The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Vol. 28, No. 1 ECONOMIC REVIEW 1992 Quarter 1 Recent Behavior of Velocity: Alternative Measures of Money 2 by John B. Carlson and Susan M. Byrne Commodity Prices and P-Star 11 by Jeffrey J. Hallman and Edward J. Bryden The Causes and Consequences of Structural Changes in U.S. Labor Markets: A Review by Randall W. Eberts and Erica L. Groshen FEDERAL RESERVE BANK OF CLEVELAND 18 ECONOMI C REVI EW 1992 Quarterl Vol. 28, No. 1 Recent Behavior of Velocity: Alternative Measures of Money 2 by John B. Carlson and Susan M. Byrne Changes in the structure of the U.S. financial industry over the last dec ade have raised questions about the reliability of M2 as the primary guide for monetary policy. Although the simple ratio of economic activ ity to M2 — that is, M2 velocity — indicates nothing unusual, the rela tionship between velocity and interest rates has been disrupted in recent years. This appears to be related to a breakdown in money demand in 1988, which could in turn be linked to the restructuring of depositories. The authors examine the velocities of two alternative monetary aggre gates, but find that, like M2, these measures are not impervious to finan cial change. Commodity Prices and P-Star 11 by Jeffrey J. Hallman and Edward J. Bryden The P-Star (P*) model forecasts inflation by exploiting the stability of M2 velocity and the tendency of the real economy to operate near its potential. For a given stock of M2, P* is the price level that would pre vail if velocity were at its mean and real income equaled potential out put. The ratio of the actual price level (P) to P* can be considered an indicator of how the current money stock will affect inflation over the next several years. Over shorter horizons, other factors may be ex pected to influence the inflation rate. This paper shows how the P* model can be modified to include information about the recent behav ior of commodity prices. This modified model yields more accurate short-run inflation forecasts while still retaining the property that, over longer horizons, only money matters. The Causes and Consequences of Structural Changes in U.S. Labor Markets: A Review 18 by Randall W. Eberts and Erica L. Groshen During the initial stages of the expansion of the 1980s, wage growth remained relatively subdued. Even as the economy picked up steam later in the decade, tight labor markets did not drive up wages to the extent that past experience would have suggested. In an effort to find out what was behind this unusual wage restraint, the Federal Reserve Bank of Cleveland held a two-day conference in October 1989 on the causes and consequences of structural changes in U.S. labor markets. This article provides an overview of those proceedings. Economic Review is published quarterly by the Research Depart ment of the Federal Reserve Bank of Cleveland. Copies of the Review are available through our Public Affairs and Bank Relations Depart ment, 1-800-543-3489. Coordinating Economist: James B. Thomson Advisory Board: David Altig Erica L. Groshen William P. Osterberg Editors: Tess Ferg Robin Ratliff Design: Michael Galka Typography: Liz Hanna Opinions stated in Economic Review are those of the authors and not necessarily those of the Federal Reserve Bank of Cleveland or of the Board of Governors of the Federal Reserve System. Material may be reprinted provided that the source is credited. Please send copies of reprinted material to the editors. ISSN 0013-0281 Recent Behavior of Velocity: Alternative Measures of Money by John B. Carlson and Susan M. Byrne Introduction The unusual weakness of the M2 monetary aggre gate over the past year or so has raised concerns about implications for the economy and has brought into question the reliability of this meas ure as a guide for policy. These concerns height ened last summer as initial reports indicated that M2 declined in the third quarter, leaving it around the lower bound of its target range. His torically, such sharp slowdowns in money growth have been associated with subsequent weakness in economic activity.1 By contrast, growth in the narrower M l meas ure has been robust, having increased almost 9 percent in 1991. Growth in the monetary base has also been strong, driven to some extent by the transitory foreign demand for U .S . currency during the Gulf War. Moreover, Poole (1991) and Motley (1988) have proposed alternative measures of money that suggest monetary policy is not as stringent as it might appear. 1 Although revisions to the data revealed that initial reports under stated M2 growth during the summer and for the year, the revised figures were still inexplicably sluggish. We recognize, however, that the associa tion between money growth and economic activity does not imply causal http://fraser.stlouisfed.org/ ity. For a discussion of this issue, see Carlstrom and Gamber (1990). John B. Carlson is an economist and Susan M. Byrne is a senior research assistant at the Federal Reserve Bank of Cleveland. The authors gratefully acknowledge helpful discussions with David Altig, W illiam Gavin, William Poole, and James Thomson. From time to time, policymakers reexamine the potential usefulness of alternative measures of money as policy guides. Indeed, in the mid1980s, the Federal Open Market Committee (FOMC) abandoned M l as its primary policy tar get in favor of M2. One basis for forsaking M l is most clearly evident in the marked change in the historical pattern of its velocity, defined as the ratio of nominal income to M l (see figure 1). Over much of the postwar period, M l velocity increased steadily along a trend rate of 3 percent. In the early 1980s, however, this measure became substantially more variable with no clear trend. The dismption in the historical pattern was attrib uted largely to financial innovation in conjunction with deregulation and disinflation.2 M2 velocity, on the other hand, appeared to be unaffected by these events. Although the measure has always varied systematically with interest rates, it is essentially trendless both before and after the early 1980s (see figure 1). In fact, since the founding of the Federal Reserve in 1913, M2 ■ Federal Reserve Bank of St. Louis ■ 2 For a discussion of the breakdown of M1 velocity and its implica tions for monetary targeting, see Poole (1988). For a brief summary of the effects of financial innovation, deregulation, and disinflation on M1 and its velocity, see Judd and Scadding (1982) and Carlson (1989). 3 FI GURE 1 been trending up while its opportunity cost has been falling. Yet, history suggests that velocity should be declining, at least in the short mn. Un derstanding this anomaly is, of course, important for interpreting the aggregate’s recent weakness. This article examines the factors that some analysts believe account for the unusual behav ior of M2 and its velocity. We also discuss alter native measures of money recently proposed by Poole, and estimate a demand function for an expanded measure of M2. The analysis suggests that part of the anomalous behavior of M2 veloc ity is related to the ongoing restmcturing of the savings and loan (thrift) industry. Although the implications for the long m n are unclear, M2 velocity is likely to remain higher over the near term than one might otherwise expect given the level of its opportunity cost. M1 and M2 Velocity Ratio SOURCE: Board of Governors of the Federal Reserve System. F I G U R E 2 M2 Velocity and Opportunity Cost Ratio I. Velocity and the Demand for Money Percent SOI JRCE: Board of Governors of the Federal Reserve System. and nominal income have grown at approxi mately the same rate, suggesting the existence of a relatively simple and enduring relationship between the two. On the surface, there is little basis for believ ing that M2 velocity has behaved unusually over the past few years; after all, it is currently close to its trendless long-run average. What is unusual, however, is that around 1989, an appar ent break occurred in the relationship between M2 velocity and the opportunity cost of the aggre gate, defined as the difference between the mar ket interest rate and the rates paid on M2 instru ments (see figure 2).3 Since then, M2 velocity has ■ 3 More precisely, the interest rate paid on M2 instruments is the weighted average of the component rates, where the weights are relative shares. The market rate is the weighted average of yields on the threemonth Treasury bill and the three-year Treasury note, with weights being shares of both the non-tim e deposit (zero-maturity) and small time de posit components of M2. For further discussion of this opportunity cost measure, see Carlson and Parrott (1991). http://fraser.stlouisfed.org/ Federal Reserve Bank of St. Louis The concept of velocity is central to discussions of monetary policy7, largely because it affords the Federal Reserve a straightforward and relatively nontechnical language that Congress and the pub lic can easily understand.4 The fact that M2 veloc ity has been trendless makes it easy to convey to the public why the M2 aggregate might te viewed as a reliable guide over the long term. As the simple ratio of income to money, however, velocity em bodies some complex structural relationships. Economic explanations for the behavior of velocity have generally focused on the existence of a “stable” money demand function. The notion of stability typically implies that this function should have relatively few arguments, and that it should include some measure of spending or economic activity (see Friedman [1956] ).5 If the function were to require knowledge about a large number of variables in order to pin it down, the simple relationship between money and eco nomic activity would be less predictable. For many years, economists were confident that reasonably stable money demand specifica tions could be estimated for narrow definitions of money. Many specifications were based on the inventory-theoretic models of Baumol (1952) ■ 4 Since 1978, the FOMC has been required by law to report to Congress on its annual monetary objectives. The Committee's progress is reviewed at midyear and again at the beginning of the following year, when the next set of objectives is reported. ■ 5 For a recent comprehensive survey of the empirical literature spawned by Friedman’s restatement, see Judd and Scadding (1982). and Tobin (1956). One theoretical result of such models was that the income elasticity of cash bal ances is less than one, implying that the velocity of money would rise secularly. This seemed to square with estimates of in come elasticity associated with conventional specifications of M l demand. Estimates of inter est elasticity, however, were much lower than the theoretical models predicted.6 Although Hoffman and Rasche (1989) recently obtained more substantial estimates of the long-run inter est elasticity of M l, it is doubtful that stable short-run specifications for M l demand exist. The evidence suggests that changes in the struc ture of the financial industry have affected M l demand in too many ways to pin down. Because M2 velocity appeared to be imper vious to the financial changes of the 1970s and early 1980s, attention turned toward finding stable short-mn specifications for M2 demand. One of the most promising was developed by Moore, Porter, and Small (1990), hereafter referred to as MPS. They applied econometric techniques that enabled them to take account of the stationarity of M2 velocity and found that a reasonably stable M2 demand specification could be estimated, at least through 1988. Since then, however, their model has overpredicted M2 growth, raising questions about whether M2 velocity has drifted upward. MPS specified M2 demand in an errorcorrection framework, noting two advantages to this approach.s First, error-correction regressors — entered as first differences in the levels — are more likely to be stationary and are much less colinear than they would be as undifferenced regres sors. Second, the long- and short-run money demand relationships are clearly distinguished. The long-run money demand function is specified as cost.9 Since one might expect M2 opportunity cost to be essentially trendless in the long run, M2 velocity would also be trendless. Thus, although equation (1) may specify a partial equilibrium for the level of M2, the implied long-ain general equi librium for its velocity is essentially a constant. The second component of the MPS model is a dynamic equation based on an error-correction adjustment specification: u = a + bet_ j + w V + 1 di As,~i + X / A y , i=0 /= 0 q n + X X g ijA x i,t-j+en i= 17=0 where et_ j is the deviation of money from its long-run equilibrium value (derived from equa tion [1]) and £t is white noise. Adjustment speed is determined by changes in the lagged values of M2 and in the current and lagged values of the opportunity cost and scale variable. The general form of the model allows other variables to affect adjustment speed (both current and lagged val ues). These variables, which need not affect longRin equilibrium money balances, include any factors that influence the adjustment process.10 Equation (2) essentially specifies the short-run convergence process of M2 to its equilibrium val ue. When the coefficient on the error-correction terni is negative, convergence is ensured. Substi tuting (1) into (2) yields (3) A mt = a - b a - b $st_ x+ b (m t_ l - yt_ j) 11 +E (1) V ci Am t~i i= 16 mt - a + yt + ps, + et , + X i= 1 + S di Ast~i /=0 w where mt - log (M2), yt - log (n o m in al G N P), and 6'= log (opportunity cost). Note that the unitary' coefficient on nominal GNP ensures that equation (1) also specifies a relationship in which long-Rin velocity varies only with opportunity7 ■ c; Amt_ i= 1 q fi ,+ X n X g i j A x i t _ j + e, i= 1 j =o ■ 6 For a possible explanation of this discrepancy, see Poole (1988). 7 Although Hendry and Ericsson (1990) have found “stable” speci fications, these generally include many explanatory variables and hence are not convincingly useful for predicting changes in the simple link be tween M1 and economic activity. 9 MPS include a time index as a regressor to estimate any drift in M2 velocity directly. Although they find the coefficient to be significant, the drift is negligible (around 0.03 percent per year). ■ ■ 8 Earlier advocates of this framework include Baba, Hendry, and Starr (1988). ■ 10 MPS also specify a set of error-correction models for determin ing interest rates paid on the components of M2. They find that many bank deposit rates adjust relatively slowly to changes in money market in terest rates. However, because their specifications are not very durable, we w ill focus only on the demand for M2 given the opportunity cost, not on how the opportunity cost is determined. hmqqh II. The MPS Specification and Thrift Restructuring CP Specification The CP specification and estimated coefficients are A mt= -.053 - .0095,_ j - .138 ( mt_ , - y ,_ x) (4.44) (4.60) (5.13) + ,245Aw/_ j - .007As, - .007As,_ j (3.08) (3.32) (3.39) + .186Ac,+ .241A xt_ ! + .031REGDUM + e, (2.87) (3.30) (7.38) Adj. R 2 = .74; SEE = .0040; est. period = 1964:IQ to 1986:IVQ, where .s' is a measure of opportunity cost, c is personal con sumption expenditures, x is thrift deposits (including other checkables, money market deposit accounts [MMDAs], savings deposits, small and large time deposits, and term repurchase agreements [RPs]), and REGDUM is a qualitative variable that equals zero in all quarters except 1983:IQ, when it equals one.a Because thrift restructuring has been ongoing since 1988, and because we seek to avoid high influence points given the sub stantial changes in the industry since that time, equation (4) is estimated before the thrift crisis (1964:IQ to 1986:IVQ) and simulated through 1990. All parameters are significant at the 5 percent level or better. a. Following MPS, we present results that approximate log(s) using a firstorder Taylor series expansion w hen the opportunity cost is less than 0.5. W e also estimate the model using the simple log o f opportunity cost. Although the simple measure improves the in-sample fit, out-of-sample simulations are less favorable. Nevertheless, the usefulness of the Taylog transformation remains an open issue, though beyond the scope o f this study. MPS estimate a version of equation (3) over the 1964:IQ to 1986:IIQ period. They find that their specification is relatively stable, despite the advent of both deregulation and, perhaps more signifi cantly, disinflation. Beginning in 1988, however, it begins to overpredict M2 growth. The implications of this overprediction for velocity depend on what parameters of the M2 demand function may be changing. If any of those in the long-run specification (equation [1]) have changed, then M2 velocity will likely fluctuate around a new, higher equilibrium level. If, on the other hand, the error-correction process is misspecified, the divergence between interest rates and velocity could be temporary. Carlson and Parrott (1991), hereafter CP, pro pose a specification of equation (3) that includes the change in thrift deposits as a determinant in the error-correction equation (see box l) .11 They argue that this change is a proxy for depositpricing effects that, though related to the thrift restructuring, are not adequately captured in the measure of opportunity cost. This implicitly as sumes that the effects of restructuring influence the adjustment of M2 to its equilibrium level, but do not affect the equilibrium level itself. These effects may arise when failing thrifts are liquidated and time deposit contracts are abrogated. Because many of these deposits were contracted at rates substantially higher than those paid in recent years, holders of these deposits realize a sharp drop in their returns when contracts are nullified. Since the historical series on time-deposit yields records only the rate paid on new contracts, it understates this recent decline. Thus, the measure of opportu nity cost is inadequate. For holders of abrogated contracts, opportunity cost has increased; in contrast, measured opportunity cost has fallen in recent years. The CP specification is estimated before 1988 to avoid high influence points given the collapse in thrift deposits thereafter. (Out-of-sample sim ulations after 1988 account for most of the short fall evident in the MPS model. ) The results are consistent both with the hypothesis that thrift restructuring has played a major role in the recent M2 weakness, and with the belief that this realign ment will not significantly alter long-mn velocity. It is important to note that the CP specification does not examine the potential for effects on the equilibrium level of velocity. Unfortunately, the data are not of sufficient duration to discriminate convincingly between long- and short-run effects. Nevertheless, the depository restructuring hypoth esis is consistent with previous anomalies in the relationship between interest rates and velocity. For example, M2 velocity appeared to be un usually low in the mid-1980s, given the level of its opportunity cost (see figure 2). Soon after the advent of deregulation, many analysts specu lated that M2 velocity would shift downward.12 It was believed that deregulation left the deposi tories in a better position to compete for funds to ■ 11 For an alternative approach, see Duca (1991). ■ 12 See Hallman, Porter, and Small (1989). H F I G U R E 3 MZM and M2E Billions of dollars (ratio scale) NOTE: M2E equals M2 plus institution-only MMMFs. MZM equals M2E minus small time deposits. Shaded areas represent recessions. Estimated trough date for 1990-91 recession is 1991 :IIQ. SOURCE: Board of Governors o f the Federal Reserve System. expand their market share of credit; hence, many argued that deposits included in M2 would in crease as a share of the nation’s portfolio. This in aim implied that M2 velocity would fall. III. Alternative Measures of Money Historically, when money demand specifica tions have broken down, analysts have found that the problem is often reflective of the partic ular definition of money being used. Over time, financial innovations occur, resulting in new in struments that have properties similar to more than one asset. For example, money market mutual funds (MMMFs), first offered in the early 1970s, have characteristics of both transactions deposits and mutual funds. Moreover, when regulations change, such as the elimination of Regulation Q, the range of assets for which deposits are substitutable can be substantially affected. Hence, financial innovation and dereg ulation can blur the functional distinctions be tween the monetary aggregates. Poole ( 1991) recently identified three func tional components of M2: 1) traditional trans actions balances (currency plus checkable deposits) that are defined as M l, 2) savings bal ances that can be converted without notice into transactions balances (such as MMMFs and state ment savings accounts at banks), and 3) small time deposits (defined as certificates of deposit denominated in amounts of less than $100,000) that can be converted into transactions balances (without penalty) only upon maturity. Although M2 has served well until recently, Poole questions its longer-term durability as the appropriate measure of money. He proposes two alternative aggregates. The first, based on a principle advanced by Friedman and Schwartz (1970), views money as a “temporary abode of purchasing power.” To satisfy this principle, Poole advocates including all instruments avail able with zero maturity. Thus, he would broaden M l to include all savings balances that can be immediately converted into transactions balances (hereafter called MZM).13 Poole also advocates expanding the M2 meas ure (M2E hereafter) to include MMMFs available to institutions only. He notes that these instal ments allow institutions to earn interest on check able accounts in the face of the long-standing and still-effective prohibition of interest payments on demand deposits. The time series of the two meas ures are illustrated in figure 3Prior to 1980, MZM velocity seemed to be trending up, although at a slower rate than that of M l. Since 1983, however, MZM’s velocity has appeared to be the more stable of the two series ■ 13 Although Motley (1988) proposed a measure of zero-maturity instruments, the logical antecedent to this measure is Friedman and Schwartz's M2 aggregate, which consists of all commercial bank deposits (demand plus time and savings). FI GURE 4 M1 and MZM Velocity Ratio SOURCE: Board of Governors o f the Federal Reserve System. FI GURE 5 M2E Velocity and Opportunity Cost Ratio 1.80 1964 1969 1974 1979 1984 1989 SOURCE: Board of Governors o f the Federal Reserve System. (see figure 4).14 Poole recognizes that the stabil ity of MZM velocity (manifest only since dereg ulation) does not provide a sufficient empirical basis for choosing this aggregate over the broader alternatives. Nevertheless, he prefers it because, as a comprehensive measure of assets that serve as a temporary abode of purchasing power, MZM should be durably linked to spend ing. Moreover, he essentially argues that the trend in MZM velocity prior to 1980 was largely a consequence of Regulation Q, which distorted the competition between time deposits and non regulated depository assets. Without Regulation Q, banks have much less incentive for developing regulatory avoidance schemes, such as automatic transfer accounts, that distort the relationship between measured transac tions deposits and spending.1’ Also, it seems rea ■ 14 Because there is no empirical basis for assuming that MZM velocity has been stable, we do not estimate a demand function below. sonable to assume that without interest-rate regulation, banks will treat small time deposits much more like managed liabilities, enabling them to compete more directly in capital markets. Thus, the volume of small CDs will be more indicative of changes in the competitive positions of depositories than of monetary conditions. It also seems less likely that nondepository competitors w ill have the same incentives to in vest in financial innovations that seek to compete directly with depository savings instruments. For example, the explosive growth of MMMFs was due in large part to the inability of depositories to compete for funds on the same footing with liquid instmments offering market rates of return. In the absence of binding constraints, it is unlikely that we will see the same burst of financial innovation as occurred under Regula tion Q; hence, one might expect a more stable link between zero-maturity instruments and economic activity. Since it remains to be seen whether the principle guiding the choice of MZM will lead to an empirically more robust measure of money, however, Poole recom mends that M2 and MZM be given equal weight in policy deliberations. The velocity of M2F appears to have charac teristics that suggest its relationship to the econ omy is less disrupted by regulatory change than that of M2 (see figure 5). Indeed, M2E velocity has been falling in recent years roughly com mensurate with the decline in opportunity cost. This more consistent pattern suggests that over the whole sample period, the demand for M2E has been relatively more stable than the demand for M2. Nevertheless, before it returned to a more consistent relationship with interest rates, M2E velocity was still unusually low over most of the 1980s. As suggested above, this could reflect the unsustainable attempt by depositories to in crease their market share once they were freed from the regulatory constraints that limited the types of loans they could make. Perhaps the best example of this was in the thrift industry. By the early 1980s, the rising cost of funds, reflecting accelerating inflation, had left many thrifts that were holding relatively low-yielding mortgages insolvent. Kane (1989, p. 4) argues that, with nothing to lose, these “zombie” institu tions attempted "to grow out of their problems by undertaking long-shot lending and funding activities” that essentially renewed and ex- ■ 15 Although the existence of reserve requirements on transactions deposits leaves some incentive intact, the effects of most potential avoidance schemes would probably be internalized in zero-maturity assets. FI GURE 6 first step in resolving the insurance crisis, and to close zombie thrifts. The weakness in deposit growth since 1988 is to some extent an unw ind ing of the unsustainable depository share of credit markets. Nontransactions Deposits Billions of dollars 3,000 IV. The Demand for M2E 2,000 1,000 1986 1987 1988 1989 1990 1991 1992 SOURCE: Board o f Governors of the Federal Reserve System. FI GURE 7 Simulated and Actual M2E: Based on Equation (4) Billions of dollars 4,000 Simulated^’ " 3,600 We estimate two variations of the velocity speci fication (equation [31) using the M2E measure.17 The first regression includes a temporary inter cept shift variable embodying the hypothesis that the unsustainable expansion of depositories affected equilibrium velocity in the 1980s. It pre sumes that the overextension of depository intermediation was financed largely by time deposits, which are closer substitutes for capital market instalments than are money market se curities. This hypothesis would explain why a large part of the ainoff of nontransactions de posits at thrifts did not find its way back to other depositories, but was instead transferred to non depository investment vehicles (see figure 6). The first specification does not include the thrift-change variable proposed by CP. The esti mated coefficients are ^^^^Actual (4) 3,200 - A mt= -.076 - .012s(_ j - . 1 (5.13) (5.25) 8 9 yt_ j) (5.62) 2,800 + ,421Aw/_ 1 - .008As,- ,005As/_ 1 2,400 1985 1 1 1 » -* 1 » 1 « 1 -i-l ■.l i i i 1 i i » 1 i i i_ 1986 1987 1988 1989 1990 1991 SOURCE: Authors’ calculations. panded the lost bets of the past. To finance this expansion, thrifts offered a premium on depos its, leading to a sharp increase in the depository component of M2E (and M2) relative to income, thereby decreasing velocity.16 With the under standing that such instruments were federally guaranteed, depositors were all too willing to provide the funds. As the decade unfolded, however, it became clear that this strategy was not sustainable. Beginning in 1989, Congress and the Bush administration officially recognized the insolvency of both a large portion of the savings and loan industry and the thrift deposit insurance fund. In August of that year, they allocated funds as the ■ 16 As CP note, such a premium is not adequately incorporated in measured yields. Thus, measured M2E opportunity cost probably over states true opportunity cost. This would explain why M2 velocity appears to be low relative to its measured opportunity cost. http://fraser.stlouisfed.org/ Federal Reserve Bank of St. Louis (6.91) (4.27) (1.92) 1992 + ,292Ac/+ .005DBUDUMt_1 (4.27) (2.81) + .02AREGDUM + e, ( 5 -6 1 ) Adj. R 2 = .72; SEE = .0043; est. period = 1964:IQ to 1989:IVQ, where 5 is a measure of M2E opportunity cost, c Ls personal consumption expenditures, DBUDUM is the temporary intercept shift variable, and REGDUM is a qualitative variable accounting ■ 17 Although Hoffman and Rasche (1989) find a stable long-run relationship between real M 1, interest rates, and real income, they ques tion the existence of a stable short-run specification for M1 demand. Hendry and Ericsson (1990) do find stable short-run specifications for the narrow measure, but raise a number of issues that are beyond the scope of this paper. We focus on the short-run demand for M2E, which has the virtue of a trendless velocity over the past 30 years. 9 F I G U R E where x denotes thrift institution deposits (includ ing other checkables, MMDAs, savings deposits, small and large time deposits, and term RPs). The coefficient on the thrift proxy is statistically signifi cant, but somewhat smaller than in the CP specifi cation. This suggests that depository restructuring is an important and continuing factor, at least in the short tun. Out-of-sample simulations of M2E demand tend to underpredict M2E over most of the past three years, but the bias has been small (see figure 8). Thus, although not immune to the structural change, the measure would seem to warrant a closer look. 8 Simulated and Actual M2E: Based on Equation (5) Billions of dollars V. Conclusion SOURCE: Authors' calculations. for the introduction of nationwide N OW accounts. DBUDUM and REGDUM equal zero in all periods except 1981 :IVQ- 1988:IIQ and 1983:IQ, respectively, when they equal one. The estimated coefficient on DBUDUM is positive and statistically significant. This is con sistent with the hypothesis that equilibrium ve locity was temporarily low in the 1980s. Though the model has reasonably good in-sample prop erties, out-of-sample simulations indicate that it overpredicts M2E growth in 1991 (see figure 7). The 1991:HIQ drop in M2E (and the sharp rise in its velocity) is greater than two standard deviations of its predicted value based on insample experience.18 One explanation for the shortfall in M2 is that the savings and loan restructuring peaked in the summer months of 1991- Thus, the sec ond regression extends equation (4) to include the change in thrift deposits as a regressor in the error-correction specification: (5) Am, = -.079 - .O lls ,^ - ■ 194(mt_ i -yt_ l) (5.64) (5.07) (6.10) + ,271Am,_1 - .008As, - .OOôAs,^ (3.77) (4.05) (2.56) + .240Ac, + .004DBUDUM t_ l (3.62) (3.18) Changes in the structure of the U.S. financial in dustry have justifiably brought into question the reliability of M2 as a guide for monetary policy. The aggregate’s appeal as an intermediate pol icy guide has been largely due to its relatively stable and simple relationship with income and interest rates. Over most of the past 30 years, this stability was manifest in the behavior of M2 velocity, which, though influenced by interest rates, ultimately reverted to a trendless mean. Although M2 velocity, by itself, indicates noth ing unusual, its relationship with interest rates has been disrupted in the last few years. This appears to be related to a breakdown in M2 demand after 1988, which probably reflects to some extent the restructuring of depositories. We examine the velocities of two alternative measures of money: MZM and M2E. O f these, M2E holds the most promise, because its veloc ity appears to be least affected by the events of recent years. Moreover, velocity specifications of money demand seem to be more durable for the M2E measure than for M2. Nonetheless, we must stress the tentative nature of any conclusions based on the analysis above. Unfortunately, money demand theory has not ad vanced to a state in which empirical hypotheses are sharply defined and testable. This perhaps re flects the tension arising from the idea that if money demand is to be useful for policy, it should have relatively few detenuinants. O n the other hand, as Judd and Scadding (1982) note, the fundamental source of the insta bility of money demand has been the excessive growth in money. They argue that the failure of + .027REGDUM + •180A *,_•, + £, (6.37) (3.31) Adj. R 2 = .75; SEE = .0041; est. period = 1964:IQ http://fraser.stlouisfed.org/ to 1989:IVQ, Federal Reserve Bank of St. Louis ■ 18 We recognize that statistical tests comparing M2 and M2E may not be very meaningful. However, from a monetary targeting point of view, it is much more persuasive if one can demonstrate an empirical basis for believing that the velocity of the targeted aggregate is relatively stationary. monetary policy to restrain inflation led to the high market interest rates that, in combination with regulatory restraints, induced much of the financial innovation disrupting the relationship between M l and the economy. Similarly, one might argue that rising inflation was the funda mental source of the unsustainable expansion — and ultimate collapse — of the thrift industry. References Baba, Yoshihisa, David F. Hendry, and Ross M. Starr. “U.S. Money Demand, 1960-1984,” Dis cussion Papers in Economics, No. 27, Nuffield College, Oxford, England, January 1988. Baumol, William J. "The Transactions Demand for Cash: An Inventory Theoretic Approach,” Quarterly Jo u rn a l o f Economics, vol. 66 (November 1952), pp. 545 - 56. Carlson, John B. “The Stability of Money Demand, Its Interest Sensitivity, and Some Implications for Money as a Policy Guide,” Federal Reserve Bank of Cleveland, Economic Revieu', vol. 25, no. 3 (1989 Quarter 3), pp. 2 -13_______ , and Sharon E. Parrott. “The Demand for M2, Opportunity Cost, and Financial Change,” Federal Reserve Bank of Cleveland, Economic Review, vol. 27, no. 2 (1991 Quar ter 2), pp. 2-11. Carlstrom, Charles T., and Edward N. Gamber. “Does the Fed Cause Christmas?” Federal Reserve Bank of Cleveland, Economic Com mentary, January 1, 1990. Duca, John V. “The Case of the ‘Missing M2’,” Federal Reserve Bank of Dallas, mimeo, December 1991. Friedman, Milton. “The Quantity Theory of Money — A Restatement,” in Milton Friedman, ed., Studies in the Quantity Theory’ o f Money. Chicago: University of Chicago Press, 1956, pp. 3-21. ______ , and Anna Schwartz. Monetary Statistics o f the United States: Estimates, Sources, Methods. New York: National Bureau of Economic Research, 1970. Hallman, Jeffrey J., Richard D. Porter, and David H. Small. “M2 per Unit of Potential GNP as an Anchor for the Price Level,” Board of Gover nors of the Federal Reserve System, Staff Studies 157, April 1989Hendry, David F., and Neil R. Ericsson. “Model ing the Demand for Narrow Money in the United Kingdom and the United States,” Board of Governors of the Federal Reserve System, International Finance Discussion Papers No. 383, July 1990. Hoffman, Dennis, and Robert H. Rasche. “LongRun Income and Interest Elasticities of Money Demand in the United States,” National Bureau of Economic Research, Working Paper No. 2949, April 1989. Judd, John P., and John L. Scadding. “The Search for a Stable Money Demand Function: A Sur vey of the Post-1973 Literature,”Journal o f Economic Literature, vol. 20, no. 3 (Septem ber 1982), pp. 993-1023. Kane, Edward J. The S&L Insurance Mess: How D id It Happen? Washington, D.C.: The Urban Institute, 1989. Moore, George R., Richard D. Porter, and David H. Small. “Modeling the Disaggregated Demands for M2 and Ml: The U.S. Experience in the 1980s,” in Peter Hooper et al., eds., Financial Sectors in Open Economies: Empirical Analysis an d Policy Issues. Washington, D.C.: Board of Governors of the Federal Reserve System, 1990, pp. 21-105. Motley, Brian. “Should M2 Be Redefined?” Federal Reserve Bank of San Francisco, Economic Re view, Winter 1988, pp. 33 - 51. Poole, William. “Monetary Policy Lessons of Recent Inflation and Disinflation,” Jo u rn a l o f Economic Perspectives, vol. 2 (Summer 1988), pp. 73-100. ______ . Statement before the Subcommittee on Domestic Monetary Policy of the Committee on Banking, Finance, and Urban Affairs, U.S. House of Representatives, November 6, 1991. Tobin, James. “The Interest-Elasticity of Transac tions Demand for Cash,” Review o f Economics a n d Statistics, vol. 38, no. 3 (August 1956), pp. 241-47. Commodity Prices and P-Star by Jeffrey J. Hallman and Edward J. Bryden Introduction A recent article by Hallman, Porter, and Small (1991), henceforth referred to as HPS, presented the P-Star (P*) indicator of future inflation. The HPS models exploit the stability of two long-run relationships: that between M2 and nominal out put, and that between actual and potential real output. Despite paying no attention to other possible influences on inflation, such as com modity prices or interest rates, the HPS models produced better forecasts of the GNP implicit price deflator over the 1970s and 1980s than did a number of alternatives, including both univar iate ARIMA models and the published forecasts of several econometric consulting firms. Most economists believe the quantity theory relationship that underlies the P* model holds only in the long mn, if at all. In the short run, standard economic theory predicts that any fac tors affecting aggregate supply or aggregate demand may also affect the price level. Com modity prices may give early indications of shifts in either supply or demand; if so, augmenting the P* models to include information about the recent behavior of commodity prices may be expected to improve both fit and forecasting perfomiance, Jeffrey J. Hallman Is an economist and Edward J. Bryden is a research assistant at the Federal Reserve Bank of Cleveland. particularly over short horizons. This paper shows that these results can indeed occur. After briefly reviewing the P* idea, we show how the P* approach can be extended to incor porate commodity price data. Two notable results are obtained: First, the resulting models outperform the HPS models in fit and forecast ing ability, although the improvement is not large. Second, the relative significance of the P* and commodity price terms depends on the sampling frequency of the data. The two terms are equally significant in explaining inflation at quarterly and annual frequencies, but inflation in the biennial version of the model depends exclusively on the monetary (P*) term. This ac cords well with the orthodox view that, while commodity market developments may signifi cantly influence inflation in the short run, in the long run only money matters. I. The P* Model The HPS work was motivated by the observa tion that, rather than trending up or dow n over the past 35 years, V (the velocity of M2, defined as GNP divided by M2) has simply fluctuated around its average value of 1.65 (V*). During Ej FI GURE 1 M2 Velocity, Real GNP, and Potential Output Ratio the same period, real GNP (Q ) has usually been within 5 or 6 percent of potential output (Q*). Both of these regularities can be seen in figure 1. Using lower-case letters to denote natural loga rithms, HPS found that they could not reject the hypothesis that the velocity gap (v- v *) and output gap (q * - q) are covariance stationary.1 A stationary series has a mean to which it tends to return infinitely often. More important, as the forecasting horizon lengthens, the optimal fore cast of a stationary series tends to approach the series mean.2 The relation between real output, the price level as measured by the implicit price deflator (P), money, and velocity is given by the quan tity equation: Billions of dollars (1) SOURCE: Board o f Governors o f the Federal Reserve System. F I G U R E 2 p + q - m2 + v. As the velocity and output gaps are stationary, so too is the price gap (p - p *), where P* is defined as M2* X V* / Q*. The reason is that the price gap is simply the sum of the velocity and output gaps. Since V* is constant, the economic interpretation of stationarity of the price gap is that, given M2 and potential output, the price level that can be supported in the long a m is P*. Historically, inflation has usually accelerated when P* has exceeded P, and slowed when the reverse was true. Letting rc denote the rate of change of the implicit price deflator, HPS used a quarterly model of the form Long-Run Aggregate Supply and Demand 4 (2 ) A n t= a ( p [_ 1- p * _ l) + Y J P /A tc,_ y 7=1 Price level to exploit this regularity in forecasting inflation. The lagged Arc terms in equation (2) reflect the fact that quarterly changes in the inflation rate are often partly reversed in subsequent quarters, perhaps due to measurement error. Lagged Arc terms are unnecessary when the model is esti mated using annual data. The model then be comes (3) Ant = a ( p l_ 1-p*_1), where the subscript t is now an annual index (measured in every fourth quarter) rather than a quarterly index as in equation (2). Estimates of the models in equations (2) and (3), and a bien nial version (one observation from the fourth Output SOURCE: Authors’ calculations. ■ 1 A series yt is covariance stationary if, for all k and j, E{yl+ k ) = m, and cov( yw ,y ,) = c o v (y t+ k , jt y t+ k ) . ■ 2 Optimal here means the forecast that minimizes the expected variance of the forecast error. 13 TABLE 1 Estimates of the Models Coefficients «1 a2 Annual'1 Biennial0 ß4 R2 Standard Error Durbin Watson -.12 (-1.3) .32 1.56 1.96 .38 1.19 2.36 .70 .97 1.50 .36 1.51 1.97 .56 1.00 2.34 .84 .77 2.25 Price Gap Models -.64 (-7.4) -.145 (-4.2) -.214 (-4.2) -.47 (-4.7) -.27 (-2.7) -.31 (-5.6) Price Gap Models with Commodity Price Inflation Frequency Quarterlya -.107 (-3.0) Annualb -.171 (-3.5) -.37 (-7.4) Biennial0 ß3 Pi Frequency Quarterly3 Regression Statistics -.039 (-2.9) -.061 (-3.5) -.015 (-1.8) -.70 (-8.1) -.30 -.55 (-5.5) -.34 (-3.5) -.15 (-1.8) (-2.3) a. Estimated from 1960:IIQ to 1990:IIQ; 121 observations. b. Estimated from 1961 to 1990; 30 observations. c. Estimated from 1962 to 1990; 15 observations. NOTE: T-statistics are indicated in parentheses. SOURCE: Authors’ calculations. quarter of every other year) of equation (3) are given in table 1. The economic intuition underlying the P* model can be illustrated using the long-run ag gregate supply and long-run aggregate demand diagram in figure 2. Stationarity of the velocity and output gaps indicates the shape of the curves. The stationarity of velocity means that V is con stant at V* along the long-run aggregate demand curve. But because money is also held fixed along the curve, long-mn aggregate demand is repre sented as the hyperbola fonned by the locus of points where P X Q = M X V*. The stationarity of the output gap means that the long-mn aggregate supply curve is vertical at Q*. Changes in the money stock shift the longm n aggregate demand curve by proportionate amounts. A 4.5 percent increase in M2, for ex ample, shifts the demand curve up (or equiva lently, to the right) by 4.5 percent. That is, the shifted curve is the locus of (P, Q) pairs whose product is now 4.5 percent greater than before the shift. Changes in potential output are drawn as shifts in aggregate supply to the left or right. A 2.5 percent increase in Q* shifts the vertical line 2.5 percent to the right. In the example shown, the result of both shifts taken together is a 2 percent rise in prices accompanied by a 2.5 percent increase in real output. Although the above analysis shows how money and potential output determine prices in the long run, it may have less application in the short run. The velocity and output gaps are sta tionary, but they are not identically zero at all times. Short-mn aggregate demand may not be a hyperbola, and short-mn aggregate supply may not be vertical. Furthermore, in constructing a measure of potential output, it is customary to restrict it to grow smoothly over time. The best that can be hoped for such a measure is that it will correctly capture the trend in potential output. This catalog of short-run omissions and in adequacies is reason to hope that the P* model can be enhanced. The inclusion of variables that reflect information about the location or movements of the short-run aggregate supply and demand curves may improve the model’s performance. One such candidate for inclusion is a measure of commodity prices. FI GURE 3 Implicit Price Deflator, P*, and the Commodity Price Index Index SOURCES: Board of Governors o f the Federal Reserve System, and the Commodity Research Bureau. FI GURE 4 Inflation Rates for the Implicit Price Deflator and the Commodity Price Index SOURCES: Board o f Governors o f the Federal Reserve System, and the Commodity Research Bureau. II. Commodity Prices and the Deflator Figure 3 plots the logarithms of P, P*, and the Commodity Research Bureau’s (CRB) spot market price index for 22 commodities, labeled p c in the figure. While both p and p c have in creased over the years, a simple long-run rela tionship between their levels is not apparent. Rather, there has been a decline in (pc - p), the log of real commodities prices. There is no reason to expect a simple relation ship to hold between the levels of a commodity price index and the implicit price deflator. The deflator is meant to measure the price of all of the economy’s outputs, while most of the 22 CRB commodities are primarily used as inputs to vari ous production processes. Even if firms price on a pure markup basis, output prices can diverge from commodity prices if the relative prices of other, noncommodity inputs change. One can easily imagine that continuing technological progress in agriculture, for example, will result in a downward trend in the price of wheat relative to the price of labor. Trends in the real prices of commodities are simply trends in the prices of commodities relative to the prices of other goods, so we should not be surprised to find them in the data. While it is easy to see how a trend in (p c- p) might arise from such forces as technological progress, it is much harder to imagine a scenario in which the inflation rates n and nc would trend apart indefinitely. Not only would this imply a trend in the real price of commodities, but it would require a trend in the trend, resulting in explosive real commodity prices. That kind of relative behavior is not evident in the data, so there is an implied long-run relationship be tween the inflation rates of prices in general and commodity prices. The top panel of figure 4 plots quarterly infla tion rates for the GNP deflator and the com modity price index. The commodity inflation index is divided by three to put it on the same scale in the plot. The relationship between the two indices is clearer in the lower panel, where inflation rates are calculated over four quarters, resulting in a smoother plot. Notice that in keep ing with the downward trend in real commodity prices evident in figure 3, the arithmetic mean of the commodity inflation index is lower than the arithmetic mean of GNP inflation. Finally, figure 5 shows that both (p - p *) and (the change in / / over four quarters) foreshadow subsequent changes in the inflation rate. FI GURE 5 III. P* with Commodity Prices Inflation Predictions by the Price Gap and by Commodity Price Changes The figures argue for a specification of the dynamic relationship between p and p c that preserves a long-run relationship in which the commodity inflation rate is less than GNP de flator inflation. Just as the term (pt_ x—p*_ j) in equations (2) and (3) enforces a long-run rela tionship between p and p*, the introduction of a term rn) with a negative coeffi cient will enforce a similar relationship between the inflation rates for P and Pc. Here, m denotes the mean difference of the two inflation rates, so that the real commodity inflation rate (tu4 - n4 ) enters as a deviation about its mean. The result ing equation for the quarterly model is Percent change (4) A n l= a l(pt_ x- p * _ l) + a 2(7iu _ 1 -71 - rn) 4 + X P /A ic ,_ y , 7=1 SOURCES: Board of Governors o f the Federal Reserve System, and the Commodity Research Bureau. and the implied equilibrium has both p = p* and a trend of -rn in the real price of commodities. The annual and biennial versions of this model take the form (5) A n t= a 1(pt_ ]- p * _ 1) + a 2(A4p t_ l- A 4p ct_ l- rn) + p, An,_v T A B L E 2 Significance of the Error-Correction Terms T-statistics Frequency (P ~P*) (n - if ) Quarterly -3.0 -2.9 Annual -3.5 -3.5 Biennial -7.4 -1.8 SOURCE: Authors’ calculations. The biennial version does not require the lagged An term. Estimates for all three versions may be found in table 1. In all of the estimated models, both a coeffi cients are significantly less than zero, and the equations using the price gap model with com modity price inflation fit better than those that do not use the commodity price term. Table 2 shows an interesting comparison of the error-correction t-statistics taken from table 1. The quarterly model predicts inflation over the coming quarter, the annual model over the coming year, and the biennial model over two years. At the low est frequency, the explanatory power of the com modity inflation term is minor; nearly all of the explanatory power of the model comes from the price gap tenn. This result confirms that w hile supply shocks reflected in ( n - if ) affect inflation in the short run, over longer horizons money (reflected in [p - p*}) is more important. If both velocity and real output eventually return to their long-mn values, supply shocks not accommodated Ea TABLE 3 Year-yithead Forecasting Performance Quarterly ( p -P*) Actual Predicted Error 1970 5.17 1971 1972 6.09 4.42 5.41 4.94 1973 1974 8.25 9.96 1975 1976 1985 1986 1987 1988 8.35 5.74 6.82 7.98 8.87 9.88 8.76 5.06 3.64 3.42 2.94 2.58 2.96 4.14 1989 1990 3.73 3.98 Year 1977 1978 1979 1980 1981 1982 1983 1984 1991 Mean forecast error Mean absolute forecast error Root mean squared forecast error 6.39 5.42 8.88 9.69 7.97 Error Predicted Error -0.24 5.52 4.99 6.38 -0.35 1.10 -1.96 7.14 1.15 -1.97 2.82 -1.97 0.70 -2.08 5.05 8.94 3.19 1.02 9.25 7.75 6.08 7.47 7.95 7.90 8.17 -0.91 -2.01 1.09 -1.34 7.49 8.49 6.92 3.70 3.29 3-31 3.44 0.13 -0.38 -0.86 3.79 3.46 -0.83 0.68 - 0.66 0.25 4.39 3.74 Annual (p - P *) and (7t - 7 t C) Predicted -2.23 0.09 0.69 1.05 2.39 0.27 -1.85 -0.06 6.73 7.30 7.82 Annual ( p -P*) 3.03 6.73 3.02 2.94 3.12 3.26 3.75 3.76 4.59 3.80 3.80 0.01 1.001 1.28 0.75 0.51 0.92 1.98 0.59 -1.67 0.62 0.48 -0.18 -0.68 -0.80 0.37 -0.86 0.19 0.11 1.006 1.24 5.39 6.49 7.45 8.10 9.14 7.81 7.52 6.98 8.57 8.10 8.00 6.27 4.28 5.13 3.11 2.60 3.69 4.44 4.68 3.57 3.57 0.79 -1.86 -0.80 -2.07 -0.70 1.00 0.30 1.78 0.76 -1.20 -0.64 -1.71 -0.17 -0.02 -0.74 -0.30 -0.96 0.41 -0.27 0.999 1.15 NOTE: Entries represent the annual growth rate o f the GNP implicit price deflator over fourth-quarter to fourth-quarter periods ending in the indicated year. SOURCE: Authors’ calculations. by the monetary authority will have no long-run effect on the overall price level. In the short run, they can influence the rate at which P converges to P*, but the fit of the biennial equation indi cates that the effect dissipates within two years. Table 3 compares the out-of-sample forecast ing performance of some of these models. The first three columns showr the performance of the model in equation (1), while the remaining columns are for the annual models in equations (3) and (5). The forecasts reported in the table are not truly out of sample for two reasons. First, the estimates of v* and q* used in constructing the p* series are actually based on the full sample of ob servations. However, all of the models use this p* series, so none of them has an unfair advantage. This “cheating” would be important if we were http://fraser.stlouisfed.org/ comparing the perfonnance of P* models with Federal Reserve Bank of St. Louis other models, as was done in HPS.3 Second, the year-ahead forecasts made from the quarterly model cheat in the sense that we assume the values taken by p* over the intervening three quarters were known. This is not true of the year-ahead fore casts for the annual models making up the rest of the table, since all of their right-hand-side variables are lagged at least once. The root mean squared forecast error for the annual price gap model (equation [3]) is slightly smaller than for the quarterly version, showing that the short-run dynamics modeled by the lagged Attterms in the quarterly model are not ■ 3 In their paper, HPS did try estimates of q* and v* based only on information that would have been available to a forecaster operating in real time and found that it made little difference to the forecasting per formance of their models. WÈÈÊÊÊÈM T A B L E 4 f§ Encompassing Tests Independent Variable (right-hand side) Dependent Variable (left-hand side) e (i —eP* e ci eP' eq e ci —e CP e cP e? e cP e p* —e cp T-statistic _ 1.08 0.01 2.23 1.18 2.24 1.63 For 1991, the quarterly and annual price gap models predict inflation rates of 3-1 and 3.8 per cent, respectively. Augmenting the annual model with the commodity price tenn yields a somewhat lower prediction of 3.6 percent. The low predic tion of the quarterly model reflects the unusually slow money growth in the first three quarters of this year. If we assume that P, P*, and Pc all grew at a 2 percent annual rate for the fourth quarter of 1991, the inflation rate for all of 1991 would be 3.2 percent. Both annual models (equations 131 and [51) would then forecast 1992 inflation at 2.1 percent. SOURCE: Authors’ calculations. IV. Conclusion important for forecasting a year ahead. The results for the annual models (equations [3] and [51) show that adding the commodity price tenn yields smaller forecast errors. Another way to compare competing forecasts is to ask whether the forecast error from a given model can be ex plained (encompassed) by the forecast of another model, as elucidated in Chong and Hendiy (1986). To make this comparison, let e ' and e 2 represent the forecast errors made by models 1 and 2. Then examine the t-statistics for the coefficients a 1and or in the regressions e) = v}(e ) - ef), e f = a 2( e f - e j) . If a 1is significantly different from zero, model 2 encompasses model 1, as it contains useful in formation (for forecasting purposes) that is not in model 1. If a 1is significant but a 2 is not, the encompassing is one-way; that is, model 1 is en compassed (by model 2) but is not encompassing. Mcxlel 2 is then clearly better on statistical grounds. The statistics in table 4 show that the annual model (equation [5]) with commodity prices holds just such a relationship with the quarterly and annual price gap models, while neither of the latter en compasses the other. Although the comparisons in tables 3 and 4 show that commodity price information can im prove the forecasting performance of the price gap model, the t-statistic comparison in table 2 indicates that the improvement will be less at longer horizons. Commodity prices themselves are notoriously difficult to forecast (as implied by efficient markets theory), so it is probably im practical to make multistep forecasts from mod els like equations (4) and (5) to predict at longer horizons. The P* approach to forecasting inflation exploits the long-ain tendencies of output to return to potential and velocity to return to its mean. However, other factors may also influence the inflation rate over shorter horizons. Incorporat ing such influences into the HPS model can be accomplished simply by including additional terms that measure the lagged discrepancy be tween the actual inflation rate and the rate that would be predicted based on the relationship between inflation and the new factor. The commodity inflation data used here are only one of many possible augmentations. Infla tion as measured by a wholesale price index, for example, would be one reasonable addition; another might be an expectations measure de rived from the yield curve. So long as these addi tional factors are related to the inflation rate, rather than to the price level, the latter will continue to be determined only by money and potential out put. If potential output is taken to be exogenous, then ultimately, only money matters. References Chong, Yock Y., and David F. Hendry. “Econo metric Evaluation of Linear Macro-Economic Models,” Review o f Economic Studies, vol. 53, no. 4 (August 1986), pp. 671-90. Hallman, Jeffrey J., Richard D. Porter, and David H. Small. “Is the Price Level Tied to the M2 Monetary Aggregate in the Long Run?” Ameri can Economic Review, vol. 81, no. 4 (Septem ber 1991), pp. 841-58. The Causes and Consequences of Structural Changes in U.S. Labor Markets: A Review by Randall W. Eberts and Erica L. Groshen Introduction Despite apparently tight labor markets, wage infla tion in the late 1980s was much lower than most observers anticipated. The W all StreetJournal quoted one noted economist as saying, “The most interesting phenomenon in the United States to day is the existence of enormous labor shortages in some areas accompanied by no upward pres sure on wages.”1The article went on to state that the reasons for this phenomenon challenge the assumptions about the relationship between wage changes and general price changes that we formed during the 1960s and 1970s. Several explanations were offered at that time for the slow nominal wage growth seen during the second half of the decade. Chief among the factors cited by labor-market analysts and the media was a reversal in labor-management psy chology about wage increases, brought on in part by slow productivity growth, a severe eco nomic downturn, and increased foreign compe tition. The common perception was that during the 1970s, workers, with the consent of manage ment, felt entitled to automatic wage increases that were at least in line with inflation. The http://fraser.stlouisfed.org/ ■ 1 See Uchitelle (1987). Federal Reserve Bank of St. Louis Randan w. Eberts is an assistant vice president and economist and Erica L. Groshen is an economist at the Federal Reserve Bank of Cleve land. The authors would like to thank William Osterberg and James Thomson for helpful suggestions. demand for “3 percent plus cost of living” was a common refrain around many negotiating tables. This mind-set evaporated as workers suffered massive job losses during the twin recessions of the early 1980s, and as managers faced mounting foreign competition that eroded U.S. firms’ market share and placed downward pressure on domes tic prices. Instead of focusing on wage increases, negotiations became centered on wage conces sions in exchange for job security. In addition to a change in the psychology of wage-setting behavior, institutional changes were also cited as possible causes of sluggish wage growth. Mitchell (1989), in comparing the wage pressures of the 1980s with those of the 1960s, concludes that recent changes in labormarket institutions have pushed wage-setting in a more competitive direction. With the declines in the proportion of workers in the union sector and in big firms, jobs are less likely to be cushioned from labor-market forces by union contracts and bureaucratic personnel practices. Changes in demographics, particularly the greater participation of women in the labor force, were also said to figure into the moderate wage growth witnessed during the 1980s. To the extent that women are less attached to the labor force than are men, they may provide a 19 TABLE 1 Economic Conditions in Previous Decades 1960s Expansion Quarters3 Condition Average annual percentage change in Average hourly earnings, private business sector Compensation per hour index Consumer Price Index Output per hour, private business sector Real GNP, 1982 dollars Average level of: Unemployment rate Unemployment rate, male, age 25 and up Capacity utilization 1970s Recession Quarters15 ____________ 1980s__________ Expansion Quarters Recession Quarters0 Expansion Quarters 6.93 7.38 7.66 3.39 6.36 9.01 8.45 -0.38 8.35 6.94 1.82 9.10 10.04 0.18 4.26 3.41 2.41 4.21 0.18 3.47 -0.25 3.65 4.06 5.37 2.99 6.42 3.78 8.17 5.81 7.02 2.31 87.80 80.99 80.78 75.93 79.99 5.21 3.68 1.70 5.39 a. 19 6 l:IQ to 1969:IVQ. h. 1970:IQ to 1970:IVQ and 1973:IVQ to 1975:IQ. c. 1980:IQ to 1980:IIQ and 1981:IIQ to 1982:IVQ. SOURCES: U.S. Department of Commerce, Bureau of Economic Analysis; U.S. Department o f Labor, Bureau of Labor Statistics; and Board of Governors of the Federal Reserve System. buffer by filling vacancies during tight labor markets and by leaving the labor force during slack periods. The questions facing policymakers and analysts during this period were twofold: What was really behind the apparent change in wage behavior, and was the shift permanent or tempo rary? In October 1989, the Federal Reserve Bank of Cleveland sponsored a conference on the causes and consequences of structural changes in U.S. labor markets. Several prominent labor economists were asked to provide a careful and comprehensive analysis of some of the important developments that took place during the 1980s. The research focuses on a range of labor-market behaviors and industrial relations practices that could explain the macroeconomic relationship between unemployment and wages, and also on the effects of this relationship on output and employment stability. Four of the six papers deal with alternative compensation practices (fringe benefits and lump-sum and profit-sharing pay ments) and the structure of union contracts. The remaining studies examine how changing laborforce demographics and increased pressure from international competition have affected wages. I. Comparisons across the Last Three Decades Was wage behavior different during the 1980s than in the preceding two decades? This brief sec tion argues that this may indeed have been the case. Many analysts have noted that nominal wage growth during the expansions of the 1980s fell far short of that experienced during the upturns of the 1970s and even of the 1960s (table 1). And the same relatively low growth rates are also evident for the broader measure of compensation per hour, which includes fringe benefits, a growing component of employee compensation. This sluggish response alone might tempt one to conclude that fundamental changes in the structure of wage determination and worker compensation during the 1980s dampened the upward pressure on wages. However, leaping to that conclusion ignores differences in eco nomic conditions across the past three decades. Although observers in the 1980s generally per ceived labor markets to be extremely tight (par ticularly during 1988 and early 1989), typical measures of labor-market tightness do not sup port this view. In fact, the minimum unemploy ment rate during the expansions of the 1980s TABLE 2 Explaining Annual Percentage Changes in Nominal Average Hourly Earnings 1960s 1970s 1980s 0.465 (0.63) 6.022 (7.43) 0.473 (2.11) Consumer Price Indexa 0.887 (2.07) 0.082 (1.18) 0.325 (5.70) Unemployment rateb -0.018 (-1.74) 0.027 (1.84) Capacity utilization rateb 0.045 (0.63) 0.150 (2.66) 0.051 (6.09) 0.142 Labor productivity11 0.286 (2.07) -0.221 (-2.70) GNP implicit price deflator*1 0.271 (0.77) 0.138 (1.12) (0.03) 0.498 (5.60) -0.387 (-1.10) -0.674 (-1.80) -0.138 (-0.43) 0.89 0.52 0.99 Intercept Recession0 R2 (6.19) 0.002 a. Year-over-year change, lagged one quarter. b. Year-over-year change. c. Variable equals one for quarters marked by recession. NOTE: Observations are quarterly, and percentage changes are year over year. Separate regressions were run for each decade. T-statistics are in parentheses. SOURCE: Authors’ calculations. (5.2 percent) was higher than that of the up turns of the previous two decades (3.4 percent during the 1960s and 4.8 percent during the 1970s). Moreover, the maximum rate of capacity utilization was lower in the 1980s expansions (84.4 percent) than during those of the 1960s and 1970s (91-6 percent and 87.3 percent, respec tively). Thus, it is not clear whether the slow wage growth of the 1980s stemmed from struc tural changes in wage-setting practices or simply from differences in business conditions. One way to partially disentangle these effects is to ask the conceptual question, What would have happened to wages if the expansions of all three decades had shared the same economic conditions and differed only in the relationship between wages and changes in the economic en vironment? We use a simple econometric tech nique to estimate the wage behavior separately for each of the last three decades. These estimates, which summarize the link between wages and economic conditions in each decade, are then used to simulate the net nominal wage change that would have taken place if wages had re sponded to similar conditions. We follow a variant of the wage-change model used recently by Wachter and Carter (1989) and earlier by Gordon (1982).2 Annual changes in average hourly nominal earnings are explained econometrically by annual changes in the unemployment rate, capacity utilization, labor productivity (measured by output per hour), the GNP implicit price deflator, and the Consumer Price Index (CPI, all items for urban workers). Other specifications of the wage-change model are possible, and many have been posited. Our simple five-variable specification is based on the premise that wages reflect both pressures in the labor market and inflation expectations. We use the CPI to measure expected price inflation. Changes in the unemployment and capacity utilization rates are assumed to proxy for shifts in the tightness of labor and product markets. Labor productivity changes measure workers’ contribution to production and, consequently, employers’ ability to grant higher wages. And finally, the GNP implicit price deflator captures shifts in producer prices, which also reflect employers’ ability to pay higher wages. We estimate these relationships separately for each decade using quarterly observations. We also include a variable that takes the value of one during quarters marked by national recessions to account for business-cycle effects. Because our main purpose is to demonstrate wage behavior under similar economic condi tions, we do not dwell on the estimates of in dividual coefficients. Nevertheless, we note that most of the variables in table 2 appear to have the expected effect on nominal wage changes: Higher nominal wage increases are generally associated with stepped-up inflation expecta tions, increased capacity utilization, labor pro ductivity gains, and higher producer prices. Although the positive relationship between changes in nominal wages and unemployment rates in the 1970s and 1980s is perhaps surpris ing, it is consistent both with periods of stagfla tion during the earlier decade and with the long, gradual recovery of the 1980s, when wage and price increases moderated and unemploy ment fell. The net effects of these differences in the relationship between nominal wage changes and changes in economic conditions are shown ■ 2 We present this specification simply as an illustration of the changes sensed by analysts and practitioners during the 1980s. Individual coefficient estimates from this exercise should be interpreted cautiously because of the short time periods involved. 21 T A B L E II. Why the Slow Wage Growth in the 1980s? 3 Simulations of Annual Nominal Hourly Earnings Changes Explanatory variables (economic conditions) Structure (relationship between conditions and wages) --------------------------1960s 1970s 1980s 1960s 5.41 6.17 3.65 1970s 9.52 7.51 6.97 1980s 6.43 6.65 4.24 NOTE: The values are the average annual percentage changes in nom inal hourly earnings during the decade. Simulations were performed by multiply ing the explanatory variables for a given decade by the coefficients for the appropriate decade. The values on the diagonal (that is, for the same decade) are identical to the actual annual wage changes. SOURCE: Authors’ calculations using estimates from table 2. in table 3. The bottom row is of primary interest. The first entry in that row is the average annual nominal wage change that would have taken place in the 1980s if labor had had the same relationship to economic conditions then as in the 1960s. In this hypothetical case, wages would have increased an average of 6.43 percent annually in the 1980s. Subjecting the wage be havior that prevailed during the 1970s expan sions to 1980s economic conditions yields a slightly higher annual growth rate of 6.65 per cent. Both of these figures substantially exceed the 4.24 percent average annual increase that actually took place during the 1980s. It is also interesting to note that if wages had had the same relationship to economic conditions during the 1960s as they did in the 1980s, wage growth would have been considerably lower in the earlier decade than it actually was (3.65 per cent versus 5.41 percent). The same holds true for the 1970s. The actual annual wage increase was 7.51 percent, compared with 6.97 percent when the 1980 wage structure is used. This simple analysis suggests that something dampened the relationship between wages and economic conditions during the 1980s, such as changes in unemployment rates and in price levels. The papers summarized below explore the various shifts that have taken place and consider their implications for both wage behavior and the performance of the U.S. economy. The explanations explored at this conference for the slow wage growth of the 1980s can be grouped around three phenomena: increased international competition, changes in wage-setting practices, and demographic shifts. Increased International Competition The first category considers increased competition within product markets, particularly that resulting from greater penetration of foreign imports into U.S. domestic markets. Under this scenario, pres sures to keep prices in line with those of foreign competitors would moderate wage increases. Susan Vroman and Wayne Vroman address this issue in “International Trade and Money Wage Growth in the 1980s.” Their focus on international trade as a significant contributor to sluggish wage growth is well supported by the events of the 1980s. The U.S. economy has become increasingly open to foreign trade with respect to both imports and exports. As imports further penetrate our product markets, one would also expect labor markets to become more competitive, constraining domestic nominal wage growth. The authors present two sets of estimates to test this hypothesis. The first is based on a timeseries analysis of a modified Phillips curve, which shows the trade-off between nominal wage growth and unemployment. The second is based on a longitudinal study of more than 2,000 collective bargaining agreements in the man ufacturing sector between 1959 and 1984. Both sets of estimates show that developments in inter national trade in the 1980s contributed to the slow down in money-wage inflation, with nonpetroleum import prices and real nonpetroleum import share registering the most significant effects. The authors are quick to point out, however, that international trade accounted for only a small part of the slow down, at most 18 percent in selected years. This contribution would have been even less significant for the private business sector as a whole, since in ternational trade should have the largest impact on manufacturing, a sector that directly involves only one-fifth of U.S. workers. Vroman and Vroman place the estimated trade effect into perspective by exploring other possible explanations for the modest wage growth of the 1980s. Most important among these are inflationary expectations and the composi tion of unemployment. The authors conclude that, of the factors considered, the reduction in inflationary' expectations during the latter half of the decade was the primary factor in the slower nominal wage growth. The unusually high rate of unemployment among prime-age males was also found to exert a restraining effect on money-wage growth, equaling the impact of international trade. Changes in Wage-Setting Practices The second class of explanations relates to institu tional changes in wage-setting practices. These in clude alternative forms of compensation, such as lump-sum payments, profit sharing, and fringe benefits. Also covered are changes in the structure of labor union contracts, such as contract dura tion, cost-of-living indexation, and the emphasis on job security over wage growth. As documented below, workers have increas ingly received compensation in forms other than cash wages. For instance, the percentage of workers receiving lump-sum or profit-sharing payments has risen over the last decade. Fringe benefits as a share of total compensation has also increased, although a slightly smaller pro portion of workers are now covered by pen sions and health care benefits— the two largest components of this fomi of payment. Moreover, the prevalence of cost-of-living indexation fell dur ing the 1980s, while contract length grew. A theme shared by all four papers summarized in this sec tion is that developments in wage-setting pro cesses may have reduced the trade-off between wage inflation and unemployment. Lump-Sum Payments. In “Lump-Sum Payments and Wage Moderation in the Union Sector,” Linda Bell and David Neumark examine the growth of lump-sum payments in union firms in an effort to de termine whether the spread of this alternative com pensation arrangement contributed to the decline in wage growth during the past decade. Lump-sum payments can reduce wage inflation in at least three ways. First, they may signal a change in the labormanagement environment toward either a strength ened management stance or a worker preference for that form of payment. Second, they may simply reflect an accounting change as certain labor costs are shifted out of wages and salaries and into other forms of compensation. Third, they may increase labor-market flexibility by tying compensation more directly to worker productivity and to firm profits. Profit sharing provides employers with a method for responding to shocks in the product market beyond simply adjusting employment. Since profit-sharing payments are not counted as part of an employee’s base salary, adjustments in either direction can be made quickly in response to changing business conditions. This increased flexibility has led some economists, including Weitzman (1986), to advocate profit sharing as a means of stabilizing employment and output. To analyze these effects, Bell and Neumark ex amine more than 5,000 contracts negotiated in 1,200 private-sector establishments between 1975 and 1988. Within this sample, they find a dramatic jump in the number of contracts with lump-sum payment provisions. Indeed, between 1983 and 1984, the proportion of workers signing such con tracts skyrocketed from 5.9 percent to 69-5 per cent! The authors present evidence that this surge resulted from unions’ preference for this alterna tive form of compensation. Applying the Phillips-curve framework to the trade-off between nominal wage increases and un employment, Bell and Neumark find that the prev alence of lump-sum payments is associated with reduced wage growth. They estimate that a 10percentage-point rise in the share of workers cov ered by lump-sum contracts pushes the annual rate of wage inflation down 0.3 to 0.4 percentage point. The authors then reject all but one of the aforementioned explanations for this dampening effect. They dismiss the accounting explanation of a shift from base wages by showing that lump sum payments also reduce the percentage in crease in firms' total labor costs. Likewise, they find little support for the flexibility explanation. In fact, their estimates are inconsistent with the hypothesis: Firms offering lump-sum payments ex hibit less labor-cost flexibility in response to changes in demand for their products. The authors conclude that the labor-management environment must have changed during the 1980s. Profit Sharing. Douglas Kruse explores the effect of a second form of nonwage payment— profit sharing— on wage growth. In “Profit Sharing in the 1980s: Disguised Wages or a Fundamentally Different Form of Compensation?” he points out that even this somewhat narrow type of com pensation takes several different fomis, includ ing profit-related bonuses, deferred pension plans, or some combination of the two. Results of his study show a steady growth in deferred profit sharing, as the percentage of the private wage and salary work force with such coverage rose from 13-3 percent in 1980 to 18.4 percent in 1986. Although this still represents a relatively small share of the total labor force, the covered workers appear to be concentrated in industries that have historically demonstrated downwardly rigid wage behavior, such as manufacturing. Kruse concentrates on increased labor flexibil ity to explain the negative relationship between profit sharing and wage growth— a relationship that is similar to the one between lump sums and wages described by Bell and Neumark. He re views the empirical literature on the connection between profit sharing and employment stability and finds little agreement among the studies. Kruse also pursues his own empirical test using deferred pension plans as a measure of profit shar ing. His analysis yields some support for the posi tion that firms do not view profit-sharing payments as part of the short-mn cost of labor, but rather as a distribution of profits to labor after other costs (including base labor costs) have been taken into account. In this way, a company’s employment decisions are not influenced by profit-sharing pay ments, since these are not considered part of base wages. For 586 publicly traded U.S. companies, Kruse notes little trade-off between higher profitsharing payments and employment. On the other hand, he does find the expected trade-off be tween base wages and employment. The author concludes that profit sharing is not simply “dis guised wages,” but a more flexible form of employee compensation. Fringe Benefit Coverage. In "The Decline of Fringe-Benefit Coverage in the 1980s,” Stephen Woodbury and Douglas Bettinger suggest that compensation became more flexible during the last decade because a lower percentage of workers received employer-based health in surance coverage and pension plans. The share of workers included in employer-provided pen sion plans dropped from 60 percent in 1979 to 55 percent in 1988. During the same period, the per centage of workers covered by employer-provided group health insurance plans shrank slightly, from 74 percent to 72 percent. As a result, the ratio of employer costs for these two fringe benefit pack ages to wages and salaries edged down. These statistics suggest that the moderate wage growth in the 1980s was not necessarily due to large off setting increases in benefit coverage. However, the reduced coverage may have led to more flex ible compensation. Woodbury and Bettinger’s primary purpose is to provide a detailed analysis of the determi nants of fringe benefit coverage. They conclude that the decline in coverage during the 1980s resulted both from the decrease in marginal tax rates on personal income during the middle of the decade and from the steady drop in union rep resentation throughout the decade. Dwindling manufacturing employment, shifts in occupa tional mix, and aging of the work force had little to do with the decrease in coverage, according to the authors. The most significant detenninant was the low ering of marginal tax rates in 1986, which induced workers to trade fringe benefits for increased wages. However, their willingness to substitute wages for fringes was not uniform across all types of voluntary benefits. Woodbury and Bettinger estimate that workers were more willing to trade wages for employer-provided pensions than for employer-provided health coverage. The authors interpret the decline in benefit coverage as a ten dency for a reduction in the fixed component of worker compensation, which can be seen as a move toward a more “spot market” type of pay. Union Contracts. Wage moderation in the 1980s was disproportionately concentrated in the union sector, which experienced lower wage growth in the latter half of the decade (14.2 per cent) than did the nonunion sector (23-9 percent). In contrast, in every year between 1976 (when data first became available) and 1982, union wage hikes outpaced nonunion wage changes. After 1982, when the economy began to recover from the high unemployment brought on by the twin recessions that inaugurated the decade, many unions placed job security above wage growth as the top priority in their bargaining rounds. This reordering is certainly evident in unions’ nominal wage increases. In “Indexation and Contract Length in Union ized U.S. Manufacturing,” Mark Bils examines two changes in the stmcture of labor contracts that could have led to slower wage growth in the union sector: reductions in indexation and shorter contract length. For all union contracts settled in the private sector, the proportion of workers with inflation escalator clauses fell from an average of 55.2 percent between 1980 and 1983 to 36.8 per cent between 1984 and 1988. However, the length of contracts remained the same over the decade, averaging slightly more than 31 months. An extensive body of theoretical literature supports the view that the length of contracts and the inclusion of indexation reflect the de gree of uncertainty facing workers and employ ers. To explore this proposition, Bils examines a detailed longitudinal set of major collective bar gaining agreements reached between 1955 and 1985 in the manufacturing sector. His results con tradict the generally accepted prediction that increased uncertainty will shorten contracts. Rather, he finds that contracts are longer in industries that face more uncertainty (durable goods, for example). Bils suggests that these results are consistent with the notion that longer contracts are written in order to reduce strikes. With respect to indexing, he finds that the per centage of contracts with cost-of-living escalator clauses is positively related to increases in both in flation and inflation uncertainty. This is consistent with the generally accepted view that escalator clauses protect workers from unanticipated pricelevel changes. Consequently, consumers’ lower inflation expectations during the latter half of the 1980s could explain the lower nominal wage growth at that time. This finding is in accord with Vroman and Vroman’s results. Demographic Shifts In “Gender Differences in Cyclical Unemploy ment,” Sanders Korenman and Barbara Okun con sider the effect of female participation in the work force on cyclical unemployment. It may be that women provide a pool of workers who move freely into and out of the labor force (depending on the stage of the business cycle), since they are historically less attached to it than are men. Such a procyclical participation pattern of a large group of workers would weaken the effectiveness of un employment rates as a measure of labor-market tightness. Consequently, fluctuations in wages and in standard measures of unemployment rates asso ciated with business cycles wrould be dampened. The major issue that Korenman and Okun ex plore, therefore, is whether women are indeed less attached to the labor force than are men. Their analysis shows that although women are still less attached, their connection grew during the 1980s. These results might suggest that cyclical unemployment should rise in response to in creased labor-force attachment among women, but further analysis shows no association be tween these two factors. The authors attribute this to the disparate distribution of the sexes across industries and occupations. W om en’s employment is disproportionately concentrated in growth industries that demonstrate little cycli cal fluctuation, while men are concentrated in industries with the opposite characteristic. Thus, although Korenman and Okun do not rule out the possibility that the increase in female labor supply during the 1980s reduced wage growth by lowering labor-market tightness, their find ings suggest little, if any, change in the long-run cyclical behavior of the economy as a result of this phenomenon. III. Implications for Macroeconomics Two prominent macroeconomists, Olivier Blanchard and Finn Kydland, were invited to the conference to comment on whether the find ings of the papers summarized above alter the way in which labor markets figure into their view of the workings of the macroeconomy. In particular, we were interested in whether the trend toward more flexibility and risk sharing in wage-setting practices would alter their theories and policy recommendations. Blanchard’s remarks focus primarily on the macroeconomic implications of lump-sum bonuses and profit sharing. He sees both schemes as ways of lowering the risk of bank ruptcy among firms, and notes an interesting tension between the implications of increased risk sharing in labor contracts and of recently introduced financial arrangements, such as highyield junk bonds. The former generally provides greater stability by reducing the likelihood of bankmptcy, while the latter raises the chances of a firm going under. Blanchard argues that the reduction in bankmptcy risks has had three macroeconomic effects: 1) stabilization of employment in the short run, 2) alteration of the factors determining labor mobil ity, and 3) modification of the Phillips-curve speci fication. The first effect results from the simple fact that firms will not be as likely to close their doors during downturns and, through wage adjustments, will be able to retain workers longer. This should reduce employment swings during business cycles. The second effect is related to labor adjust ments that follow sectoral shocks. If wages are rigid, then declining employment is the only sig nal that leads workers to leave hard-hit sectors. However, if wages vary, then both they and job security enter into a worker’s decision. Finally, an increase in wage flexibility breaks the link be tween tightness in the labor markets (as measured by unemployment rates) and price inflation. Kydland frames his remarks in terms of im plications for business-cycle theory. He notes that many researchers interested in this line of inquiry have changed their methodology from the system-of-equations approach popular in the 1960s to one based on the neoclassical growth model. Under the former framework, models are constmcted around equations that describe aggregate economic behavior, such as wage rates, unemployment rates, household con sumption, and business investment. In contrast, the approach based on the neoclassical growth model stresses the use of empirical knowledge to obtain parameter estimates for technology, pref erences, and institutional arrangements. These parameter estimates provide realistic calibra tions for simulation models intended to mimic, and thus explain, macroeconomic phenomena. Consequently, this transition to the use of the neoclassical growth model as the basis of macroeconomic analysis is important in detemrining how questions are posed and data are organized. Kydland finds that research presented in this volume is, for the most part, organized around the fonner methodology— that is, based on estimates of aggregate behavioral equations. He stresses that in order to bridge the gap, questions posed in the business-cycle framework will have to be translated into the behavioral-equation framework, and vice versa. Therefore, business-cycle researchers may have to ask slightly different questions or else organ ize the information presented at this conference in a different way if they are to incorporate these find ings into their research. IV. Conclusion The research presented at this conference under scores the thinking of many observers and market analysts who, during the latter half of the 1980s, perceived that developments were taking place in labor markets that altered certain basic relation ships between wage behavior and economic perfonnance. These essays suggest that the increased adoption of more-flexible pay schemes during the latter half of the decade led to lower labor costs, perhaps to more flexibility for firms in their em ployment decisions, and, in general, to more sta bility in employment (at least in the short run). Thus, evidence indicates that these more flexible pay schemes might be able to accommodate rela tively lower unemployment rates without igniting serious wage inflation. Although some observers argue that this in creased flexibility, which stemmed from the adop tion of lump-sum payments and profit-sharing arrangements, is simply a way to obscure wage concessions, the research presented here finds little support for this view. The fairly widespread acceptance of these alternative compensation practices by both workers and managers suggests that the shift in the relationship between labor markets, unemployment, and price inflation observed in the 1980s may extend well into the 1990s. This structural change, along with other changes noted at the conference, may be wel comed by policymakers attempting to contain in flation while simultaneously stabilizing output. References Bell, Linda, and David Neumark. “Lump-Sum Payments and Wage Moderation in the Union Sector,” in Randall W. Eberts and Erica L. Groshen, eds., Structural Changes in U.S. Labor Markets. Aunonk, N.Y.: M.E. Sharpe, Inc., 1991. pp. 45-62. Bils, Mark. “Indexation and Contract Length in Unionized U.S. Manufacturing,” in Randall W. Eberts and Erica L. Groshen, eds., Structural Changes in U.S. Labor Markets. Armonk: N.Y.: M.E. Sharpe, Inc., 1991, pp. 145-71. Blanchard, Olivier Jean. “Macroeconomic Impli cations,” in Randall W. Eberts and Erica L. Groshen, eds., Structural Changes in U.S. Labor Markets. Armonk: N.Y.: M.E. Sharpe, Inc., 1991, pp. 201 -05. Gordon, Robert J. “Inflation, Flexible Exchange Rates, and the Natural Rate of Unemploy ment,” in Martin N. Baily, ed., Workers, Jobs, a n d Inflation. Washington, D.C.: The Brook ings Institution, 1982, pp. 89- 155. Korenman, Sanders, and Barbara Okun. “Gender Differences in Cyclical Unemployment,” in Ran dall W. Eberts and Erica L. Groshen, eds., Struc tural Changes in U.S. Labor Markets. Armonk: N.Y.: M.E. Sharpe, Inc., 1991, pp. 177-95. Kruse, Douglas. “Profit Sharing in the 1980s: Disguised Wages or a Fundamentally Dif ferent Fonn of Compensation?” in Randall W. Eberts and Erica L. Groshen, eds., Structural Changes in U.S. I.abor Markets. Armonk: N.Y.: M.E. Sharpe, Inc., 1991, pp. 67-99. Kydland, Finn E. “Macroeonomic Implications,” in Randall W. Eberts and Erica L. Groshen. eds., Structural Changes in U.S. Labor M ar kets. Armonk: N.Y.: M.E. Sharpe, Inc., 1991, pp. 207-13. Mitchell, Daniel J.B. “Wage Pressures and Labor Shortages: The 1960s and the 1980s,” Brook ings Paper's on Economic Activity, vol. 2 (1989), pp. 191-231. Uchitelle, Louis. “Wage Increases Are Sluggish despite a Scarcity of Workers,” The W all StreetJournal, September 1, 1987, p. Al. Vroman, Susan, and Wayne Vroman. “Interna tional Trade and Money Wage Growth in the 1980s," in Randall W. Eberts and Erica L. Groshen, eds., Structural Changes in U.S. Lahor Markets. Armonk: N.Y.: M.E. Sharpe, Inc., 1991, pp. 13-40. Wächter, Michael L., and William Carter. “Norm Shifts in Union Wages: Will 1989 Be a Replay of 1969?” Brookings Papers on Economic Ac tivity, vol. 2 (1989), pp. 233-64. Weitzman, Martin L. “Macroeconomic Implica tions of Profit-Sharing,” in Stanley Fischer, ed., NBERMacroeconomics A n n u a l 1986. Cambridge, Mass.: MIT Press, 1986. Woodbury, Stephen A., and Douglas R. Bettinger. “The Decline of Fringe-Benefit Coverage in the 1980s,” in Randall W. Eberts and Erica L. Groshen, eds., Structural Changes in U.S. Labor Markets. Armonk: N.Y.: M.E. Sharpe, Inc., 1991, pp. 105-38. First Quarter Working Papers ■ 9119 Estimating a Firm's AgeProductivity Profile Using the Present Value of Workers’ Earnings Current Working Papers of the Cleveland Federal Reserve Bank are listed in each quarterly issue of the Economic Review. Copies of specific papers may be re quested by completing and mail ing the attached form below. Single copies of individual papers will be sent free of charge to those who request them. A mailing list service for personal subscribers, however, is not available. Institutional subscribers, such as libraries and other organiza tions, will be placed on a mail ing list upon request and will automatically receive Working Papers as they are published. ■ 9121 Rising Inequality in a Salary Survey: Another Piece of the Puzzle ■ 9201 Binomial Approximation in Financial Models: Computational Simplicity and Convergence ■ 9203 Post-Louvre Interven tion: Did Target Zones Stabilize the Dollar? by Erica L. Groshen by Anlong Li by Laurence J. Kotlikoff and Jagadeesh Gokhale ■ 9120 A Dynamic Analysis of Recent Changes in the Rate of Part-time Employment by Donald R. Williams ■ 9122 A Different Kind of Money Illusion: The Case of Long and Variable Lags by Michael F. Bryan and William T. Gavin ■ 9202 Wagner’s Hypothesis: A Local Perspective by Randall W. Eberts and Timothy J. Gronberg by Richard T. Baillie and Owen F. Humpage ■ 9204 Bank Performance and Regional Economic Growth: Evidence of a Regional Credit Channel by Katherine A. Samolyk ■ 9205 The Gold Standard as a Rule by Michael D. Bordo and Finn E. Kydland Please complete and detach the form below and mail to: Research Department Federal Reserve Bank of Cleveland P.O. Box 6387 Cleveland, Ohio 44101 Check item(s) requested Send to: Please send the following Working Paper(s): □ 9119 □ 9122 □ 9203 □ 9120 □ 9201 □ 9204 □ 9121 □ 9202 □ 9205 Please print Name Address City State Zip E9 Structural Changes in U.S. Labor Markets Conference Proceedings Now Available The papers in this book, “Struc tural Changes in U.S. Labor Mar kets: Causes and Consequences," edited by Randall W. Eberts and Erica L. Groshen, were presented and discussed at a conference held at the Federal Reserve Bank of Cleveland in October 1989. The purpose of the conference was to identify and analyze recent de velopments in personnel policy and worker compensation prac tices, which may have led to less wage inflation during the 1980s and may continue to affect wage behavior in the 1990s. Also con sidered were possible conse quences that these changes might have on the formulation of macroeconomic policy. The contributors— academic and research economists in labor eco nomics— provide a comprehen sive assessment of the current state of the wage-setting process in the U.S. labor market. ■ Profit Sharing in the 1980s: Disguised Wages or a Fundamentally Different Form of Compensation? ■ Indexation and Contract Length in Unionized U.S. Manufacturing Contents: ■ Overview by Randall W. Eberts and Erica L. Groshen ■ International Trade and Money Wage Growth in the 1980s by Susan Vroman and Wayne Vroman Comments: Louis Jacobson ■ Lump-Sum Payments and Wage Moderation in the Union Sector by Mark Bils Comments: Michael L. Wächter by Douglas Kruse Comments: Sharon P. Smith ■ Gender Differences in Cyclical Unemployment ■ The Decline of Fringe-Benefit Coverage in the 1980s by Sanders Korenman and Barbara Okun Comments: Katherine G. Abraham by Stephen A. Woodbury and Douglas R. Bettinger Comments: David Lewin by Linda Bell and David Neumark ■ Macroeconomic Implications by Olivier Jean Blanchard and by Finn E. Kydland Comments: Ken Ross Ordering Information Structural Changes in U.S. Labor Markets: Causes and Consequences Randall W. Eberts and Erica L. Groshen, editors 1991 352 pages: tables, figures, index ISBN 0-87332-825-6 Hardcover, $39.95 For orders and inquiries, please contact: M.E. Sharpe, Inc. 80 Business Park Drive Armonk, NY 10504 Credit card orders accepted. Call toll-free 1-800-541-6563 FAX: 914-273-2106