine if it is likely that such a relationship would have been generated if mean output growth were really equal under inflation or deflation. That is, we test the statis- FED ER A L RESERVE B A N K OF ST. L O U I S 31 Australia 1 + Sweden 1 Germany 1 + Norway 1 +. UK 1 + Italy 1 ..Denmark 2 + Germany 2 .Sweden 2 1.5 - Is It a Coincidence That Output Growth Is Lower During Periods of Deflation? Denmark 1 + Germany lx Australia 1 + ,c + , , Canada }+ +Swed‘!n > UK 1 +Norwayl 1.5 + Denmark 2+ 3.5 2.5 +US2 Australia 2 +US1 SEPTEMBER/OCTOBER 1 9 95 Ta b le 1 In te rn a tio n a l O u tp u t G ro w th an d Inflatio n Statistics U n d e r Rising a n d Fa llin g Prices Unconditional Statistics A ustralia Canada D e n m a rk G e rm a n y Italy N o rw a y Sw eden United K ingdom United States M ean Inflation M ean O utp u t G row th 0 .2 3 0 .5 9 0 .1 4 0.61 0 .7 4 0.71 0 .6 6 0 .0 8 0 .2 6 3 .1 0 3 .9 0 3 .2 0 2 .6 6 1 .4 5 2 .1 7 2 .7 3 1 .8 8 4 .0 3 Proportion of Years Prices are Rising (percent) 5 1 .1 6 6 7 .4 4 4 8 .8 4 5 8 .1 4 5 8 .1 4 5 8 .1 4 6 0 .4 7 5 1 .1 6 4 6 .5 1 | Ta b le 2 Rising Prices M ean Inflation M ean O utp u t G row th 3 .3 7 2 .5 7 1 .6 7 2 .6 9 3 .6 0 3 .3 5 3 .0 7 1 .8 3 2 .8 0 3 .6 0 3 .9 8 3.51 2 .7 9 2 .5 8 2 .6 8 3 .4 7 2 .1 6 4 .0 3 Falling Prices M ean O u tp u t G ro w th M ean Inflation -3 . 0 6 -3 .5 1 -1 . 8 6 -2 .2 9 -3 . 2 4 -2 .9 6 -3 .0 2 -1 .7 6 -2 . 9 3 2 .5 7 3 .7 5 2.91 2 .4 8 -0 .1 3 1 .4 5 1 .6 0 1 .5 8 4 .0 3 o Tests of the E q u a lity of M e a n O u tp u t G ro w th U n d e r Inflatio n v s . D efla tion Test of Equality of Mean Output Growth Between the Subperiods 1 8 7 0 -9 6 and 1 8 9 7 -1 9 1 3 A ustralia Canada D e n m a rk G e rm a n y Italy N o rw a y Sw eden U .K . U .S . A g gre g a te Test of Equality of Mean Output Growth Conditioned on Inflation or Deflation Test Statistic p -va lu e Test Statistic p -va lu e 1 .4 4 1 5 .4 5 0 .3 2 0 .0 0 3 .1 9 0 .8 2 0 .6 6 0 .0 9 0 .1 6 11.01 0 .2 3 0 .0 0 0 .5 7 1 .0 0 0 .0 7 0 .3 7 0 .4 2 0 .7 7 0 .6 9 0 .2 8 1 .0 6 0 .0 5 0 .3 6 0 .1 0 7 .3 6 1 .5 2 3 .5 0 0 .3 3 0 .0 0 1 5 .3 2 0 .3 0 0 .8 2 0 .5 5 0 .7 6 0.01 0 .2 2 0 .0 6 0 .5 6 1 .0 0 0 .0 8 the overall mean output growth for all nine countries for the second period is the same as the overall mean output growth for the first period. The p-value from such a test is 0.28 (see the third column, last row of Table 2), which strongly suggests that it is very possible that the data were generated by processes with equal means. That is, for only two countries could we conclude that aggregate mean output growth in the second period was statistically significantly higher than 0.1 or 0.05 are usually interpreted as meaning that we can reject the idea that the means are the same. A lower p-value means that it is less likely that the means are the same. These tests of equality of means reject the idea that the conditional means are equal for Canada and Italy, but not for the other countries if our criterion for rejection is a p-value less than 0.1. If we pool the observations from all the countries, we can test the hypothesis that F ED E R A L RE S E RV E B A N K OF S T . L O U I S 32 SEPTEMBER / OCTOBER 1 9 9 5 Ta b le 3 Ta b le 4 Fit of Asym m etric V s. Sym m etric M odels of Prices and O utput Tests of Lin e a r Forecasting A b ility of Price Changes a n d O u tp u t G ro w th Preferred Model Under the Akaike Criterion Schwarz Criterion Australia Canada D e n m a rk G e rm a n y Italy N o rw a y Sw eden UK US sym m e tric sym m e tric asym m etric a sym m e tric sym m e tric sym m e tric sym m e tric a sym m e tric sym m e tric Granger Causality Statistics (p-value) Test that Price Changes Test Th a t O utput G row th Do Not H elp Forecast Does Not H elp Forecast O utput G row th Price Changes sym m e tric sym m e tric a sym m e tric sym m e tric sym m e tric sym m e tric sym m e tric sym m e tric sym m e tric than the mean of output growth in the first period. Columns four and five of Table 2 present results of similar tests for equality of means for the data in Figure 4. For Italy and Sweden, we reject the idea that the mean of output under inflation was the same as that during deflation. For this test, however, aggregating the observations across countries leads to the conclusion that output growth was signifi cantly lower in a statistical sense during peri ods of deflation. The p-value for the test of that hypothesis is 0.08 (see the fifth column, last row of Table 3). 2 0 .5 3 7 (0 .0 0 0 ) 7 .3 6 7 (0 .0 1 0 ) Canada 3 .8 7 5 (0 .0 1 8 ) 8 .8 4 8 (0 .0 0 0 ) D e n m a rk 1 .8 4 8 (0 .1 7 2 ) 5 .7 6 8 (0 .0 0 7 ) G e rm a n y 5 .4 9 3 (0 .0 2 4 ) 0 .3 4 3 (0 .5 6 2 ) Italy 2 .2 6 2 (0 .1 1 9 ) 0 .0 6 1 (0 .9 4 1 ) N o rw a y 1 .3 4 7 (0 .2 5 3 ) 7 .2 4 5 (0 .0 1 0 ) Sw eden 0 .2 1 0 (0 .6 4 9 ) 5 .0 7 1 (0 .0 3 0 ) US 2 .0 2 0 (0 .1 3 0 ) 1 .4 4 3 (0 .2 4 8 ) UK 1 .3 4 6 (0 .2 5 3 ) 2 .0 8 9 (0 .1 5 6 ) own lagged values. VARs are a commonly used, general method of modeling the dynamic relationship between macroeconomic variables. In the first system o f equations, we treat positive and negative price changes as two different variables and allow them to influence output growth (and each other) differently.11 In the second system, we treat price changes as one variable, forcing positive and negative changes to have mirror-image effects on out put growth. Then we examine which model fits the data better. We judge the fit of the systems accord ing to two commonly used criteria: the Akaike information and the Schwarz infor mation criteria. These measures of the fit of the two models on the Backus and Kehoe data are shown in Table 3. The results indi cate that the Akaike criterion favors the asymmetric model for Denmark, Germany and the United Kingdom, but the Schwarz criterion favors it only for Denmark. For the other countries, the simpler symmetric model Do Price Changes Have an Asymmetric Effect on Output? The previous analysis described the period-by-period relationship between aver age output growth and average price changes conditioned on the sign of the price changes. M acroeconomic variables, however, influ ence each other not ju st contemporaneously, but also over time. The symmetry of the dynamic relationship between output growth and price changes is important, because an essential implication of the idea that deflation is harmful to output is that output reacts asymmetrically to price changes over time. To explore this issue, we again break the price changes into positive and negative changes so that we can fit two systems of regression equations (called vector autore gressions, or VARs) in which we regress output growth and price changes on their Australia NK OF ST. L OU I S 33 11 The three variables in the system ore output growth, positive price changes (INFLDP) and negative price changes (D EFD P), where INFLDP= DP, if DP> 0 = 0, otherwise DFFDP = DP, if DP < 0 = 0, otherwise and DP is the rate of change of prices. REVIEW Se p tem b er / O ctober 1 9 9 5 PRICE LEVEL VS. INFLATION TARGETING Price stability has attracted a lot of attention lately. Unfortunately, the im portant choice between inflation and price level targeting has been neglected. Either would lead to a lower and more stable inflation rate than we have observed over the past 25 years, but there is a fundamental distinction between the two. Price level targeting “corrects” past errors in monetary policy, while inflation targeting ignores them. To make this distinction more concrete, consider the problem of a monetary authority with an inflation target of zero to 2 percent in which the 1995 inflation rate is 3 percent, 1 percentage point above the target range. In choosing monetary policy for 1996, the authority will aim, as usual, for an inflation rate of zero to 2 percent. It will not try to make up for past errors. In contrast, if the same monetary authority has targeted a static price level (zero percent inflation on average) and observes 1 percent inflation, it will have to try to reduce the price level by 1 percent in the years ahead. This difference makes price level targeting a long-run commitment in ways in which inflation targeting is not. There are three major consequences of this divergence between the two. First, the average rate of inflation over a long horizon can be predicted very well under a price level targeting regime; it is less certain under an inflation targeting regime.1 Advocates of price level targeting often point to the greater certainty of the price level (average inflation rate) in the long run as an advantage. As the accompanying chart shows, uncertainty about the future price level associated with an inflation targeting range of zero to 2 percent increases as the time horizon grows. In contrast, the level of uncertainty asso ciated with a price level target is constant (and sm all), even over long time horizons. For example, an investor evaluating the real return on, or the present value of, a project can do so much more easily because the price level can be predicted over long periods. Second, an important theoretical advantage of the long-run nature of price level targeting is that by being a multi-period commitment, it does not suffer from the time-inconsistency problem described by Barro and Gordon (1 9 8 3 ). In their model, a monetary authority has an incentive to produce a one-time monetary stimulus that results in a burst o f output 1 The expected prediction error for future overage inflation would go to zero under a price level targeting regime os the time horizon increases, while it would remain constant under an inflation targeting regime. is favored.12 These tests provide mixed evi dence on the hypothesis that price changes have an asymmetric effect on output for the countries considered here. test directly whether the deflation itself was the cause of lower growth, we can test whether the falling price level helped to forecast it. Such a test of linear forecasting ability is called a test of Granger causality. If price changes improve the forecasts of output growth, they are said to “Granger-cause” output growth. The idea is that if a falling price level causes lower growth, then it should precede output growth and be useful in forecasting it. Note, however, that if a third factor is causing both deflation and lower growth, this statistical procedure can find that deflation helps fore cast lower growth, even when it is not the cause of lower growth. Does Deflation Forecast Lower Output Growth? 12 Because the Akaike and Schwarz criteria ore non-nested model selec tion criteria, they ore not formal statistical tests and do not hove "significance levels." Instead, they informally test for statistical signifi cance by penolizing more complex models. Previously, we showed that, under the gold standard, output growth tended to be lower than average during periods of deflation. Then we showed at least some evidence in favor of the hypothesis of an asymmetric dynamic relationship between price changes and output growth. Although we cannot 34 growth and inflation. Price Le ve l V s . In fla tio n Ta rg e tin g Because the public under stands this incentive, it Price Level reacts in such a way that the authority inflates each period but fails to increase output. A price level target solves the time-inconsistency problem by requiring the monetary authority to correct past errors. The authority has no incentive to stimulate the economy with a little inflation, because it would then have to reduce the price level back to the target level. Therefore, a price level target should be more credible than an inflation target.2 A third major difference motivates the subject of this article. A static price level target requires the monetary authority to reduce the price level in response to surprise increases. While an inflation rate target may produce occasional reductions in the price level accidentally, they will be rare if the average inflation rate is high relative to the volatility in inflation. In contrast, under a static price level target, price changes will be negative roughly half the time. A hybrid of targeting inflation and targeting a static price level is targeting a small upward trend in the price level. Such a system has the long-term predictability of a static price level target but does not require the monetary authority to correct past upward deviations in the price level with deflation. 2 This argument assumes that even anticipated deflations will be as costly as the benefit gained from the initial inflation. To test whether price changes improve the forecasts of output growth, we first fore cast output growth using only its own lags. Then we add lagged price changes as another explanatory variable to see if their inclusion improves the forecasts. The second column of Table 4 displays the test statistic and p-value (significance level) from the tests that price changes do not Granger-cause (help forecast) output growth. For Australia, Canada and Germany, we reject the null hypothesis that lagged values of price changes do not improve the forecasts of output growth. In other words, the data suggest that price changes do help forecast output growth for three countries in this period. We should emphasize that rejec tions of Granger causality tests are a neces sary but not sufficient condition to determine that output growth is not “caused” by price changes. Once again, the data provide us with mixed results on the idea that price changes have an asymmetric effect on output. We can also investigate whether output growth helps forecast price changes in this system. Econom ic commentators commonly suggest that price pressures (or the lack thereof) are due to the level of output growth, employment, capacity utiliza tion or some other real variable. The test sta tistics and p-values from the tests that output 35 REVIE W SEPTEMBER/ OCTOBER 13 See Stable Money: A History of the Movement by Irving Fisher, 1 9 3 4 . growth does not help forecast future price changes are in the third column of Table 4. These statistics indicate that output growth does help forecast price changes for Australia, Canada, Denmark, Norway and Sweden. Although these results do not shed light directly on a possible asymmetric response of output to price changes, they are consis tent with traditional Phillip’s curve explana tions of inflation. From 1931 to the trough of the Depression, the price level fell by 20 percent to 3 0 per cent in countries that stayed on the gold standard, while falling less than 2 percent in Sweden (from 100 in September 1931, when the Riksbank started targeting the price level, to 98 .4 in October 1933). Unlike a gold standard, price level targeting permits control of the price level through the money supply. Caveat Emptor CONCLUSION Because we have only a small sample, the predictive power of one variable on another must be very strong for tests for Granger causality to find a relation. Weaker but important relations may not be found at all. Statisticians would say that tests o f Granger causality may have “low power.” Another complication is that both price and output changes may result from some third factor, which has been left out of the analysis. No matter how confident we are that we understand how these economies functioned 100 years ago, we must be cautious about using historical data to answer policy questions today. For example, economic structures such as the wage-setting mechanism, the degree of flexibility o f the labor market and credit allo cation mechanisms— all of which may influ ence how changes in the money supply translate to changes in the price level— have changed a great deal in the last century. Even methods of data collection are much different now. Finally, we remind the reader that the economists who observed this episode first-hand believed that deflation was a disruptive factor causing lower output growth. Many recommended a price level target as a remedy for that problem .13 Presumably, the finite sample variance of the price level would be much different under a price level targeting regime than it was under the gold standard. Some evidence in favor of this view can be found by com paring Sweden’s experience with prices during the Great Depression with that of countries that stayed on the gold standard. A number of countries, including New Zealand, Canada and the United Kingdom, have recently announced explicit target ranges for inflation. Such a policy has also been suggested for the United States. Others have suggested that we target the price level instead of the rate of inflation. One potential reason to oppose this sugges tion is that such a policy would necessitate that the monetary authority reduce the level of prices, that is, deflate the economy, to off set any transient, positive shocks to the price level. The historical association between deflation and bad econom ic performance has led some econom ists to reject price level targeting as bad policy. We find that lower output growth was associated with periods o f deflation in nearly all the countries examined. For a m ajority of the countries, the dynamic relationship between price changes and output growth appeared to be symmetric, and price changes did not help forecast output growth. There is more evidence, however, that output growth forecasts price changes. Ultimately, a final conclusion about the desirability o f a price level target requires more complete econom ic modeling than we have attempted. W hat we have presented are some simple facts about deflation and output that are touted as reasons to reject a particular type of price stability. Economists who support price level targeting must make the case that the temporary periods of deflation necessary to maintain long-term price stability would be fundamentally different than those observed under the gold standard. Sweden left the gold standard in 1931 and began to target the consumer price index. NK o r 36 ST. LOUIS HNIIEN S EPTEMBER/OCTOBER 1 9 9 5 REFERENCES Fischer, Stanley. "Long-Term Contracts, Rational Expectations, and the Optim al M oney Supply R ule," Journal of Political Economy (February 1 9 7 7 ), pp. 1 9 1 -2 0 5 . Bockus, David K ., and Patrick J. Kehoe. "International Evidence on the Historical Properties of Business Cycles," The American Economic Review (Septem ber 1 9 9 2 ), pp. 8 6 4 -8 8 . Fisher, Irving. Stable Money: A History of the /Movement. Adelphi Com pany, 1 9 3 4 . Barro, Robert J. "Inflation and Economic G row th," Bank of England Quarterly Review (M a y 1 9 9 5 ), pp. 1 6 6 -7 6 . _ _ _ _ _ _ _ . The Purchasing Power of Money. M acm illan, 1 9 1 1 . _ _ _ _ _ _ _ and David B. Gordon. "Rules, Discretion and Reputation in a M odel of M onetary Policy," Journal of Monetary Economics (Ju ly 1 9 8 3 ), pp. 1 0 1 -2 1 . Friedm an, M ilton. "Th e Role of M onetary Policy," The American Economic Review (M arch 1 9 6 8 ), pp. 1 -1 7 . Greasley, David. "British W ages and Incom e, 1 8 5 6 -1 9 1 3 : A Revision," Explorations in Economic History (April 1 9 8 9 ), pp. 2 4 8 -5 9 . Bem anke, Ben S. "Th e Macroeconom ics of the Great Depression: A Com parative Approach," Journal of Money, Credit, and Banking (February 1 9 9 5 ), pp. 1 -2 8 . _ _ _ _ _ _ _ . "British Econom ic Growth: The Paradox of the 1 8 8 0s and the Tim ing of the Clim acteric," Explorations in Economic History (October 1 9 8 6 ) pp. 4 1 6 -4 4 . _ _ _ _ _ _ _ and Harold Jam es. "Th e Gold Standard, Deflation, and Financial Crisis in the Great Depression: An International Com parison," in R. Glenn Hubbard, ed.. Financial Markets and Financial Crises. University of Chicago Press, 1 9 9 1 , pp. 3 3 -6 8 . H um e, David. Of Money in Political Discourses. 1 7 5 2 . Lucas, Robert E ., Jr. "Expectations and the Neutrality of M o n e y," Journal of Economic Theory (April 1 9 7 2 ), pp. 1 0 3 -2 4 . Blinder, Alan S. "W h y are Prices Sticky? Prelim inary Results from an Interview S tudy," The American Economic Review (M a y 1 9 9 1 ), pp. 8 9 -9 6 . M cCollum , Bennett. T. Monetary Economics: Theory and Policy. M acm illan, 1 9 8 9 . Capie, F. H ., and Alan W ebber. >1 Monetary History of the UK: 18701982, Vol. 1 Data: Surveys and M ethods. Allen ond Unw in, 1 9 8 5 . O hanian, Lee E., Alan C. Stockm on and Lutz Kilian. "Th e Effects of M onetary and Real Shocks in a Business Cycle M odel with Som e Sticky Prices," Journal of Money, Credit, and Banking (forthcom ing). Crafts, N. F. R., S . J. Leybourne and Terence C. Mills. "Th e Climacteric in Late Victorian Britain and France: A Reappraisal of the Evidence," Journal of Applied Econometrics (April/June 1 9 8 9 ), pp. 1 0 3 -1 7 . Taylor, John B. "Aggregate Dynam ics and Staggered Contracts," Journal of Political Economy (February 1 9 8 0 ), pp. 1 -2 3 . Craig, Ben. "Are W ages Inflexible?," Federal Reserve Bank of Cleveland Economic Commentary (April 1, 1 9 9 5 ). W ynne, M ark A. "Sticky Prices: W h a t is the Evidence?" Federal Reserve Bank of Dallas Economic Review (first quarter 1 9 9 5 ), pp. 1 -1 2 . Feinstein, Charles H. "W ages and the Paradox of the 1 8 8 0 s: Com m ent," Explorations in Economic History (April 1 9 8 9 ), pp. 2 3 7 -4 2 . _ _ _ _ _ _ _ . National Incom e, Expenditure and O utput over the United Kingdom : 1 8 5 5 -1 9 6 5 . Cam bridge University Press, 1 9 7 2 . IS 37 n SEPTEMBER/ OCTOBKR 1 9 9 5 Federal Reserve Bank of St. Louis W orking Paper Series 9 5 - 0 0 6 B - James B. Bullard and John Duffy, “On Learning and the Stability of Cycles.” orking papers from the Federal Reserve Bank of St. Louis contain preliminary results of staff research and are made available to encourage com m ent and discussion, as well as invite sug gestions from other researchers for revision. The views expressed in the working papers are those of the authors and do not necessarily reflect official positions of the Federal Reserve Bank of St. Louis, the Federal Reserve System or the Board of Governors. For published papers, contact the author or the publication listed. Single copies of working papers are available by writing to: W 9 5 - 0 0 7 A - Cletus Coughlin and David C. Wheelock, “Lessons from the United States and European Community for the Integration of High and Low Incom e Econom ics.” 9 5 - 0 0 8 A - Rowena A. Pecchenino and Patricia S. Pollard, “The Effects of Annuities, Bequests, and Aging in an Overlapping Generations Model of Endogenous Growth.” 9 5 - 0 0 9 A - Michael R. Pakko, “Tariffs and Asset Market Structure: Some Basic Comparative Dynamics.” Federal Reserve Bank of St. Louis Research Department PO. Box 442 St. Louis, MO 63166-0442 9 5 - 0 1 O A - William T. Gavin and Finn E. Kydland, “Endogenous Money Supply and the Business Cycle.” 1 9 9 5 W O R K IN G PAPERS 9 5 - 0 1 1 A - Guyhan Kim, “Exchange Rate Constraints and Money Control in Korea.” 9 5 - 0 0 1 A - Patricia S. Pollard and Rowena A. Pecchenino, “Dependent Children and 9 5 - 0 1 2 B p Donald S. Allen and Michelle T. Aged Parents: Funding Education and Social Security in an Aging Economy.” Gyles, “Trade, Growth and Capital: A Case Study of Jam aica.” 9 5 - 0 0 2 A - Chris Neely, Amlan Roy and Charles Whiteman, “Identification Failure in the Intertemporal Consumption CAPM.” 9 5 - 0 1 3 A - David C. Wheelock and Paul W. Wilson, “W hy Do Banks Disappear: The Determinants of U.S. Bank Failures and Acquisitions.” 9 5 - 0 0 3 A - Michael Dueker and Patrick K. Asea, “Nonmonotonic Long Memory Dynamics in Black-Market Premia.” FORTHCOMING: The Jou rnal o f C om parative Economics. 9 5 - 0 1 4 A - Michael Dueker and Andreas M. Fischer, “Inflation Targeting in a Small Open Economy: Empirical Results for Switzerland.” 9 5 - 0 0 4 A - Joseph G. Haubrich and Joseph A. Ritter, “Commitment as Irreversible 9 5 - 0 1 5 A - Joseph G. Haubrich and Joseph A. Investment.” Ritter, “Dynamic Commitment and Imperfect Policy Rules.” 9 5 - 0 0 5 A - Michael Dueker and Andreas M. Fischer, “Identifying Austria’s Implicit Monetary Target: An Alternative Test of the ‘Hard Currency’ Policy.” 9 5 - 0 1 6 A - William R. Emmons, “Interbank Netting Agreements and the Distribution of Bank Default Risk.” 38 MVHH SEPTEMBER/OCTOBER 1 9 95 1 9 9 4 W O R KIN G PAPERS 9 4 -0 1 1A - James B. Bullard and John Keating, “Supemeutrality in Postwar 9 4 - 0 0 1 B - Peter S. Yoo, “The Baby Boom and Economic Growth.” Econom ies.” 9 4 -0 1 2A 9 4 -0 0 2 A - - James B. Bullard and Steven H. Russell, “Monetary Steady States in a Low Peter S. Yoo, “Age Distributions and Returns of Financial Assets.” Real Interest Rate Economy.” 9 4 -0 0 3 A - Peter S. Yoo, “Age Dependent Portfolio Selection.” 9 4 -0 1 3A 9 4 -0 0 4 A - Joseph A. Ritter, “The Transition From Barter to Fiat Money.” PUB LISHED: The Am erican Econom ic Review (March 1995). 9 4 - 0 1 4B 9 4 -0 0 5 A - - James B. Bullard and John Duffy, “Learning in a Large Square Economy.” - James B. Bullard and John Duffy, “A Model of Learning and Emulation with Artificial Adaptive Agents.” 9 4 -0 1 5B - Michael J. Dueker, “Markov Switching in GARCH Processes and Mean Reverting Stock Market Volatility.” FORTH COMING: Jou rn al o f Business and Economic Statistics. (Published as “Mean Reversion in Stock Market Volatility.”) Sangkyun Park, T h e Bank Capital Requirement and Information Asymmetry.” 9 4 -0 0 6 A - Richard G . Anderson and Kenneth A. Kavajecz, “A Historical 9 4 -0 1 6B - Michael J. Dueker, “Compound Volatility Processes in EMS Exchange Rates.” Perspective on the Federal Reserves Monetary Aggregates: A Tim eline.” PUB LISHED: this Review (March/April 1994). 9 4 -0 1 7 A - Michael J. Dueker and Daniel L. Thornton, “Asymmetry in the Prime Rate and Kenneth A. Kavajecz, “The Evolution of the Federal Reserves Monetary Aggregates: A Timeline.” Firm s’ Preference for Internal Finance.” 9 4 -0 0 7 C Proske, “Are There Adverse Real Effects from 9 4 -0 1 8 A - Richard G . Anderson and William G . Dewald, “Replication and - John A. Tatom and Dieter Monetary Policy Coordination? Some Evidence from Austria, Belgium and the Netherlands.” Scientific Standards in Econom ics a Decade Later: The Impact of the JM C B Project.” PUBLISHED: this Review (November/December 1994). 9 4 - 0 1 9 A - Michael R. Pakko, “Reconciling International Risk Sharing with Low CrossCountry Consumption Correlations.” 9 4 -0 0 8 B - Christopher J. Neely, “Target Zones and Conditional Volatility: The Role of Realignments.” 9 4 -0 2 0 B - Christopher J. Neely, “Realignments o f Target Zone Exchange Rate Systems: W hat Do We Know?” 9 4 - 0 0 9 A - Christopher J. Neely, Dean Corbae and Paul Weller, “The Distribution 9 4 -0 2 1 A of Target Zone Exchange Rates Under Alternative Realignment Rules.” - David C. Wheelock and Paul W. Wilson, “Productivity Changes in U. S. Banking: 198 4 -9 3 .” 9 4 - 0 10A - Christopher J. Neely, “A 9 4 -0 2 2 A Reconsideration of the Properties of the Generalized Method of Moments in Asset Pricing Models.” - Alison Butler and Michael Dueker, “Product Cycles, Innovations and Relative Wages in European Countries.” 39 R EVIE W SEPTEMBER/OCTOBER 94-023A - Sangkyun Park, “Market Discipline by Depositors: Evidence from Reduced Form Equations.” 1995 1 9 9 3 W O R K IN G PAPERS 93-001A - Patricia S. Pollard, “Macroeconomic Policy Effects in a Monetary U nion.” 94-024A - Byung Chan Ahn, “Monetary Policy and the Determination o f the Interest Rate and Exchange Rate in a Small Open Economy with Increasing Capital Mobility.” 93-002A - David C. Wheelock and Paul W. Wilson, “Deposit Insurance, Regulation, and Efficiency.” FORTHCOMING: The Review o f Econom ics and Statistics. 94-025A - Sangkyun Park, “Banking and Deposit Insurance as a Risk-Transfer M echanism.” 1 9 9 2 W O R K IN G PAPERS “Characterizing Cross-Country Consumption Correlations.” 9 2-001A - John A. Tatom, “The P-Star Model and Austrian Prices.” PUBLISHED: E m pirica (no. 1, 1992). 94-027A - Michael Dueker and Richard 92-002A - David C. Wheelock and Subal C. Startz, “M aximum-Likelihood Estimation of Kumbhaker, “The Slack Banker Dances: Deposit Insurance and Risk-Taking in the Banking Collapse of the 1920s.” PUB LISHED: Explorations in Econom ic History (July 1994). 94-026A - Michael R. Pakko, Fractional Cointegration with an Application to the Short End of the Yield Curve.” 94-028A - James Bullard and John Duffy, “Using Genetic Algorithms to Model the Evolution o f Hetrogeneous Beliefs.” 92-003A - Daniel L. Thornton, “The Market’s Reaction to Discount Changes” W hat’s Behind the Announcement Effect?” 94-029A - Donald S. Allen, “W hy Does Inventory Investment Fluctuate So Much During Contractions?” 92-004A - Daniel L. Thornton, “Why Do T-Bill Rates React to Discount Rate Changes?” PUBLISHED: Jou rn al o f Money, Credit, and Banking (November 1994), pp 839-50. Donald S. Allen, “A M ulti-Sector Inventory M odel.” FORTHCOMING: Jou rn al o f Econom ic B ehavior and Organization. 92-005A - John A. Tatom, Heinz Gluck and Dieter Proske, “Monetary and Exchange Rate 94-030A - Michael Ulan, William Dewald Policy in Austria: An Early Example of Policy Coordination.” PUBLISHED: Econom ic Policy Coordination in an Integrating Europe. Bank of Finland, 1992. and James Bullard, “U.S. Official Forecasts of Group of Seven Econom ic Performance, 1976-90.” PUBLISHED: this Review (March/April 1995). 92-006A - John A. Tatom, “Currency Appreciation and ‘Deindustrialization’: A European Perspective.” 9 4 -0 3 1 A - Peter S. Yoo, “The Baby Boom and International Capital Flows.” 94-032B - Daniel Thornton, “The Information Content of Discount Rate Announcements: W hat’s Behind the Announcement Effect?” 9 2 -0 0 7A - David C. Wheelock, “Government Policy and Banking Instability: Overbanking in the 1920s.” PUBLISHED: Jou rn al o f Econom ic History (December 1993). (Published as “Government Policy and Banking Market Structure in the 1920s.”) 40 REVIE W SEPTEMBER/OCTOBER 9 2 - 0 0 8 A - Michael T. Belongia and Dallas S. Batten, “Selecting an Intermediate Target Variable for Monetary Policy W hen the Goal is Price Stability.” 1991 W O R KIN G PAPERS 91 -0 0 I D - Mark D. Flood, “Market Structure and Inefficiency in the Foreign Exchange Market.” PUBLISHED: Jou rnal o f International Money and Finance (April 1994). 91 -0 0 2 A - James B. Bullard and Alison Butler, “Nonlinearity and Chaos in Economic Models: Implications for Policy Decisions.” PUBLISHED: Econom ic Jou rn al (July 1993). 91 -0 0 3 A - James B. Bullard, “Collapsing Exchange Rate Regimes: A Reinterpretation.” 91 -0 0 4 A - James B. Bullard, “Learning Equilibria.” PUBLISHED: Jou rn al o f Economic Theory (December 1994). 91 -0 0 5 B - David C. Wheelock and Subal C. Kumbhaker, “W hich Banks Choose Deposit Insurance? Evidence of Adverse Selection and Moral Hazard in a Voluntary Insurance System.” PUBLISHED: Jou rn al o f Money, Credit, and Banking (February 1995). 91 -0 0 6 A - David C. Wheelock, “Regulation and Bank Failures: New Evidence from the Agricultural Collapse of the 1920s.” PUBLISHED: Jou rn al o f Econom ic H istory (D ecem b er 1 9 9 2 ) . 41 REVIE W SEPTEMBER/OCTOBER 1995 COM ING IN THE NEXT ISSUE OF REVIEW: “Antitrust Issues in Payment Systems N etw o rks” (1 9 9 5 spring symposium) A n titru s t a n d P a ym e n t Technologies D ennis W . C arlton a n d A la n S . F rank el Discussants: Ja m e s S. M cA n d re w s N icholas Econom ides S h a re d A T M N e tw o rk s — Th e A n titru s t D im e nsion D o n a ld I. B a k e r P a ym e n t System s a n d A n titru s t: Can the O p p o rtu n itie s fo r N e tw o rk C o m p e titio n Be R e co g n ize d ? D a v id A . B alto u is 42 A ll nonproprietary and nonconfidential data and programs for the articles published in Review are available to our readers and can be obtained from three sources: l.F R E D (F e d e ra l R e se rve Economic 2 . Research D e p a rtm e n t, Federal D a ta ), an electronic b u lle tin b o a rd service. You can access FRED by dialing (3 1 4 ) 6 2 1 -1 8 2 4 through a modem-equipped personal computer. For a free brochure on FRED, please call (3 1 4 ) 4 4 4 -8809. Reserve Bank of St. Louis, Post Office Box 4 42, St. Louis, MO 63166 -0 4 4 2 . 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. 3 . In te r-u n iv e rs ity C onsortium lo r Political and Social Research (ICPSR ). Member institutions can request data through the CDNet Order facility. Nonmembers should write to: ICPSR, Institute for Social Research, P.O. Box 1248 Ann Arbor, Michigan 4 8 1 0 6 , or call (3 1 3 ) 763-5010. F e d e r a l R e s e r v e B a n k o f S t . L o u is Post O ffic e B ox 4 4 2 St. Louis, M is s o u ri 6 3 1 6 6 - 0 4 4 2 recycled paper