A Long History of FOMC Voting Behavior.

Henry W Chappell, Jr. [*]

Rob Roy McGregort [+]

We devise and apply a method for estimating, monetary policy reaction functions for individual members of the Federal Open Market Committee (FOMC) of the Federal Reserve. Our method uses members' votes on the monetary policy directive in FOMC meetings as the key source of data on individual preferences. The analysis provides a ranking by preference for ease for 84 FOMC members who served during the 1966-1996 period.

1. Introduction

In the literature on monetary policymaking, one frequently encounters references to "the central banker," "the monetary authority," or "the policymaker." The choice of language is not accidental. It accurately conveys two related messages about the state of that literature. One is that monetary policymaking is typically modeled in an institutionally naive manner. The second is that modelers implicitly assume that policy is made by a single decision maker, who presumably has stable preferences. For some purposes, this modeling environment may be appropriate, but if one wishes to describe real-world policymaking, it is surely inadequate.

In the United States, the most important monetary policy decisions are made by the Federal Reserve's Federal Open Market Committee (FOMC). This committee is composed of the seven members of the board of governors, including the chairman and five of the twelve district Federal Reserve Bank presidents. Governors are appointed to staggered 14-year terms by the president, while bank presidents are selected by bank directors and periodically rotate on and off of the FOMC. [1] The FOMC guides monetary policy through its adopted directives, which must be approved by a majority vote at regularly scheduled meetings. Clearly, the monetary authority (the FOMC) is not a single agent with immutable preferences. It is a committee whose decisions reflect an aggregation of preferences over a diverse and changing membership.

There have, of course, been studies investigating the internal dynamics of decision making at the fed. Many of these have analyzed patterns of voting on the monetary policy directive within the FOMC, concentrating on differences in the voting behavior of various groups of committee members. [2] In the past, differences among the governors and the bank presidents have led to widely publicized conflicts within the FOMC over whether to respond more actively to a sluggish U.S. economy. As reported by David Wessel in the Wall Street Journal of June 26, 1992,

President Bush's appointees--Mr. Mullins, Mr. Lindsey, and Susan Phillips--are said to be ready to push [interest] rates down again. They have at least a couple of allies among the presidents of Fed district banks.[ldots] They also face substantial opposition.[ldots] Two of President Reagan's appointees to the Fed, Wayne Angell and John LaWare, seem firmly opposed to further rate cuts.[ldots] What's more, two of the five district-bank presidents who currently vote on monetary policy--newcomer Jerry Jordan of Cleveland and Thomas Melzer of St. Louis-- [ldots] worry that the Fed may already have gone too far [in easing monetary policy]. (p. A7)

Such anecdotal evidence concerning the monetary policy decision process, along with the evidence provided by the dissent voting record, suggests that focusing on groups of FOMC members and their influence over monetary policy may sometimes obscure important differences among individuals within these groups. It also shows that an improved understanding of monetary policymaking requires investigating the motivations of specific individuals.

In this paper, we document the voting behavior of individual FOMC members over the 1966-1996 period. Specifically, our results provide a ranking by preference for ease for 84 individuals who served on the FOMC during this period. To accomplish our objectives, we estimate parameters of individual FOMC members' monetary policy reaction functions. These reaction functions describe empirical relations between macroeconomic conditions and individuals' desired settings of a monetary policy instrument. We use members' votes on the monetary policy directive as the key source of data in revealing individuals' desired policies.

The Memoranda of Discussion for FOMC meetings through March 1976 and transcripts of committee meetings held since March 1976 reveal information about members' desired policy actions that is more detailed than that revealed by voting records alone. [3] Unfortunately, the fed has not yet released transcripts for the 1978-1981 period, and transcripts for the period since 1993 are still subject to a five-year holding period prior to release. Using the published voting record therefore allows us to consider a longer sample period than would be possible based on the Memoranda and transcripts. Further, changes over time in Federal Reserve operating procedures affect the character of the committee's monetary policy discussions. For example, under the federal funds rate targeting procedure used during the 1970s, FOMC members frequently stated the target level of the funds rate that they preferred to see adopted. Under the borrowed reserves targeting procedure used during the 1980s, FOMC members typically stated the t arget level of reserve borrowings that they preferred to see adopted. Differences such as these affect our ability to consistently extract comparable information about committee members' positions on the appropriate course for monetary policy from the textual record. Inevitably, then, some degree of subjectivity will be involved in trying to code members' positions in detail. The use of voting records helps us avoid any systematic biases that this subjectivity might create.

We begin by presenting our model in the second section. The empirical specification and the data set employed in the analysis are described in the third section. Results are discussed in the fourth section, and conclusions are offered in the final section.

2. A Model of FOMC Decision Making

In this section we describe a model of FOMC decision making that permits reaction function parameters to vary across committee members and that can be estimated using available macroeconomic time series and FOMC voting records. Our approach offers three fundamental advantages over previous studies of monetary policymaking. First, our empirical model links policy outcomes to the reaction functions of FOMC members, providing microfoundations for aggregate reaction functions. Second, specifying differences across members as differences in reaction function parameters implicitly controls for the state of the economy and for prevailing policy when evaluating members' voting records. Third, differences across FOMC members can be interpreted as differences in desired settings of a policy instrument, which are more meaningful indicators of policy preferences than dissent voting frequencies. Our presentation proceeds by discussing the specification of policy preferences of individual FOMC members, the connection betw een those preferences and resulting policy outcomes, and issues associated with estimation. [4]

Individuals' Reaction Functions

We assume that individuals occupy N positions on the FOMC (excluding the position of the chairman). [5] Each of the N members is assumed to have a desired interest rate reaction function of the following form: [6]

[[r.sup.*].sub.it] = [[alpha].sub.0] + [[[sum].sup.k].sub.k=1] [[alpha].sub.k][D.sub.kit] + [[[sum].sup.J].sub.j=1] [[beta].sub.j][X.sub.jt] + [e.sub.it], i = 1, [ldots], N; t = 1, [ldots], T. (1)

The dependent variable, [[r.sup.*].sub.it], is member i's desired federal funds rate for the inter-meeting period following meeting t. This variable is unobserved. The independent variables, [X.sub.jt], j = 1, [ldots], J, are those that vary over time but not across members. Among these are forecast values of macroeconomic variables of concern to the fed (e.g., inflation, unemployment, and economic growth). The remaining independent variables, [D.sub.kit], k = 1, [ldots], K, vary across both members and time and can include dummy variables indicating specific individuals who served on the FOMC during the chosen sample period. [7]

We specify a similar desired interest rate reaction function for the chairman (who is indicated by the position index 0):

[[r.sup.*].sub.0t] = [[delta].sub.0] + [[[sum].sup.M].sub.m=1] [[delta].sub.m] [C.sum.mt] + [[[sum].sup.J].sub.i=1] [[beta].sub.j][X.sub.jt] + [e.sub.0t], t = 1, [ldots], T. (2)

Equation 2 contains a series of dummy variables, [C.sub.mt], m = 1,[ldots], M, indicating specific chairmen (e.g., Bums, Miller, Volcker, and Greenspan). We assume that coefficients of the [X.sub.jt] are identical for chairmen and other members.

Error terms for the reaction functions in Equations 1 and 2 are assumed to be identically distributed normal random variables, which are uncorrelated over time but correlated across individuals at a given meeting:

E([e.sub.it])= 0, E([[e.sup.2].sub.it]) = [[sigma].sup.2], for i = 0, [ldots], N, t = 1, [ldots], T,

E([e.sub.it][e.sub.js]) = 0 for t [neq] S, E([e.sub.it][e.sub.js]) = [rho][[sigma].sup.2] for i [neq] j,

where [rho] measures the correlation between different individuals' contemporaneous reaction function error terms.

Monetary Policy Choices and FOMC Voting

Monetary policy directives are adopted by a majority vote of the FOMC at regularly scheduled meetings of the committee. The discussion of monetary policy usually begins with a report from the board staff, which covers economic forecasts and presents alternative policy options. Members then offer comments in the course of the policy go-around, which is followed by general discussion. The chairman, or a member designated by the chairman, then proposes wording for the policy directive. The actual directive may be Purposefully vague, but we assume that it embodies an implicit target for the federal funds rate. [8] The chairman plays a key role in the development of the directive in two ways. First, in shaping the wording of the directive, the chairman may be able to tilt policy in a direction he favors. Second, and perhaps more importantly, the chairman may influence the content and emphases of the staff report that is initially presented to the committee. [9] In view of these powers, we model the chairman as pr oposing a target for the federal funds rate on which the committee votes. [10] Our model reflects both the consensus-building and agenda-setting roles of the chairman by specifying that his proposed funds rate, [r.sub.t], be a weighted average of his own desired rate and the mean desired rate of all other members:

[r.sub.t] = [[[gamma]r.sup.*].sub.0t] + (1 - [gamma])[[[bar{r}].sup.*].sub.t], 1/12 [leq] [gamma] [leq] 1.0, (3)

where

[[[bar{r}].sup.*].sub.t] = [lgroup][frac{1}{N}][rgroup] [[[sum].sup.N].sub.i=1] [[r.sup.*].sub.it], (4)

and [gamma] is the weight attached to the chairman's desired interest rate. In practice, the chairman's proposal is put to a formal vote only when its adoption is assured. We therefore regard the proposed (and approved) interest rate, [r.sub.1] as the funds rate observed in the post-meeting period.

Once the chairman has proposed a federal funds rate target to the committee, a formal vote follows. Members can then dissent in favor of ease, dissent in favor of tightness, or assent. The discrete variable [V.sub.it], referring to the vote by member i in meeting t, is defined to equal -1, 1, or 0 in these three cases. [11] Although the ultimate outcome is generally not in doubt when the formal voting takes place, members do occasionally dissent as a matter of conscience. Because members are also under pressure to present a united external front, dissents apparently occur only when disagreements are acute. [12] Accordingly, we assume that a member dissents only when the difference between the proposed federal funds rate and his desired federal funds rate exceeds a threshold level [lambda] [greater than] 0:

if [r.sub.t] - [[r.sup.*].sub.it] [greater than] [lambda], then [V.sub.it] = -1 (i.e., dissent favoring ease); (5a)

if [r.sub.t] - [[r.sup.*].sub.it] [less than] -[lambda], then [V.sub.it] = 1 (i.e., dissent favoring tightness); (5b)

if -[lambda] [leq] [r.sub.t] - [[r.sup.*].sub.it] [leq] [lambda], then [V.sub.it] = 0 (i.e., assent). (5c)

Condition 5a says that if the proposed rate exceeds his desired rate by more than [lambda] units, then member i will dissent favoring ease. Similarly, Condition 5b says that if the proposed rate is less than his desired rate by more than [lambda] units, then member i will dissent favoring tightness. Finally, Condition 5c says that if the difference between desired and proposed rates is less than [lambda] in absolute value, then member i assents.

Estimation of the Model

Substituting Equations 1, 2, and 4 into Equation 3 yields a reduced form equation for the post-meeting federal funds rate:

[r.sub.t] = [gamma][[delta].sub.0] + (1 - [gamma])[[alpha].sub.0] + [gamma] [[[sum].sup.M].sub.m=1] [[delta].sub.m] [C.sub.mt] + [[[sum].sup.J].sub.j=1] [[beta].sub.j][X.sub.jt] + (1 - [gamma]) [[[sum].sup.K].sub.k=1] [[alpha].sub.k][[bar{D}].sub.kt] + [u.sub.t], (6)

where

[[bar{D}].sub.kt] = [lgroup][frac{1}{N}][rgroup] [[[sum].sup.N].sub.i=t] [D.sub.kit], [u.sub.t] = [[gamma]e.sub.0t] + (1 - [gamma])[[bar{e}].sub.t], and [[bar{e}].sub.t] = [lgroup][frac{1}{N}][rgroup] [[[sum].sup.N].sub.i=1] [e.sub.it].

This equation explains the post-meeting interest rate as a function of exogenous economic variables and variables describing the composition of the committee. Its reduced from coefficients can be estimated by ordinary least squares (OLS). Note that estimates of the structural coefficients of the economic variables in the reaction function (i.e., the [[beta].sub.j]S) are obtained directly from this regression.

Once again considering voting behavior, we substitute Equations 1 and 6 into Conditions 5a-5c to obtain

if [gamma]([[delta].sub.0] - [[alpha].sub.0]) + [gamma] [[[sum].sup.M].sub.m=1] [[delta].sub.m] [C.sub.mt] + (1 - [gamma]) [[[sum].sup.K].sub.k=1] [[alpha].sub.k] [[bar{D}].sub.kt] - [[[sum].sup.K].sub.k=1] [[alpha].sub.k][D.sub.kit] + [v.sub.it] [greater than] [lambda], then [V.sub.it] = -1; (7a)

if [gamma]([[delta].sub.0] - [[alpha].sub.0]) + [gamma] [[[sum].sup.M].sub.m=1] [[delta].sub.m] [C.sub.mt] + (1 - [gamma]) [[[sum].sup.K].sub.k=1] [[alpha].sub.k] [[bar{D}].sub.kt] - [[[sum].sup.K].sub.k=1] [[alpha].sub.k][D.sub.kit] + [v.sub.it] [less than] -[lambda], then [V.sub.it] = 1; (7b)

if -[lambda] [leq] [gamma]([[delta].sub.0] - [[alpha].sub.0]) + [gamma] [[[sum].sup.M].sub.m=1] [[delta].sub.m] [C.sub.mt] + (1 - [gamma]) [[[sum].sup.K].sub.k=1] [[alpha].sub.k][[bar{D}].sub.kt] - [[[sum].sup.K].sub.k=1] [[alpha].sub.k][D.sub.kit] + [v.sub.it] [leq] [lambda], then [V.sub.it] = 0, (7c)

where [v.sub.it] = [[gamma]e.sub.0t] + (1 - [gamma])[[bar{e}].sub.t] - [e.sub.it]. Conditions 7a-7c characterize a reduced from ordered probit model, which can be estimated using voting data pooled over members and time. [13] The model explains members' voting decisions as a function of variables indicating the identity or characteristics of a particular voting members (the [D.sub.kit]), as well as variables describing the composition of the full committee (the [[bar{D}].sub.kt]).

Profit models generally require a normalization in order to identify the scale of a latent propensity, which determines values of an observable discrete variable. This is usually accomplished by setting the variance of the probit model error term equal to 1.0. In our model, the latent propensity is measured in interest rate units, and its scale can be identified via the following cross-equation restrictions implied by the model:

[[[sigma].sup.2].sub.u] = [[gamma].sup.2][[sigma].sup.2] + [(1 - [gamma]).sup.2] [[frac{[[sigma].sup.2]}{N}] + [frac{N - 1}{N}[[rho][sigma].sup.2]] + 2[gamma](1 - [gamma])[[rho][sigma].sup.2] and (8a)

[[[sigma].sup.2].sub.v] = [[gamma].sup.2][[sigma].sup.2] + [(1 - [gamma]).sup.2] [[frac{[[sigma].sup.2]}{N}] + [frac{N - 1}{N}[[rho][sigma].sup.2]] + [[sigma].sup.2] - [2[rho][gamma].sup.2][[sigma].sup.2] - 2(1 - [gamma]) [[frac{[[sigma].sup.2]}{N}] + [frac{N - 1}{N}][[rho][sigma].sup.2]]. (8b)

Equation 8a can be solved for [[sigma].sup.2] and the resulting expression substituted into Equation 8b so that [[[sigma].sup.2].sub.v] is expressed as a function of [[[sigma].sup.2].sub.u], N, [rho], and [lambda]. An estimate of [[[sigma].sup.2].sub.u] can be obtained from the OLS regression estimating Equation 6; N is observable; and, as we discuss later, values for [rho] and [gamma] can be obtained from external sources. Consequently, [[[sigma].sup.2].sub.v] is identified and no normalization is required. In our empirical work, we proceed by estimating Equation 6 by OLS and the ordered probit model (Condition 7) by the maximum likelihood method. By imposing cross-equation restrictions, we are able to identify all of the model's parameters. [14]

One econometric complication should be noted: The reduced form probit model error terms, the [v.sub.it], are correlated across members at time t. Consequently, the statistical properties of the reported estimates are uncertain. To investigate these properties, we have evaluated our technique with a Monte Carlo study. Our coefficient estimates appear to be consistent and reported standard errors are very close to true standard errors calculated in the Monte Carlo experiments. [15]

3. Data Considerations

The variables [D.sub.kit] in the ordered probit model (Condition 7) include dummies for 84 individual members of the FOMC who voted five or more times on the monetary policy directive during the 1966-1996 period. [16] Remaining members are represented in the intercept. The variables [C.sub.mt] indicate the identity of the current chairman, and dummies for Arthur Burns, G. William Miller, Paul Volcker, and Alan Greenspan are included, with William McChesney Martin captured by the intercept.

Independent variables included among the [X.sub.jt] in the interest rate Equation 6 are comparable to those used in Chappell, Havrilesky, and McGregor (1993). The lagged federal funds rate, [r.sub.t-1], defined as the average federal funds rate between meetings t -- 1 and t, is included to account for inertia in the movement of interest rates. Lagged money growth, [M.sub.t-1], defined as the growth rate of M1 over the two months preceding the meeting, is included because of its presumed link to future inflation and because it has often been an explicit intermediate target. [17] Because the Federal Reserve placed greater emphasis on its money growth targets during the period from October 1979 through September 1982, we define a dummy variable for this period and include in the reaction function an interaction of this dummy variable with money growth. [18]

The remaining variables included among the [X.sub.jt] are two-quarter-ahead forecasts of macroeconomic goal variables thought to influence monetary policy: [[dot{P}].sub.t], the annualized percentage inflation rate calculated from the implicit price deflator for real Gross National Product (GNP); [U.sub.t], the civilian unemployment rate; and [Y.sub.t], the annualized percentage growth rate of real GNP. [19] These forecasts for 1966-1993 are taken from the projections presented to the FOMC in the Greenbook prepared for each committee meeting by the Federal Reserve Board staff. Since the FOMC's Greenbooks are available only with a five-year lag, these projections were supplemented for the 1994-1996 period with projections of the same variables from the Fairmodel maintained by Ray C. Fair of Yale University. The actual post-meeting interest rate, [r.sub.t], is measured as the average federal funds rate between meetings t and t + 1.

We recognize that the desired policies of particular individuals are likely to respond to economic conditions in different ways. We also recognize that individuals' policy preferences might change over time. Unfortunately, limitations of both our data and our modeling methods prohibit us from being very general in permitting coefficients to vary over time and members. We can, however, investigate two generalizations of possible importance. First, we investigate a possible structural break. Second, we explore differences between district bank presidents and governors.

Following the FOMC meeting of October 6, 1979, the Federal Reserve announced a change in operating procedures that was intended to give the committee better control over money growth in its efforts to slow the rate of inflation. This change reflected a shift in focus--indeed, one that still prevails today--toward stabilizing inflation and away from stabilizing unemployment. We attempt to capture this shift in focus by creating a dummy variable for the period following October 1979 (POST79) and augmenting the reaction function with interactions of this dummy variable with each of the economic variables in the specification. This generalization allows us to account for a structural break associated with the change in procedure and focus.

Previous studies have also noted differences in the desired policy preferences of governors and bank presidents. On average, bank presidents appear to favor a more anti-inflationary policy stance. [20] Although this difference in behavior might be adequately captured by individual intercepts, it is also possible that governors and bank presidents respond differently to individual economic variables. To permit this possibility, we include in the model interactions between a dummy variable for district reserve bank presidents and each of the economic variables. These interactions are included among the [D.sub.kit] (consisting of variables that are permitted to vary over members and time).

The data set used in the empirical analysis includes the previously described variables and individual voting records linked to 319 regular meetings of the FOMC over the 1966-1996 period. [21] The votes of the chairman are excluded, as are observations associated with vacancies, absences, and uncodable dissents. The final sample contains 3339 voting observations; among these, there are 85 dissents for ease (2.5%), 177 dissents for tightness (5.3%), and 3077 assents.

4. Empirical Results

Table 1 provides estimates of the empirical model, including OLS estimates of Equation 6 and ordered probit estimates of Conditions 7a-7c. Identification of the scale of the ordered probit model requires values of [rho] and [gamma]. We have estimated the model for alternative values of both parameters. To conserve space, we report results for [rho] = 0.70 and [gamma] = 0.50. We base our choice of these values on results from Chappell, McGregor, and Vermilyea (1998), who estimate that during the l970s, Arthur Burns' voting weight in the monetary policy decision process was about 0.50. The average pairwise correlation between the residuals from individual reaction functions estimated by Chappell, McGregor, and Vermilyea (1998) was about 0.70. [22] Results for other values of [rho] and [gamma] are qualitatively similar.

The results for the economic variables offer support for the view that the fed attempts to stabilize the business cycle. The coefficient on the growth rate of real GNP is positively signed and statistically significant, suggesting that the fed tightened when output growth accelerated and eased when it slowed. Similarly, the coefficient on the unemployment rate is negatively signed and statistically significant, suggesting that the fed tightened when unemployment decreased and eased when it increased. The results for the interaction terms suggest that the fed attached significantly more weight to output growth and to inflation following the October 1979 change in operating procedures. The inflation result, in particular, is consistent with our expectations. The lagged federal funds rate is also highly significant, reflecting substantial inertia in the path of the funds rate; however, the significant negative coefficient on the interaction term involving the funds rate suggests that the fed has placed less emp hasis on interest rate stability since October 1979. Lagged money growth is significant and positive for bank presidents but is significant for governors only during the October 1979-September 1982 period when the fed was placing greater emphasis on its money growth targets.

The constant term reported in Table 1 is the composite intercept appearing in Conditions 7a-7c, but the remaining parameter estimates for Conditions 7a-7c are of the structural parameters [lambda] (the dissent threshold parameter) and the [[alpha].sub.k]S (individual-specific intercepts). [23] The point estimate of [lambda] appears to be large, but this is not surprising given the paucity of dissenting votes in the sample. Estimates of the individual-specific intercepts reveal considerable variation in preferences across committee members.

For comparisons within the group of bank presidents or within the group of governors, these intercept shifts are directly indicative of preferences for ease or tightness (larger intercepts implying a desire for higher interest rates). For example, among the governors, Henry Wallich's reaction function indicates a preference for an interest rate higher than that of Martha Seger by 2.73 interest rate points.

Across the two groups, intercepts are not directly comparable, because we have specified that the two groups respond differently to changes in economic conditions. To allow comparisons across the two groups, we use the estimates in Table 1 and mean values of the economic variables to compute average desired interest rates for each individual who voted five or more times during the 1966-1996 period. Based on these average desired interest rates, Table 2 ranks committee members from easiest to tightest. For comparison, each member's net ease dissent frequency and rank by this measure are also provided. [24]

In Table 2, the ease tendencies of most governors are apparent, as are the tightness tendencies of most reserve bank presidents. Rankings by average desired interest rates are similar to rankings by net ease dissent frequencies. The correlation coefficient between members' average desired interest rates and dissent frequencies is -0.8705 (the negative correlation reflecting the fact that low values of the individual average desired interest rates are associated with high net ease dissent frequencies). By either measure, Governors Martha Seger, Preston Martin, and Nancy Teeters are the most ease-oriented FOMC members, while Reserve Bank Presidents Mark Willes, Darryl Francis, William Ford, Lee Hoskins, and Alfred Broaddus are among the most tightness-oriented members. There are, however, other comparisons that illustrate advantages of our ranking method over rankings based on dissent voting frequencies.

Consider Reserve Bank Presidents David Eastburn of Philadelphia and Karen Horn of Cleveland. Eastburn dissented for ease more often than Horn; however, Horn's ease votes came at a time in the 1980s when policy was already expansionary, whereas Eastburn's came in the mid-1970s, when policy was more restrictive. Accordingly, our method ranks Horn as more ease-oriented than Eastburn.

Even more informative are the cases of Governors Laurence Meyer, Alice Rivlin, and William Sherrill. None of these governors dissented during the 1966-1996 period, and so they tie for 17th in the ranking by net ease dissent frequency. Meyer and Rivlin, however, are judged as ease-oriented because they never dissented for tightness during a period in the 1990s when policy was relatively expansionary. By contrast, Sherrill did not dissent for ease during periods in the 1970s when policy was relatively restrictive; as a consequence, he is judged as tightness-oriented.

Our method also permits statistical tests of reaction function differences between individual FOMC members within the governors group and within the bank presidents group. Wald tests of equal coefficients are appropriate for this purpose. Although standard errors for most individual intercept estimates are large, a number of significant (at the 0.10 level) differences are revealed. Consider first the members of the board of governors. For example, Martha Seger prefers significantly lower interest rates than 30 other governors, while Henry Wallich prefers significantly higher interest rates than 28 others. Governor Wayne Angell, whom our method ranks more toward the middle of the ease-to-tightness spectrum, prefers significantly higher interest rates than 18 others, but significantly lower interest rates than 2 others.

Comparisons among Reserve Bank presidents are especially revealing. Table 2 shows that most of the presidents prefer a more conservative policy; however, Cleveland fed President Horn prefers significantly more stimulative monetary policy than all other Reserve Bank presidents, except Edward Boehne, Eastburn, Anthony Solomon, Jerry Jordan, and Robert McTeer. Further, we can use our method to investigate differences between the presidents of any particular Federal Reserve Bank. For example, Chicago Fed President Robert Mayo differs significantly from fellow Chicago Fed Presidents Charles Scanlon and Silas Keehn, with Mayo preferring lower interest rates than Scanlon and higher interest rates than Keehn. Significant differences also occur between Minneapolis Fed Presidents Willes and Hugh Galusha, Bruce MacLaury, and Gary Stern, with Willes favoring higher interest rates than Galusha, MacLaury, and Stern.

Finally, we note one particularly interesting feature of our results. During the 1966-1996 period, seven women--Governors Seger, Teeters, Rivlin, Susan Phillips, and Janet Yellen and Reserve Bank Presidents Horn and Cathy Minehan--have served on the FOMC. Our method ranks six of these women among the thirteen most ease-oriented members. In the past, concerns have been raised about the underrepresentation of women on the board of governors and as presidents of district Federal Reserve Banks. Given the small number of women in the sample, any conclusions are, at best, speculative. Nevertheless, our results suggest that greater representation of women in the fed's monetary policy decision process could have an important effect on policy outcomes.

The empirical results presented in this section illustrate the value of looking at the behavior of individual FOMC members. They also suggest that the individual composition of the committee should not be ignored when studying the fed's monetary policy decisions. In fact, our results allow us to conduct a test of the hypothesis that changes in the composition of the FOMC may have an important influence on monetary policy outcomes. The term [[[sum].sup.k].sub.k] = 1 [[alpha].sub.k] [[bar{D}].sub.kt] in Equation 6 captures effects on the interest rate related to changes in the composition of the committee. In our estimation, we used estimates from Conditions 7a-7c to construct [[[sum].sup.k].sub.k] = 1 [[alpha].sub.k] [[bar{D}].sub.kt] which was then included in the OLS estimation of Equation 6. A positive coefficient on this term would indicate that the composition of the committee does matter in interest rate determination.

The results, unfortunately, fail to support our hypothesis; the coefficient on [[[sum].sup.k].sub.k] = 1 [[alpha].sub.k] [[bar{D}].sub.kt] is correctly signed but is not significantly different from zero. This result at first appears to be at odds with evidence presented by Chappell, McGregor, and Vermilyea (1998), who find strong links between committee preferences and resulting monetary policy outcomes during the Burns era. The negative result here probably reflects the limitations of [[[sum].sup.k].sub.k] = 1 [[alpha].sub.k] [[bar{D}].sub.kt] as an indicator of individuals' preferences at a particular moment in time. By measuring differences across members with a small number of non-varying parameters, we may capture members' broad tendencies, but we cannot account for idiosyncratic movements in individual preferences over time. If these idiosyncratic movements are large, the poor performance of [[[sum].sub.k].sub.k] = 1 [[alpha].sub.k] [[bar{D}].sub.kt] in the reaction function is both understandab le and compatible with the proposition that members' preferences might nevertheless be important in determining policy outcomes. Further research should aim at obtaining better data on members' preferences and summarizing those preferences more completely.

5. Conclusions

Using a method that permits the estimation of monetary policy reaction functions for individuals, we have documented the considerable diversity in policy preferences of FOMC members over the 1966-1996 period. As with other studies, our evidence suggests that there are systematic tendencies that distinguish governors from bank presidents; however, we also find exceptions within each of these groups. Accurately describing these differences across individuals is a first step in the process of carefully describing the forces underlying policy choices made by the FOMC. Eventually, such descriptions might be used to ensure accountability of policymakers and to promote the adoption of institutions that will produce desirable policy choices.

(*.) Department of Economics, University of South Carolina, Columbia, SC 29208, USA; E-mail chappell@darla. badm.sc.edu.

(+.) Department of Economics, University of North Carolina at Charlotte, Charlotte, NC 28223, USA; E-mail [e-mail]; corresponding author.

We acknowledge the able research assistance of Matthew Birmingham, Ron Gill, Susan Harden, Mike Nelson, and Paul Prochaska. Janice Boucher, Charles Evans, John Gildea, Siu-Ki Leung, Raymond Lombra, Pedro Portugal, and seminar participants at Duke University and the University of South Carolina have provided helpful comments. This paper has also benefited from earlier work undertaken jointly with the late Thomas M. Havrilesky. Financial support was provided by National Science Foundation grants SES-9122322, SES-9121941, SBR-9422850, and SBR-9423095.

Received August 1998; accepted August 1999.

(1.) On the latter point, the president of the New York Federal Reserve Bank is an exception. The New York president is a permanent member of the committee.

(2.) See, for example, Puckett (1984), Woolley (1984), Laney (1990), Havrilesky and Gildea (1991a, b, 1992), Chappell, Havrilesky, and McGregor (1993), and McGregor (1996). In a series of related theoretical analyses, Waller (1989, 1992a, b) has investigated the role of partisan appointments in the policymaking procedures of central banks.

(3.) Chappell, Havrilesky, and McGregor (1997) provide a content analysis of the Memoranda of Discussion for the February 1970-March 1976 period. Chappell, McGregor, and Vermilyea (1998) use data from the Memoranda and transcripts to analyze monetary policy decision making during Arthur Burns' tenure as fed chairman (February 1970-February 1978).

(4.) The discussion draws from our earlier paper, Chappell, Havrilesky, and McGregor (1993).

(5.) If there are no absences or vacancies, N = 11.

(6.) Most empirical studies show that, at a minimum, the fed has attempted to influence short-term interest rates over most of the period studied in this paper. In reaction function contexts, Beck (1982) and Fair (1984) provide confirming evidence. Cook and Hahn (1989), by examining interest rate reactions to changes in fed target rates, show that the fed influences not only the federal funds rate, but also longer term interest rates. Bernanke and Blinder (1992) provide evidence that movements in the funds rate are primarily a consequence of fed policies rather than money demand shifts, and they also argue that the funds rate is the best predictor of future movements of real macroeconomic variables.

(7.) Over an extended sample period, many specific individuals will occupy the N positions on the committee. In our notational scheme, member i refers to the position number of a member, that is, i [epsilon] {1, [ldots], N}, not the identity of a particular individual. Particular individuals' intercept shifters will be included in the [D.sub.kit] and referenced by the subscript k.

(8.) Lombra and Moran (1980) and Karamouzis and Lombra (1989) provide detailed descriptions of the policymaking process within the Federal Open Market Committee (FOMC).

(9.) Woolley (1984) has suggested that the chairman may also gain some leverage over the committee through his functions as a liaison between the fed and the outside world and as an internal allocator of fed resources among individual governors.

(10.) The chairman has a formal vote, but because none has ever dissented, we regard his voting as superfluous.

(11.) "The "Record of Policy Actions" for FOMC meetings briefly describes members' reasons for dissenting Votes. In all but a few cases, these explanations can be coded to indicate dissents favoring ease or tightness. For example, at the FOMC meeting held on February 10.-11, 1987, St. Louis Fed President Thomas Melzer dissented because be "favored some tightening of reserve conditions.[ldots] Finally, looking ahead, be pointed out the potential for a further rise in inflationary expectations, and accordingly, he believed that prompt action toward restraint might avert the need for more substantial tightening later" ["Record of Policy Actions of the Federal Open Market Committee," Federal Reserve Bulletin 73 (June 1987), P. 451]. Melzer is accordingly coded as having dissented for tightness. Similarly, at the meeting held on May 19, 1987, Governor Martha Seger dissented "because she did not want to lean on the side of any tightening of reserve conditions beyond the firming that had occurred since the March mee ting. She was concerned that the degree of reserve pressure prevailing recently [ldots] represented a risk to an already weak economic expansion" ["Record of Policy Actions of the Federal Open Market Committee," Federal Reserve Bulletin 73 (September 1987), p. 716]. Seger is accordingly coded as having dissented for ease.

(12.) Anecdotal evidence concerning motives for dissent and pressures for consensus are presented in Havrilesky and Schweitzer (1990) and Chappell, Havrilesky, and McGregor (1993). The former study also develops a net utility maximizing model of FOMC dissent voting.

(13.) McKelvey and Zavoina (1975) describe the ordered probit model and provide an application to roll-call voting in the U.S. Congress.

(14.) Specifically, we first obtain estimates of [lambda] and [[alpha].sub.k], k = 1, [ldots], K, from the ordered probit model (Condition 7), employing an arbitrary value for [[[sigma].sup.2].sub.u] (hence, an arbitrary value for [[[sigma].sup.2].sub.v]) to determine the scale of the underlying latent variable. We then incorporate estimates of the [[alpha].sub.k]S into an ordinary least squares (OLS) estimation of Equation 6, obtaining estimates of the [[beta].sub.j]S. This estimation also provides an estimate of the true [[[sigma].sup.2]sub.u], which we then use to correctly recalibrate the scale of the estimated [[sigma].sub.k]S in Conditions 7a-7c and (1 - [gamma]) in Equation 6.

(15.) On average, true Monte Carlo standard errors exceed our reported standard errors by 0.43%. The Monte Carlo study is based on an analysis of estimations using 30 sets of data created by random simulations of a true ordered probit model incorporating error correlations across members within meetings. We employed large data sets (100 meetings, 10 voting members, hence 1000 voting observations), so results characterize asymptotic properties of the estimator.

(16.) The choice of five votes as the criterion for including a dummy variable for a specific individual is somewhat arbitrary; however, it is difficult to obtain reliable estimates for individuals who do not vote often. The criterion of at least five votes seems to offer a reasonably inclusive and tractable solution.

(17.) Because the FOMC placed greater emphasis on M2 and less on Ml after October 1982, we have also estimated Equation 6, replacing M1 growth with M2 growth. We find that M1 growth yields a better fit than M2 growth and therefore report results only for the former.

(18.) This procedure is consistent with that of Fair (1984), who argues that the fed's behavior can be modeled using an interest rate reaction function, provided that proper account is taken of the significantly greater importance attached to lagged money growth after October 1979.

(19.) Our forecasts are generally calculated as an average of forecasts for the current quarter and the upcoming quarter. For several meetings where only the current quarter forecast is available, that measure is employed.

(20.) See Puckett (1984), Woolley (1984), Laney (1990), Havrilesky and Gildea (1991a), Chappell, Havrilesky, and McGregor (1993), and McGregor (1996). In contrast, Tootell (1991) questions the existence of a significant difference.

(21.) Our sample does not include telephone meetings that have been held between regularly scheduled FOMC meetings. From time to time, a directive was modified, but votes on new directives were not taken during telephone meetings.

(22.) Chappell, McGregor, and Vermilyea (1998) analyze the monetary policy decision process during the Burns era using data on individual FOMC members' desired federal funds rates. These data were extracted from the Memoranda of Discussion and transcripts of FOMC meetings held during the Burns era.

(23.) The intraequation restrictions relating coefficients of the [[bar{D}].sub.kt] and the [D.sub.kit] variables in Conditions 7a-7c are imposed in the estimation.

(24.) The net ease dissent voting frequency is defined as dissents for ease less dissents for tightness, divided by total votes. A positive net ease dissent frequency indicates that the member dissented for ease more often than for tightness, while a negative dissent frequency means that the member dissented for tightness more often than for ease.

References

Beck, Nathaniel. 1982. Presidential influence on the Federal Reserve in the 1970's. American Journal of Political Science 26:415-45.

Bernanke, Ben, and Alan Blinder. 1992. The federal funds rate and the channels of monetary transmission. American Economic Review 82:901-21.

Chappell, Jr., Henry W., Thomas M. Havrilesky, and Rob Roy McGregor. 1993. Partisan monetary policies: Presidential influence through the power of appointment. Quarterly Journal of Economics 108:185-218.

Chappell, Jr., Henry W., Thomas M. Havrilesky, and Rob Roy McGregor. 1997. Monetary policy preferences of individual FOMC members: A content analysis of the Memoranda of Discussion. Review of Economics and Statistics 79:454-60.

Chappell, Jr., Henry W., Rob Roy McGregor, and Todd Vermilyea. 1998. Models of monetary policy decision-making: Arthur Burns and the Federal Open Market Committee. Unpublished paper, University of South Carolina.

Cook, Timothy, and Thomas Hahn. 1989. The effect of changes in the federal funds rate target on market interest rates in the 1970s. Journal of Monetary Economics 24:331-51.

Fair, Ray C. 1984. Specification, Estimation, and Analysis of Macroeconomic Models. Cambridge, MA: Harvard University Press.

Havrilesky, Thomas M., and John Gildea. 1991a. The policy preferences of FOMC members as revealed by dissenting votes. Journal of Money, Credit, and Banking 23:130-8.

Havrilesky, Thomas M., and John Gildea. 199 lb. Screening FOMC members for their biases and dependability. Economics and Politics 3:139-50.

Havrilesky, Thomas M., and John Gildea. 1992. Reliable and unreliable partisan appointees to the board of governors. Public Choice 73:397-417.

Havrilesky. Thomas M., and Robert Schweitzer. 1990. A theory of FOMC dissent voting with evidence from the time series. In The Political Economy of American Monetary Policy, edited by Thomas Mayer. New York: Cambridge University Press, pp. 197-210.

Karamouzis, Nicholas, and Raymond Lombra. 1989. Federal reserve policymaking: An overview and analysis of the policy process. In Carnegie-Rochester Conference Series on Public Policy 30, edited by Karl Brunner and Allan H. Meltzer. Amsterdam: North-Holland, pp. 7-62.

Laney, Leroy O. 1990. The FOMC dissent voting record: Reserve bank presidents versus the governors. Unpublished paper, First Hawaiian Bank.

Lombra, Raymond, and Michael Moran. 1980. Policy advice and policymaking at the Federal Reserve. In Monetary Institutions and the Policy Process (Carnegie-Rochester Conference Series on Public Policy 13), edited by Karl Brunner and Allan H. Meltzer. Amsterdam: North-Holland, pp. 9-68.

McGregor, Rob Roy. 1996. FOMC voting behavior and electoral cycles: Partisan ideology and partisan loyalty. Economics and Politics 8:17-32.

McKelvey, Richard, and William Zavoina. 1975. A statistical model for the analysis of ordinal level dependent variables. Journal of Mathematical Sociology 4:103-20.

Puckett, Richard. 1984. Federal Open Market Committee structure and decisions. Journal of Monetary Economics 14:97-104.

Tootell, Geoffrey M.B. 1991. Are district presidents more conservative than board governors? New England Economic Review September/October:3-12.

Waller, Christopher J. 1989. Monetary policy games and central bank politics. Journal of Money, Credit, and Banking 21:422-31.

Waller, Christopher J. 1992a. A bargaining model of partisan appointments to the central bank. Journal of Monetary Economics 29:411-28.

Waller, Christopher J. 1992b. The choice of a conservative central banker in a multi-sector economy. American Economic Review 82:1006-12.

Wessel, David. 1992. Money talks: Fed officials, meeting next week, are facing policy confrontation. The Wall Street Journal, 26 June, pp. A1 and A7.

Woolley, John. 1984. Monetary Politics. New York: Cambridge University Press.

    Monetary Policy Reaction Functions with Individual-Specific Effects
Variable                                Coefficient  Standard Error
Ordinary Least Squares Estimates of (6)
[r.sub.t-1]                               0.9994 [*]     0.0381
[[dot{M}].sub.t-1]                        0.0142         0.0102
[hat{P}]                                  0.0435         0.0539
[hat{Y}]                                  0.0632 [*]     0.0198
[hat{U}]                                 -0.1043 [*]     0.0588
[r.sub.t-1] [cdotp] POST79               -0.1500 [*]     0.0457
[[dot{M}].sub.t-1] [cdotp] POST79        -0.0154         0.0131
[hat{P}] [cdotp] POST79                   0.1464 [*]     0.0772
[hat{Y}] [cdotp] POST79                   0.1522 [*]     0.0371
[hat{U}] [cdotp] POST79                  -0.0587         0.0647
[[dot{M}].sub.t-1] [cdotp] D7982          0.1230 [*]     0.0138
[[[sum].sup.K].sub.k=1]
[[alpha].sub.k][[bar{D}].sub.kt]          0.1100         0.2944
Ordered Probit Estimates of Bank
President Interactions
[r.sub.t-1] [cdotp] BANKPRES             -0.0290         0.0282
[[dot{M}].sub.t-1] [cdotp] BANKPRES       0.0274 [*]     0.0101
[hat{P}] [cdotp] BANKPRES                 0.0516         0.0748
[hat{Y}] [cdotp] BANKPRES                 0.0026         0.0253
[hat{U}] [cdotp] BANKPRES                -0.0038         0.0529
Ordered Probit Estimates
of Individual Intercepts
INTERCEPT [[gamma]([[delta].sub.0]
- [[alpha].sub.0])]                      -0.0332         0.2655
BURNS                                     0.1915         0.1914
MILLER, G.W.                             -0.7910 [*]     0.3596
VOLCKER                                  -0.7072 [*]     0.2714
GREENSPAN                                -0.7036 [*]     0.2959
ANGELL                                    0.1601         0.5699
BALLES                                   -0.2818         0.5899
BAUGHMAN                                  0.2408         0.5623
BLACK                                     0.6140         0.4775
BLINDER                                  -0.5882         0.5068
BOEHNE                                   -0.9781 [*]     0.5364
BOPP                                      0.2516         0.4803
BOYKIN                                    0.3690         0.5314
BRIMMER                                   0.2939         0.5304
BROADDUS                                  1.0795 [*]     0.6238
BUCHER                                   -0.6657         0.5738
CLAY                                      0.4974         0.5198
COLDWELL                                  0.3834         0.5385
CORRIGAN                                 -0.6741         0.4725
DAANE                                     0.2972         0.5423
DEBS                                      0.1950         0.5367
EASTBURN                                 -0.8426         0.5563
ELLIS                                     0.0271         0.4738
FORD                                      1.4465 [*]     0.6545
FORRESTAL  -0.6396     0.4788
FRANCIS     1.3843     0.5500
GALUSHA    -0.1908     0.4827
GARDNER     0.0260     0.4905
GRAMLEY    -0.0367     0.5925
GUFFEY     -0.2305     0.5672
HAYES       1.0604 [*] 0.4855
HEFLIN     -0.0542     0.4941
HELLER     -0.4603     0.4932
HICKMAN     0.6248     0.5400
HOENIG     -0.0414     0.6251
HOLLAND    -0.1021     0.5995
HORN       -1.4588 [*] 0.5997
HOSKINS     1.2812 [*] 0.5414
IRONS       1.0669 [*] 0.6132
JACKSON     0.2951     0.5615
JOHNSON    -0.8934     0.5775
JORDAN     -1.0142     0.7936
KEEHN      -0.6101     0.4804
KELLEY     -0.5849     0.5840
KIMBREL     0.2419     0.4981
LAWARE      0.1906     0.5065
LILLY      -0.4582     0.6199
LINDSEY    -0.6621     0.6559
MACLAURY    0.4269     0.5718
MAISEL     -0.5552     0.5431
MARTIN, P. -1.7662 [*] 0.5939
MAYO       -0.1640     0.4772
MCDONOUGH  -0.6454     0.4855
MCTEER     -0.8403 [*] 0.5060
MELZER      0.4313     0.5596
MEYER      -1.1267 [*] 0.5478
MINEHAN    -0.3884     0.5043
MITCHELL   -0.2606     0.5388
MORRIS     -0.3216     0.5791
MOSKOW     -0.3884     0.5043
MULLINS    -0.5737     0.5096
PARRY       0.2059     0.5409
PARTEE     -0.9539 [*] 0.5310
PATTERSON  -0.0017     0.4688
PHILLIPS   -0.6912     0.5059
RICE       -1.2877 [*] 0.5700
RIVLIN     -1.1267 [*] 0.5478
ROBERTSON   0.1694     0.5307
ROBERTS    -0.3498     0.5682
ROOS        0.3480     0.6695
SCANLON     0.6615     0.5351
SCHULTZ    -0.6687     0.5507
SEGER      -2.0832 [*] 0.5380
SHEEHAN    -0.3868     0.5938
SHEPARDSON  0.9551     0.6182
SHERRILL    0.0446     0.5207
SOLOMON    -0.7367     0.6349
STERN       0.1088     0.5411
SWAN       -0.0188     0.4789
SYRON      -0.6715     0.5135
TEETERS     1.7509 [*] 0.5678
TIMLEN     -0.6814     0.4975
TREIBER     0.5451     0.6068
VOLCKER     0.3133     0.5263
WALLICH     0.6491     0.5361
WAYNE      -0.1822     0.4888
WILLES      1.4983 [*] 0.5978
WINN        0.3543 [*] 0.5035
YELLEN      0.7414 [*] 0.5099
[gamma]     1.9121 [*] 0.0490
(*.)Significant at the 0.10 level or better.
           Federal Reserve Governors and Bank Presidents Ranked by
        Average Desired Interest Rates and Dissent Voting Frequencies
                                   Rank by
                          Average  Average   Rank by  Net Ease
                          Desired  Desired  Net Ease   Dissent
                          Interest Interest  Dissent  Frequency
Member     Type             Rate     Rate   Frequency    (%)
SEGER      Governor         4.72       1        1       29.63
MARTIN, P. Governor         5.04       2        3       16.13
TEETERS    Governor         5.05       3        2       18.00
RICE       Governor         5.52       4        7        6.35
HORN       Bank president   5.54       5        8        5.26
MEYER      Governor         5.68       6       17        0.00
RIVLIN     Governor         5.68       6       17        0.00
PARTEE     Governor         5.85       8        8        5.26
JOHNSON    Governor         5.91       9       15        2.86
JORDAN     Bank president   5.98      10       12        4.35
BOEHNE     Bank president   6.02      11       16        2.13
YELLEN     Governor         6.06      12       17        0.00
PHILLIPS   Governor         6.11      13       17        0.00
SCHULTZ    Governor         6.14      14       17        0.00
BUCHER     Governor         6.14      14        4       10.00
LINDSEY    Governor         6.14      14       17        0.00
EASTBURN   Bank president   6.15      17        5        8.57
MCTEER     Bank president   6.16      18       17        0.00
BLINDER    Governor         6.22      19       17        0.00
KELLEY     Governor         6.22      19       17        0.00
MULLINS    Governor         6.23      21       17        0.00
MAISEL     Governor         6.25      22        6        6.74
SOLOMON    Bank president   6.26      23       17        0.00
TIMLEN     Bank vice        6.31      24       17        0.00
             president
CORRIGAN   Bank president   6.32      25       17        0.00
SYRON      Bank president   6.32      25       17        0.00
HELLER     Governor         6.34      27       17        0.00
LILLY      Governor         6.35      28       11        5.00
MCDONOUGH  Bank president   6.35      28       17        0.00
FORRESTAL  Bank president   6.36      30       17        0.00
KEEHN      Bank president   6.39      31       17        0.00
SHEEHAN    Governor         6.42      32       10        5.26
MITCHELL   Governor         6.54      33       13        3.20
MINEHAN    Bank president   6.61      34       17        0.00
MOSKOW     Bank president   6.61      34       17        0.00
ROBERTS    Bank president   6.65      36       17        0.00
MORRIS     Bank president   6.67      37       17        0.00
HOLLAND    Governor         6.70      38       14        2.94
BALLES     Bank president   6.71      39       53       -2.08
GUFFEY     Bank president   6.76      40       58       -4.35
GRAMLEY    Governor         6.77      41       66       -7.69
GALUSHA    Bank president   6.80      42       17        0.00
WAYNE      Bank president   6.81      43       17        0.00
GARDNER    Governor         6.83      44       17        0.00
MAYO       Bank president   6.83      44       17        0.00
SHERRILL   Governor         6.85      46       17        0.00
HEFLIN     Bank president 6.94 47 17   0.00
HOENIG     Bank president 6.95 48 62  -6.25
ANGELL     Governor       6.96 49 69  -9.38
ROBERTSON  Governor       6.97 50 50  -1.02
SWAN       Bank president 6.98 51 17   0.00
PATTERSON  Bank president 6.99 52 17   0.00
ELLIS      Bank president 7.02 53 17   0.00
BRIMMER    Governor       7.10 54 51  -1.79
JACKSON    Governor       7.10 54 54  -2.56
DAANE      Governor       7.10 54 52  -2.00
STERN      Bank president 7.10 54 69  -9.38
LAWARE     Governor       7.19 58 72 -11.32
DEBS       Bank vice      7.19 58 17   0.00
             president
PARRY      Bank president 7.20 60 67  -8.00
BAUGHMAN   Bank president 7.24 61 56  -4.17
KIMBREL    Bank president 7.24 61 59  -4.69
BOPP       Bank president 7.25 63 17   0.00
COLDWELL   Bank president 7.25 63 61  -5.68
             and governor
VOLCKER    Bank president 7.31 65 64  -6.67
ROOS       Bank president 7.34 66 74 -13.64
WINN       Bank president 7.35 67 68  -8.47
BOYKIN     Bank president 7.36 68 76 -15.63
MACLAURY   Bank president 7.42 69 56  -4.17
MELZER     Bank president 7.43 70 79 -22.58
WALLICH    Governor       7.45 71 78 -18.03
CLAY       Bank president 7.49 72 55  -3.64
TREIBER    Bank vice      7.54 73 60  -5.00
             president
BLACK      Bank president 7.61 74 77 -17.91
HICKMAN    Bank president 7.62 75 65  -7.32
SCANLON    Bank vice      7.66 76 62  -6.25
             president
SHEPARDSON Governor       7.76 77 71 -11.11
HAYES      Bank president 8.06 78 75 -15.53
IRONS      Bank president 8.06 78 73 -13.33
BROADDUS   Bank president 8.07 80 81 -37.50
HOSKINS    Bank president 8.28 81 82 -43.75
FRANCIS    Bank president 8.38 82 80 -28.57
FORD       Bank president 8.44 83 84 -62.50
WILLES     Bank president 8.49 84 83 -54.55