The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
3 Strategic Considerations in Monetary Policy with Private Information: Can Secrecy Be Avoided? 17 The German Monetary Unification (Gmu): Converting Marks to D-Marks 37 The Effects of Financial Innovations on Checkable Deposits, M l and M2 THE FEDERAL A KI SI KM JZnXSKoi A r S T .I/H IS 1 Federal R eserve Bank of St. Louis R e v ie w July/August 1990 In T his Issue . . . Much recent criticism o f the Federal Reserve System has focused on the secrecy that surrounds monetary policy decisions. Some observers have suggested that the Fed disclose the Federal Open Market Commit tee’s policy decisions immediately, to prevent it from concealing useful information from the public and to make it more accountable for its actions. In the first article o f this Review, “ Strategic Considerations in Monetary Policy with Private Information,” Seonghwan Oh and Michelle R. Garfinkel illustrate w hy a central bank may not be able to make credible announcements about its policy decisions precisely even if it w ould be better o ff by doing so. The lack o f credibility results from the public’s knowledge that the central bank gains from "surprising” the public with its monetary policy actions. Oh and Garfinkel show, however, the central bank can make announcements that imprecisely reveal its private information. These imprecise or "noisy” an nouncements will be credible, only if constraints can be imposed on the central bank that limit its policy independence. Hence, the authors argue that, if limiting policy flexibility is costly, the advantages of avoiding secrecy in monetary policy—even partially—must be carefully weighed against the cost o f doing so. * * * In the second article in this Review, Peter Bofinger provides a detailed look at “The German Monetary Unification (Gmu): Converting Marks to D-Marks.” The German monetary union represents the first step in uniting tw o countries with widely disparate econom ic systems and con ditions. Bofinger describes these differences and shows how they in fluence the choice o f the specific conversion rates used to determine the DM values o f East German financial claims and incom e and salary payments previously valued in East German marks. Bofinger also shows how the debate over the “appropriate” conversion rate was related to a variety o f important concerns facing both East and West Germans. Among these w ere the resulting debt burdens that East German firms would face, the associated wealth transfers between East and West Germany, the impact on the level o f West German govern ment debt and the possible effect on the West German inflation rate. Bofinger concludes that the actual conversion rates chosen limit the wealth transfer from West to East Germany to a relatively small amount and make it unlikely that West German inflation will accelerate. * * * JULY/AUGUST 1990 2 Conventional wisdom holds that the introduction o f new interestbearing checkable deposits, especially in the early 1980s, had a substan tial effect on M l and its relationship to econom ic activity. Equally wide ly shared is the view that these accounts and the subsequent introduc tion o f new m oney market deposit accounts had little or no effect on the broader monetary aggregates, such as M2. In the third article in this Review, “The Effects o f Financial Innovations on Checkable Deposits, M l and M2,” John A. Tatom examines this financial innovations hypothesis. As Tatom explains, the hypothesis implies that the growth o f these new checkable deposits should have reduced the turnover o f total checkable deposits and boosted the demand both for checkable deposits relative to currency holdings, and for M l. The growth in m oney market balances should not have affected the composition or demand for M l and M2. Tatom finds that these innovations did not have the statistically signifi cant effects predicted by the financial innovations hypothesis. In par ticular, new interest-bearing checkable deposits had no effect on the turnover rate o f total checkable deposits, the demand for total checkable deposits relative to currency, or the demand for M l. Also to the contrary, Tatom finds that M l and M2 demand w ere both affected by the introduction o f money market balances. Tatom concludes that analysts o f financial innovation effects generally have focused on the w rong innovation and the w ron g monetary ag gregate. His results indicate that the principal influence o f financial in novations has been the substantial effect o f m oney market balances on the demand for M2. FEDERAL RESERVE BANK OF ST. LOUIS * * * 3 Seonghwan Oh and Michelle R. Garfinkel Seonghwan Oh, an assistant professor of economics at the University of California, Los Angeles, was a visiting scholar at the Federal Reserve Bank of St. Louis. Michelle R. Garfinkel is a senior economist at the Federal Reserve Bank of St. Louis. Scott Leitz provided research assistance. Strategic Considerations in Monetary Policy with Private Information: Can Secrecy Be Avoided? _1_ 11K FEDERAL RESERVE System has been criticized often for the secrecy that surrounds monetary policy. In particular, many observers have questioned the desirability o f the Fed’s practice o f not disclosing the decisions o f the Federal Open Market Committee (FOMC) im mediately following its meeting. This criticism has been heightened recently by legislation in troduced in the House o f Representatives, p ro posing, among other things, that the Fed release the contents o f the FOMC’s directives immediate ly after each meeting rather than with a sevenweek delay.1 The econom ic rationale behind this proposal is that the Fed’s maintained secrecy limits the in formational content o f prices in financial markets and thereby detracts from the markets’ ability to allocate resources efficiently. If, for example, 1Lee Hamilton and Byron Dorgan, HR2735-the Federal Reserve Reform Act of 1989. See Hamilton (1989) for a brief discussion of the key changes in the structure of the Fed proposed by this legislation. As discussed by Goodfriend (1986), however, legislation proposed in this spirit is not new. the FOMC voted to maintain its current policy stance but subsequently added reserves to the banking system as a technical and temporary action, market participants might mistakenly in terpret such an action as a fundamental change in policy. According to this view, without im mediate disclosure o f the FOMC's policy direc tive, confusion about the Fed’s intentions can add to the variability o f market interest rates. Those w ho are skeptical o f the value o f this legislation argue that immediate disclosure of the FOMC's directive would complicate the im plementation o f monetary policy.2 For example, the markets’ response to announcements could generate large changes in interest rates that, ac cording to this view, would be excessive and destabilizing. 2See, for example, Mooney (1989), Rosenbaum (1989) and Uchitelle (1989). Also, see Goodfriend (1986) for an in teresting and useful critique of the arguments made for maintained secrecy at the Fed. JULY/AUGUST 1990 4 In the context o f a relatively simple gametheoretic model o f monetary policy, in which the Central Bank would expect to be better off if it had no private information, this article shows w hy the Central Bank cannot reveal its private information credibly and precisely. The Central Bank might be able to reveal this infor mation partially through imprecise or noisy an nouncements. From the Central Bank’s perspec tive, however, such announcements are not costless, nor can they rem ove secrecy from poli cy perfectly. Hence, the analysis illustrates that, even if the Central Bank perceived monetary policy secrecy as undesirable, fully eliminating it might not be feasible. STRATEGIC MONETARY POLICY: THE BASIC MODEL To address issues o f secrecy in monetary policy, it is helpful to study a model o f monetary policy that specifies the objectives and constraints faced by a Central Bank. Given the particular specification, the m odel provides a fram ework for analyzing various strategies for the Central Bank and, in turn, for predicting which strategy is optimal for the Central Bank. The model, a slight variant o f Canzoneri (1985), builds on a simple specification o f the econom y.3 Output is given by (1) y ,= y ° + ( p ,- w t), w here y „ pt and w, denote, respectively, the logarithms o f output, prices and nominal wages in time t; y ” denotes the log o f output that cor responds to the "natural" rate o f unemployment. In this model, the natural level o f output is the one that w ould prevail with a steady rate of inflation. The public attempts to specify wages so as to minimize deviations o f output from its natural level. Accordingly, it wants to set w t= pt. But, in this model, prices are not known at the time wages are set. Hence, wages are set to satisfy (2) w, = p|, 3The model is intended only to be an illustration, not a complete characterization of the economy. Canzoneri’s (1985) model resembles that of Barro and Gordon (1983) except that it provides a role for the Central Bank to react to shocks. As will be evident below, this model does not imply that the first-best policy is a constant money growth rule. Rather, it is a contingent money growth rule. See Cukierman (1986) for a helpful review of this relatively new literature on central bank behavior. FEDERAL RESERVE BANK OF ST. LOUIS where p ' denotes the public’s expectation, as described below, o f the log o f the price level conditional on information available to the public at the beginning o f period t. By combining equa tions 1 and 2, output can be expressed as follows: (3) y t=y° + (nt-Tite), where nt= p , - p t_, is the actual rate o f inflation in time t; rc' = p ' - p t_, denotes the public’s expec tation o f inflation. Equation 3 captures the notion that the longrun Phillips curve, w hich is the relationship (trade-off) betw een inflation and unemployment, is vertical. On average, unemployment and, con sequently, output are independent o f both ex pected and actual inflation. In any period, h ow ever, unanticipated inflation can create a wedge between output and its natural level. Specifical ly, the existence o f contracts that fix nominal wages for a specific period means that actual output can depart from its natural level if people underestimate or overestimate the future rate o f inflation.4 The effect o f unanticipated infla tion on output is only temporary. In this model, it lasts only one period. The variance o f output implied by equation 3 is simply the variance of the market’s inflation forecast error. The following simple variation o f the quantity theory equation describes how prices are deter mined in each period given monetary policy: (4) p t= mt- y n+ vt, where v, denotes an innovation to m oney de mand and mt denotes the log o f the m oney supply in time t. Taking the first-difference o f equation 4 and rearranging shows how monetary policy affects inflation: (5) n, = g ,- d „ w here g t= mt- m t_, is the growth rate of money, the Central Bank’s policy instrument, and <5t= vt. , - v t denotes a random disturbance. This disturbance, w hich is bounded betw een 4That unanticipated inflation can drive output above its natural level would also be implied by the Lucas-type (1973) supply curve. The important feature of this equa tion—that output, on average, will be independent of inflation—assumes that the public forms expectations ra tionally. The assumption that the elasticity of output with respect to unanticipated inflation is equal to one is used to simplify the notation and does not affect the qualitative results discussed below except where noted. 5 - D and +D, is assumed to have a zero uncon ditional mean and a finite, constant variance, a As revealed by equation 5, the Central Bank’s control over inflation is imperfect; inflation depends not only on monetary policy but on the disturbance to m oney demand. Thus, equation 5 implies that the public’s expectation fo r inflation in time t equals the difference between its ex pectation o f m oney grow th in time t, g', and its expectation o f d„ d'. Secrecy arises in this model because, in con trast to the public, the Central Bank has a (non trivial) forecast o f the disturbance to m oney de mand.5 The Central Bank’s “private" forecast, dt ■■Et{c5t}, satisfies Fig u re 1 S e q u e n c e of E vents in P erio d t. dt dt + l -»-------(------ j------j------------------------------------- 1----------*- time t T t t+1 77e il t 9 a, (6) dt= dt+ £„ where Et{ • } denotes the Central Bank’s expec tation, based on information available to it at the beginning o f period t, before wage con tracts are signed. The Central Bank's forecast error, c„ has an expected value o f zero, a finite variance, a and no correlation with the Central Bank’s forecast. The assumption that the forecast is independent o f the forecast error implies ol = o2 d+ ol, where a\ is the variance o f the private forecast. Although this forecast is made just before wages are set, the markets’ expectation o f dt equals zero without any meaningful announce ments by the Central Bank. W hen the Central Bank does not attempt to convey its private in formation, 7i*= g'. The public observes dt after policy is implemented w hen nt is realized. The public, however, cannot infer from that obser vation what the Central Bank’s forecast had been. Similarly, it cannot identify the Central Bank’s forecast error. (See figure 1 which sum marizes the sequence o f events during any period t.) Nevertheless, people understand the Central Bank's objectives as described below and its constraints subject to the unknown dis turbance dt; they incorporate that understand5That the public does not have a forecast of d, implies d; = 0, providing that the Central Bank does not com municate to the public its own forecast. Note that it is not crucial that the public has no forecast of the disturbance to money demand. Provided that the Central Bank’s forecast is private, the following analysis is relevant. Fur thermore, the private information could be in terms of a forecast about a supply shock or the Central Bank’s preferences. The qualitative results to follow would not be affected. Also, it should be noted that the present model differs from Canzoneri’s (1985) model in that the timing of the forecast here is such that, if the Central Bank released this information, it could be used by the public. The ing into their expectations o f m oney growth and, accordingly, their wage specification. Following Canzoneri (1985), the analysis assumes that the Central Bank has tw o goals: output and inflation stabilization. Its expected lifetime utility in period t = 1 is given by (7a) Uj = 1 / T ‘ E, {u ,}, 0 < (3 < 1 t= 1 where (7b) u t= —(yt —y*)2—f(nt—n*)2, f > 0 . 13 is the Central Bank’s discount factor.6 The parameter f is the weight the Central Bank places on its objective o f stabilizing inflation around its target level, rt*, relative to its objec tive o f hitting its target for the log o f output, y*. These targets are given and fixed parameters. The Central Bank’s inflation target need not be zero. But its objective to stabilize inflation is consistent with the public's objective to forecast ing future inflation correctly. In other words, by minimizing the variability o f inflation, the Central Bank minimizes the variance o f the pub lic's inflation forecast error. The Central Bank's assumed sequence of events, shown in figure 1, is necessary for the analysis of imprecise announcements below. 6Note that equation 7 implies that the Central Bank is infinitely-lived. This assumption is only important for the discussion of reputational considerations below. This discussion would be qualitatively the same if, instead, the Central Bank lived only a finite number of periods, T, pro vided that T is not known with certainty. In this case, p would reflect the Central Bank’s chances of survival as well as its time preference. See Grossman and Van Huyck (1988), for example. JULY/AUGUST 1990 6 objective to stabilize output, however, is consis tent with the public’s objective to forecast infla tion correctly only if the Central Bank's target for output equals the natural level. But, in this case, the interesting issues revolving around monetary policy secrecy do not arise. As in m uch o f this literature, then, the pre sent analysis assumes that y* > y ”. That is to say, the Central Bank prefers output to exceed the public's target. Possible interpretations o f this assumption could stem from either social welfare or self-interest considerations.7 It is only im por tant for the present analysis that the natural output level or the public’s target for output be different than the Central Bank’s (given) target. This assumption implies that the Central Bank does not have enough instruments to reach its tw o goals, giving rise to a credibility problem in policy as illustrated below. Using equations 3, 5 and 7b and noting that the public's expectations for inflation, n‘, equals g '- d ', the Central Bank’s utility in period t can be written as (8) ut= —(gt —g[—dt+ d' —A*)2—f(gt —d, —ti’ )2, where A * = y * - y " > 0 and 6\= 0, without any an nouncem ents by the Central Bank about its private forecast. The Central Bank’s problem is to choose gt to maximize the expected value of its lifetime utility, after the markets set wage grow th equal to expected inflation, ti'. The solu tion depends on h ow the Central Bank treats the markets’ expectations. 7See Barro and Gordon (1983) and Canzoneri (1985) for a discussion of possible social-welfare interpretations of this assumption. These interpretations build on existing distor tions in the economy. For example, the existence of large unions that keep real wages too high or the use of income taxes that influence labor decisions depress average out put (or the natural level) below the “ potential” level (or that level considered desirable from a social-welfare perspective). Although these distortions could be modeled explicitly here, the associated modifications would add un necessary complexity to the model without providing much insight into the issues at hand. But see Cukierman (1986) for a useful critique of the social-welfare interpretation. Cukierman (1986) also provides an extensive discussion of a political interpretation. For example, although the Central Bank might be an independent institution, it might feel compelled, in order to preserve its existence or in dependence, to react to signals by the fiscal authority. The fiscal authority might be motivated to stimulate the economy to enhance its chances for re-election. 8ln fact, the same outcome would be obtained if the Central Bank’s forecast were not known by the public until after FEDERAL RESERVE BANK OF ST. LOUIS The First-Best Solution To see w h y the Central Bank might want to disclose its private information (that is, its forecast o f the m oney demand disturbance), consider the benchmark case wherein the Cen tral Bank recognizes the impact it can have on the markets’ expectations and dt is public infor mation. Furthermore, assume that the Central Bank can make binding commitments to pursue an announced policy. In this case, it chooses g„ subject to the restriction that expectations are consistent with its policy, to maximize its ex pected lifetime utility. Because o f the stationary (time-independent) nature o f the model, this maximization problem reduces to a sequence o f one-period problems, in which the Central Bank chooses gt to maximize its expected one-period utility, shown in equation 8, fo r each period t. Given the constraint that gt= g', creating sur prise inflation in an effort to increase output above its natural level is precluded. Rather, the Central Bank commits itself to the following policy: (9) g, = n* + d„ where gt= g ' for all t. Note equation 9 implies that, on average, inflation would be equal to the Central Bank’s target rate. Because the policy fully accommodates the part o f the disturbance to m oney demand predicted by the Central Bank, n' = ti* and wage grow th is set equal to ti*.8 The Central Bank’s expected one-period utility in this regime can be found using equations 8 wages were set, so that d; still equaled zero. Because the Central Bank fully accommodates dt, expected inflation, n;, is independent of d, in this regime. This is not to say that the Central Bank has no preferences about maintaining the privacy of its forecast. As will become obvious, the Central Bank wants to reveal its private forecast so that it can obtain this outcome. Whether the Central Bank should accommodate disturbances to the economy is a matter of controversy. In this model, its motive to react to d, is com patible with the public’s interests. The public prefers the Central Bank to react to its forecast, because such reac tions minimize the variance of the public’s forecast error. An argument against such a policy, for example, would be that it is destabilizing because the Central Bank’s forecasts are inaccurate. As shown below, however, even if its private forecasts are fairly accurate (provided that oJ^O), the Central Bank might not find it desirable to react to its forecast. (Given o5, however, the more accurate the forecast, the less likely the Central Bank would be willing to sacrifice flexibility in policy.) The alternative argument against flexibility in policy in this paper builds on the Cen tral Bank’s credibility problem. 7 and 9, with the assumptions that gt'= g t and d‘ = d t: (10) Et{ti} = - ( l + f)oe2-A * 2, for all t, where, as defined previously, a] denotes the finite variance o f the Central Bank’s forecast error. It is equal to the variance o f in flation and output in this regime. The contingent policy in equation 9 is referred to as the firstbest solution since it yields the highest utility to the Central Bank among those policies that are consistent with the public’s expectations. As demonstrated by Kydland and Prescott (1977), however, the policy in equation 9 is not “dynamically consistent.” That is, given the pub lic’s expectations, the Central Bank has an incen tive to deviate from the first-best policy. Specifi cally, given 7i'=n“, the Central Bank would rather implement the following policy: and to disclose its private information truthful ly. Without being able to exploit the dependence o f its actions on the public’s actions, the Central Bank chooses g, to maximize its expected oneperiod utility, shown in equation 8, as if it were not trying to influence g ' or d'. Before the Central Bank sets g„ the public specifies wage growth equal to its expectations o f inflation. Because the public understands the Central Bank’s maximization problem, it forms g ' by taking an (unconditional) expectation of the Central Bank's first-order condition given by (12) - 2(gt- g ' - d t + d '- A ') - 2 f ( g , - dt- 7i*) = 0, for each t. Even though the Central Bank observes d, before the public form s its expecta tions, without any announcements, d '= 0. Since the public’s expectation o f gt equals g' and its expectation o f dt equals d', g'=7i* + A*/f. (11) gCH=n* + dt + A,/(l + f). If the Central Bank could create surprise infla tion with the policy shown in equation 11, it could augment output above the natural level to approach its target.9 Such a “cheating” policy would increase the Central Bank’s expected oneperiod utility by A’ 2/(l+ f). The M yopic Solution But, even if the Central Bank could break its commitment to follow the first-best policy, cheat ing w ould be impossible as long as people can not be fooled. That is, rational people will always anticipate the Central Bank’s incentive to cheat, if it cannot make binding commitments. To consider another solution, one that is more likely to emerge as the equilibrium outcom e when the Central Bank has private information, suppose the Central Bank ignores any impact that it could have on the public's expectations. This is not to say that the Central Bank actually fails to understand the impact o f its actions on the public’s actions that, in turn, influence its ow n welfare. Rather, given the Central Bank’s incentive to cheat, it cannot control the public’s expectations directly unless it could som ehow be committed to follow an announced policy 9The solution in equation 11 is found by substituting g ;-d ; = n'into the Central Bank’s expected one-period utili ty function and maximizing that function with respect to g,. (See the first-order condition below in equation 12.) The Central Bank would follow the same cheating strategy if it People recognize the Central Bank’s incentive to engineer surprise inflation so as to augment output above its natural level. To protect themselves against a decline in their real wage, then, people specify higher rates o f wage grow th (equal to g') than in the first-best solu tion with commitments. Given that specification, the Central Bank’s policy, g„ w hich is referred to here as the “m yopic” solution for reasons that will becom e obvious later, is given by (13) g, = n* + d t+ A7f, for each t. With the m yopic policy, the Central Bank fully accommodates its prediction o f the m oney demand disturbance as in the first-best solution. Further, the policy shown in equation 13 validates the public's expectations, implying an average inflation rate equal to Tt'+A*/f. W hen the Central Bank acts as if it w ere ig noring the impact that it can have on the public’s expectations, the best it can do is to follow the policy shown in equation 13. This policy, however, is myopic. Because it essentially ignores the potential benefit o f reducing the public’s expectations for inflation, it generates an "inflationary bias” for the econom y. That is, had not announced its private information before wages were set. It should be noted that, since such cheating strategies are not consistent with the public’s expecta tions, they are implausible equilibrium strategies and are assumed not to be observed in equilibrium. JULY/AUGUST 1990 8 inflation, on average, exceeds the Central Bank’s target level by A7f without the benefit o f in creasing average output above the natural level. It is important to note that the inflationary bias w ould emerge even if dt w ere not private infor mation, as long as the Central Bank did not try to influence the markets’ expectations.10 The Central Bank’s expected one-period utility in this regime can be found by using equations 8 and 13 with g' = rc'+ A '/f and d' = 0: (14) E, { u } = - ( l + f)ot2- ( l + (l/f))A*2, for all t. Because the variance o f inflation and output are the same as in the first-best regime, o\, the only difference betw een equations 10 and 14, A*2/f, is the Central Bank’s one-period disutility o f the inflationary bias or, equivalent ly, the inefficiency o f taking the market’s expec tations as given. Note that the larger A* (which reflects the difference between the Central Bank’s and the public’s target for output) and the smaller f (the Central Bank’s preference for inflation stability relative to output stability), the larger is the inflationary bias. The inflationary bias is not easily avoided without the ability to make commitments. The problem stems from the Central Bank's incen tive to create surprise inflation. This incentive to cheat, given expectations, ultimately stems from the insufficient num ber o f instruments available to the Central Bank. In the present model, the Central Bank has tw o objectives with only one instrument. If it had tw o independent instruments, the Central Bank could achieve both o f its goals simultaneously.11 Alternatively, if the Central Bank “ignored” its goal o f output stabilization or f became infinitely large, then the credibility problem would disappear and 10ln this regime, as in the first-best outcome, expected infla tion is independent of d,, since d, is fully accommodated by the myopic policy. Nevertheless, because the presence of private information makes it difficult for the Central Bank to avoid the inflationary bias, as discussed below, it would like to be able to reveal its private forecast truthfully and precisely. See, however, Cukierman and Meltzer (1986) who show that the Central Bank might prefer to maintain the secrecy of its private information when it can not control the growth of the money perfectly. In their analysis, maintained secrecy about its changing preferences permits the Central Bank to engineer inflation surprises when desired. "Actual policy and expected policy are not independent in struments provided that the public is rational and forwardlooking. If it were not, however, the Central Bank would optimally announce g* = g ,-A *, where g, = n *+ d t, so that the Central Bank could systematically fool the public. If the FEDERAL RESERVE BANK OF ST. LOUIS there would be no inflationary bias in equilibri um.12 But, with an insufficient number o f in struments, the Central Bank’s incentive to sur prise the public remains, making the first-best policy dynamically inconsistent and not credible, thereby calling into question the feasibility of the first-best solution. REPUTATIONAL CONSIDERATIONS If the Central Bank did not possess any pri vate information, then a legislated rule could be imposed to force the Central Bank to follow the first-best policy. Even if it w ere not feasible to enforce such a rule, the Central Bank could recognize the importance o f its "reputation” to eliminate or mitigate the inflationary bias.13 To see w hy its reputation could be important, suppose the Central Bank announces that it will always follow the first-best policy as shown in equation 9. Further, assume that the public al ways expects the Central Bank to adhere to that policy, provided that it never has cheated in the past by having deviated from the first-best policy. Through its policy actions, then, the Cen tral Bank can maintain a reputation for not deliberately creating surprise inflation. If, however, the Central Bank w ere to cheat, then people would expect the Central Bank to continue to cheat in the future. Once having lost its reputation by cheating, the Central Bank is "punished.” Anticipating that the Central Bank will continue to cheat in the future because it has done so in the past, people will incorporate an inflationary bias into their wage specifica tion. Given this specification for expectations, public believed that announcement, the Central Bank’s ex pected one-period utility could increase to -(1 + f)aj. 12lf the objective function in (7) were interpreted as a socialwelfare function, then the analysis above suggests that ap pointing a “ conservative” Central Banker (i.e., one whose concern about pursuing a goal of inflation stability exceed ed that of society) would enhance social welfare. See Rogoff (1985) for a detailed discussion of this point. In deed, this is the thrust of Representative Stephen L. Neal’s recently proposed legislation to make price stability the ultimate objective of the Federal Reserve System (H.R. Res. 409). But also see Neumann (1990) who argues that strengthening the independence of the Central Bank could similarly help to avoid the credibility problem in monetary policy without explicitly imposing a goal of price stability on the Central Bank. 13See, for example, Barro and Gordon (1983). 9 the Central Bank can do no better than to follow the m yopic policy shown in equation 13 once having cheated. During the "punishment,” the outcom e w ould return to the m yopic solu tion that includes the inflationary bias, A*/f.14 In some cases, the Central Bank’s concern for its reputation can provide the same result as binding commitments when there is no private information. The critical condition is that the expected long-term gain from eliminating the in flationary bias must always exceed the expected short-term gain that could be realized by creat ing surprise inflation. The long-term gain is simply the present discounted disutility o f the inflationary bias, ^ P ^ ^ short-term gain is the difference betw een the expected oneperiod utility if the Central Bank w ere to cheat and the expected one-period utility from adhering to the first-best policy, A*2/(l + f) Note that as the Central Bank’s discount factor, ft, increases (that is, as it cares more about the future), the ex pected long-term gain from maintained reputa tion is m ore likely to exceed the short-term gain from cheating in the current period. Hence, as ft increases, the Central Bank’s concern for its reputation is m ore likely to support the firstbest outcome. Even if reputational considerations w ere not a perfect substitute for binding commitments to achieve the first-best outcome, they could still diminish the magnitude o f the equilibrium infla tionary bias. As long as the threat o f punish ment is sufficiently large, the Central Bank will be induced to adhere to the reputational policy 14Making this reputational mechanism effective, in the pre sent model, requires that the Central Bank is infinitelylived or has a finite but uncertain lifetime, which is consis tent with the Central Bank’s objective function shown in (7). If the Central Bank were to live a finite and certain number of periods, T, then it would always cheat in the last period, T. But, if the public expects such behavior, the period T outcome would just be the myopic solution. Along this line of reasoning, the solution unravels and the reputational mechanism cannot diminish the inflationary bias below A*/f. Alternatively, if the Central Bank were finitely lived, but its preferences were private information (e.g., the value of the parameter f), the Central Bank could “ build” credibility as an inflation-fighter by signaling with monetary policy actions. See, for example, Backus and Driffill (1985). 15Suppose, for example, that the Central Bank announces g, = k + 7i* + dt, where k is the average inflation in ex cess of the optimal rate (0 ^ k < A 7 f) and d, is public infor mation. (Note that when k = 0, this policy is simply the first-best one and when k = A'/f the policy is the myopic one.) Provided that k< A*/f, the temptation for the Central Bank to cheat, given by (A2+ fak! -2fA'k)/(1 +f), will be that involves a smaller (if not zero) inflationary bias.15 Hence, in the reputational equilibrium, cheating is never observed. The presence o f private information, however, greatly complicates this situation, influencing the possibilities for cheating. Specifically, because the public does not observe dt (the Central Bank’s private forecast) directly, it can never be certain that the Central Bank has actually imple mented the reputational policy that depends on dt. The public can easily verify that money growth equals the Central Bank’s announced reputational policy. But the public cannot be sure that the Central Bank’s announcement about d, is truthful. Indeed, as shown below, the Central Bank has an incentive to misrepre sent its private information. W H Y ARE PRECISE ANNOUNCE MENTS NOT FEASIBLE? The existence o f private information weakens the ability o f reputational considerations to achieve the efficient outcome. This can be il lustrated by showing that it is impossible to force the Central Bank to adhere to the firstbest policy, because the Central Bank cannot make credible announcements that precisely reveal its private information. Suppose that the Central Bank could be forced to adhere to a specified policy, but could not be forced to reveal its private information credibly and precisely.16 For example, the following rule might be legislated: positive but will decrease as k increases. The general con dition for reputational considerations to work is that this temptation be less than the expected gain to maintained reputation given by _ i_ _ (A2. _ f kA , 1-/3 V f ) which is also positive as long as k<A '/f. This gain also decreases as k increases. Even if the expected present discounted gain from maintained cooperation is smaller than the Central Bank’s temptation to cheat for k = 0, the reputational equilibrium inflationary bias, k, can be less than A'/f, if the temptation decreases faster than the ex pected present discounted gain as k < A '/f increases. 16That there is no separate mechanism to force the Central Bank to reveal its private information might seem puzzling. For example, in the United States, Congress or the Ad ministration could set up an agency to monitor the Central Bank’s activities and take part in formulating monetary policy, whereby the private forecasts can be revealed to the public. Why such an arrangement is not adopted is beyond the scope of this analysis. JULY/AUGUST 1990 10 (15) gt = n* + df, fo r all t, w here df denotes the Central Bank’s announcement o f its private forecast. If that an nouncem ent w ere believed by the public, the public w ould form the following expectations: g'=n* + df and n‘ = n". With these expectations, before setting its policy in period t, the Central Bank w ould announce optimally (16) d f= d , + A 7 (l + f). If the public w ere to believe the Central Bank’s announcement, the Central Bank would be able to disguise its cheating policy (shown in equa tion 11) as the first-best policy by overstating the value o f its forecast.17 In this case, the Cen tral Bank could drive output above its natural level by A*/(l + f). But, as in the case o f simple cheating, the Central Bank’s incentive to lie, w hich also fun damentally stems from its incentive to create surprise inflation, will be fully recognized; as a result, no one will believe the announcement. Given that the public cannot determine with certainty whether or not dA= d„ it can do no better than to protect itself from surprise infla tion by setting wage grow th equal to rr* + A’ /f.18 Because the Central Bank’s forecast is private information, a legislative approach depending on that information is not effective in achieving a better outcom e than the m yopic solution.19 Similarly, the Central Bank’s private information obscures the relevance o f reputational con siderations to im prove upon the m yopic out come. Although people can see whether the Central Bank has implemented its announced policy—for example, the policy show n in equa tion 15—they cannot verify that its announce 17This can be seen by substituting equation 16 into equation 15. To verify that equation 16 is the optimal announce ment, substitute equation 14 into the Central Bank’s oneperiod utility function (8) and choose d * to maximize the expected value of (8) subject to the public’s expectations g;—<5;=n*. The Central Bank would lie in the same manner if it were not necessary to make its announcement until after the policy was implemented. 18To see this, note if the Central Bank were to act on its in centive to create surprise inflation given the public’s ex pectations, it would set its policy optimally to satisfy the first-order condition in equation 12. Rearranging equation 12 and using Et(d,) = d„ one can verify the following: q = g '~ d' + fa‘ +A' + d. 1+f 1 +f Noting that gt - d, equals the Central Bank’s expectation for inflation given d „ Et (nj, and g;-<J; = n;, the expression above implies that FEDERAL RESERVE BANK OF ST. LOUIS ment truly reflects the value o f its private forecast (that is, df = dt), unless the forecast w ere always perfect (that is, £t= 0 for all t). Hence, the public cannot evaluate the Central Bank’s reputation based on past policy actions. A CONSTANT MONEY GROW TH RULE AND THE ROLE OF NOISY ANNOUNCEMENTS Although the Central Bank cannot make credi ble announcements that precisely state its private information, it can make announcements that have some informational content. In a recent study, Stein (1989) applies the w ork o f Crawford and Sobel (1982) to show that, through noisy announcements or "cheap talk,” the Central Bank can reveal its private information partially. In his application, w here the Central Bank’s private in formation concerns its objective for the target exchange rate, Stein illustrates h ow the Central Bank can make announcements o f a range in which its target falls. Because the announce ment does not state the exact value o f the Cen tral Bank's target, it is a noisy announcement. These announcements are a costless form o f communication in that no resources are used in making them. But the announcements are credi ble because the Central Bank w ould incur an implicit cost if it w ere to lie. This cost is suffi ciently large to induce the Central Bank to reveal its private information truthfully, though not precisely. An application o f Crawford and Sobel’s (1982) analysis to the present model, however, shows that noisy announcements might not be as "cheap” as Stein’s (1989) analysis would suggest. Since Et(nt)> n ; for rrj^rr' + A'/f, the Central Bank always has an incentive to create surprise inflation unless the public incorporates the inflationary bias A'/f into its wage specification. 19Garfinkel and Oh (1990a) have shown how a legislative ap proach that is independent of the Central Bank’s private forecast can achieve a better outcome than the myopic solution studied above. With a multi-period (N periods) average targeting procedure, requiring x g, = NrT, the 1 Central Bank can diminish the magnitude of the infla tionary bias that emerges in equilibrium. This procedure is not efficient, however, in that it necessarily limits the Cen tral Bank's flexibility to stabilize output and inflation. Nevertheless, it can permit more flexibility than a strict constant money growth rule. 11 In contrast to the present model, Stein’s model implies that, if it w ere possible to force the Cen tral Bank to reveal its exchange rate target truth fully and precisely, then the first-best outcom e could be obtained. Accordingly, noisy announce ments alone can easily achieve a better outcome than no announcements or complete secrecy. (otherwise desirable) reactions to the part of m oney demand disturbances predicted by the Central Bank.22 As such, this rule produces a higher variance o f the public’s forecast error for inflation and, hence, a higher variance of output than in both the first-best and myopic regimes. The credibility problem in monetary policy in the present framework, however, is slightly m ore complicated. As indicated above, even if it w ere possible to make the Central Bank reveal its private forecast truthfully and precisely, im posing an additional restriction on policy either through a legislative rule or reputational con siderations would be necessary to ensure that the Central Bank follow the first-best policy. That is, even if the public's expectations, g' and n‘, included information about d„ the Central Bank w ould have an incentive to surprise the public (according to equation 12) unless n't also w ere to incorporate the inflationary bias, A*/f.20 But the Central Bank has no motivation to reveal dt if it cannot reduce or eliminate the inflationary bias in doing so. Similarly, the Central Bank’s incen tive to create surprise inflation would not disap pear if it w ere to make noisy announcements about its private forecast and could contaminate those announcements. The Central Bank's expected utility under this regime without any announcements is given by A Constant M on ey Growth Rule Because o f this incentive to surprise the market with inflation, limiting the degree o f flex ibility permitted in monetary policy is necessary to ensure that the announcements contain some information while allowing the Central Bank to avoid the inflationary bias. In other words, a rule for monetary policy must be imposed to "tie” the hands o f the Central Bank. As indicated above, for this constraint to be effective, the rule must be independent o f the private infor mation.21 For example, legislation could require (17) Although this constant money grow th rule eliminates the inflationary bias, it precludes any 20See footnote 18. 21Whether it is possible to enforce a legislated rule is beyond the scope of this paper. Of course, reputational considerations might be able to support the same rule. To simplify the discussion, the analysis assumes that it is possible to enforce a legislated rule that does not depend on the Central Bank’s private information. 22lf there were another shock, say, in the supply equation, and the Central Bank’s information about this shock were (18) Et{u } = -( 1 + f)o2-A * 2, for all t. Expected utility in this regime will ex ceed that under the myopic regime only if A*2/ f > ( l + f)oj. This condition underscores the Central Bank’s trade-off betw een eliminating the inflationary bias and eliminating flexibility in monetary policy with the constant m oney growth rule. The larger is the inflationary bias that emerges in the myopic outcom e (that is, the smaller f and/or the larger A*), the m ore likely this condition will be satisfied. The Central Bank is less likely, h ow ever, to prefer a constant m oney growth rule over the myopic policy the larger the variance o f the com ponent o f the money demand distur bance predicted by the Central Bank, o\, which captures the expected benefit o f being able to react to dt. Because the legislated rule in equa tion 17 does not permit the Central Bank to react to its forecast o f the disturbance to money demand to stabilize inflation, the variance o f in flation and output increase to o]. Nevertheless, if the possible benefits o f maintained flexibility are not too large (that is, if o\ is small), the Cen tral Bank might prefer to be constrained not to react to its private forecast to avoid the infla tionary bias. It is important to note that, even with this rule, the Central Bank still would not precisely reveal its forecast. In particular, given g, = n*, the Central Bank would want to overstate the value o f its forecast according to (19) df = dt + A\ not private, then the legislated rule could provide flexibility to react to this shock. Moreover, not all flexibility needs to be removed from policy in this model. The constant money growth rule is not the only way to tie the hands of the monetary authority to make the announcements mean ingful. The imposition of a multi-period average targeting rule that permits some flexibility would also work; however, with this constraint, the inflationary bias would not be eliminated totally. See Garfinkel and Oh (1990a). JULY/AUGUST 1990 12 Equation 19 illustrates again that the credibility problem o f monetary policy is not easily resolved in the presence o f private information. But if dt w ere not private information, the Central Bank’s expected one-period utility with a constant m oney grow th rule w ould be hand, if the Central Bank announced that dt fell in the low er range, [-D ,a ], the market would expect a higher inflation rate equal to the dif ference between ir' and the expected value of dt given that it falls som ewhere betw een - D and a. Call this conditional expectation dL' (20) E ,{u } = - f o 2- A ' 2. Hence, the Central Bank would prefer to disclose dt under a constant m oney growth rule even though it cannot do so precisely. If d, is greater than - D but less than a, then the Central Bank’s expected one-period utility by announcing dt i [-D ,a ] must be greater than or equal to that by claiming d, t [a,D] for the form er announcement to be credible. That is, Noisy Announcements (21) - E t{(dL- d t- £ t-A *)2} > - E t{(dh- d t- £ t- A ') 2}. By making noisy announcements about its forecast, the Central Bank could enhance its ow n welfare under the rule. Given that it must follow the rule in equation 17, the Central Bank cannot actively pursue its goal to stabilize infla tion and output by reacting to dt. Making noisy announcements, as an alternative policy tool, permits the Central Bank to pursue its goal of stabilizing output. Specifically, the Central Bank could partly influence expectations by announc ing a range in which its forecast falls, thereby reducing the variance o f the public’s inflation forecast error and, in turn, reducing the vari ance o f output. The inequality in equation 21 would be reversed if dt were greater than a and less than D. Final ly, if dt= a, then the Central Bank must be indif ferent between announcing the higher and low er ranges. To take a concrete example, suppose that the Central Bank announces that d, lies either bet w een - D and a or betw een a and D.23 For any announcement to contain some information about d„ the Central Bank must perceive that lying is costly. The cost, however, cannot be directly imposed by the market upon observing d, because, as mentioned earlier, the market cannot infer the true value o f dt from that ob servation. Rather, the cost o f lying about dt is implicitly contained in how such a lie would af fect the market’s expectations about d,. Suppose the Central Bank w ere to announce that dt fell in the higher range, [a,D]. Given that announcement and the m oney grow th rule shown in equation 17, the market form s an ex pectation about future inflation. This expecta tion would equal the Central Bank’s target rate o f inflation, it*, minus the expected value o f dt given that it lies som ewhere betw een a and D. Call this conditional expectation dh. On the other 23See the appendix for a more detailed example. Also see Garfinkel and Oh (1990b). 24Again, see Garfinkel and Oh (1990b). Their analysis pro duces a somewhat surprising result: under the conditions FEDERAL RESERVE BANK OF ST. LOUIS This last condition can be used to determine the dividing point o f the distribution o f d„ a, such that for all possible values o f dt, the Cen tral Bank’s announcement is credible. The deter mination o f the dividing point from that condi tion ensures that the Central Bank will not act on its motive to lie about the range in w hich d, falls. For example, w hen dt is in the low er range, the Central Bank will not announce that d, is in the upper range. If it did so, the public’s infla tionary expectations would fall by a sufficiently large amount that, in turn, drives output too far from the Central Bank’s output target and, hence, renders lying undesirable. By making noisy announcements about its private forecast while adhering to the constant m oney grow th rule, the Central Bank can en hance its expected utility above what it would be w hen it simply follows the rule. This is not to say that the Central Bank will always choose to make noisy announcements. As illustrated with a more specific example in the appendix, the Central Bank w ould prefer to maintain full discretion and secrecy, the more it cares about inflation stability, the less the difference between its and the public’s output goals, and the more accurate the private forecast.24 The basic intuition here is essentially the same as that used w hen discussing the merits o f a that noisy announcements are more likely to be preferred by the Central Bank, the credibility problem is more severe so that, at the same time, these announcements cannot be particularly informative. 13 simple constant m oney grow th rule over those o f the m yopic policy. The presence o f private information forces the Central Bank to face a new trade-off betw een removing the inflationary bias and limiting flexibility in policy. But the money grow th rule with noisy announcements is more likely to dominate the m yopic policy than the rule by itself. Although both output and inflation will have a greater variance in the regime with noisy announcements than in the myopic regime, the variance o f output will be smaller in this regime than w hen the Central Bank simply follows a constant m oney growth rule. The elimination o f the inflationary bias possible with the constant m oney grow th rule, com bined with the slight reduction in the vari ance o f output possible with noisy announce ments, provide the main benefits that would make abandoning the m yopic policy—that is, maintaining complete secrecy with full discre tion -desirable from the Central Bank’s perspective. CONCLUDING REMARKS This article has examined the possibility o f ful ly or at least partially removing secrecy in m one tary policy. In the context o f a model in which the Central Bank has an incentive to create sur prise inflation, the Central Bank w ould like to reveal its private information, w hereby it could easily avoid an inflationary bias. The Central Bank’s private information com bined with its in centive to surprise individuals gives rise to a credibility problem in monetary policy that is nearly impossible to resolve. Neither reputational considerations nor binding commitments to force the Central Bank to adhere to the firstbest policy are effective in improving upon the myopic solution if the public never directly ob serves the Central Bank’s private information. Although the Central Bank cannot make precise announcements, it can make announce ments that partially reveal its private inform a tion. By announcing a range in w hich its forecast falls and adhering to the constant money growth rule, the Central Bank can avoid the inflationary bias and influence the market’s expectations in a discrete way to low er output variability below that generated by a simple constant m oney growth rule alone. Nevertheless, some secrecy remains. M oreover, the Central Bank might prefer to maintain complete secrecy. Unlike Stein’s (1989) result that there is always room for im prove ment with noisy announcements, in the context o f the more general model developed here, noisy announcements require constraints on flexibility that can be permitted in the conduct of monetary policy—for example, a legislated con stant m oney grow th rule. The constraints are costly if they preclude desirable reactions to disturbances in the economy. More generally, the analysis suggests that leg islation requiring the Fed to disclose the FOMC’s decisions immediately after its meeting might be o f little value. If the Central Bank has private information about the econom y that influences its decisions and has an incentive to surprise the public, it will not release this information truthfully and precisely. The Central Bank’s in centive to misrepresent its private information detracts from the value o f any information it releases. That noisy announcements can w ork in en hancing the efficiency o f monetary policy only under restrictive conditions prompts a general but m ore fundamental conclusion. In the pres ence o f private information, the Central Bank faces a trade-off between higher-than-desired average inflation and limited flexibility. Without eliminating the ultimate source o f the credibility problem —namely, that the Central Bank has too few tools to achieve its ultimate goals—this con sequence o f the strategic considerations o f m on etary policy is not easily avoided. REFERENCES Backus, David, and John Driffill. “ Inflation and Reputation," American Economic Review (June 1985), pp. 530-38. Barro, Robert J., and David B. Gordon. “ Rules, Discretion and Reputation in a Model of Monetary Policy,” Journal of Monetary Economics (July 1983), pp. 101-21. Canzoneri, Matthew B. "Monetary Policy Games and the Role of Private Information,” American Economic Review (December 1985), pp. 1056-70. Crawford, Vincent P., and Joel Sobel. “ Strategic Information Transmission,” Econometrica (November 1982), pp. 1431-51. Cukierman, Alex. “ Central Bank Behavior and Credibility: Some Recent Theoretical Developments,” this Review (May 1986), pp. 5-17. Cukierman, Alex, and Allan H. Meltzer. “A Theory of Am biguity, Credibility, and Inflation under Discretion and Asymmetric Information,” Econometrica (September 1986), pp. 1099-128. Garfinkel, Michelle R., and Seonghwan Oh. “ Strategic Discipline in Monetary Policy with Private Information: Op timal Targeting Periods,” UCLA Working Paper No. 584 (January 1990a). JULY/AUGUST 1990 14 _______ “ When and How Much To Talk: Credibility and Flexibility in Monetary Policy With Private Information,” Federal Reserve Bank of St. Louis Working Paper No. 90-004 (June 1990b). Goodfriend, Marvin. “ Monetary Mystique: Secrecy and Cen tral Banking,” Journal of Monetary Economics (January 1986), pp. 63-92. Grossman, Herschel I., and John B. Van Huyck. “ Sovereign Debt as a Contingent Claim: Excusable Default, Repudia tion, and Reputation,” American Economic Review (December 1988), pp. 1088-97. Hamilton, Lee H. “ Regulating the Federal Reserve Board,” Christian Science Monitor, October 20, 1989. Kydland, Finn E., and Edward C. Prescott. “ Rules Rather than Discretion: The Inconsistency of Optimal Plans,” Jour nal of Political Economy (June 1977), pp. 473-91. Lucas, Robert E., Jr. “ Some International Evidence on Output-lnflation Tradeoffs,” American Economic Review (June 1973), pp. 326-34. Mooney, Richard E. “ Don’t Fiddle with the Fed,” New York Times, October 3, 1989. Neumann, Manfred J.M. “ Precommittment to Rules in Monetary Policy: A Comment,” in Michael T. Belongia, ed., Monetary Policy on the 75th Anniversary of the Federal Reserve System, Proceedings of the Fourteenth Annual Economic Policy Conference of the Federal Reserve Bank of St. Louis (Kluwer Academic Publishers, forthcoming). Rogoff, Kenneth. “ The Optimal Degree of Commitment to an Intermediate Monetary Target,” Quarterly Journal of Economics (November 1985), pp. 1169-89. Rosenbaum, David E. “ Little Chance Seen for Bills on Fed Rein,” New York Times, October 10, 1989. Stein, Jeremy C. “ Cheap Talk and the Fed: A Theory of Im precise Policy Announcements,” American Economic Review (March 1989), pp. 32-42. Uchitelle, Louis. “ Moves On in Congress to Lift Secrecy at the Federal Reserve,” New York Times, August 24, 1989. Appendix Why Not Complete Secrecy?—An Example This appendix illustrates with a simple exam ple h ow noisy announcements w ork and under what conditions the Central Bank would prefer to use them rather than not reveal anything about its private forecast. For simplicity in what follows, suppose that 6t has a uniform distribu tion bounded by - D and D.1 Consider the sim plest example w here there is only one dividing point, a1( over that distribution.2 Then, given an announcement by the Central Bank, say, that dt falls in the low er range, [ - D.a^, and (17) in the main text, the public will form expectations ac cording to the following: w hich dt falls, it must always be indifferent b e tween announcing the ranges, [-D ,a ,] and [a,,D] w hen dt= a r - D + a, (Al) ti'( - D , a,) = i t -------- ----- • D +a - Et(( — - —L - a, - £ t-A *)2], With this influence on the public’s expecta tions, it is important to ensure that the Central Bank will announce the correct range. For ex ample, if d, e [ —D,a ,], the Central Bank should not announce dt £ [a,,D], To guarantee that the Central Bank will not misrepresent the range in 1Hence, the probability that dt = d, where d is any possible realization of d„ is the same for any value of d: 1/2D. The distributions of d, and t, are not specified here. They need only be independent random variables with zero means and finite variances that sum to the variance of d,. See Crawford and Sobel (1982) for a more general analysis of the noisy announcement equilibrium. FEDERAL RESERVE BANK OF ST. LOUIS Formally, this condition, called the "arbitrage condition,” is written as (A2) Et(u([-D ,a,], dt)] = E,fc([a,,D], dt)], or equivalently, —D + a - E t[ ( ---- -— L - a .- t .- A * ) 2} = w here dt= a,. For this condition to be satisfied, at must equal -2A*. The basic idea here is that, given that the Central Bank must follow the constant m oney growth rule, its incentive to lie depends on its 2See Garfinkel and Oh (1990b) for a derivation of a more general noisy announcement equilibrium of size n in this framework. (In this particular example, with n = 2, a0= - D and a; = D.) 15 forecast. The dividing point a,, determined from the arbitrage condition, implies that if the Cen tral Bank w ere to overstate the value o f its fore cast, when dt< a j, it would have to do so by an amount so large that it is too costly to lie. nouncements with any number o f dividing points (greater than or equal to 1) will always be better than a simple constant m oney growth rule provided that the first step is greater than - D .4 Note that the dividing point is such that for d ^ a , the announcement is m ore precise—that is, informative. M ore generally, w hen there are n steps, the subintervals becom e longer as they move away from the low er bound. For example, consider w hen there are tw o dividing points, at and a2. In this case, the arbitrage condition re quires aj= - D /3 - 4 A ” and a2= D /3 -4 A ‘ . The length o f the first interval from - D to a1 equals 2D /3-4A *; the length o f the next interval equals 2D/3; and, the length o f the last interval equals 2D/3 + 4A*. W hen the disturbance is smaller (closer to -D ), the Central Bank's incentive to overstate the value o f the forecast is smaller. In addition, the Central Bank’s utility under this regime can be greater than that under the m yopic regime. In the present example, this condition is given by Although a constant money growth rule is not first-best in that it does not permit (otherwise desirable) reactions to the Central Bank’s fore casts o f money demand disturbances, it does eliminate the inflationary bias. W hen the Cen tral Bank also makes noisy announcements, it can enhance its expected welfare above that with a simple constant m oney grow th rule. With only one dividing point, its expected one-period utility is given by: (A3) u = - ( 6 A*2+ D 2(f + |))/3, which is always greater than the Central Bank’s utility w hen it simply follows a constant m oney grow th rule, provided that A *<D /2.3 Note that this condition will be satisfied by the require ment that a, > - D . M ore generally, noisy an- 3See Garfinkel and Oh (1990b), who show that, for a general noisy announcement equilibrium of size n, the Central Bank’s expected one-period utility is given by u, = -(A *2(n2+ 2) + D2(f+ 1/n 2))/3. Under the specifications for the distribution of <5t, the one-period expected utility for the Central Bank is -(1 +f) D2/3 - A * 2 when it follows a simple constant money growth rule. This can be easily verified by either using the above expression for expected utility with n = 1 or by using equation 18 and noting that the variance of a random variable which has a uniform distribution bounded by x, and x2 is given by (X j-x ^ /1 2 . 4This no-nonsense condition is automatically satisfied by the requirement that the partition equilibrium of size n is feasible. See Garfinkel and Oh (1990b). 5ln the myopic regime, the Central Bank’s one-period ex pected utility is - (1 +f)(1 - a 2)D2/3 -(1 +(1/f))A'2 since, by the definition of a] 1 - a 2= o,2/oj. (A4) D2(a2(l + 4 f)-3 (1 -cr))/12 < A2( l - f ) /f , where a1=a\lal with 0 < a < l . 5 The parameter a captures the degree o f accuracy o f the Central Bank's forecast. As a approaches 1, the Central Bank’s forecast is generally m ore accurate. The condition in (A4) is weaker than that for a strict rule to dominate the myopic policy. Nev ertheless, this condition is quite strong, reflecting the idea that, although the inflationary bias can be avoided, the resulting loss o f flexibility in this regime can be costly. In fact, when the monetary authority's forecast is extremely ac curate (that is, a approaches 1), a sufficient con dition for the myopic policy to dominate the constant m oney grow th rule with noisy an nouncements and one dividing point, a,, is simp ly that f > l . If the Central Bank cares more about inflation stability than about output stabil ity (and its forecast is extremely accurate), then it will not prefer noisy announcements, with a,, over the myopic policy. W hen a is close to 1, it can be shown that, given that f > l , noisy an nouncements with any number o f partitions will not be desired by the Central Bank.6 Nevertheless, noisy announcements might en hance the Central Bank’s utility if f < l . Even if a strict constant m oney growth rule without 6See Garfinkel and Oh (1990b). The intuition here, as discussed in the main text, follows simply from the trade off between the benefits of reducing the inflationary bias and the benefits of maintained flexibility. Assuming that a is sufficiently close to 1, the larger is f, the smaller is the inflationary bias that emerges in the myopic regime and the smaller is the benefit of avoiding the inefficiency of that bias relative to the expected costs of not reacting to money demand disturbances. In the case that the elasticity of output with respect to unanticipated inflation were not equal to 1, the sufficient condition for the Central Bank to prefer the myopic policy is that f be greater than the square of that elasticity. The smaller that elasticity, the greater the likelihood of the Fed preferring the myopic policy. For example, if the elasticity were equal to 1/2, then f > 1/4 would imply that the myopic policy dominates the constant money growth rule with noisy announcements. JULY/AUGUST 1990 16 any announcements does not dominate the m yo pic policy, there can be room fo r improvement with noisy announcements and the strict rule. In the case o f one dividing point, there can be room for improvement provided that f < 1/2 even w hen the private forecasts are extremely accurate (that is, a is close to one). More generally, the condition in equation A4 implies that noisy announcements are m ore likely to be preferred over no announcements with full flex FEDERAL RESERVE BANK OF ST. LOUIS ibility in monetary policy, the less accurate the Central Bank's forecast (when there is a smaller desire for flexibility in monetary policy). Fur ther, the larger the difference betw een the out put goals o f the Central Bank and the public and the smaller the Central Bank’s relative pre ferences for inflation stability, the Central Bank is less likely to prefer complete secrecy over noisy announcements. 17 Peter Bofinger Peter Bofinger, an economist at the Landeszentraibank in Baden-Wiirttemberg, Germany, was a visiting scholar at the Federal Reserve Bank of St. Louis. Scott Leitz provided research assistance. The German Monetary Unification (Gmu): Converting Marks to D-Marks1 TJLHE MONETARY and econom ic unification of the East and W est German econom y is a task without precedent in peacetime econom ic histo ry. It not only merges tw o countries with strong ly divergent income and productivity levels, but also unifies tw o economies with radically dif ferent econom ic structures—the German Dem o cratic Republic's (GDR) centrally planned econ o my and the Federal Republic o f Germany’s (FRG) "social market econom y.” Although conventional wisdom calls for gradualism in the process of monetary and econom ic unification o f capitalist economies and in the transition process from socialism to a market econom y, in the German case, these tasks will be accomplished virtually overnight.2 The legal basis for the unification process is the treaty ratified by the East and West German 1At the present time, discussion and analysis of the Gmu has appeared primarily in German newspapers. I do not quote these articles explicitly in the paper. Publications of the Deutsche Institut fuer Wirtschaftsforschung, Berlin, of Norbert Kloten, Karl Otto Poehl, Helmut Schlesinger and Horst Siebert provided valuable insights and analysis. I have profited from many discussions with Norbert Kloten and members of the Research Department of the Federal Reserve Bank of St. Louis and of the Volkswirtschaftliche Abteilung der Landeszentraibank in Baden-Wuerttemberg. parliaments on June 21, 1990, which took effect on July 1, 1990. The agreement outlines the principles for monetary union, the econom ic and social community o f the tw o states and the fiscal reform o f East Germany. The arrangements for monetary union, which involved the replacement o f the East German “Mark” (M) by the West German "Deutsche Mark" (DM), established the rates at which East German financial stocks and flows would be converted from their Mark values to D-Mark values. A 1M:1DM rate was applied to East Ger man wages, salaries, rents, leases and pensions. Savings accounts o f GDR citizens w ere con verted at a 1M:1DM rate up to a limit o f M 4,000 (approximately $2,425 at the current DM/$ exchange rate) for persons betw een 15 and 59 years o f age. The corresponding limit 2For the standard arguments, see Committee for the Study of Economic and Monetary Union (1989), (“ Delors Commit tee” ). In the debate on economic transformation of socialist countries see, for instance, Daviddi and Espa (1989). There seems to be, however, a growing awareness that partial reforms generate only limited success; see Roe and Roy (1989). JULY/AUGUST 1990 18 was M 6,000 for older persons and M 2,000 for younger persons. A 2M:1DM rate was used to convert all other financial assets and liabilities o f GDR residents. Mark assets held by individuals w ho live outside the GDR w ere converted at a 3M:1DM rate. The legal fram ew ork for the econom ic com munity betw een the tw o states and for the transformation o f East Germany’s econom ic order involves nearly a complete adoption o f the FRG’s econom ic laws and regulations by East Germany. These changes include the restoration o f private property and competition in East Germany and the free movement of goods, services, labor and capital between East and West Germany. In addition, social welfare, pensions, unemployment and health insurance programs similar to those in West Germany were introduced into East Germany; any deficits in these new programs will be financed tem porarily by the FRG. Pensions in East Germany w ere converted to DM values based on net East German incomes; an East German w orker can receive a maximum pension of 70 percent o f his or her net incom e after 45 years o f employ ment. The agreement also guarantees that the DM value o f East German pensions cannot fall below their form er Mark equivalents. Under the agreement, the East German government will abolish its old system o f high tax levies on state enterprises and introduce, in stead, a system o f income and value added taxes consistent with those o f West Germany. Future debt issues by the GDR government must be issued directly via the Deutsche Bundesbank or with its approval. The FRG will finance two thirds o f the East German deficits from 1990 through 1994. For this purpose, a “German Uni ty” fund o f DM 115 billion was launched; it will be financed by a combination o f bond issues (DM 95 billion) and expenditure reductions in the FRG central government budget (DM 20 billion). 3See the detailed report of the Institute for International Finance (1990). “Data for the GNP are not available. 5The lower figure is an estimate of the Kiel Institute for World Economics; the higher figure was estimated by the five leading economic research institutes of the Federal Republic in their report of April 12, 1990. 6At the present DM/Dollar exchange rate of about 1.65 DM per Dollar, this equals $9,000 to $13,000. For comparison, 1989 per capita income in the United States was $21,000. FEDERAL RESERVE BANK OF ST. LOUIS GOALS AND PROBLEMS OF THE GMU The paper starts with a short analysis o f the econom ic situation in East Germany after the fall o f the Wall. It tries to identify both the goals o f the East German people and those of the W est German government w hich together have led to the present unification o f both Germanys. A brief outline o f the reform s necessary to transform the East German econom y are discussed first. The rest o f the paper focuses on monetary unification, certainly the most con troversial issue in the debate over unification. The East German Econom y After the Wall Fell The deep econom ic malaise o f the East Ger man econom y provides a good example o f the general failure o f the centrally planned econom ic systems o f Eastern Europe.3 Prior to W orld W ar II, the part o f Germany that now makes up the GDR was essentially as developed as those regions w hich now constitute the FRG. Data for 1936, for example, show that per capita income was 993 Reichsmark in the East and 996 Reichsmark in the West. Today, o f course, it is not as easy to assess the relative per capita incomes o f the tw o Germanys. The GDR’s administratively-set domestic prices and exchange rates do not accurately reflect its econom ic conditions; consequently, “official” data, w hen available, must be treated with skepticism. For example, the East German Statistical Office recently published the first o f ficial income estimate for East Germany; it reported that GDP was M 353 billion for 1989.4 Most West German estimates o f the GDR’s 1989 GNP range from M 280 billion to M 300 billion.5 If a 1M:1DM conversion rate is used with these estimates, the GDR’s 1989 per capita income was somewhere between DM 15,000 and DM 21,000,6 only about half o f that estimated for 19 Table 1 Basic Data for East and West Germany (1988) Unit East Germany West Germany Area thousands of sq. miles 41,768 96,094 Population thousands 16,675 61,715 Employment Agriculture1 Manufacturing1 Services1 Trade and transport1 thousands % of total % of total % of total % of total 8,594 11 47 25 18 27,306 5 40 36 19 GNP Mark/D-Mark billion (1989) 280-350 2,260 GNP per capita Mark/D-Mark (1989) Gross monthly salary Mark/D-Mark 1,250 3,192 Net monthly salary Mark/D-Mark 1,050 2,153 Monthly social security retirement benefits Mark/D-Mark 450 1,597 49 100 17,000-21,000 Labor productivity as a percent of West German labor productivity 36,600 1Data are for 1987. SOURCES: Official statistics of the GDR compiled by the Deutsche Bundesbank and Deutsches Institut fuer Wirtschaftsforschung. the FRG (DM 36,600) in 1989 (table 1). Thus, despite East Germany’s good educational system, its per capita income and, by proxy, its labor productivity, is estimated to be, at best, only half that o f West Germany.7 Of course, these comparative productivity figures are likely to prove misleading if used to predict what might occur after unification takes place; for example, GDR products previously produced and sold under a central plan designed to achieve autarky may not be able to com pete effectively with goods that can now be imported from the West. The econom ic disparity between the FRG and the GDR is further demonstrated by the ex tremely high environmental pollution in East Germany, its obsolete infrastructure, outdated manufacturing plants and the generally poor quality o f its housing stock. Another indication H'he 50 percent estimate for the GDR’s relative labor pro ductivity was made by the Deutsches Institut fuer Wirt schaftsforschung, Berlin, in 1987 for the year 1983. It is nearly identical to Collier’s (1985) estimate of 54 percent and to cross-country comparisons (see Cornelsen and o f the disparity between the tw o Germanys is shown by the relatively high proportion o f total employment devoted to agriculture and manufacturing in the GDR (58 percent) com pared to that in the FRG (45 percent); indeed, the GDR's current proportion o f employment in agriculture and manufacturing is roughly iden tical to that w hich prevailed in the Federal Republic over 20 years ago.8 In the past, the large difference in living stan dards between the tw o German states could be maintained only by the GDR’s actions to close its borders with the W est and prohibit virtually all unauthorized movement o f labor, capital, goods and services betw een East and West Ger many. Since the border became permeable in autumn 1989, m ore than 2000 East German citizens have moved into West Germany daily; Kirner (1990)). However, the Kiel Institute for World Economics estimates that GDR labor productivity is only about 35 percent of West German levels. 8See Gerstenberger (1990). JULY/AUGUST 1990 20 as a result, betw een then and the first several months o f 1990, the GDR’s population decreased by about 500,000 persons. This massive exodus was possible only because the West German constitution grants citizenship status to all East Germans. Among other things, this allowed East Germans w ho m oved to West Germany to obtain immediate social benefits (unemployment benefits, retire ment insurance and aid to the disadvantaged) that are tied to West German income levels. FRG unemployment payments, for example, are about 68 percent o f West German net incomes; in comparison, net incomes in the GDR are only about one-third o f that in the FRG.9 The substantial difference between West German unemployment benefits and East German in com e levels explains, in part, the massive migra tion o f East German workers. However, these specific incentives w ere eliminated on July 1, 1990, w hen the social community between both states was established. From that date, all social benefit payments to East Germans will be based on East German income levels, not on those in W est Germany. The migration o f many skilled w orkers to the FRG caused the econom ic situation in the GDR to substantially deteriorate. Since November 1989, GDR industrial production and em ploy ment has decreased and most East German enterprises have been unable to fulfill their pro duction plans. By the end o f April 1990, in dustrial production was 4.5 percent below its level one year before, and the number of employed persons had fallen by 4.6 percent. Shortages o f goods and services produced grow ing social unrest in East Germany. The Disparate Goals o f East Ger mans and West Germans Given the circumstances described above, the goals o f the GDR population are quite evident: They want to improve their relatively low stan dard o f living as quickly as possible. Given the disappointing econom ic results associated with socialism, they w ere generally unwilling to ex periment with a system part-way betw een socialism and capitalism. They chose, instead, Pensioners moving to West Germany received an average pension of DM 1,121 (1988), more than twice the average East German pension in Marks. 10ln fact, almost all West German economists as well as the Bundesbank preferred a more gradual approach involving FEDERAL RESERVE BANK OF ST. LOUIS immediate and complete integration with the Federal Republic o f Germany even though they knew that it would require total restructuring of the East German econom ic and political systems. The extreme political uncertainty in the GDR after the Wall fell, the obvious desire o f the East German population to unify both countries and the massive outflow o f East Germans into West Germany, w hich aggravated housing p ro blems in the FRG, left little room for political maneuvering in West Germany and little time to find a solution that w ould satisfy both East and W est Germans. Legally, o f course, West Ger many could not oppose rapid unification; the W est German constitution (Article 23 o f the “Basic Law") explicitly permits the East German states to join the Federal Republic without re quiring the consent o f either the W est German Government or its Parliament. This excluded a variety o f possible partial solutions and gradual approaches.10 Therefore, the main task facing West Ger many was to design a unification strategy that would restore the confidence o f East Germans in the future prospects o f East Germany and, at the same time, be compatible with the chief in terests o f W est Germans. Consequently, the debate in West Germany focused on the possi ble costs o f the unification process. Among the costs mentioned were: 1. The possible increase in the West German inflation rate, 2. The prospects o f either higher taxes or higher interest rates (due to increased FRG borrowing) resulting from increased FRG expenditures for East Germany, and 3. The wealth transfer from W est Germans to East Germans associated with the replace ment o f Mark-denominated savings and cur rency in the GDR by DM-denominated m on etary assets. Once the actual conversion rates are chosen, it is possible, albeit tentatively, to assess the im pact o f monetary unification on matters that concern the East and West Germans. The ten- either flexible or fixed exchange rates between the two currencies as an intermediate stage during the period of economic transformation in the GDR. 21 tative nature o f the assessment is chiefly due to the absence o f reliable data on the East German econom y and to the simultaneity o f the monetary and political integration with the transformation o f the East German economy. The short-term focus o f the analysis should not lead to the impression that the risks and p ro blems associated with unification o f the two Germanys are either substantial or pervasive. The strong overall consensus in both German states is that the long-term prospects o f unifica tion are positive and that East Germany has the potential to repeat the “econom ic miracle” achieved by West Germany from the 1950s to the present.11 REAL SECTOR REFORM Although this paper focuses primarily on monetary unification, a brief discussion o f the econom ic reform s necessary in the real sector o f the GDR econom y is needed. The Gmu itself will not improve the econom ic situation in East Germany substantially; it can provide, however, a sound monetary fram ew ork for an overall restructuring o f the GDR’s econom ic and legal system. A cornerstone o f real sector reform in the GDR will be the introduction o f free-market pricing and production. Previously, most pro duction and prices had been set by government agencies in accordance with their central plans. One consequence o f this system—as in many other socialist countries—was that these prices had been held essentially unchanged for years despite changes in demand and cost condi tions.12 For example, the GDR’s official index for consum er prices has shown virtually no m ove ment over the entire post-W orld W ar II era. Moving to a market-based econom y will re quire a number o f changes. First, the current pricing structure is distorted by large subsidies for some industries, especially food and energy (their subsidies totaled M 50 billion in 1988, about one-third o f total private expenditures in the GDR) and heavy taxes on other industries, primarily consum er durable goods (the tax total ed M 43 billion in 1988). These distorting in fluences on prices will have to be reduced. Second, the central planning approach to pric ing and production must be replaced by the usual market mechanisms that determine these decisions in free-market economies. Not only must prices be set by market conditions rather than by government bureaucrats, but also the extensive system o f state-owned enterprises must be privatized as well. In order fo r market prices and wages to successfully provide the signals for reallocating resources, the traditional "soft budget constraint” o f state-owned enter prises has to be replaced by the “hard budget constraint” o f profits, losses and, if necessary, strict bankruptcy laws.13 Third, the “Kombinate,” which are con glomerates o f GDR firms that produce similar products, have created an extremely high degree o f horizontal concentration in the GDR economy; this has contributed to the GDR price inflexibility discussed previously. Consequently, price reform requires that these "Kombinate” be dismantled as soon as possible. However, even if this is not immediately forthcoming, the in troduction o f the freely convertible D-Mark will create a m ore competitive environment because it will significantly open up the GDR's econom ic relations with West Germany and the rest of the world. Thus, while there are many open questions concerning specific details o f how the divergent legal systems will be reconciled and how privatization will be achieved, there is wide ac ceptance that these are the central elements of real sector reform and that they will take place.14 DETERMINING THE EAST GERMAN-WEST GERMAN MONETARY CONVERSION RATE: PRINCIPLES AND PROBLEMS As noted previously, the major controversy over unification focused on monetary unification—that is, h ow to determine the rates at which GDR financial stocks and flows denominated in Marks would be converted into their appropriate D-Mark values. The main reason for the intensive debate was that none 11See Institut der deutschen Wirtschaft (1990). 13See Sokil and King (1989). 12See the survey conducted by Commander and Coricelli (1990). 14See e.g. “ Reform” (1990). JULY/AUGUST 1990 22 o f the existing exchange rates betw een the Mark and the D-Mark seemed relevant for determining the D-Mark value o f GDR financial stocks and flows after unification. In this respect, the Gmu is quite different from the fo r mation o f a monetary union between tw o or m ore market economies. For instance, the ap propriate conversion rate was easily determined w hen Saarland, w hich had becom e independent from Germany after the W orld W ar II, was unified with the FRG in 1959. In this instance, Saarland’s financial flows and stocks, which had been denominated in French Francs prior to unification, w ere simply converted to their DM values at the prevailing market exchange rate between the Franc and the D-Mark. In the case o f Gmu, however, all existing ex change rates w ere either highly distorted or essentially devoid o f econom ic significance. The same criticism applies to the m acroeconom ic data that might otherwise have been used to calculate an "equilibrium exchange rate” on the basis o f the traditional exchange rate models.15 The Flaws with Using Existing Ex change Rates f o r Conversion After the fall o f the Wall, the one market ex change rate betw een Mark and D-Mark was the DM price for Mark bank notes that had been il legally "exported” from the GDR to West Ger many. However, this rate, which is shown in figure 1, is not representative o f the underlying fundamental relative price o f Marks in terms of DMs for several reasons. First, it was subject to speculative influences which made it very volatile.16 Second, it reflected demands by East Germans for certain goods (e.g., consum er elec tronics and coffee) that w ere highly taxed in the GDR; table 2 shows that the Mark prices o f these products in the GDR w ere about five times higher than their D-Mark prices in West Ger 15See Frenkel and Goldstein (1986), Williamson and Miller (1987). 16lt varied from 16:1 (November 17, 1989) to 3:1 after the definitive conversion rate for non-GDR residents had become public. 17See Wolf (1985b, pp. 215). 18See Cornelsen and Kirner (1990). 19Until 1989, this rate was also used for the “ forced ex change” (“ Zwangsumtausch” ) of DM 25 for West Ger mans who wanted to visit East Germany. From the begin ning of 1990, West German travelers could exchange DMarks at a 1:3 rate against Marks. FEDERAL RESERVE BANK OF ST. LOUIS many. Third, the arbitrage (flow) o f subsidized East German products to the West has remained relatively weak due to transaction costs and trade restrictions. Another possible candidate for the "true ex change rate” to use for conversion purposes might have been the so-called “Devisenrentabilitaet” (foreign exchange profitability) o f GDR exports in terms o f their DM equivalent. This rate is calculated by dividing the Mark value of the aggregate GDR exports by their DM revenue when they are sold to West Germany. In 1989, this ratio, w hich was used by the GDR govern ment for all internal conversion calculations, was 4.4 Marks per DM. Again, however, this ratio does not indicate what the market ex change rate would be. First, the domestic prices o f many GDR export products w ere artificially high due to taxes imposed by the GDR; conse quently, the numerator o f the ratio is heavily influenced by tax policy, not econom ic values. Second, export decisions w ere made by the GDR government primarily to obtain foreign ex change to finance its imports. It is evident that this non-market allocation process, which is typical o f centrally planned econom ies,17 is not representative o f market-based trade; among other consequences, it can lead to exports with very low profitability.18 A third alternative is the official 1M:1DM ex change rate set in the past by the East German Government.19 Like all such official exchange rates established in socialist countries, this was an arbitrary rate used primarily as an accoun ting unit w hich embodies no useful econom ic information relevant to determining the rate to use for Gmu conversion purposes.20 All foreign exchange transactions w ere conducted at flexi ble (implicit) exchange rates w hich w ere the ratios o f the internal Mark price to the world market DM price o f each product.21 20The same criticism applies to the exchange rates agreed to by the East and West German governments in December 1989 when they established a fund to exchange bank notes for travel: Each East German citizen was entitl ed to purchase up to 100 DM at a 1DM:1M rate and an additional 50 DM at a 1DM:5M rate. 21See Wolf (1985b). 23 Figure 1 Exchange Rate for Mark Banknotes (M per DM) Daily values for bid and ask prices October 2, 1989-April 4, 1990 Percent Percent 1989 The Problems with Using Exchange Rate Models to Deter mine the Conversion Rate The lack o f reliable market exchange rates to use in setting the Gmu conversion rate might tempt one to consider using one or m ore tradi tional exchange rate models to calculate an ap propriate "equilibrium exchange rate.” Several approaches to exchange rate determination ap pear in the international econom ics literature; 1990 among the alternative approaches are purchas ing pow er parity (PPP), structural exchange rate models and the so-called "underlying balance ap proach’’.22 A detailed discussion o f these approaches is beyond the scope o f this paper. However, they have limited usefulness in the Gmu context because they w ere developed primarily to ex plain exchange rate fundamentals in economies with open financial markets. All variants o f PPP, 22See, for instance, Frenkel and Goldstein (1986). JULY/AUGUST 1990 24 Table 2 Consumer Prices of Selected Goods and Services in the GDR and the Federal Republic of Germany (1985)_______________ Good/Service Potatoes (5 kg) Tomatoes (1 kg) Rye Bread (1.5 kg) Beef (1 kg) Chocolate (100 g) Coffee (250 g) Jeans (men’s) Brown Coal (50 kg) Radio/Cassette Recorder Color TV Rent (1 bedroom) Electricity (75 kwh) Haircut (man) Railway Ticket (50 km) Price in (M) GDR M4.05 4.40 0.93 9.80 3.85 25.00 135.00 3.51 1,160.00 5,650.00 75.00 7.50 1.90 4.00 Price in (DM) FRG DM5.32 2.10 4.54 19.45 0.89 5.25 59.90 19.40 199.95 1,199.00 390.00 29.30 11.25 9.20 M price as a percent of the DM price 76 210 21 50 433 476 225 18 580 471 19 26 17 43 SOURCE: Materialien zum Bericht zur Lage der Nation, 1987, pp. 513, 516, 732-735. for example, rely on the “law o f one price” holding in integrated and competitive markets.23 The relative version o f PPP, developed by Gustav Cassel to determine equilibrium ex change rates after W orld W ar I,24 relates the re quired exchange rate adjustment betw een the currencies o f tw o countries with different infla tion rates. However, this procedure requires the existence o f an unbiased base period in the past, a condition which clearly is not met in the GDR setting. The absolute version of PPP avoids the base period problem by defining an equilibrium ex change rate as the ratio o f the price o f a stan dard market basket o f goods in one currency to the price o f the same basket in another curren cy. Thus, the consumption basket o f an average GDR household could provide one basis for ab solute PPP calculations o f an appropriate Mark/D-Mark exchange rate. In 1985, for exam ple, the goods and services w hich made up this basket (excluding rents) had a DM equivalent value w hich was 10 percent higher than their 23See Cassel (1918, p. 413): “ As long as anything like free movement of merchandise and a somewhat comprehen sive trade between the two countries takes place, the ac tual rate of exchange cannot deviate very much from this purchasing power parity.” 24See Dornbusch (1987). http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis Mark price.25 If w e assume that the GDR price level has remained unchanged while the market basket’s DM equivalent value has risen at the in flation rate in West Germany since 1985, the price differential w ould be about 15 percent in 1989. Thus, an absolute PPP exchange rate based on consum er prices (for a given market basket o f goods and services) w ould be 1.15DM:1M or a 1DM:0.9M conversion rate. However, due to high subsidies and the existence o f a "monetary overhang” (explained later in the paper), in dicative o f an excess demand for goods in the GDR, the econom ic relevance o f such calcula tions is severely limited.26 Because most structural exchange rate models (e.g., those based either on the monetary ap proach with fixed or flexible prices or on the portfolio balance approach) require either short term or long-term PPP to hold, they are beset with the same conceptual drawbacks as the sim ple PPP calculations already discussed. In addi tion, they presume that people are able to engage in unlimited arbitrage betw een financial ^Including rents the difference was 25 percent. 26An alternative PPP measure, the “ Devisenrentabilitaet,” yields an equilibrium exchange rate of 1DM:4.4M. However, the problems of this specific measure have already been discussed. 25 markets in the respective countries; this condi tion did not exist in the GDR prior to the Gmu.27 W H AT FINANCIAL STOCKS AND FLOWS WERE CONVERTED? In order to clarify the issues associated with the Gmu, this section presents a brief discussion o f the main items whose values w ere converted from Marks to DMs in the process o f monetary unification. Throughout the paper, a distinction will be made between financial stocks and financial flow s. Conversion o f the Stock o f GDR Monetary and Financial Assets and Liabilities The stock o f financial assets in the GDR is represented by the consolidated balance sheet o f its banking system presented in table 3. In contrast to how these accounts would be drawn up in the United States, the loans and liabilities o f the central bank ("Staatsbank”) must be add ed to the state-owned commercial banks on a consolidated basis. As is typical in most central ly planned economies, the GDR did not permit direct financial transactions betw een enterprises and households ("dichotomized money sup ply”).28 Therefore, this consolidated balance sheet presents a comprehensive picture of the stock o f all financial assets and liabilities in East Germany. The principal items on the asset side o f the banking system w ere loans to state-owned enterprises, housing (chiefly state-owned), direct credits to the government, and claims on for eigners. Loans to households w ere negligible, making up less than 1 percent o f all bank assets. In contrast, such loans represent about 23 per cent o f bank assets in the Federal Republic. Savings o f private households are the important liability of the GDR’s banking The consolidated balance sheet prior to Gmu also shows a considerable amount most system. the of 27This also excludes the application of the “ underlying balance approach to exchange rate assessment,” which was developed by the International Monetary Fund. Accor ding to Williamson and Miller (1987, p. 10), who have elaborated this method, the “ fundamental equilibrium ex change rate” is defined as the rate “ which is expected to generate a current account surplus or deficit equal to the underlying capital flow over the cycle, given that the coun- foreign liabilities. However, the bulk o f these foreign liabilities (M 96 billion) was simply an accounting item ("Richtungskoeffizient”) arising from the GDR’s practice o f valuing its foreign assets and liabilities at a 1DM:4.4M exchange rate rather than at its “official” 1DM:1M ex change rate. After Gmu, o f course, the DM denominated foreign debt o f the GDR will be valued at its face value. The revaluation o f foreign assets and liabilities also reduced the amount o f external claims (from M 45 billion to DM 36 billion) and the debt o f the government (from M 61 billion to DM 12 billion). After revaluation o f foreign assets and liabili ties and the overall 1DM:2M conversion o f all domestic items, except for the limited 1DM:1M conversion o f savings, the liabilities o f the GDR banking system (DM 246 billion) exceeded its assets (DM 220 billion) by DM 26 billion. This difference was created by the asymmetric con version o f the left and the right side o f the con solidated balance sheet produced by an effective 1DM:1.4M conversion rate o f total savings. To equilibrate their balance sheets, East German banks w ere given interest-bearing government assets from an equalization fund established by the GDR for this purpose. Except for this fund, the post-Gmu balance sheet shows that the net bank debt o f the actual GDR government sector is relatively small (DM 7 billion). Conversion o f Financial Flows The 1DM:1M conversion rate for financial flows determined the D-Mark equivalent for Mark-denominated wage and rent contracts in existence prior to July 2, 1990. Although these contracts can (and undoubtedly, will) be renego tiated after this date, a legal transformation of existing contractual obligations from their pre vious Mark payments into DM payments (“rekurrenter Anschluss”) was required.29 For all new contracts and those old contracts for which payments could be adjusted immediately or on short notice, the conversion rate was irrelevant. GDR pensions w ere treated somewhat differ- try is pursuing ‘internal balance’ as best as it can and not restricting trade for balance of payments reasons.” 28See Wolf (1985a). 29Poole (1990) emphasizes, on purely economic grounds, that “ any attempt to convert prices of goods and services from OM (Ostmark) to DM through central direction can only cause great difficulty.” JULY/AUGUST 1990 26 Table 3 Consolidated Balance Sheet of the Banking System of the GDR as of May 31 1990 Assets 1. Lending to domestic borrowers Total Government of which Lending in connection with the revaluation of external liabilities Claims on the government from the initial provision of notes and coins in 1948 Enterprises Housing sector Individuals (excluding housebuilding loans) 2. External claims (a) CMEA countries (b) Western industrial and developing countries 3. Participations 4. Other assets Total Balancing item Total DM M Conver billion sion rate billion Liabilities M Conver DM billion sion rate billion 397.4 60.6 1. Deposits from domestic non-banks Total Government 249.9 10.8 31.2 4.9 231.7 102.6 2.5 45.0 17.4 — 2:1* — — 115.8 51.3 2:1 1.3 — — 3.1 36.3 8.7 27.6 1:1 2:1 1.1 1.5 446.6 — 219.6 — — 26.4 446.6 1.81:1 — 2:1 156.6 5.4 — 2:1 2:1 — 27.6 1.1 180.7 12.3 246.0 Enterprises Individuals Giro and savings balances of individuals Residents Non-residents Life insurance 57.0 182.1 165.6 2.3 14.2 2. External liabilities 152.5 (a) CMEA countries 1.1 (b) Western industrial and developing countries 55.0 (c) Provisions for external liabilities (“ Richtungskoef fizienten” )1 96.4 3. Currency in circulation (excluding the banks’ cash holdings) 13.6 4. Accumulated profits/ reserve funds/guarantee funds 5. Other liabilities Total Balancing item Total 2.05:1' — 27.8 123.4 1.44:1“ 115.2 2.05:1' 1.1 2:1 7.1 _ — b 55.6 0.6 55.0 f — 2:1 6.8 23.4 1:1 23.4 7.2 2:1 446.6 — 446.6 — — 1.81:1 3.6 246.0 — 246.0 1These are actually liabilities of the banking sector to the government, which might also be shown in liabilities item 1. In this table, they are shown in connection with the external liabilities of the GDR because the item may also be regarded as a kind of “ value adjustment” for the external liabilities, which are otherwise put at too low a value in GDR Mark. ’ Conversion of a balance of M 24.5 billion, which results after offsetting the lending from the revaluation of external liabilities (M 31.2 billion) and claims arising from the initial provision of notes and coins in 1948 (M 4.9 billion) against provisions for external liabilities (“ Richtungskoeffizienten” ) to the same amount. ‘ External claims (assets item 2 (b)) and external liabilities (liabilities item 2(b)) are here still valued at the accounting rates of the end of 1989. The market rates of June 30, 1990 are to be used for the final conversion. The amounts shown will then presumably be somewhat lower (liabilities item 2(b) also includes foreign currency deposits from residents). 'Conversion rate for balances of non-residents arising on and after January 1, 1990 3:1, otherwise 2:1. dConversion rate of 1:1 for M 2,000 x 3.2 million = DM 6.4 billion; M 4,000 x 10.1 million = DM 40.4 billion and M 6,000 x 3.0 million = DM 18.0 billion yields a total of DM 64.8 billion; the remainder (M 100.8 billion) was converted at a rate of 2:1. 'Balances as at the end of 1989 amounting to M 2.1 billion were converted at 2:1, the remainder at 3:1. 'Partly offset against lending in connection with the revaluation of external liabilities (M 31.2 billion) and claims arising from the initial provision of notes and coins in 1948 (M 4.9 billion); the arithmetical remainder (M 60.1 billion) was used to reduce the balancing item. FEDERAL RESERVE BANK OF ST. LOUIS 27 ently. As already mentioned, the social union adjusted the GDR pension system to make it consistent with West German standards; among other things, this meant that the DM value o f GDR pensions was not less than their previous Mark value. Implications o f the Conversion Rate f o r Price Stability in Germany For West Germans, w ho traditionally have placed a very high social value on price stabili ty, concern over the implications o f the Gmu on the inflation rate played a predominant role in the choice o f the conversion rate. The existence o f an excess supply o f money, which, by Walras’ Law, reflects rationing on goods and labor markets,30 is a widely acknowledged o c currence for centrally planned economies, in cluding the GDR.31 Many observers expected that a flat 1M:1DM conversion o f the East Ger man m oney stock would produce a rise in the price level and, hence, a transitory increase in the measured rate o f inflation for the integrated German currency area after the Gmu. about 13 percent o f W est Germany’s GNP, or some measure o f potential output determined by relative labor productivity estimates. This lat ter method32 uses the proportion o f the East Germany population to West German population (about 26 percent) and the estimated average GDR labor productivity relative to that in West Germany (about 50 percent) to obtain a relative GDR potential production o f about 13 percent, w hich is identical to the relative GNP differen tial noted above. Use o f a lower estimate o f the GDR productivity differential, for instance, the 30 percent estimate o f the Kiel Institute for W orld Economics, reduces the GDR's potential production to only about 10 percent o f that in the FRG. The expected impact o f monetary unification on the German inflation rate can be determined as follows: First, estimate a hypothetical GDR money stock that would be compatible with stable DM prices in the GDR; second, compare this hypothetical m oney stock with the actual DM money stock o f the GDR after conversion. If the actual money stock exceeds the hypotheti cal one, the conversion could produce a tem porary increase in inflation in both Germanys; otherwise, the conversion does not have infla tionary implications. This approach can be used to determine the ”non-inflationary" conversion rates for different monetary aggregates; various estimates are shown in table 4. Applying the West German ratio between potential output to the stock of currency yields a conversion rate o f about 1M:1DM for East German currency holdings. To calculate a non-inflationary M l money measure for the GDR requires determining the "m o neyness” o f the various GDR deposit categories. If the "Spargiro” (M 69.0 billion) and deposits of enterprises are essentially demand deposits and the "Buchsparen” (M 90.7 billion) are essentially the same as traditional savings deposits included in M3, the pre-Gmu GDR M l m oney stock was about one-third of that in West Germany. Thus, the non-inflationary conversion rate for the GDR’s M l m oney stock would lie in the 1DM:2.4M to 1DM:3.3M range. Using the M3 money stock, the non-inflationary conversion rate would lie within 1DM:1.5M and 1DM:2M.33 To accomplish the first step requires calculating the East German m oney demand after the Gmu. To do this, o f course, one has to estimate the velocity o f m oney and potential nominal production o f the GDR econom y. The following estimates are based on the assumption that both relative and absolute DM prices in the GDR as well as the GDR’s velocity o f m oney will be identical to their W est German counterparts after unification. GDR potential production can be estimated either by using its GNP, w hich is Comparing East and West German money stock measures is always problematical because there is a much wider spectrum o f financial op portunities available to West German investors. Their savings in long-term time and savings deposits, bank savings bonds and other financial instruments issued by banks, which are called "monetary capital” and not included in M3, are larger than the M3 m oney stock. Adding these financial assets to the West German M3 money stock yields a liquidity stock measure (L). 30See Commander and Coricelli (1990), p. 3. 31See Sokil and King (1989). 32lt which was suggested by the President of the Kiel In stitute for World Economics, Horst Siebert. 33ln West Germany, the money stock M3 includes currency in circulation (excluding banks’ cash balances), sight deposits, time deposits with a maturity of less than four years and savings deposits at statutory notice. JULY/AUGUST 1990 28 Table 4 Conversion Rates on the Basis of a Non-lnflationary Money Stock for the GDR (1989)_______________________________________ Currency M14 M35 L<s Actual values Non-inflationary values for the GDR GDR' FRG (M billion) (DM billion) A2 B3 (DM billion) 17.0 146.6 252.0 252.0 146.9 450.6 1255.5 2738.3 19.1 59.9 167.0 364.2 14.7 45.1 125.6 273.8 Conversion rates A B 1:0.9 1:2.4 1:1.5 1:0.7 1:1.2 1:3.3 1:2.0 1:0.7 ’ Values for the GDR include deposits of enterprises and households with the banking system. 2Assuming that GDR potential output is 13.3 percent of FRG potential output. 3Assuming that GDR potential output is 10 percent of FRG potential output. 4Currency in circulation and domestic non-banks’ sight deposits. 5M1 plus domestic non-banks' time deposits and funds borrowed for less than 4 years plus savings deposits at statutory notice. 6M3 plus saving deposits at agreed notice, long-term time deposits, bank savings bonds and other financial instruments held by private households and enterprises with banks. SOURCE: Deutsche Bundesbank, Monatsberichte, Jahresbericht 1989 der Staatsbank der DDR. In contrast, savings deposits and currency are the only financial stores o f value available in East Germany.34 Thus, for East Germany, L and M3 are identical. Using the L measure, the noninflationary conversion rate would be about 1DM:1M. However, in order to make the con verted East German L measure truly com parable to the West German L, about 50 per cent o f East German savings would have to be “frozen” for about four years. in a range between 1DM:1.5M to 1DM:2M. While the latter value was recom m ended by the Bundesbank, the political compromise reached between the two governments led to an average conversion rate o f about 1DM:1.7M. While the estimated non-inflationary conversion rates shown in table 4 are subject to considerable uncertainty, the final conversion program chosen for the Gmu seems unlikely to produce any substantial inflationary impact on prices in the new DM currency area. As table 4 shows, an assessment o f the infla tionary impact associated with the Gmu depends on which monetary aggregate is regarded as the one linked most closely to inflation. Most econom etric estimates for the Federal Republic show a very stable relationship between the m oney stock, M3, and inflation (and nominal GNP); this is the reason why the Bundesbank uses M3 as its main inflation indicator and as its central intermediate monetary policy target.35 Taking M3 as the benchmark money stock for non-inflationary purposes suggests that the con version rate for the GDR money stock should lie The Effect o f the Velocity Assump tion on the Non-lnflationary Con version Estimates 34ihis aspect was emphasized by the East German Central Bank (“ Staatsbank” ) in an official statement of April 3, 1990. FEDERAL RESERVE BANK OF ST. LOUIS The non-inflationary conversion calculations described above assumed that the velocity of the appropriate m oney stock in the GDR is iden tical to that in the FRG. Some observers in the Federal Republic have argued that countries with higher per capita income levels have dif ferent monetary velocities from those in lessdeveloped countries. Figures 2 and 3 show the results of a cross-country analysis comparing 35See Deutsche Bundesbank (1989a), Schlesinger and Jahnke (1987). 29 Figure 2 Velocity of M1 and GNP per Capita for OECD Countries Velocity Velocity 8 8 5000 10000 15000 20000 25000 GNP per capita per capita nominal incomes and velocities for the M l and M3 m oney stocks in OECD coun tries in 1987. The figures indicate that per capi ta incom e has no significant influence on the velocity o f money. However, the marked intercountry differences in the velocity o f m oney serves as yet another reminder that the noninflationary GDR m oney stock calculations are subject to considerable uncertainty. Implications o f the Gmu on the “Competitiveness” o f East German Enterprises The potential impacts o f the Gmu on the unemployment rate in the GDR and its econom ic growth prospects w ere another im portant determinant o f the conversion rate. While West Germans w ere concerned about the possible fiscal costs o f unemployment payments to East Germans, the East Germans, as noted earlier, w ere primarily interested in the p ro spects for employment and for raising their standard o f living as quickly as possible. These prospects depend fundamentally on how com petitive the GDR firms will be after the central planning process is dismantled and the econom y o f the GDR is opened up to w orld markets. While it is evident that the conversion o f enter prise debt has a direct impact on the financial structure and capital costs o f firms in East Ger many, the implications o f the conversion rate for wages in the GDR are m ore difficult to evaluate. JULY/AUGUST 1990 30 Figure 3 Velocity of M3 and GNP per Capita for OECD Countries Velocity 2.5 Velocity 2.5 5000 10000 15000 20000 25000 GNP per capita The Impact o f the Conversion o f Enterprise Debt The 1DM:2M conversion rate for financial stocks determines the debt burden and interest payments of enterprises after the unification process.36 It has important consequences for the costs o f capital and for the projected privatiza tion o f East German firms. The latter, an essen tial element o f the process o f econom ic transformation, requires that the firms to be privatized must have a positive net worth. For enterprises that will remain under state ow ner ship, the ratio o f their DM equity to their total 36Again, the conversion rates which were put forward in the debate varied widely, ranging from a 100 percent debt relief which was recommended by the five leading German FEDERAL RESERVE BANK OF ST. LOUIS DM assets after conversion will play a key role in determining whether they can survive with a "hard budget constraint,” i.e. without subsidies from the government. Because there is no data on the debt-equity ratios o f East German enterprises, it is difficult to assess the implications o f the 1DM:2M con version rate on their financial situation and on their interest payments. To get a rough estimate o f the sustainability o f alternative debt burdens, however, the proportion o f GDR potential out put to FRG potential output can be used; as already mentioned, estimates vary between 10 economic research institutes in their report of April 12, 1990, over the 1:2 rate, which was proposed by the Bundesbank, to a full 1:1 conversion. 31 and 13 percent. According to Bundesbank (1989) statistics, the net financial debt o f West German enterprises (excluding housing and the financial sector) was DM 681.5 billion and the book value o f their non-financial assets totaled DM 1096.5 billion in 1988. The FRG figures indicate that a 1DM:1M conversion for the debt o f GDR enter prises w ould have produced a relatively large DM 175 billion net debt (see table 3), about 26 percent o f the West German level.37 If East Ger man firms’ nonfinancial assets are w orth about 10 to 13 percent38 o f that for West German firms, the right side o f their balance sheets would have exceeded the left side (DM 110-140 billion) by huge amounts, even if West German firms' balance sheets contain extensive hidden reserves. Without further debt reduction, privatization o f virtually all East German firms would have been impossible. According to West German bankruptcy law, w hich requires bankruptcy proceedings if a firm has negative net worth, most East German firms would have had to be declared bankrupt. Of course, an outright cancellation o f all GDR enterprise debt would have avoided these p ro blems. However, this "solution” was dismissed for tw o reasons: First, firms with permanent net debt levels are com m on in all industrialized countries; second, it w ould have led to a huge increase in the government debt as described later in this paper. The debt reduction39 achiev ed by the 1DM:2M conversion rate places the average ratio o f equity to assets for GDR enter prises in a range between 20 and 37 percent, which should allow the privatization o f at least some firms. By comparison, the average equity to asset ratio is about 20 percent in West Ger many and 50 percent in the United States.40 Because the results are quite sensitive to the estimate o f the value o f real assets in the GDR, it is difficult to assess whether the interest burden o f East German firms will be similar to the West German enterprise sector or whether it will be significantly higher. In addition, it is not yet clear whether East German firms will have to pay market-determined interest rates on their debt after monetary unification. 37A high debt burden is regarded as a typical concomitant of the central planning mechanism, which gives enterprises automatic bank credits inducing large hoardings of inven tories or camouflaging cost overruns, waste and sales in the black market (Grossman 1989, p. 31). 380n the basis of a Cobb-Douglas production function and assuming an identical elasticity of output with respect to W A G E C O N V E R S IO N The second determinant o f East German firms’ post-conversion competitiveness are the DM wages they will have to pay. While neither the GDR nor the FRG government should deter mine wages after the transition to a market economy, their treaty established a wage con version rate to define the financial obligations o f existing contracts for the time immediately after July 2, 1990. However, the actual conver sion rate chosen has implications for wage levels and competitiveness only if nominal DM wages in the GDR after conversion are inflexible dow nw ard and if initial DM wages are set “too high” com pared with labor productivity. These considerations would have called for a conversion rate that reduced average wages below the level indicated by the GDR's average productivity. The advantage o f this low starting level for wages is that it would have allowed w orkers and firms in East Germany to renegotiate their contracts more easily after Gmu. This w ould have enabled them to establish a wage structure m ore closely mat ching sectoral productivity differentials than the prior GDR wage structure, in w hich wages w ere relatively uniform regardless o f productivi ty differences. To evaluate the competitiveness o f East Ger man firms after a 1DM:1M wage conversion, their labor productivity relative to that in com parable West German firms must be compared with their relative DM wages. These com parisons would require information on the pro ductivity o f individual firms or, at least, in dividual sectors in the GDR after July 2, 1990. Unfortunately, such sectoral data are not available at all; moreover, estimates o f labor productivity in the GDR after the transition to a market econom y are very difficult to determine ex ante. However, the experience following the West German currency reform in June 1948 shows that large productivity gains can be achieved rather quickly; these gains arise from better incentives associated with the market process and increased availability o f inputs. In capital as in West Germany, Alexander and Gagnon (1990) estimate the level of the East German capital stock to be 10.4 percent of the West German capital stock. 39A strategy of recapitalization is now also suggested for other Eastern European countries. See Hinds (1990, p. 44). 40See Bank for International Settlements (1989, p. 86). JULY/AUGUST 1990 32 the GDR, where such incentives are lacking, shortages o f specific inputs are often reported to have led to significant decreases in output and productivity.41 On the other hand, the far-reaching restruc turing o f production processes will not be possi ble without some tem porary output disrup tions.42 If these positive and negative effects roughly cancel each other in the first few months after conversion, the GDR’s productivity should reach about 50 percent o f that in West Germany, w hich is consistent with past estimates made by the Deutsches Institut fuer Wirtschaftsforschung. Before the conversion took place, the average monthly salary o f a w orker was M 1250 in the GDR and DM 3192 in West Germany. With the 1DM:1M conversion o f the initial nominal wages, monthly wage costs (including employers’ contributions to social security) for East German firms would be about 37 percent o f West German wages.43 Thus, the average DM wage level in the GDR after the conversion is not so high relative to the average productivity differential betw een GDR and FRG w orkers that it would preclude future wage negotiations. However, the initial wage differential cannot be held constant by the government after the Gmu. Therefore, the medium-term outlook for employment as well as for foreign and West German investment in the GDR will depend mainly on the rate o f subsequent wage in creases in the GDR. If these exceed the growth o f productivity in the GDR, employment and in vestment in GDR firms will fall. 41According to a survey of the Institut der deutschen Wirtschaft, about one third of all GDR employees had to sus pend their work for two or more hours per day because of shortages and defective machines. 42ln the past, all decisions on investment, production and sales were made by the central planning bureaucrats; managers of firms were mainly responsible for technical operations. 43The 1DM:2M conversion rate proposed by the Deutsche Bundesbank differs less from the 1DM:1M rate chosen than one might assume at first glance. In its calculations, the Bundesbank assumed that all subsidies would be removed before conversion, requiring an increase in Mark wages of about 25 to 30 percent to compensate for this ef fect. If these new Mark wages were then converted at a 1DM:2M rate, the effective conversion rate between initial East German Mark wages and DM wages after the Gmu FEDERAL RESERVE BANK OF ST. LOUIS Implications o f the Conversion Rate f o r GDR Real Incom es and Labor Migration Because the unification process has been driven primarily by the desire o f East Germans to improve their standard o f living, the effects o f monetary unification on the FRG-GDR real in com e differential w ere intensively discussed in both East and West Germany. However, since wages will be renegotiated after the Gmu, monetary unification will have only a short-term impact after July 2, 1990. An estimate o f the change in East German real incomes resulting from the Gmu can be calculated by assuming that nominal wages in the GDR will remain constant after conversion and after the various subsidies are abolished. The basis for comparing pre- and post-Gmu real incomes in the GDR is the consumption basket of an average GDR household that was discuss ed earlier. The abolition of trade restrictions and product-specific taxes and subsidies will produce price structures and a price level in East Ger many similar to that in West Germany. Thus, the Gmu will cause a "one-shot" consum er price increase o f about 15 percent for the unchanged GDR consum er goods basket.44 In addition, the increase in social security contributions, due to the introduction o f the West German social security system into East Germany, will reduce the average net monthly incom e o f an East Ger man w orker from M 1050 to DM 983 after unification. Together with the one-shot price ad justment in consum er goods, real incomes in the GDR will be reduced by about 21 percent.45 would have been about 1DM:1.2M. Including the employer’s contribution to social security, the initial labor costs in the GDR would have been about one third of the West German level if the 1DM:2M conversion rate had been used. ■^This change from Mark prices to D-Mark prices has no ef fect on the overall German inflation rate which is measured on the basis of the DM equivalent of goods and services. 45These orders of magnitude show that conversion rates for GDR incomes considerably above 1DM:1M, for instance, 1DM:2M or 1DM:3M, would have strongly increased the movement of workers from East to West Germany. Assum ing constant consumption patterns, a 1DM:2M (1DM:3M) rate would have reduced GDR real incomes by 57 percent (70 percent) compared to their pre-Gmu levels. 33 The above calculations overstate somewhat the negative welfare implications o f unification. Households will adjust to the price changes by purchasing more o f the goods with relatively cheaper DM prices and less o f those whose prices rose m ore because they had been heavily subsidized in the past. The prospective adjust ment o f the previous Mark price structure to the DM price structure is indicated in table 2. As no detailed data on consumption patterns of East Germans are available, the quantitative relevance o f this substitution effect is difficult to evaluate.46 The same comment applies to the positive welfare effects attributed to prospective quality improvements in available consumer goods; after the Gmu, East Germans will be able to buy W est German products which, on average, are o f better quality than their East German counterparts. On balance, the real incom e o f East Germans and the real income differential between East and W est Germany will remain essentially un changed immediately after the Gmu, with real net incomes in the East about 50 percent lower than in the West. This result reflects the fact that monetary unification by itself can only create a fram ework for real sector reform. Significant improvements in East German living standards will only be generated by better allocation o f their resources and increased in vestment. Thus, the incentive for East German workers, especially skilled workers, to move to the Federal Republic o f Germany remains at least as strong as it was before the Gmu. However, the prospect o f a rapid and wideranging restructuring o f the GDR econom y has already im proved the motivation o f East Ger mans to remain in the GDR and contribute to its econom ic recovery. The num ber o f GDR citizens moving to the FRG, which reached a monthly peak o f 133,000 in November 1989, fell to only 19,000 by May 1990. The Gmu Wealth Transfer Between East and West Germany The Gmu will result in a wealth transfer from West Germany to East Germany.47 The mechanisms and the quantitative effects o f this wealth transfer, however, remain uncertain. 46An analysis of the Deutsches Institut fuer Wirtschaftsforschung comes to the result that private households can compensate the price effect by reducing their con sumption of foods by 10 percent. To examine this issue, even if a definitive answer is not forthcoming, it is useful to start with an example o f a hypothetical currency unification between tw o market economies, e.g., between France and West Germany. Suppose that the DM is to be replaced by the Franc and that the current market exchange rate (1DM = 3FF) will be used to convert all DM financial and real stocks and flows in their Franc equivalent. In this case, there is no transfer o f real wealth; simply multiplying all DMark prices by three does not reallocate wealth within Germany nor between France and the Federal Republic.48 Redistribution o f wealth bet w een creditors and debtors in both countries could occur only if that currency unification leads to unexpected changes in inflation and if some debtors or creditors had been expecting a parity adjustment. In this case, the net transfer between the tw o countries w ould then be deter mined by creditor/debtor relations between France and Germany and by the direction of the change in expectations. In the Gmu case, there is no wealth transfer between GDR residents and West Germans due to unexpected exchange rate variations because there w ere virtually no financial linkages bet ween individuals or enterprises in both coun tries prior to the Gmu. The asymmetric conver sion o f assets and liabilities, however, transfers GDR debt to the FRG (see shaded insert). Before the Gmu, the aggregate wealth o f the East Ger man econom y consisted o f its aggregate real assets and its aggregate net foreign claims (debts); domestic financial claims and liabilities simply cancel out in the aggregation process. Because monetary unification has no implica tions for the GDR’s foreign claims and liabilities, it can increase the wealth o f the GDR only if its domestic financial assets, which are mainly sav ings, are converted at a higher rate than its domestic liabilities. In a closed econom y, even this asymmetric conversion would have no aggregate effect on the econom y’s wealth; the gap between assets and liabilities in the consolidated banking system would have to be filled by government 47See, for instance, Poole (1990). " I t is assumed that a procedure for an equitable distribution of seignorage can be devised. JULY/AUGUST 1990 34 Mechanics of the Wealth Transfer Effected by Gmu The total wealth (W,) o f each econom ic agent in the GDR is the sum o f its real wealth (RW,) plus its net monetary wealth (NM,): Canceling all intra-GDR financial claims and liabilities (RC GDR = R LGDR) yields total wealth before Gmu: (4a) RW = RRW + R(CF - LF) (1) W, = RW, + NM, Net monetary wealth is the sum o f claims on other GDR residents (C,GDR) and on foreigners (C/) minus liabilities against GDR residents (LjGI5R) and foreigners (L;F): The asymmetric conversion o f intra-GDR financial liabilities and claims (RC ,GDR> RL GDR) requires the creation o f an equalization item (E) which leads to: (5) R C GI>R = RL GDR + E (2) NM, = C GDR + C,F - Li - L,r Substituting (2) in (1) yields: (3) W, = RW, + (C GDR - L™11) + (C,F - LiF) If this equalization item is regarded as a financial liability o f West Germany, (5) can be substituted in (4): (6) RW = RRW + R (CF - LF) + E. Total wealth o f the GDR is the sum o f in dividual total wealth: (4) RW = RRW + R(CGDR - L GDR) + R(CF - L,F) Comparing 4a and 6 shows that the wealth transfer, which is directly associated with Gmu, depends on the amount o f this equalization item. bonds.49 In the case of Gmu, the gap is closed by bonds w hich are issued by equalization funds established by East Germany. While these bonds are formally a debt o f the East German government, they can actually be regarded as a financial obligation o f West Germany. This con clusion is based on the wide-ranging financial support that West Germany agreed to provide to the East German public sector and the pro spect o f rapid political unification. The wealth transfer directly produced by Gmu is thus iden tical to the amount o f bonds needed to equalize the consolidated balance sheet o f the East Ger man banking system after the Gmu.50 At the moment, most GDR firms are ow ned by the state. To the extent that these assets are transferred to a com m on German government, the net wealth transfer arising from the money stock conversion will be reduced. The same result would occu r if these firms are sold at market prices and the proceeds are then used to repay part o f the GDR government debt. This latter option is presently being discussed in the Federal Republic. A second determinant of the wealth transfer betw een East and West Germany is the distribu tion of the GDR’s real wealth after conversion. The consolidated balance sheet o f the GDR’s banking system in table 3 shows that the 1DM:2M conversion o f the GDR enterprise sec- 49This was the case in the West German currency reform of 1948. “ Gmu would have indirect wealth effects if it contributes to non-competitive wages and if these wages are inflexible FEDERAL RESERVE BANK OF ST. LOUIS The Direct Impact o f Gmu on Ger man Government Debt downward, which would require unemployment benefits from West Germany to East Germany. 35 tor’s net debt and the limited 1DM:1M conver sion o f savings (including currency) implies a DM 26 billion (5.3 percent) increase in the Ger man central government debt. The impact of this asymmetric conversion would have been even higher if it w ere not for the "Richtungskoeffizient” discussed previously. Using an assumed 8 percent interest, this ad ditional debt will increase the German govern ment's interest payments by DM 2.1 billion, about 0.7 percent o f its total expenditure. A uniform 1DM:1M conversion o f enterprise debt, savings and currency would have produced a DM 76 billion increase in government debt. If this debt w ere borne mainly by West German tax payers, this would have been identical to a wealth transfer o f DM 1230 from each West German—in the form of an interest-bearing and non-repayable IOU—and would have provided each East German with an additional DM 4560. This example illustrates why the 1DM:1M conversion rate for savings was controversial in West Ger many after it had become evident that a 1DM:1M rate for enterprise debt was impracticable. transfer from the West Germans to East Germans has been widely overestimated. The ultimate out come of unification will be determined by the productivity of East German firms, the real in come necessary to encourage East German workers to remain in the GDR and the actual wage and income levels that will be achieved in East Germany. Monetary unification has only have a short-term impact on the initial wages and incomes in the GDR. Because the conversion rates are compatible with the more pessimistic estimates of the pro ductivity differential between East and West Ger many, they do not appear to have produced the problem of too-high initial GDR wage levels and possible downward-stickiness of wages in the face o f some initial unemployment pressures. Whether the prospects provided by the economic and social community o f the two states and the farreaching financial assistance offered to East Ger many by the West German government will suf fice to keep skilled workers in the GDR remains open to question. REFERENCES SUMMARY The set o f conversion rates chosen for the Gmu has important implications for the debt burden of East Germany’s enterprise sector, for the wealth transfer between both German states and for the level of West German government debt. The 1DM:2M conversion rate for enterprise debt may cause some financial difficulties for many GDR firms, but it will also lay the groundwork for the privatization of the more profitable enterprises. This result is a necessary precondition for the GDR’s transition to a market economy. The ceil ings for the 1DM:1M conversion of savings limit the wealth transfer from West Germany to East Germany to a relatively small amount. The same applies to the required increase in German government debt and its interest payments. A (transitory) rise in the inflation rate of the common German currency area is unlikely after the Gmu. The post-conversion money stock in the GDR seems to be roughly compatible with the GDR money demand at the new DM prices. The medium- and long-term impacts of monetary unification on the competitiveness of GDR firms, on unemployment and relative living standards in East Germany, and on the wealth Alexander, Lewis S., and Joseph E. Gagnon. “The Global Economic Implications of German Unification,” International Finance Discussion Papers, No. 379 (Board of Governors of the Federal Reserve System, April 1990). Bank for International Settlements. 59th Annual Report (Basle, Switzerland, 1989). Cassel, Gustav. “Abnormal Deviations in International Ex changes,” Economic Journal (December 1918), pp. 413-15. Collier, Inwin L. “The Estimation of Gross Domestic Product and its Growth Rate for the German Democratic Republic,” World Bank Staff Working Papers, No. 773, Washington, D.C., 1985. Commander, Simon, and Fabrizio Coricelli. “ Levels, Rates and Sources of Inflation in Socialist Economies: A Dynamic Framework,” paper prepared for a seminar in Laxenburg, Austria, March 6-8, 1990. Committee for the Study of Economic and Monetary Union. “ Report on Economic and Monetary Union in the European Community,” Luxembourg, 1989. Cornelsen, Doris, and Wolfgang Kirner. “Zum Produktivitaetsvergleich Bundesrepublik-DDR,” Deutsches Institut fuer Wirtschaftforschung-Wochenbericht, April 5, 1990, pp. 172-74. Daviddi, Renzo, and Efisio Espa. “The Economics of Rouble Convertibility: New Scenarios for the Soviet Monetary Economy,” Banca Nazionale del Lavoro Quarterly Review (December 1989), pp. 441-65. Deutsche Bundesbank. “The Monetary Union with the Ger man Democratic Republic,” Monthly Report of the Deutsche Bundesbank (July 1990). _______ “The Deutsche Bundesbank. Its Monetary Policy Instruments and Function,” Deutsche Bundesbank Special Series No. 7 (Frankfurt, 1989a). JULY/AUGUST 1990 36 _______ . “ Enterprises’ Profitability and Financing in 1988,” Monthly Report of the Deutsche Bundesbank, (November 1989b). Dornbusch, Rudiger. “ Purchasing Power Parity,” in John Eatwell, Murray Milgate, and Peter Newman, eds., The New Palgrave: A Dictionary of Economics (London, 1987), pp. 1075-88. Frenkel, Jacob, and Morris Goldstein. “A Guide to Target Zones,” IMF Staff Papers, Vol. 33 (December 1986), pp. 633-73. Gerstenberger, Wolfgang. “ Das zukuenftige Produktionspotential der DDR—ein Versuch zur Reduzierung der Unsicherheiten,” ifo-schnelldienst (July 1990), pp. 13-22. Grossman, Gregory. “ Monetary and Financial Aspects of Gorbachev’s Reform,” in Christine Kessides et al., eds., Financial Reform in Socialist Economies (Washington, D.C.: World Bank, 1989), pp. 28-46. Hinds, Manuel. “ Issues in the Introduction of Market Forces in Eastern European Countries,” paper presented at an Economic Institute of the World Bank Conference in Warsaw, Poland, March 12-13, 1990. Institut der deutschen Wirtschaft. “ Sozialvertragliche Ausgestaltung der Deutsch-Deutschen Wahrungsunion,” Report for the State of Lower Saxony from March 12, 1990. Institute of International Finance. “ Building Free Market Economies in Central and Eastern Europe: Challenges and Realities” (Washington, D.C., 1990). Poole, William. “The German Democratic Republic: Economic Goals, Constraints and Monetary Reform,” Shadow FEDERAL RESERVE BANK OF ST. LOUIS Open Market Committee: Policy Statements and Position Papers, March 18-19, 1990. “ Reform der Wirtschaftsordnung in der DDR und die Aufgaben der Bundesrepublik, Stellungnahme einer deutschdeutschen Arbeitsgruppe,” Deutsches Institut fuer Wirtschaftforschung-Wxhenbericht, February 8, 1990, pp. 65-71. Roe, Alan, and Jayanta Roy. “Trade Reform and External Adjustment: The Experience of Hungary, Poland, Portugal, Turkey, and Yugoslavia,” Economic Development Institute of The World Bank Seminar Report, No. 16, Washington, D.C., 1989. Schlesinger, Helmut, and Wilfried Jahnke. “ Geldmenge, Preise und Sozialprodukt,” Jahrbuecher fuer Nationaloekonomie und Statistik, Vol. 205, pp. 410-26. Sokil, Catherine, and Timothy King. “ Financial Reform in Socialist Economies: Workshop Overview,” in Christine Kessides et al., eds., Financial Reform in Socialist Economies (Washington, D.C.: World Bank, 1989). Williamson, John, and Marcus H. Miller. “Targets and In dicators: A Blueprint for the International Coordination of Economic Policy,” Policy Analyses in International Economics, Number 22, September 1987. Wolf, Thomas A. “ Economic Stabilization in Planned Economies,” IMF Staff Papers, Vol. 32 (March 1985a), pp. 78-129. _______ "Exchange Rate System and Adjustment in Planned Economies,” IMF Staff Papers, Vol. 32 (June 1985b), pp. 211-47. 37 John A. Tatom John A. Tatom is an assistant vice president at the Federal Reserve Bank of St. Louis. Kevin L. Kliesen provided research assistance. The Effects of Financial Innovations on Checkable Deposits; M l and M2 D URING THE EARLY 1980s, several new types o f financial assets w ere authorized by Congress and included in the definitions of various monetary aggregates. The principal new accounts w ere NOW accounts, which w ere authorized nationwide in January 1981, and money-market deposit and super-NOW ac counts, w hich becam e available in Decem ber 1982 and January 1983, respectively. Their growth and inclusion in monetary aggregates gave rise to increased uncertainty in explaining movements in the monetary aggregates and questions about the relationship o f the m one 1These uncertainties have been a continuing source of con cern for the Federal Open Market Committee (FOMC). This concern has focused primarily on M1. See Hafer (1986) and Nuetzel (1987) for discussions of uncertainties associated with M1. In 1981, when the authority to offer interest-bearing checkable deposits was extended nation wide, the FOMC announced targets for the old M1-type measure that excluded such new deposits and for an M1-type measure that added these so-called other checkable deposits. See Tatom (1982) and Thornton (1982) for an analysis of the 1981 developments and their effects on monetary policy; the latter article discusses the evolution of the current M1 measure following the 1980 redefinitions discussed in Hafer (1980). In 1983, the FOMC refrained from targeting on M1 and indicated a greater reliance on M2. See Hafer (1985) for a discussion of the effects of 1983 innovations on policy deliberations. tary aggregates to various measures o f econom ic perform ance.1 The widely accepted view is that these finan cial innovations have rendered M l less useful, or even useless, as a monetary policy target.2 The related view—that the broader aggregate M2 has been unaffected by these innovations and therefore remains a useful target—is almost as widely shared. While an apparent change in the linkage between M l and econom ic perfor mance in the 1980s has buttressed the impres sion that financial innovations distorted M l and Mascaro and Marlow (1989), Friedman (1988), Haraf (1986), Hetzel (1989), Hetzel and Mehra (1989), Judd and Trehan (1987), Judd, Motley and Trehan (1988), Keeley and Zimmerman (1986), Kopcke (1987), Porter and Offenbacher (1984), Mehra (1989), Roth (1987), Siegel (1986), Simpson (1984) and Wenninger (1986). In short, this view is widespread. Earlier studies disputing these claims in clude Cook and Rowe (1985), Gavin (1987), Hein (1982), Jordan (1984) and Tatom (1982, 1983a, 1983b). These studies follow an earlier theoretical and empirical tradition which suggested the ineffectiveness of deposit rate regula tions. This literature includes such works as Barro and Santomero (1972), Bradley and Jansen (1986), Cox (1966), Frodin and Startz (1982), Kareken (1967), Benjamin Klein (1970, 1974), Michael Klein (1974), Saving (1971, 1977, and 1979), Santomero (1974), Startz (1979) and Tatom (1971). 2Some examples are: Hafer (1984), Barnett (1982), Spindt (1985), Morris (1982), Cox and Rosenblum (1989), Darby, JULY/AUGUST 1990 38 impaired its usefulness, few quantitative studies have assessed the actual effects of financial in novations on the monetary aggregates. This paper first describes the financial innova tions hypothesis that M l, but not M2, has been significantly affected by the introduction and grow th o f these new assets. It then assesses the validity o f this hypothesis by examining whether the turnover rate for checkable deposits, cur rency preferences, and M l and M2 demand (velocity) have been affected as the hypothesis suggests.3 MONETARY AGGREGATES AND FINANCIAL INNOVATIONS Table 1 shows the components of M l and M2 in 1989. M l consists o f currency in the hands o f the public, demand deposits, other checkable deposits and travelers checks. Other checkable deposits include accounts on which financial in stitutions can make explicit interest payments. During the 1970s, a few states authorized interest-paying negotiable order o f withdrawal (NOW) accounts. In 1978, checkable accounts with automatic transfer from interest-paying savings accounts (ATS) w ere authorized by the Federal Reserve System. As figure 1 shows, the share o f other check able deposits in total checkable deposits (demand and other checkable deposits) rose from about 10 percent in late 1980 to m ore than 25 percent by the end o f 1981, the first year that nation wide NOW accounts w ere authorized. This share continued to rise, in part because o f the intro duction o f super-NOW accounts (interest-bearing other checkable deposits with unregulated inter est rates) in early 1983. By 1989, other check able deposits had risen to $278.5 billion, nearly half o f total checkable deposits and about 36 percent o f M l. M2 is the sum o f M l, saving and small time deposits at all financial institutions, overnight (and continuing contract) repurchase agreements issued by all commercial banks, overnight Eurodol lars issued to U.S. residents by foreign branches o f U.S. banks and m oney market accounts (MM), 3Numerous other financial innovations have occurred over the past several decades. This article focuses solely on the introduction of the principal new types of monetary assets that are included in the monetary aggregates. Moreover, the analysis is limited solely to the effects of FEDERAL RESERVE BANK OF ST. LOUIS Table 1 M1 and M2 in 1989 (billions of dollars) Components Currency Demand deposits Other checkable deposits Travelers checks M1 Money market mutual funds component1 Money market deposit account balances Savings Small time Overnight Eurodollars2 and repurchase agreements M2 Amount $217.5 280.4 278.5 7.3 $783.7 $276.3 475.0 410.0 1,105.5 79.1 $3,129.53 General purpose and broker-dealer funds. Eurodollar deposits issued to U.S. residents by foreign branches of U.S. banks. Components do not add to total because of rounding. which include both general purpose and brokerdealer m oney market mutual funds (MMMF) and money market deposit accounts (MMDA). Money market deposit accounts, which have unregu lated interest rates, w ere authorized at the same time as super-NOW accounts and became avail able in Decem ber 1982. Within the first tw o quarters o f 1983, they had grow n to 17 percent o f M2 (figure 2). Some o f this growth apparently came at the expense o f m oney market mutual fund accounts, since the total share o f m oney market accounts, MMDA and MMMF, rose by less than 17 percentage points; the share of total m oney market balances, rose from 10 per cent to about 24 percent o f M2 at the time. Since there is little difference between MMDAs and MMMFs, which becam e available in 1978, they are grouped together here as money mar ket accounts. The share o f MM in M2, called s22 below, rose to nearly 25 percent o f M2 by 1989 (see table 1 and figure 2). these innovations on M1 and M2; it ignores the effects on broader aggregates or on differently weighted aggregates, like the divisia or turnover-weighted aggregates. These other measures are discussed by Barnett (1982) and Spindt (1985). 39 Figure 1 Share of Other Checkable Deposits in Total Checkable Deposits Seasonally Adjusted Quarterly Data Percent o I....... 1970 Percent 60 ...................................................................................................1 o 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 1989 Figure 2 Share of Money Market Instruments in M2 Percent 30 Percent 30 Quarterly Data Money Market Balances1 I Money Market Deposits 1974 75 76 77 78 79 80 81 82 83 84 85 86 87 88 'Money market balances include both the money market deposit account and money market mutual fund components of M2, which are not seasonally adjusted. 1989 JULY/AUGUST 1990 40 THE FINANCIAL INNOVATIONS HYPOTHESIS The financial innovations hypothesis described here focuses primarily on the effects o f the grow th o f these new assets on M l. According to this hypothesis, the introduction o f interestbearing checking accounts made depositors m ore willing to hold savings balances in their checking instead o f their savings accounts. Thus, the grow th o f other checkable deposits, especially nationwide NOW accounts in 1981 and super-NOW accounts in 1983, was expected to boost total checkable deposits and M l and raise the interest elasticities of their demands.4 In addition, movements o f funds from savings to checkable deposits w ere expected to take place among com ponents o f M2, so that the total demand for M2 was unaffected by shifts to other checkable deposits. Similarly, the shift of funds into MMDAs was expected to flow from other com ponents o f M2, especially MMMFs; thus, the expansion of MMDAs was not ex pected to boost M2.5 One implication o f this hypothesis is that the growth o f MMDAs, or of MMMFs earlier, did not affect the demand for M l, its use or its composition.6 If these assets provide transaction services that are substitutes for total checkable deposits, however, then shifts to these balances should reduce the de mand for total checkable deposits relative to currency holdings, or raise the currency ratio. Such shifts would also reduce the overall de mand for M l. W hether m oney market innova tions had any significant effects is also tested below. The surge in the share o f MMs in M2 in early 1983 was associated with a sharp rise in M2 4Rasche (1988a) cites several studies which argue that financial innovations lowered the interest elasticity of money demand. More recent proponents of a financial in novations effect argue for an increase in this elasticity. Rasche (1987, 1988a and 1988b) has provided evidence for a rise in the interest elasticity of M1 demand, but he does not link this to financial innovations. Friedman (1988), Moore, Porter and Small (1988), Carlson (1989), Mehra (1989) and Poole (1988) also have pointed to the rise in the interest elasticity of M1 demand, although for different reasons. The first four studies suggest that this effect arose from financial innovations, while Poole suggests that it is not a recent development; instead, only its recognition is recent. 5See Thornton (1983). In late 1982, the FOMC anticipated that maturing all-savers certificates and the impending in troduction of MMDAs would temporarily boost M1 and, to a lesser extent, M2. The FOMC decided in October 1982 to set no short-run objective for M1, but to place greater FEDERAL RESERVE BANK OF ST. LOUIS growth from a 9.1 percent rise in the four quarters o f 1982 to a 16.6 percent annual rate in the first half o f 1983. While this movement ran counter to the financial innovations hypo thesis, many thought that it was transitory and carried little implication for future econom ic perform ance.7 Testing the Financial Innovations Hypothesis In this article, the financial innovations hypo thesis is tested by examining w hether these new assets have influenced the use, composition or demand for total checkable deposits, M l and M2 as predicted. If total checkable deposits and M l are boosted by inflows o f savings into other checkable deposits, then the total checkable deposit turnover rate—the ratio o f debits on total checkable deposits to total checkable de posits—should be inversely related to the share o f other checkable deposits in total checkable deposits (si = OCD/TCD). Similarly, the desired ratio o f the currency com ponent o f M l to the total checkable deposit com ponent also should be inversely related to s i.8 W hen the effects o f other checkable deposits on M l and M2 are investigated, the innovations measures used are their ratios to M l ( s l l = OCD/M1) and to M2 (sl2 = OCD/M2), respectively. If M l is increased by an inflow o f savings into other checkable deposits, then the demand for M l, given its other determinants, must be posi tively related to s l l . According to the financial innovations hypothesis, the impact of money market balances, measured relative to M l (s21 = MM/MI), on M l demand is zero. Similarly, if the hypothesis is correct, the demand for M2 should weight on M2. There was no indication that M2 would rise relative to M1, especially by as much as it did. 6Some analysts, however, point to the similarities between super-NOW and money market accounts; the latter offer limited checking services and unregulated interest rates. They suggest that money market balances are close substitutes for M1. See Cox and Rosenblum (1989) and Motley (1988), for example. 7For example, the FOMC’s initial target range for M2 an nounced in February 1983 called for M2 growth in the 7-to-10-percent range from the February-March average to the fourth quarter of 1983. This range was viewed as com parable to the 1982 range of 6 to 9 percent, allowing for a further boost to M2 due to new MMDAs. Hafer (1985) discusses these developments and their effects on the FOMC deliberations in detail. 8The appendix to this article presents a more formal discus sion of the tests of the effects of financial innovations. 41 be unrelated both to other checkable deposits, measured by sl2, and to m oney market bal ances, measured by s22. Testing f o r a Shift in the Interest Rate Elasticity The effect o f other checkable deposits on the interest elasticity o f each relationship also is ex amined. The financial innovations hypothesis in dicates that the weighted average cost o f hold ing total checkable deposits and M l and the in terest elasticity o f various monetary linkages are functions o f the relative size o f other checkable deposit balances. The implication is that the relevant interest elasticity rose, on average, after the introduction o f other checkable deposits. Under the financial innovations hypothesis, the rise in the interest elasticity is a function o f si, the relative size o f other checkable deposits. Thus, if P0 is the interest elasticity before the introduction o f other checkable deposits (that is, when s i is zero), then following this innovation the interest elasticity becom es ft* = p 0+pi si. In the log-linear relationships estimated below, the interest elasticity following the advent o f other checkable deposits is found from the ft coefficients in the expression: [30 Ini + /3,(sl Ini); the interest elasticity is (30 plus p l weighted (multiplied) by the average value o f si. In a first-difference equation, the appropriate expres sion is: P0 Alni + /}, A(sl Ini). W hether the in terest elasticity has increased as a result o f this financial innovation is indicated by the sign and statistical significance o f /?,. In summary, in this study the financial in novations hypothesis is rejected if: (1) measures o f other checkable deposit innovations have no significant effect on the M l-related variables and their interest elasticities, (2) these same measures have a significant effect on the size or interest elasticity o f M2 demand, or (3) mea sures o f m oney market innovations have any significant effect on the use, com position or de mand for M l or the demand for M2. These re lationships are examined below. FINANCIAL INNOVATIONS AND THE DEPOSIT TURNOVER RATE The turnover o f other checkable deposits, their debits per dollar o f deposits, is low er than the turnover o f demand deposits. For example, in May 1989, the annual rate o f debits per dol lar o f demand deposits at banks outside New York (where demand deposit turnover is nearly seven times larger) was 467.5; turnover on ATS and NOW accounts at commercial banks was only 18.2 times per year, m uch closer to the 3.6 rate on savings deposits at commercial banks.9 The similarity betw een the turnover o f ATS and NOW balances and that on saving deposits is sometimes taken as evidence to support the financial innovations hypothesis. The hypothesis says that other checkable de posits include balances that would have been held in savings or other non-M l balances before interest-bearing checking accounts becam e avail able. As these savings flow ed into other check able deposits, the turnover o f total checkable deposits should have fallen, and its interest elas ticity should have been altered. Figure 3 shows the natural logarithms o f the turnover rate for demand deposits and total checkable deposits (demand, ATS and NOW balances) since 1970. Turnover has a strong up w ard trend; for example, the turnover rate of demand deposits m ore than doubled from 1970 to early 1979. The tw o measures began to de viate in late 1978, w hen ATS accounts w ere in troduced, reflecting the low er turnover rates for ATS and NOW balances. The upward surge o f demand deposit turnover, especially in 1981, suggests that low er turnover deposits w ere switched from demand deposits to the new ac counts. M ore important, however, the turnover rate for total checkable deposits rose in 1981, counter to the decline predicted by the financial innovations hypothesis. Overall, the turnover rate for total checkable deposits looks more like a continuation o f the 1970-78 demand deposit turnover series than does the demand deposit turnover series itself. 9These data are available in the Federal Reserve statistical release, G.6, Debits and Deposit turnover at Commercial Banks. Debits on ATS and NOW accounts, like those on demand deposits, typically are third party payments; debits on savings, on the other hand, typically are in-bank withdrawals. Moreover, deposit turnover is substantially larger for business accounts than individuals; only the lat ter, however, can legally hold NOW and ATS accounts. JULY/AUGUST 1990 42 Figure 3 Demand Deposit and Total Checkable Deposit Turnover Seasonally Adjusted Monthly Data Logarithm 7.0 Logarithm 7.0 Demand deposit turnover Total checkable deposit turnover 1970 71 72 73 74 75 76 77 78 79 Deposit turnover measures are velocity mea sures; as such, they are related to the same fac tors, like interest rates and income, that influ ence the demand for money. Higher interest rates, by increasing the cost o f holding check able deposits, should reduce the quantity of these deposits demanded and increase their turnover rates. As incom e rises, the demand for these deposits should rise; whether the turn over rate rises or falls, however, depends on whether debits rise m ore or less than the de mand for checkable deposits. The continuous annualized grow th rate o f monthly total check able deposit turnover, CDT, was estimated as a function o f the continuous annualized rates of increase o f the three-month Treasury bill rate, R, and real personal income, y, for the period January 1979 to January 1989. The financial innovations hypothesis indicates (1) that a rise in s i should significantly reduce the turnover o f total checkable deposits and (2) that a rise in m oney market balances, measured here by a rise in the ratio o f m oney market balances to total checkable deposits, s2, should not affect it. This was tested by adding current http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis 80 81 82 83 84 85 86 87 88 1989 and up to 12 lagged values o f the annualized first-differences o f s i and s2, labeled D sl and Ds2, respectively, to the turnover equation; ad ding lagged effects beyond one month, h ow ever, was uniformly unnecessary. The estimate for total checkable deposit turn over that contains the most statistically signifi cant innovations term is: (1) CDTt= 13.00 - 0.043R, + 0.110R,_, (5.22) (-1 .3 6 ) (3.62) - 1.013yt_, + 0.227Dsl._, (-2 .4 5 ) (0.76) Pt = 0.255 (2.80) P2 = 0.244 (2.68) R2 = 0.15 S.E. = 29.255 D.W. = 2.00 (The numbers in parentheses in the equation estimates reported here are t-statistics.) The results in equation 1 show that the share o f other checkable deposits has not significantly depressed the turnover o f checkable deposits; instead, the estimated effect is positive, but 43 statistically insignificant.10 This result is counter to the financial innovations hypothesis. If financial innovations increased the interest elasticity o f total checkable deposits turnover, then the coefficients on the interest rate terms (R„ Rt_,) in equation 1 should be related to si. To test whether these coefficients have increas ed with the rise o f the share o f other checkable deposits in total checkable deposits, the an nualized change in the product (sljlnR,) for the current and past month are added to equation 1. The sum o f these coefficients is positive, 0.03, but it provides no significant explanatory pow er to the equation. The F-statistic for testing whether these coefficients are zero is F2112 = 0.04, well below the critical value (5 percent) o f 3.08. Thus, financial innovations, as defined here, have had no significant effect on the interest elasticity o f total checkable deposit turnover. Again, this result is counter to the financial in novations hypothesis. FINANCIAL INNOVATIONS AND THE CURRENCY-DEPOSIT RATIO The currency ratio, the ratio o f currency held by the public to its total checkable deposits, is a principal determinant o f the m oney multiplier (the ratio o f a monetary aggregate to the ad justed monetary base). M oreover, it is the prin cipal channel through w hich financial innova 10Either the current or first-lagged value of Ds1 is strongly and positively statistically significant when added to an identical equation for demand deposit turnover growth. When both current and lagged Ds1 values are included, however, neither is statistically significant. The standard error of the estimate is lower when the current value is us ed instead of the lagged value. The coefficient on the cur rent value is 1.025 (t = 3.49). The result in equation 1 is unaffected by regressing the growth rate of total debits on the same right-hand-side variables and on the growth rate of total checkable deposits; the coefficient on Ds1t_, is 0.282 (t = 0.95) in this case. Finally, when equation 11 in the appendix is estimated using the nonlinear least squares method, neither f nor gd is significantly different from zero. The estimates of f and gd are 0.005 (t = 0.01) and 0.021 (t = 0.37), respectively. The turnover rate for deposits, excluding demand deposits in New York (and their debits) was also examin ed. Its growth rate is white noise and is independent of in terest rates or real personal income. It is also not signifi cantly correlated with the current or lagged values of the changes in the financial innovation shares. For example, the correlation coefficient for the growth rate of turnover of total checkable deposits, excluding New York demand deposits, and the first lagged change in s1 is 0.023. This insignificnat correlation rejects the implication of the finan cial innovations hypothesis that this correlation is significantly negative. tions can affect the link betw een Federal Re serve actions and the monetary aggregates.11 The desired ratio o f currency to total checkable deposits is the outcom e o f a portfolio decision based on the relative costs and benefits of holding each means o f payment. If total check able deposits now include a larger com ponent o f savings balances than they did earlier, then the increase in the share o f other checkable deposits in total checkable deposits should have lowered the currency ratio. In addition, if money market accounts are a substitute for checkable deposits included in M l, then the introduction and spread o f money market holdings should have reduced total checkable deposits relative to currency holdings and raised the currency ra tio.12 According to the financial innovations hy pothesis outlined above, however, this latter ef fect should be zero. Figure 4 shows quarterly data on the ratio of the currency and the checkable deposit com ponents o f M l. This ratio does not decline in early 1981 or early 1983 when the largest boosts in savings held in other checkable deposits pre sumably w ould have occurred. Nor does the currency ratio rise in early 1983 when m oney market accounts surged. A modified time series model is used to test the effects o f these shifts on the currency ratio. The grow th rate o f the currency ratio can be described as a first-order autoregressive time "T he adjusted monetary base is described in Gilbert (1980 and 1987). A recent analysis of the behavior of the multiplier and its determinants can be found in Burger (1988). 12The effect of nationwide NOW accounts on the currency ratio is tested in Tatom (1982). A model of the demand for currency and demand deposits is used to test whether other checkable deposits lowered desired currency holdings relative to total checkable deposits. The tests re ject the financial innovations hypothesis. Rasche and Johannes (1987) show that the 1981 shift to NOW ac counts included a shift of savings to these accounts equal to about the 27.5 percent of such funds in the first four months of 1981. While this proportion also was suggested by the staff of the Federal Reserve Board, they suggested that it would have a continuing effect and applied it for all of 1981. Rasche and Johannes, on the other hand, argue that this shift significantly, but only temporarily, reduced the currency ratio and raised the money multiplier. They find no evidence that the shift to other checkable deposits or money market accounts had a permanent effect on the currency ratio or the multiplier. See Rasche and Johannes (1987, pp. 60-69). JULY/AUGUST 1990 44 Figure 4 Currency/Deposit Ratio1 Seasonally Adjusted Quarterly Data Percent 45 Percent 45 40 40 35 35 30 30 ■ n 25 20 25 I I I I I ■ + i . » i ’ 1 V 1 S 1 1 1 1959 61 63 65 67 69 71 73 75 77 1Ratio of Currency Component to Checkable Deposit Component of M1 1 79 • fl 81 it ? 83 1 85 I I 87 1 20 1989 series process; tw o other factors also have had a major impact on the currency ratio over the past 15 years and they are controlled for in the following estimates.13 The first factor is energy prices, which rose sharply in 1973-74 and in 1979-81 and fell sharply in 1986. A rise in ener gy prices raises expenditures that use currency relatively m ore than it raises expenditures that rely m ore heavily on checkable deposits. Thus, the currency ratio rises when energy prices in crease.14 The second factor is the transitory ef fect o f the credit control program in 1980, which temporarily boosted currency demand relative to checkable deposits in the second quarter o f the year. Credit limitations increase the use o f currency, especially in transactions that would otherwise be facilitated by retail credit.15 Finally, the current and past quarter's three-month T-bill rates are included to examine the interest rate elasticity o f the currency ratio; longer lags for the interest rate variables are not statistically significant. 13Rasche and Johannes (1987) argue for the superiority of a time series model over a structural approach like that used in Tatom (1982); the modifications here are made to in clude the sizable known effects of the two energy price shocks and to test whether the currency ratio’s interest elasticity was affected. checkable deposits. A related argument is that a change in the mix of personal consumption expenditures toward nondurable purchases raises the currency ratio. See Dotsey (1988). 14Tatom (1985) provides evidence that money demand is af fected by energy price increases. The currency-ratio effect may arise, at least in part, through gasoline purchases that affect currency demand more than the demand for FEDERAL RESERVE BANK OF ST. LOUIS The model o f the currency ratio, k, estimated for the period III/1959 to IV/1989 is shown in the first column o f table 2. The dependent vari able, kt, is the annualized continuous rate of growth o f the currency ratio. The annualized 15The effect of the credit control program on the money stock is discussed in Tatom (1982) and Hein (1982). Also see Wallace (1980) for an analysis of the effects of credit controls on currency demand. 45 continuous rate o f increase o f the relative price o f energy resources, p', is measured by the ratio o f the producer price index for fuel, p ow er and related products to the implicit price deflator for business sector output. The creditcontrol variable, D80, equals one in the second quarter o f 1980, negative one in the third quar ter o f 1980, and zero otherwise. These indepen dent variables are generally strongly statistically significant in the estimates shown in table 2.18 W hen current and lagged (up to four) values o f D sl or Ds2 w ere added to the model, only the estimate with the current-quarter change in s i (Dsl), shows a statistically significant innova tions effect; it is reported in the second column o f table 2. Although, the negative coefficient on D s lt is not statistically significant at a 5 percent level in a two-tail test, it is significantly negative using a one-tail test o f the negative effect pre dicted by the hypothesis.17 No other individual or group o f current or lagged changes o f the financial innovations variables are as signifi cant.18 These results suggest that growth in other checkable deposits has significantly low ered the currency ratio, w hich is consistent with the financial innovations hypothesis.19 This effect is weak, however, and is quite sen sitive to the exclusion o f only one observation— the second quarter o f 1981. W hen this quarter is omitted, the coefficient on D sl falls in ab solute value to -0 .0 7 3 , and its t-statistic falls to -0 .8 9 , which is far from statistical significance even with a one-tail test. Thus, the significant result for D sl, in table 2 is spurious. The largest rise in the s i measure occurs in 1/1981 not in the second quarter; the omission o f the 1/1981 observation, however, does not affect the signif icance o f Dsl,. The decline in the significance of D sl when the 11/1981 observation is omitted 16The F-statistic for a Chow test of the stability of the equa tion using the first and second half of the whole sample period is F s.n i =0.64, well below the 5 percent critical value of 2.30. Thus, the stability of the currency ratio estimate cannot be rejected. 17ln earlier versions of this article, this effect was insignifi cant even with a one-tail test. For example, before the February 1990 benchmark revisions, the estimate for the period 111/1959 to 111/1989 had a coefficient of — 0.101 (t = -1.24). The critical t-statistic value for a onetail test is about 1.65. The significance of the rest of the results reported here was not so affected. The nonlinear least-squares estimate of equation 17 in the appendix (when g equals zero) yields essentially the same result as in the text; in particular, the point estimate of f is 0.1324 (t = 1.83). The g parameter is set equal to zero in this estimate because it is not significantly different from zero when freely estimated. Table 2 Tests for the Ratio of Currency to Total Checkable Deposits (k) Dependent Variable: 400Alnk Period: 111/1959 to IV/1989 Constant 0.496 (1.70) 0.750 (2.32) 0.419 (1.37) K-, 0.503 (7.28) 0.477 (6.81) 0.503 (7.20) R, 0.013 (1.95) 0.011 (1.75) 0.014 (2.02) R,-, 0.025 (3.83) 0.028 (4.16) 0.023 (3.28) p ;- 0.058 (3.13) 0.059 (3.21) 0.053 (2.81) D80 11.292 (4.73) 10.783 (4.53) 11.398 (4.75) -0.135 (-1 .7 9 ) Ds1, -0.011 (-0 .3 9 ) D(s1, InRJ 0.037 (1.30) D(s1,_, InR,.,) R1 0.55 0.56 0.55 S.E. 3.052 3.023 3.054 D.W. 2.11 2.10 2.13 -1.01 -0 .9 2 -1 .2 0 h 18For example, the coefficient on Ds2 is 0.013 when added to the equation in the first column, and its t-statistic is only 1.19. 19Although Rasche and Johannes find a significant transitory decline in the currency ratio in early 1981, this is not found in the error in either the first or second quarter of 1981 for the first equation in table 2. This difference may arise because they use monthly, seasonally unadjusted data, while seasonally adjusted quarterly data are used here. In the form estimated, their four-month long reduc tion corresponds to one observation here. The tests here cannot readily determine whether such a brief transitory effect of financial innovations took place. JULY/AUGUST 1990 46 does not occur from a decline in the variance of D sl; the standard deviation of D sl rises from 0.076 to 0.082 when the 11/1981 observation is omitted. The significant result in table 2 arises from a spurious decline in the currency ratio in n/1981, when s i growth was relatively large. The third column in table 2 examines whether the interest elasticity o f the desired currency ratio increased in absolute value as a result of financial innovations. The results show a posi tive, but statistically insignificant, change in the interest elasticity. Neither interaction term is in dividually statistically significant, and the test statistic that they are jointly zero, F2114 = 0.91, is not statistically significant. Therefore, the hypothesis that financial innovations raised the interest elasticity o f the currency ratio is rejected. FINANCIAL INNOVATIONS AND MONEY DEMAND The evidence above on financial innovations influence on total checkable deposit turnover and the currency ratio rejects the financial in novation hypothesis. These results do not ad dress the m ore familiar literature on M l de mand or the velocity problem; nor do they ex amine the implications o f the financial innova tion hypothesis for M2. Figure 5 shows the income velocity o f M l and M2 measured by the ratio o f nominal gross na tional product to M l and M2, respectively. Movements in velocity inversely reflect m ove ments in m oney demand. The velocity o f M l has a strong positive trend until 1981, while M2 velocity does not appear to have a noticeable trend either before or after 1981. These velocity 20Both Hetzel and Mehra (1989) and Judd, Motley and Trehan (1988) take this view; indeed, the central issue in the money demand literature, according to these papers, seems to be, first, whether the recent shifts and instability of M1 demand are permanent or will disappear after some transition to a deregulated environment, and second, if the breakdown in M1 demand is only transitory, whether its statistical properties will dominate those of M2 demand when M1 demand “ settles down.” Judd, Motley and Trehan are more optimistic about a return to normal than Hetzel and Mehra. More recently, Hetzel (1989) and Mehra (1989) provide arguments intended to reinforce their view. Carlson and Hein (1980), Hafer (1981) and Tatom (1983a) report evidence on the breakdown of the M2-GNP link after 1977, however. Tatom (1983b) and Darby, Poole, et al. (1987) provide a fuller treatment of the potential causes and consequences of the change in the behavior of M1 velocity. 21Rasche (1988a) extends his 1987 M1 analysis to M2, M3 and broader measures. FEDERAL RESERVE BANK OF ST. LOUIS patterns often are cited as evidence that the de mand for M l, but not for M2, becam e less stable in the early 1980s, supporting the finan cial innovations hypothesis.20 The Demand f o r Ml Rasche provides a model o f the demand for M l and other monetary aggregates, w hich he argues has been stable for a long time.21 He ex plains that the shift in M l velocity behavior is a "shift in the drift” attributable to a change in the systematic components o f velocity that are impounded in the mean o f the growth rate specification or in the trend o f the level o f ve locity.22 Rasche also finds evidence that the in terest elasticity o f M l demand rose after 1981. He argues, however, that the timing o f financial innovations and their purported effect on M l demand are inconsistent with the timing o f the "shift in the drift” that he finds. Rasche’s evi dence also indicates that the demand for M2 is stable. In Rasche’s model, m oney demand, that is, nominal m oney per dollar o f GNP, depends upon the interest rate (the three-month Treasury-bill rate), real income and unanticipated inflation. In quarterly estimates, real income, x, is measured by real GNP, and unanticipated inflation, P“, is measured by the residuals from an M AI model o f changes in the annualized continuous rate o f increase o f the implicit price deflator for GNP. The incom e and interest rate effects on m oney demand occur over three quarters.23 An unrestricted version o f Rasche’s M l de mand equation, estimated for the period 11/1953 to IV/1989 is; 22This argument rules out shifts in M1 velocity due to changes in its response to economic factors that deter mine it or to changes in the error structure of the random elements that affect it. These two sources are typically the basis for claims of increased uncertainty or increased in stability in a demand function. Rasche conjectures, how ever, that the shift in the drift arises from the decline in in flationary expectations or a rise in the instability of the economy, but he finds no direct evidence supporting these arguments. 23Several coefficient restrictions are tested in Rasche (1987) and used in Rasche (1988a, 1988b). These are not impos ed here because they could bias the tests of the financial innovations hypothesis. 47 Figure 5 Income Velocities of M1 and M2 Seasonally Adusted (2) M l, - GNP, = -1 .9 8 9 - 0.036[400/3(lnR,-lnR,_3)] (-4 .5 3 ) (-4 .4 0 ) - 0.517 P" - 0.703 x, (-4 .1 6 ) (-1 1 .9 7 ) + 0.407 [400/2(lnx,_,-lnx,_3)] (4.17) + 2.336 D82, - 0.141 D82.DR13, (3.28) (-6 .0 7 ) R2 = 0.68 D.W. = 1.92 S.E. = 2.679 p = 0.227 (2.74) 24Rasche (1988a) omits the first and second quarters of both 1980 and 1981 in arriving at his stability results. These quarters are included here; the adjusted R2 and standard error actually improve when these quarters are included in estimating equation 2. For the M2 results, the adjusted R2 reported below falls slightly when these quarters are in where GNP is nominal GNP, and GNP and x are the annualized continuous growth rates o f nom inal and real GNP, respectively, D82 equals one from 1/1982 on and zero earlier, and DR13, is the variable in brackets in the second term on the right-hand-side o f the equation.24 The signifi cant intercept shift (D82) changes the 2.0 per cent trend rate o f velocity increase until 1982 into a 0.35 percent trend rate o f decline subse quently; the latter rate, however, is not signifi cantly different from zero. The last term in equation 2 tests w hether the magnitude o f the interest elasticity o f m oney demand rose; accor ding to the estimate, it rose significantly in ab solute value. eluded, but no other noticeable changes occur in any of the coefficients, JULY/AUGUST 1990 48 To test whether the rise in s l l has raised M l demand, the variable 400 A sll is added to the equation. The financial innovations hypothesis predicts that its coefficient should be significant ly positive. W hen this variable is added to equa tion 2, however, its coefficient is negative, but statistically insignificant, -0 .0 6 3 (t = —0.58). This result refutes the financial innovations hypothe sis about the effect o f the growth of other checkable deposits on M l demand.25 To test w hether the rise in m oney market deposits influenced M l demand, w hich the fi nancial innovations hypothesis denies, the money market innovation measure, 400 As21, is added to the M l demand equation; the result is: (3 )M lt-G N P ,= -1 .9 1 8 - 0.035[400/3(lnRt- lnR,_3)] (-4 .4 5 ) (-4 .3 5 ) - 0.533 P" - 0.699 xt (-4 .3 4 ) (-1 2 .1 0 ) + 0.392 [400/2(lnxt_ ,- l n x t_3)] (4.08) + 2.432 D82, - 0.161 D82.DR13, (3.49) (-6 .6 3 ) - 0.034 400As21t (-2 .3 6 ) R2 = 0.70 D.W. = 1.91 S.E. = 2.636 P = 0.22 (2.67) The m oney market innovations term is signifi cantly negative; the introduction and grow th of m oney market balances has statistically signifi cantly reduced M l demand. The coefficient on the innovations term is small, however; the rise in s21 to 1, about its level currently, has re duced the demand for M l by 3.4 percent. 25The absence of an effect of s11 on M1 demand implies that the growth of other checkable deposits is offset, dollar for dollar, by reductions in M1A (M1 less other checkable deposits). A similar test of whether no other checkable deposits should be added to M1A to obtain a stable de mand is easily rejected. The proportion of other checkable deposits that must be added to M1A to obtain an ag gregate whose demand is invariant to shifts in other checkable deposits is not significantly different from 100 percent. This rejects the usefulness of M1A, or at least the hypothesis that its demand is invariant to financial innovations. 26When equation 20 in the appendix is estimated with the same non-innovation variables as in equation 2, the esti mate of f, 0.014, is not significantly different from zero http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis The proportion o f MM that are transaction balances can be estimated from the coefficient on the innovations variable. The latter coeffi cient equals - g / ( l + gs21), w here g is the share o f transaction balances in MM, according to the derivation in the appendix to this article (eq. 20). Since the mean level o f s21 is 21.85 percent during the sample period, the estimated average value of g is 3.4 percent.26 A skeptic might argue that the significance of the last tw o terms in equation 2 actually dem on strates the validity o f the financial innovations hypothesis. After all, the demand fo r M l rose and its interest elasticity increased, just as the hypothesis predicted. Rasche’s timing argument indicates this is a spurious relationship, but more formal tests are possible. A test o f w hether the rise in the interest elasticity is related to the growth of other checkable deposits rejects this skeptical view. The term (sltlnRt- s l , _ 3lnRt_3) 400/3 re lates the shift in the interest elasticity systemati cally to the share o f other checkable deposits following the financial innovations hypothesis. W hen this innovations-related shift in the in terest elasticity is used in place o f the post-1981 shift variable D82DR13 in equation 2, its t-statistic is still significant, but low er (-3 .4 4 vs. -6.07); m oreover, the equation’s standard error rises (2.80 vs. 2.68). W hen both variables are in cluded in equation 2, however, the t-statistic for the innovations-related shift variable falls to -1 .4 3 , while the t-statistic for D82DR13 re mains strongly significant (t= 4.83).27 Similarly, the hypothesis that D82 is a proxy variable fo r the sharp rise in other checkable deposits in the early 1980s is tested by com par ing the effect of A s ll on equations 2 and 3 with and without D82. W hen this is done for equa tion 2, the t-statistic for 400 A s ll, is -0 .1 0 when D82 is omitted and, as indicated above, -0 .5 8 when D82 is included. W hen both are in(t = 0.15). The estimate for g, 0.037, however, is statistical ly significant (t = 2.40). 27These tests were also conducted using equation 3 instead of equation 2. When both measures are included in the equation, the shift in the interest elasticity in 1982 remains strongly significant (t= -5.36), while the s1-related interest elasticity shift is not (t= -1.00). The coefficient (-0.031) on the money market innovations term, 400 As21t, remains significant in this case (t= -2.12). 49 eluded, however, the coefficient on D82 (2.425) is the same size as in equation 2 and it remains statistically significant (t = 3.32). The use o f A sll and the sl-related shift in the interest elasticity, in place o f the 1982 constant and interest rate shifts, also are easily rejected w hen tested joint ly. Thus, the grow th o f the other checkable deposits does not account for the significance of the last tw o terms in equation 2. Similar results are obtained w hen these same substitutions are made in equation 3 and the significance o f the money market innovations term remains unaf fected by these changes. The Demand f o r M2 The M2 m oney demand equation that uses the same set o f variables for the same period as the M l estimate is: (4) M 2,-GNP, = 1 .3 8 5 - 0.055 DR13, - 0.734 P“ (3.63) (-7 .8 2 ) (-7 .4 1 ) -0.761x, + 0.428[400/2(lnx,.1 - lnx,_3)] (-1 5 .9 5 ) (5.25) -0 .8 2 2 D82t - 0.072 D82.DR13, (-1 .3 2 ) (-3 .6 2 ) R2= 0.77 S.E. =2.177 D.W. = 1.90 P = 0.289 (3.55) Unlike the M l estimate, the M2 estimate sug gests that there was no significant shift in the M2 demand intercept after 1981. The interest elasticity o f M2 demand rose significantly after 1981, however, like that for M l demand. The financial innovations hypothesis suggests that these innovations should have had no ef fect on the demand for M2. To test the hypoth 28No attempt was made to adjust the T-bill rate for the average rate paid on the components of M2 in order to better measure the opportunity cost of M2. Rasche (1988a) notes that, in an estimate like equation 4, inferior overall results were found when such a measure is used instead of the T-bill rate. 29When D82, and D82.DR13, are added to the estimate they are not statistically significant; the coefficient on D82, is -0.894 (t= -1.39), and that for the shift in the interest elasticity is -0.033 (t= - 1.60). 30These results do not depend on the inclusion of the four quarters that Rasche omits in his study. When these quarters are omitted, the standard error falls to only 1.926 percent and the other properties of the estimate are nearly identical. The same results also obtained when all four quarters of 1983, during which the largest shifts occurred, are omitted; in particular, the t-statistic for the s22 innova tion term is 2.49. esis, the same procedure used for M l was fol lowed for M2.28 The results indicate that the contemporaneous rise in the share o f money market balances in M2 (s22) has a statistically significant effect on the demand for M2, but that no other financial innovation variable (lags o f s22 or current and up to four lagged values o f sl2) has a significant effect. M oreover, when the contemporaneous share o f money market balances is included in the equation, neither the intercept shift nor the interest elasticity shift is statistically significant. The estimate, without the insignificant variables, is:29 (5) M2, - GNP, = 1.422 - 0.052 DR13, - 0.711 P" (3.55) (-7 .7 3 ) (-8 .2 9 ) -0.802x, + 0.373[400/2(lnx,_, - lnx,.,)] (-1 8 .4 8 ) (4.67) + 0.261 400As22, (6.04) R2= 0.81 S.E. = 2.006 D.W. = 1.79 P = 0.44 (5.75) The result that the rise in the share o f money market deposits significantly raised the demand for M2 runs counter to the financial innovations hypothesis.30 According to the estimate, a 25 percent share o f m oney market deposits in M2 (nearly its share at the end o f 1989) raises M2 demand relative to GNP by about 6.5 percentage points.31 Figure 6 shows the grow th rate o f M2 mea sured over four-quarter periods since 1978 and an adjusted grow th rate that removes the effect o f shifts in m oney market funds from M2 using the estimated effect in equation 5.32 The m oney- 31The theoretical value of the coefficient on 400 As22, is Q,/(1 -g, s22), where g ] is the proportion of MM balances that are not close substitutes for the rest of M2. This ex pression is derived in the appendix to this article. The sample estimate of g^ given the sample mean value of s22 of 5.39 percent, is 25.7 percent. When equation 22 in the appendix is estimated using the nonlinear least squares method and with the same other variables as in either equations 4 or 5, the other checkable deposit in novation's coefficient is not significantly different from zero, but the money market innovation term is. Using this method, the trend shift and interest-elasticity shift again are insignificant when the money market innovation term is included. For the counterpart to equation 5 in the text, the nonlinear least squares estimate of g, is nearly the same, 24.2 percent, (t = 5.17). 32This adjustment subtracts 0.261 s22, from the logarithm of M2 to obtain a series that is independent of s22. JULY/AUGUST 1990 50 Figure 6 The Growth Rate of M 21 Percent 15.0 Seasonally Adusted Quarterly Data Percent 15.0 2.5 1978 79 80 81 82 'Current rise from four quarters earlier. market-induced shift in M2 demand had the greatest effect on the measured grow th rate in 1983. In other periods, the grow th rate o f M2 has been affected only slightly. The adjusted grow th rates ranged from 6.3 percent to 9.8 percent from 1980 until 1987. The sharp accel eration o f M2 growth from 1980 to 1983 and subsequent slowing can be explained by the ef fect o f financial innovations, in this case, by the grow th o f m oney market balances. The effects on M2 velocity are shown in fig ure 7. Actual M2 velocity appears to vary about its mean in figure 7. W hen adjusted for shifts arising from m oney market accounts, however, M2 velocity has a positive trend, especially since the mid-1960s. CONCLUSION The financial innovations hypothesis that the introduction and acceptance o f other checkable FEDERAL RESERVE BANK OF ST. LOUIS 1989 deposits, especially NOW and super-NOW ac counts, have seriously, and perhaps permanent ly, distorted the measurement and effectiveness o f M l, but not M2, is widely accepted today. The counterpart o f this hypothesis—that the in troduction and grow th o f m oney market assets like m oney market deposit accounts had no ef fects on M l and M2—is as widely endorsed. A systematic investigation o f this hypothesis, which focuses on the turnover rate o f check able deposits, the desired currency-deposit pre ferences o f m oney holders, and the velocity or demand fo r M l and M2, however, generally re jects its claims. The financial innovations hypothesis implies that the turnover o f total checkable deposits and the currency ratio will decline significantly as the share o f other checkable deposits rises. The analysis here indicates that the turnover of total checkable deposits was not affected by these financial innovations. There was a signifi- 51 Figure 7 The Effect of Money Market Accounts on M2 Velocity Ratio 1.85 Seasonally Adjusted Quarterly Data Ratio 1.85 1.80 1.80 1.75 1.75 1.70 1.70 1.65 1.65 1.60 1.60 1.55 1.55 1.50 1.50 1959 cant decline in the currency ratio associated with the rise in the share o f other checkable deposits in total checkable deposits, but this significance is spurious in light o f its sensitivity to the omission o f only one observation and its refutation in the other tests presented here. The introduction and growth o f other check able deposits has had no significant effect on the velocity o f (demand for) M l. While there is evidence o f a shift in M l velocity and its interest elasticity after 1981, the tests here reject the financial innovations hypothesis that these shifts w ere related to the rise in the share o f other checkable deposits in M l in the early 1980s. The introduction o f m oney market deposit ac counts and the earlier introduction o f money market mutual funds have had a significant ef fect on the demand for monetary aggregates. The expansive growth o f these new balances has had no effect on the composition o f M l or the use o f checkable deposits. The demand for M l, however, was reduced slightly because of the grow th o f money market balances. More im 1989 portant, the growth o f these balances was asso ciated with a significant rise in the demand for M2. As a result, M2 velocity was depressed by the growth o f m oney market balances. Ironically, this reduction has provided unwarranted sup port to the view that M2 velocity is stationary and M2 demand is stable. Movements in the share o f m oney market accounts have ac counted for much o f the variation o f M2 growth over the past 10 years or so. Proponents o f the view that financial innova tions have distorted M l apparently have been focusing on the w rong innovation. According to the evidence here, explicit interest-bearing ac counts have not affected the use o f checkable deposits, the composition o f M l or the demand for M l (or M2 for that matter). Instead, the growth o f m oney market balances has signifi cantly affected the aggregates, raising M2 de mand and depressing its velocity. Money market deposits also appear to provide substitute trans action services for M l, so that their growth has had a small depressing effect on the demand for M l. JULY/AUGUST 1990 52 REFERENCES Barnett, William A. “ The Optimal Level of Monetary Aggrega tion,” Journal of Money, Credit and Banking (November 1982), pp. 687-710. Barro, Robert J., and Anthony M. Santomero. “ Household Money Holdings and the Demand Deposit Rate,” Journal of Money, Credit and Banking (May 1972), pp. 397-413. Bradley, Michael D., and Dennis W. Jansen. “ Deposit Market Deregulation and Interest Rates,” Southern Economic Jour nal (October 1986), pp. 478-89. Burger, Albert E. “ The Puzzling Growth of the Monetary Ag gregates in the 1980s,” this Review (September/October 1988), pp. 46-60. Carlson, John B. “ The Stability of Money Demand, Its In terest Sensitivity, and Some Implications for Money as a Policy Guide,” Federal Reserve Bank of Cleveland Economic Review (Quarter 3, 1989) pp. 2-13. Carlson, Keith M., and Scott E. Hein. “ Monetary Aggregates as Economic Indicators,” this Review (November 1980), pp. 12 - 21 . Cook, Timothy Q., and Timothy D. Rowe. “Are NOWs Being Used as Savings Accounts?” Federal Reserve Bank of Richmond Economic Review (May/June 1985), pp. 3-13. Cox, Albert H., Jr. Regulation of Interest Rates on Bank Deposits (University of Michigan, 1966). Cox, W. Michael, and Harvey Rosenblum. “ Money and Infla tion in a Deregulated Financial Environment: An Overview,” Federal Reserve Bank of Dallas Economic Review (May 1989), pp. 1-19. Darby, Michael R., Angelo R. Mascara and Michael L. Marlow. “ The Empirical Reliability of Monetary Aggregates as Indicators, 1983-87,” Ecomonic Inquiry (October 1989), pp. 555-85. _______ . “ The Money-GNP Link: Assessing Alternative Transaction Measures,” this Review (March 1984), pp. 19-27. _______ . “ Much Ado About M2,” this Review (October 1981), pp. 3-18. _______ . “ The New Monetary Aggregates,” this Review (February 1980), pp. 25-32. Haraf, William S. “ Monetary Velocity and Monetary Rules,” The Cato Journal (Fall 1986), pp. 641-62. Hein, Scott E. “ Short-Run Money Growth Volatility: Evidence of Misbehaving Money Demand?” this Review (June/July 1982), pp. 27-36. Hetzel, Robert L. “ M2 and Monetary Policy,” Federal Reserve Bank of Richmond Economic Review (September/October 1989), pp. 14-29. Hetzel, Robert L., and Yash P. Mehra. “ The Behavior of Money Demand in the 1980s,” Journal of Money Credit and Banking (November 1989), pp. 455-63. Jordan, Jerry. “ Financial Innovation and Monetary Policy,” in Financial Innovations: Their Impact on Monetary Policy and Financial Markets, Proceedings of a Conference held at the Federal Reserve Bank of St. Louis, October 1 and 2, 1982 (Kluwer Nijhoff, 1984), pp. 135-50. Judd, John P., and Bharat Trehan. “ Portfolio Substitution and the Reliability of M1, M2 and M3 as Monetary Policy Indicators,” Federal Reserve Bank of San Francisco Economic Review (Summer 1987), pp. 5-29. Judd, John P., Brian Motley and Bharat Trehan. “ The De mand for Money: Where Do We Stand?” unpublished paper presented at the Western Economics Association Meetings, Los Angeles, June 1988. Kareken, John H. “ Commercial Banks and the Supply of Money: A Market Determined Demand Deposit Rate,” Federal Reserve Bulletin (October 1967), pp. 1699-712. Darby, Michael R., William Poole, David E. Lindsey, Milton Friedman, and Michael J. Bazdarich. “ Recent Behavior of the Velocity of Money,” Contemporary Policy Issues (January 1987), pp. 1-33. Keeley, Michael C., and Gary C. Zimmerman. “ Deposit Rate Deregulation and the Demand for Transactions Media,” Federal Reserve Bank of San Francisco Economic Review (Summer 1986), pp. 47-62. Dotsey, Michael. “ The Demand for Currency in the United States,” Journal of Money, Credit and Banking (February 1988), pp. 22-40. Klein, Benjamin. “ Competitive Interest Payments on Bank Deposits and the Long-Run Demand for Money,” American Economic Review (December 1974), pp. 931-49. Friedman, Benjamin M. “ Lessons on Monetary Policy from the 1980s,” Journal of Economic Perspectives (Summer 1988), pp. 51-72. ________ “ The Payment of Interest on Commercial Bank Deposits and the Price of Money: A Study of the Demand for Money,” unpublished Ph.D. dissertation, University of Chicago, 1970. Frodin, Joanna H., and Richard Startz. "The NOW Account Experiment and the Demand for Money,” Journal of Bank ing and Finance (June 1982), pp. 179-93. Gavin, William. “ M1A—M.I.A.?” Federal Reserve Bank of Cleveland Economic Commentary (July 1, 1987). Gilbert, R. Alton. “A Revision in the Monetary Base,” this Review (August/September 1987), pp. 24-29. _______ . “ Revision of the St. Louis Federal Reserve's Ad justed Monetary Base,” this Review (December 1980), pp. 3-10. Hafer, R.W. “ The FOMC in 1985: Reacting to Declining M1 Velocity,” this Review (February 1986), pp. 5-21. _______ . “ The FOMC in 1983-84: Setting Policy in an Uncertain World,” this Review (April 1985), pp. 15-37. FEDERAL RESERVE BANK OF ST. LOUIS Klein, Michael A. “ Deposit Interest Prohibition, Transactions Costs, and Payments Patterns: A Theoretical Analysis,” Metroeconomica (November/December 1974), pp. 144-52. Kopcke, Richard W. “ Financial Assets, Interest Rates and Money Growth,” New England Economic Review (March/April 1987), pp. 17-30. Mehra, Yash P. “ Some Further Results on the Source of Shift in M1 Demand in the 1980s,” Federal Reserve Bank of Richmond Economic Review (September/October 1989), pp. 3-13. Moore, George R., Richard D. Porter and David H. Small. “ Modeling the Disaggregated Demands for M2 and M1 in the 1980s: The U.S. Experience,” in Financial Sectors and Open Economies: Empirical Analysis and Policy Issues, Pro 53 ceedings of a May 1988 Conference of the Board of Gover nors of the Federal Reserve System, forthcoming. Motley, Brian. “ Should M2 be Redefined?” Federal Reserve Bank of San Francisco Economic Review (Winter 1988), pp. 33-51. Morris, Frank E. “ Do the Monetary Aggregates Have a Future as Targets of Federal Reserve Policy?” New England Economic Review (March/April 1982), pp. 5-14. Nuetzel, Philip A. “ The FOMC in 1986: Flexible Policy for Uncertain Times,” this Review (February 1987), pp. 15-29. Poole, William. "Monetary Policy Lessons of Recent Inflation and Disinflation,” Journal of Economic Perspectives (Sum mer 1988), pp. 73-100. Porter, Richard D., and Edward K. Offenbacher. “ Financial Innovations and Measurement of Monetary Aggregates,” in Financial Innovations: Their Impact on Monetary Policy and Financial Markets, Proceedings of a Conference held at the Federal Reserve Bank of St. Louis, October 1 and 2, 1982 (Kluwer Nijhoff, 1984), pp. 49-98. Rasche, Robert H. “ Demand Functions for U.S. Money and Credit Measures,” Paper 8718, Department of Economics, Michigan State University (May 1988a) to be reprinted in Financial Sectors and Open Economies: Empirical Analysis and Policy Issues (Board of Governors of the Federal Reserve System, forthcoming). _______ . “ Monetary Policy and Financial Deregulation in the United States," Kredit und Kapital (Heft 3, 1988b), pp. 451-68. _______ “ M1 Velocity and Money Demand Functions: Do Stable Relationships Exist?” in Karl Brunner and Allan H. Meltzer, eds., Empirical Studies of Velocity, Real Exchange Rates, Unemployment and Productivity, Carnegie-Rochester Series on Public Policy (North Holland, Autumn 1987), pp. 9-88. Rasche, Robert H., and James M. Johannes. Controlling the Growth of the Monetary Aggregates (Kluwer, 1987). Roth, Howard L. “ Has Deregulation Ruined M1 as a Policy Guide?” Federal Reserve Bank of Kansas City Economic Review (June 1987), pp. 24-37. Santomero, Anthony M. “A Model of the Demand for Money by Households,” Journal of Finance (March 1974), pp. 89-102. Saving, Thomas R. “ Money Supply Theory with Competitive ly Determined Deposit Rates and Activity Charges,” Journal of Money, Credit, and Banking (February 1979), pp. 22-31. _______ . “A Theory of The Money Supply With Competitive Banking,” Journal of Monetary Economics (July 1977), pp. 289-303. _______ . “ Transactions Costs and The Demand for Money,” American Economic Review (June 1971), pp. 407-20. Siegel, Diane F. “ The Relationship of Money and Income: The Breakdowns in the 70s and 80s,” Federal Reserve Bank of Chicago Economic Perspectives (July/August 1986), pp. 3-15. Simpson, Thomas D. “ Changes in the Financial System: Implications for Monetary Policy,” Brookings Papers on Economic Activity (1: 1984), pp. 249-65. Spindt, Paul A. “ Money is What Money Does: Monetary Ag gregation and the Equation of Exchange,” Journal of Political Economy (February 1985), pp. 175-204. Startz, Richard. “ Implicit Interest on Demand Deposits,” Journal of Monetary Economics (October 1979), pp. 515-34. Tatom, John A. “ Interest Rate Variability and Economic Performance: Further Evidence,” Journal of Political Economy (October 1985), pp. 1008-18. _______ . “ Money Market Deposit Accounts, Super NOWs and Monetary Policy,” this Review (March 1983a), pp. 5-16. _______ . “Alternative Explanations of the 1982-83 Decline in Velocity,” in Monetary Targeting and Velocity, Federal Reserve Bank of San Francisco, Conference Proceedings, December 4-6, 1983b, pp. 22-56. _______ . “ Recent Financial Innovations: Have They Dis orted the Meaning of M1?” this Review (April 1982), pp. 23-35. _______ . “ Transactions Cost and the Supply of Real Average Demand Deposits,” unpublished Ph.D. disserta tion, Texas A&M University, 1971. Thornton, Daniel L. “ The FOMC in 1982: Deemphasizing M1,” this Review (June/July 1983), pp. 26-35. _______ . “ The FOMC in 1981: Monetary Control in a Changing Financial Environment,” this Review (April 1982), pp. 3-22. Wallace, Neil. “ Integrating Micro and Macroeconomics: An Application to Credit Controls,” Federal Reserve Bank of Minneapolis Quarterly Review (Fall 1980), pp. 16-29. Wenninger, John. “ Responsiveness of Interest Rate Spreads and Deposit Flows to Changes in Market Rates,” Federal Reserve Bank of New York Quarterly Review (Autumn 1986), pp. 1-10. JULY/AUGUST 1990 54 Appendix A Formal Statement of the Hypotheses Tested The financial innovations hypothesis, as presented and tested in this paper, states that the introduction and grow th o f other checkable deposits, OCD, distorted the measurement of both total checkable deposits and M l, but left the overall demand for M2 unaffected. M ore over, according to this hypothesis, the introduc tion and grow th o f m oney market balances (MM) had no effect on M2. Instead, the grow th o f these balances came at the expense o f other non-M l funds within M2, so that it had no ef fect on total checkable deposits, M l demand, or the composition o f M l. The hypothesis suggests that some fraction, f, o f other checkable deposits is not held as total checkable transaction balances and that money market deposit balances do not yield transaction services or are not held as part o f total check able transaction balances. Thus, the amount of total checkable deposits, TCD, that are "truly" transaction balances equals (1-fsl) TCD, where s i is the share o f other checkable deposits in total checkable deposits. If some proportion, g, o f MM are also transaction balances, then the total MM com ponent o f transaction balances can be written as gs2, w here s2 is the ratio o f MM to TCD. Total transaction balances, TTB, can be defined as: (1) TTB - ( l - f s l + gs2) TCD. In this framework, the financial innovations hypothesis is that l > f > 0 and g = 0. Prior to financial innovations, s i and s2 w ere zero and TTB equaled TCD. The effective quan tity o f M l was C + TCD, w here C is the curren cy com ponent o f M l. The effective quantity o f M l, designated M l*, when si and s2 are not zero, is C + TTB, or Ml-fOCD + gMM. If s l l is defined to be the ratio (OCD/M1) and s21 is de fined to be the ratio (MM/MI), then (2) M l* = M l ( l - f s l l + gs21). Since M l equals (1 + k) TCD, w here k is the ratio o f currency to total checkable deposits, s l l equals s l/(l + k) and s21 equals s2/(l + k). 1Since M2 = (1 + k + t) TCD, where t is the ratio of the nonMl components of M2 to M2, the ratios s12 and s22 are simply (1 + k + t)“ ‘ times s1 and s2, respectively. FEDERAL RESERVE BANK OF ST. LOUIS An effective quantity o f M2, called M2*, can be defined similarly. W hether or not certain proportions o f OCD and MM balances are ap propriately considered part o f TTB and M l*, they are definitionally part o f M2. This is the central reason that the hypothesis claims that M2 is unaffected by these innovations. If, h ow ever, some fraction o f these new deposits are not close substitutes for M2, then the effective quantity o f M2, M2*, should exclude these frac tions o f the new deposits. In particular, if some fractions, f o f other checkable deposits and gt o f MM balances, are held for non-M2*-related reasons, then shifts in holdings o f these funds will boost M2 relative to M2*, that is, (3) M2* = M 2 ( l - f 1s l 2 - g 1s22), where sl2 is the ratio o f other checkable depos its to M2 and s22 is the share o f m oney market balances in M2. According to the financial in novations hypothesis, the grow th o f other check able deposits or MM involves substitutions within M2 and does not affect its total; therefore, M2 equals M2* and gi and f ± equal zero.1 The hypothesis is tested below using the rela tionships in equations 1-3. In particular, tw o im portant econom ic variables, the turnover rate for total checkable deposits and the currency ratio, relate debits and currency holdings, respectively, to desired holdings o f checkable transaction balances. Movements in other check able deposits or m oney market deposits have predictable or systematic effects on the ratio of checkable transaction balances to observed total checkable deposits and, therefore on debits or currency holdings relative to total checkable deposits. Similarly, growth in these new assets affects the relationship o f M l* and M2* to their observed counterparts and, therefore, systemati cally affect the relationship o f the observed ag gregates, M l and M2, to the factors that influ ence the demands for M l* and M2*, respectively. The hypothesis also suggests that the interest elasticity of demand for transaction balances, 55 M l and M2 have been affected by financial in novations. The specific form o f the hypotheses and tests are derived below. THE TURNOVER RATE FOR TOTAL CHECKABLE DEPOSITS The turnover rate for total checkable transac tion deposits is the ratio o f debits, D, on these deposits to their total, TTB. If v, the turnover rate o f deposits held for transaction purposes is a function o f a vector o f variables, z0, then (4) D = v(z0)TTB. Substitution o f equation 1 in equation 4 yields: (5) D = v(z0) (1 - f s l +gs2) TCD. The left-hand side o f equation 5 includes any third-party debits on MM balances held for third-party payment, i.e., as checkable transac tion balances, (gMM). For simplicity, assume that debits include only third-party payments and thus exclude cash-withdrawal debits on both TCD and MM balances. If debits on money market balances, Dm, are also a function of gMM and the vector zQ above, or (6) Dm = vm(zo) gs2 TCD, then the debits measured against total checkable deposits D„ are balances (relative to TCD) will affect the turn over o f total checkable deposits. The sign o f this effect depends on whether 6 is positive, zero or negative, or whether transaction balances in MM have relatively low, the same or high turn over com pared with the weighted average turn over o f total transaction balances, v. A log-linear specification o f v(zo) is used, where zQ includes the current and past interest rate (i„ it_,) and real personal income, y „ or (9) lnv, = P0 + /?, Ini, + fi1 Ini,., + /?3 lny,. The log-linear specification o f equation 8 is (10) lnd, = P0 + (}l Ini, + P2 Ini,., + lny, + l n ( l - f s l + gds2), where vm/v is assumed constant. W hen equation 10 is differenced, the result is: (11) Alnd, = (3I Alni, + ft2 Alni,., + P3 Alny, + A l n ( l- f s l + gds2). The last variable in equation 11 is unknown because f, g and 6 are unknown. This problem is addressed indirectly in the paper.2 If f, g and 6 are constants, then (12) A ln ,(l-fs l+ gds2) = ---------- --------- dsl 1 - fs l +gds2 + ------- ^ ------- ds2. 1 - fs l +gds2 (7) Dt = D - D m = v(z0) [ l - f s l + gs2d] TCD, where 6 = (l-vm/v) and the turnover ratio for total checkable deposits is (8) d = D./TCD = v(z0)[l - fs l + gs2d]. A rise in s i reduces the turnover ratio for total checkable deposits; if f is zero, however, then movements in s i have no effect on v. If g and vm are not zero, movements o f funds into MM 2All the estimates in this article contain a term like Aln(1 —fsl + gds2) in equation 11. Estimating the constants like f and (g<J) directly by non-linear least-squares yields no differences from the result reported in the text for the financial innovations hypothesis. If f is correlated with movements in s1 or its counterpart measures below, the estimated coefficient on the share variables would be bias ed; if the correlation is positive, as proponents of the financial innovations might argue, this biases up the coeffi cient and biases the tests in favor of the financial innova tion hypothesis. The same argument applies to g. The op posite bias would arise if f and the other checkable deposit share were negatively correlated, but this is counterintuitive. There is no a priori reason to expect f or g (or f ] and g, below) to change, especially to change systematically with movements in the shares, however. The difference in the logarithm in the last term in equation 11 can be approximated using the total differential o f the expression in paren theses and replacing d sl and ds2 with Asl and As2.3 Thus, equation 11 can be written as: (13) Alnd, = p0 + /?, Alni, + P2 Alni,_, + p3 Alny, + P4 Asl. + P5 As2.> 3The coefficients on As1 and As2 involve s1 and s2. These coefficients are estimated as constants and are evaluated at the sample period average values for f, gd, s1 and s2. Note that gd is estimated from the s2 coefficient; conse quently, the hypothesis that g equals zero cannot be tested. If d equals zero (the turnover of transaction balances held in MM is the same as for the rest of such balances), then the coefficient on s2 will be zero; however, this does not imply that g is necessarily zero. JULY/AUGUST 1990 56 where /30 is an intercept which should have a value of zero, unless a significant time trend has been omitted from equation 10. Under the finan cial innovations hypothesis, P4< 0 and P5= 0. The Interest Elasticity o f the Turnover Rate The financial innovations hypothesis tested in the text implies that the interest elasticity o f m oney demand rose as a result of financial in novations. Since turnover is a velocity measure, a test is conducted o f whether the interest elasticity o f the turnover rate o f total checkable deposits rose in proportion to the growth o f si. In equations 9 and 10, this elasticity is constant and equals (Pt +P2). If Pl and P2 are functions of s i, for example, /3j=/?,”+ /},' si, and P2= P2 + P2 s i ,.,, then the terms (/}, Ini, + P2 lni,_,) in equations 9 and 10 must be replaced with (/3,° Ini, + 13/ si, + /J2°lni,_, + s l,_ 1lni,_1). In equation 13, P° replaces p2 replaces P2, and the additional terms / } / A(sl,lnit) and P2' A(sl,_jlni,_,) must be included. W hether the interest elasticity rose depends on whether P/, p2' and (P/ + P2’) are statistically significantly positive. THE CURRENCY RATIO The currency ratio is the ratio o f currency to total checkable transaction balances. Currency demand relative to total checkable transaction balances is (14) C = k(z,) TTB, where z, is a vector o f the determinants o f the desired ratio. With the advent o f OCD and MM balances, currency holding competes with all other transaction-related balances, or TTB. Sub stitution o f equation 1 in equation 14 yields (15) C = k(z,) ( 1 - f s l +gs2) TCD. W hen s i rises, currency demand declines, given TCD, z 1( f and g, if 0 < f < l . Changes in s2 have no effect on the currency ratio under the hy pothesis that g = 0. The variables in z, that determine the desired currency ratio, and are controlled for in testing the financial innovations hypothesis, include the autoregressive component, a first lag of the cur rency ratio, the current (i,) and past (it_,) in terest rate, energy prices, pe, and a credit con FEDERAL RESERVE BANK OF ST. LOUIS trol dummy variable, c. The first-difference of the log-linear form o f equation 15, with the ap propriate substitutions for z,, is: (16) Aln(C/TCD), = d0 + d, Alni, + d2 Alni,., + d3 Alnp' + d4 D80 + c55 Aln(C/TCD),_, + Aln(l - f s l + gs2), where D80 equals Ac. The last term on the right-hand-side can be approximated using the same argument used above for equations 11 and 12 since d l n ( l - f s l + gs2) equals [ - f / ( l - f s l + gs2)]dsl + [ g / ( l - f s l +gs2)]ds2. Thus, equation 16 can be written as: (17) Ain(C/TCD), = d0 + d, Alni, + d2Alni,_, + d3Alnp[ + d4 D80 + [ —f / ( l - f s l + gs2)] Asl, + [g/(l - fs l +gs2)] As2,. The financial innovations hypothesis, 0 < f < l , is tested by whether Asl, has a significant negative coefficient. The hypothesis g = 0 is tested by whether As2, has a significant coefficient. W hether the interest elasticity o f the currency ratio is affected by the grow th o f s i is also tested. The sum (d, + d2) in equation 16 or 17 is the interest elasticity o f the currency ratio. If each o f these components is a function o f si, then the interest components in k(z,) can be written as (d," + d /s l) Ini, + (d2° + d2' si) lni,_,, and [d,°+ d2°+ d / si, + d2' sl,_,] is the interest elasticity o f currency demand in this case. In the first-difference form given in equation 17, the interest rate components are replaced with d^Alni, + d2"Alni,_, + d /A ( s l, Ini,) + d2' A (sl,„,lni,_,). If financial innovations affect the interest elasticity, then 6/ and/or d2' are significantly different from zero. Since d, and d2 are negative, for the interest elasticity to becom e larger in absolute value requires that, d /, d2'< 0 and (d / + d2')< 0 . MONEY DEMAND Suppose “true" or effective M l demand, M l*, is a function of a vector o f variables z 3. Substi tuting equation 2 yields: (18) (1 - f s l l +gs21)M l = D(z2). In log-linear form, this equation can be re arranged as 57 (19) InMl = ln[D(z2)] - l n ( l - f s l l + gs21). W hen this is first-differenced and a similar substitution is made for the last term as was made in equation 13 and 17, the result is: (20) AlnMl = Aln[D(z2)] + d6 A sll, + d7 As21„ where d6 = f /( l - f s ll + g s 2 1 ) , and d7 = - g /( l - f s l l + gs21). If f is zero, then d6 equals 0. If 0 < f < l , however, then d6 is positive; that is, a rise in s l l should raise M l demand, given the variables in z 2. If g equals 0, then d7 equals 0; if g is positive, then d7 is less than zero. The variables included in z 2 are the interest rate, income and unanticipated inflation. The specification o f ln[D(z2)] also includes a shift in the interest rate elasticity o f m oney demand and a shift in the level of M l demand, where both shifts occu r in 1982. Therefore, tests are conducted to determine if these tw o com ponents o f z2 arise from financial innovations. For M2 demand, the same set o f tests are con ducted. In particular, if "true” M2 demand, M2* in equation 3, is a function o f variables z 3, E(z3), then substituting this in equation 3 yields (21) ( l - f ,s l 2 - g ,s 2 2 ) M2 = E(z 3). In the text, the z 3 vector includes the same set o f other m oney demand variables as M l, that is, z 3 equals z 2. In differenced log-linear form and using the exact differential to derive the discrete A l n ( l - f 1s l2 g 1s22), equation 21 becomes (22) AlnM2, = Ain E(z3) + d8 As22, + d9 Asl2, where d8 = [g,/(1 - f , s l 2 - g 1s22)] and d, = [f ,/(l-f,s l2 -g ,s 2 2 ) ]. Under the financial innovations hypothesis, f ,, g,, dg and d9 are all zero. The coefficients dg and d9 are positive if the proportions g, o f MM or f 1 o f OCD are positive; this result would in dicate that these proportions are not a close substitute, given z 3, for the rest o f M2. JULY/AUGUST 1990 Federal Reserve Bank of St. Louis Post Office Box 442 St. Louis, Missouri 63166 The Review is published six times per year b y the Research and Public Information Department o f the Federal Reserve Bank o f St. Louis. Single-copy subscriptions are available to the public f r e e o f charge. Mail requests f o r subscriptions, back issues, or address changes to: Research and Public Information Department, Federal Reserve Bank o f St. Louis, P.O. Box 442, St. Louis, Missouri 63166. The views expressed are those o f the individual authors and do not necessarily reflect official positions o f the Federal Reserve Bank o f St. Louis or the Federal Reserve System. Articles herein may be reprinted provided the source is credited. Please provide the Bank’s Research and Public Information Department with a copy o f reprinted material.