Attendance rates, political shirking, and the effect of post-elective office employment.

Lott, John R., Jr.

ATTENDANCE RATES, POLITICAL SHIRKING, AND THE EFFECT OF POST-ELECTIVE OFFICE EMPLOYMENT

I. INTRODUCTION

Political shirking--the degree to which a politican's actions deviate from the wishes of his constituents--can take many different dimensions. Economists (e.g., Amacher and Boyes [1978], Davis and Porter [1989], Dougan and Munger [1989], Kalt and Zupan [1984], Lott [1987d], Nelson and Siberberg [1987], and Peltzman [1984]) have debated whether shirking exists in how a politician votes when he does vote. This divergence is measured through some particular "special interest" voting ratings (e.g., American Conservative Union, Americans for Democratic Action, etc.). However, shirking can obviously take as many forms as there are outputs that a politician produces. Besides how he votes, he also produces services for his constituents involving everything from being personally available to listen to the constituents' grievances to ensuring effective handling of case work by his staff (see, for example, Fiorina [1977], Fenno [1978], and Cain et al. [1987]). (1) One dimension that has not been sufficiently examined is the extent to which politicians shirk by failing to vote on pending legislation. (2) No attention has been given to which mechanisms will make this particular type of shirking relatively more costly.

Most economists agree that opportunitic behavior by politicians is limited by the threat of reelection (e.g., Amacher and Boyes [1978], Barro [1973], Kalt and Zupan [1984], Nelson [1976], and Telser [1976]). (3) By implication, the level of shirking should be the greatest when a politician decides to leave office. However, this opportunistic behavior might be reduced to the extent that political parties or constituencies impose costs for shirking through affecting politicisns' post-elective office career or the careers of their children. Thus, a politicians' last term in elective office may not represent his "last period" in any meaningful sense.

The next section outlines the differing views on how politicians act when they no longer face the threat of reelection. Section III describes some empirical evidence on what mechanisms may prevent this last-period problem. The final section discusses the impliactions of these results.

II. THE LAST PERIOD

Whether economists view politicians as ideologues (e.g., Kau and Rubin [1979] and Kalt and Zupan [1984]) or as simply attempting to maximize political support (e.g., Barro [1973], Ferejohn [1986], Nelson [1976], and Peltzman [1984]), the threat of reelection is seen as preventing shirking. As Kalt and Zupan [1984, 298] phrase it, "it appears that the proximity of the next election inhibits ideological shirking."

Even those who do not view the last period as affecting how a politician votes argue that it will affect how often he votes. Lott [1987a; 1987b] has argued that if voters can determine and elect politicians that have ideological preferences corresponding exactly to the wishes of constituents, politicians' voting patterns should continue to reflect those values even when the cost of shirking, in terms of forgone future votes, is low. The penalty to ideologues of not votingfor their constituency in the last period when they no longer face reelection is the forgone utility from not doing what they personally value. Assuming that politicians also desire leisure, politicians should continue to vote for what they believe in, but they should vote less frequently since they no longer receive the additional return of larger future support.

Mechanisms to overcome this last-period problem have been suggested by Becker and Stigler [1974, 9-10]. They argue that opportunistic behavior can be controlled if officeholders face the loss of a pension (or, equivalently, a posted bond) in the event of opportunistic behavior in their last term. However, the actual pension system does not effectively eliminate the last period for politicians, since retirement payments are not based on cheating in the last period. Such a system may only increase the cost of cheating in earlier periods. However, since retirement annuities for Congress are calculated at only 2.5 percent of the average yearly salary for the last three years of employment times the number of years of service, even this effect is probably not very large. Another type of bond can involve the loss of valuable brand name. However, since politicians have only very limited means of selling their political brand name (Lott [1987a; 1987d]), politicians are different than firms in that the reduced value of their brand name in their last period can have only a very limited impact on their wealth.

Barro [1973] has suggested that political parties might offer future appointments to jobs as an inducement for good last-period performance. While intuitively quite believable, this proposition has never been tested. Eckert [1981] provided the interesting finding that 51 percent of all former Civil Aeronautics Board, Federal Communications Commission, and Interstate Commerce Commission commission members took private sector jobs in the related regulated industry after leaving their commission posts. However, he produced not direct evidence that the lure of private sector employment affected commission members' behavior while they were on the commission. (4) This paper tests to see how effective post-elective employment and other mechanisms are in reducing the return to one particular type of shirking through looking at how often a congressman votes when he no longer faces reelection.

III. ATTENDANCE RATES: THE EVIDENCE

When a legislator is in his last term in office, the cost to him of shirking in terms of forgone future political support is relatively low. One dimension that this shirking can take is that of not voting on pending legislation and thus not representing the preferences of his district. The lower the cost of shirking, the more he should shirk and consume leisure on the job. (5)

Retiring politicians are defined as those who neither run for reelection nor for any other political office in 1978. We assume that politicians know when they will retire. (6) Table I provides a first look at this problem by showing the mean percentage attendance rates and their standard deviations for both congressmen who retired at the end of the 95th Congress 1977-78) and for those who did not. The data were obtained from the Congressional Quarterly. Section 1 of the table reports that the voting rate for congressmen is lower in the term they retire, though it is neither statistically significantly nor below the rate for nonretiring politicians in the 95th Congress. Given that many congressmen retire due to illness and that this will also make it more costly for them to vote, section 2 of the table also gives the values for the two groups excluding those who are ill or have serious illnesses or deaths in their families. The means for both groups rise; not surprisingly, the mean for retiring politicians rises the most since they were most likely to have been ill, thus leaving an even smaller and less significant difference between the two groups.

Obviously, other determinants of congressional voting rates must be controlled for. To do this, an attempt was made to see whether the attendance rate for congressmen retiring in 1978 fell between the 95th and 94th Congresses (their last and second-to-last terms in elected office) relative to the change for nonretiring congressmen between the two periods. This was done by estimating the following regression across all congressmen who had been in office for both the 95th and 94th Congresses using ordinary least squares:

CHANGE IN ATTENDANCE RATES = [a.sub.0] + [a.sub.1] LAST TERM + [a.sub.2] CHILDREN'S OR OWN CAREERS + [a.sub.3] ILLNESS DURING 95th CONGRESS + [a.sub.4] ILLNESS DURING 94th CONGRESS + [a.sub.5] REPUBLICAN + [a.sub.6] TENURE + [a.sub.7] TENURE + [a.sub.7] AGE + [a.sub.8] ANOTHER OFFICE + [a.sub.9] CHANGE IN GENERAL ELECTION VOTE DIFFERENCES + [a.sub.10] CHANGE IN PRIMARY ELECTION VOTE DIFFERENCES + [a.sub.11] STATE.

The variables are defined as follows:

CHANGE IN ATTENDANCE RATES = the percentage point change in the total voting rates from the 94th to the 95th Congresses;

LAST TERM = dummy variable which equals one if the politician retires in 1978, zero otherwise;

CHILDREN'S OR OWN CAREERS = dummy variable that is one when a retired congressman engages in either lobbying for pay or takes a job with the government and/or when his offspring go into politics, work for the government, or lobby for clients dealing with the government, zero otherwise;

ILLNESS DURING ith CONGRESS = dummy variable that is one if the congressman was absent one day or more due to illness, zero otherwise, where i represents either the 95th or 94th Congresses;

REPUBLICAN = dummy variable that is one if the candidate is a Republican, zero otherwise;

TENURE = the congressman's tenure in office (in years);

AGE = the congressman's age (in years);

ANOTHER OFFICE = dummy variable that is one if the congressman is retiring from congress to run for another elective office, zero otherwise;

CHANGE IN GENERAL ELECTION VOTE DIFFERENCES = the change in percentage vote margins between the congressman and his challengers in the 1978 and 1976 general elections, which is zero if he did not run for reelection in 1978;

CHANGE IN PRIMARY ELECTION VOTE DIFFERENCES = the change in percentage vote margins between the congressman and his challengers in the 1978 and 1976 primary elections, which is zero if he did not run for reelection in 1978; and

STATE = a vector of state dummies excluding Wyoming.

If the last-period problem exists, the coefficient on LAST TERM should be negative. However, the fact that a retiring congressman is leaving office is an imperfect measure of whether he is really in his last period. For instance, constituencies may hire retiring politicians or their children, as liaisons to government bureaus, lobbyists, or consultants. Strong political parties may reward loyal service with appointments to government positions. If the salary that these constituencies or political parties are willing to pay politicians declines with the level of cheating, politicians will find it costly to cheat. It is assumed that side payments take the form of jobs and not direct pecuniary payments independent of any future work.

In an attempt to control for this type of payoff, twenty-five of the twenty-seven congressmen who retired at the end of the 95th Congress (see appendix) were interviewed. Since the remaining two congressmen (Olin Teague and James Burke) had died, their widows were interviewed. (7) If constituencies or political parties can impose costs upon cheating politicians after they leave office through future employment, the coefficient on CHILDREN'S OR OWN CAREERS should be positive. Politicians are assumed to have rational expectations with regards to these costs. The size of these costs will determine how much of this last-period problem can be mitigated.

Since the average voting are in the 95th Congress is below that for the 94th Congress, illness during the 95th Congress should make this negative difference larger, while illness during the 94th congress should make it smaller. Larger values of the CHANGE IN GENERAL ELECTION VOTE DIFFERENCES or CHANGE IN PRIMARY ELECTION VOTE DIFFERENCES indicate that the congressman faced easier elections in 1978 than in 1976, and should represent a lower opportunity cost of voting in terms of forgone campaign time. (8) Likewise, the coefficient on ANOTHER OFFICE should be negative if the returns to campaigning in areas in which the candidate has not run before are larger. Increased TENURE lowers the cost of not voting (i.e., shirking) if it is harder to defeat incumbents who have more political brand name (see Lott [1986; 1987a]). However, increased TENURE may also make voting more costly because of more and/or higher-quality committee assignments. A negative coefficient here implies either that the cost of shirking falls and/or the opportunity cost of forgoing other activities increases with tenure.

Finally, this paper involves no strong prior beliefs on how the opportunity cost of voting varies with AGE, nor on whether Republicans or Democrats tend to vote more frequently over time. (9) Other articles by Johnston [1978], Lott [1987a], and Welch [1974] find that the marginal cost to acquiring campaign funds varies across political parties, so possibly they face different opportunity cost of forgoing fund arising to go vote. Neither the removal of AGE and/or REPUBLICAN affects the following results. Data on these variables were obtained from The Almanac of American Politics.

Some of the change in a congressman's attendance rates may arise from how many of a year's issues are related to the economic interests (based on income, geography, etc.) of his constituents. For instance, a congressman's voters may not be interested in what part of another state is designated as a national park, and thus in years when a congressman's constituents are not directly affected by a relatively large number of votes, his attendance rate should be relatively low. I use the vector of state dummies as a proxy for these differences. The dummy variable for the fiftieth state (Wyoming) was excluded so that the intercept term represents that state.

The results for equation (1) are shown in column 1 of Table II. The results indicate that politicians do behave differently in their last period. The coefficient on LAST TERM reveals retiring politicians vote 7.8 percentage points less often than nonretiring politicians, and the coefficient is significant at the .004 level. Comparing two hypothetical Democratic politicians from California (where the state dummy is 4.6) who are sixty-five years old and have been in Congress for twenty years, the reduction in voting participation for a retiring politician equals -10.134 percentage points, which is 4.4 times greater than that for a nonretiring politician (-2.296) for whom the CHANGE IN GENERAL ELECTION VOTE DIFFERENCES equals zero and the CHANGE IN PRIMARY ELECTION VOTE DIFFERENCES equals zero. Using the total attendance levels shown in Table I, this result implies that while nonretiring congressmen are absent for one out of every ten votes, retiring congressmen are absent for one out of every five.

Interestingly, the coefficient for CHILDREN'S OR OWN CAREERS is of the predicted sign, but insignificant and not very large, representing only one-quarter of the reduction in the attendance rates that occurs in the last period. This result indicates that the future careers of a politician and/or that of his children do not seem to increase significantly the cost of shirking to retiring politicians.

All other coefficients are of the expected sign, with ILLNESS DURING 95th CONGRESS, ILLNESS DURING 94th CONGRESS, CHANGE IN PRIMARY ELECTION VOTE DIFFERENCES, and ANOTHER OFFICE being significant for two-tailed t-tests, and TENURE and AGE being significant for one-tailed t-tests. The results indicate that no significant difference exists between the change in voting rates of Republicans and Democrats. The coefficient on ANOTHER OFFICE implies that a politician reduces his voting rate by 15 percentage points if he attempts to run for another political office, because of the large returns to campaigning in areas in which the candidate has not run before. Even though the coefficients for illness in the 95th and 94th Congresses are quite significant, both coefficients are relatively small when compared with the drop in attendance associated with the last term of a congressman.

Columns 2, 3, and 4 in Table III attempt to reestimate equation (1) by replacing CHILDREN'S OR OWN CAREERS. Column 2 eliminates CHILDREN'S OR OWN CAREERS by controlling separately for the effects of a retiring politician's own career after holding elective office (OWN CAREER) and the careers of his children after he leaves office (CHILDREN'S CAREER). OWN CAREER equals one if the retired congressman either engages in lobbying for pay and/or takes a job with the government, zero otherwise. CHILDREN'S CAREER equals one if the retired congressman's children either attempt to run for public office, work for the government, and/or engage in lobbying, zero otherwise.

Column 3 replaces CHILDREN'S OR OWN CAREERS with a variable that multiplies the dummy variables CHILDREN'S CAREER and OWN CAREER together (CHILDREN'S*OWN CAREERS). This new variable equals one only when both the retiring politician and his offspring are involved in politics after the politician leaves elective office. If a shirking politician can be punished through both his own and his children's careers after he leaves office, the coefficient on this new variable should be both larger and more significant than the coefficient on CHILDREN'S OR OWN CAREERS or the coefficients on either of the two dummies separately since it only examines those politicians who face the greatest cost for cheating in terms of forgone future employment.

Column 4 replaces CHILDREN'S OR OWN CAREERS with a dummy variable (LAST TERM*AGE 65) that equals one if a congressman is retiring in 1978 and is less than sixty-five years of age, zero otherwise. If the side payments to retiring congressmen take the form of jobs and not direct pecuniary payments, it seems that the effectiveness of such payments will be reduced for older politicians whose remaining careers may be short or nonexistent. Alternatively, this variable could simply measure how long a retired, cheating politician must endure the distain of his former constituencies. If younger politicians expect to face such costs for a longer period of time, since they can expect to remain alive longer, younger politicians will find it more costly to shirk. Both explanations predict that the coefficient on LAST TERM*AGE 65 should be positive.

The results for column 2 are almost identical to the results for column 1. All of the coefficients are of the same sign, and all are equally significant. The coefficients for the dummy variables CHILDREN'S CAREER and OWN CAREER are still insignificant, though the coefficient for CHILDREN'S CAREER is significant at the .10 level for a single-tailed t-test. The size of the coefficient for CHILDREN'S CAREER implies that even if a retiring politician has a child who will be going into political life, the politician's attendance rate will decline by 3.8 percentage points during his last term in office.

The results for column 3 indicate that for those retired politicians who will pursue careers in government and/or lobbying and whose children are also involved in careers with the government, there is virtually no change in attendance rates during their last term in office. The coefficient for CHILDREN'S*OWN CAREERS is significant at the .09 level and slightly larger than what is needed to completely offset the decline in attendance rates that normally arises during a congressman's last term in elective office. This result implies that forgone future employment opportunities are large enough to eliminate any shirking in terms of attendance rates. The other variables remain unchanged.

Finally, the results given in column 4 provide evidence that congressmen who retire when they are less than sixty-five years old reudce their attendance rates by only about hafl as much as retiring congressmen who are sixty-five or older--or about 5 percentage points less. LAST TERM*AGE 65 is significant at the .10 level for a two-tailed t-test. The other variables are relatively unchanged except that AGE is now significant at the .10 level for a two-tailed t-test.

One possible objection to the preceding results is that the LAST TERM variable may be simultaneously determined with voting attendance and, as a consequence, the ordinary least squares estimates that have been presented yielded biased and inconsistent estimates. For instance, one story could be that marginal congressmen, those most likely to quit in anticipation of defeat (see Jacobsen and Kernell [1981, Chapter 5]), may face lower voting rates because they spend more time back in their districts deciding whether to run again. To some extent, this is probably being picked up by such variables as CHANGE IN GENERAL ELECTION VOTE DIFFERENCES and CHANGE IN PRIMARY ELECTION VOTE DIFFERENCES. However, if it is not, the question becomes to what degree those incumbents who decide to retire during their final term in office represent observations from the same model as those incumbents who made the decision to retire prior to their last term.

To test the preceding hypothesis column 1 was reestimated using two-stage least squares, assuming that the probability that a congressman will retire is a linear function of

LAST TERM = f(ILLNESS 95th, CONGRESS, ILLNESS DURING 94th CONGRESS, REPUBLICAN, TENURE, AGE, GENERAL ELECTION VOTE DIFFERENCE 1976, PRIMARY ELECTION VOTE DIFFERENCE 1976), (2)

where GENERAL ELECTION VOTE DIFFERENCE 1976 is the congressman's margin of victory in the 1976 general election and PRIMARY ELECTION VOTE DIFFERENCE 1976 is his margin in the 1976 primary. These results correspond with column 1 in Table III. The coefficients on both LAST TERM and CHILDREN'S OR OWN CAREERS both become slightly smaller and less significant, with their respective coefficients becoming -7.173 and 1.923 and absolute t-statistics equally 2.809 and 0.634. The rest of the variables also remain essentially unchanged. The estimates for equation 2 were

LAST TERM = -.047 (1.601) REPUBLICAN + .0279 (0.631) ILLNESS DURING 95th CONGRESS -.005 (0.143) ILLNESS 94th CONGRESS + .0076 (3.948) TENURE + .0035 (2.215) AGE - .00134 (2.600) GENERAL ELECTION VOTE DIFFERENCE 1976 - .00047 (0.884) PRIMARY ELECTION VOTE DIFFERENCE 1976 (3) [R.sup.2] = .1267 F-ratio = 7.044,

where the absolute t-statistics are shown in parentheses. Not surprisingly, AGE and TENURE have a significant positive effect on the probability of retirement, while the closeness of the most recent general election has a significant negative impact. The weakly significant coefficient on the party dummy variable provides some support for Jacobsen's and Kernell's [1983, 49-59] claim of the importance of party affiliation in determining the probability of retirement. Similar two-stage least squares estimates were made in Table III for columns 2, 3, and 4 with relatively little change in the estimates from the corresponding columns shown in Table II.

IV. IMPLICATIONS

Politicians do vote significantly less in their last period. However, assuming that political parties or constituencies are actually trying to control how often retiring congressmen vote through the control of employment upon leaving office, it seems they are having only modest success. Shirking is reduced or eliminated only when both the retiring congressmen and their offspring continue to be involved in politics after the congressmen leave office.

Of the twenty-seven retiring congressmen in the sample, eleen cases had neither the retired congressmen nor their children take any politically related jobs. Thus in only 59 percent of the cases examined was there even a chance of influencing a congressman's attendance rate through future employment. However, when these remaining sixteen cases are taken as a whole, their attendance rates are virtually indistinguishable from the original eleven.

Though shirking is eliminated for those congressmen for whom both they and their children are involved in politics after the congressmen's retirement from elective office, this subsample represents only 11 percent (three cases) of the sample of retiring politicians and indicates the limited scope available to completely eliminate opportunistic behavior through the control of postelective careers. Not only are the effects insignificant when the future careers of the politicians and their children are controlled for separately, but they are only able to offset half of the reduction in attendance rates.

While the coefficients for the dummy variable representing the careers of the politicians and their children are not significant when they are taken separately, the results suggest that both the size and significance of the effects increase when they occur simultaneously. The greater the total cost imposed upon politicians, the greater seems to be the response. (10)

Finally, it eems to be more costly for younger retiring politicians to shirk. Whether this occurs as a result of future employment opportunities or simply because young former politicians must endure the wrath of disappointed constituencies for a longer period is not obvious, Being young when retiring also did not completely eliminate shirking, but reduced the last-period problem by about 50 percent. (11)

Given that this evidence examines only one complete cohort of retiring congressmen, these results can be viewed as only suggestive. They do, however, strongly indicate the presence of a last-period problem for politicians in terms of attendance rates, and that this shirking can be affected to some extent by post-elective employment. If the theory presented here is correct, similar relationships are likely to emerge when other dimensions of constituent services such as case work are examined. Assuming that data on postelective office employment are eventually gathered for the 96th Congress, an obvious extension of this work would be see what effect such employment has on the measures of constituency service examined by Cain et al. [1987]

(1) Some circumstantial evidence of constituent services' importance is provided by Florida [1977, chapter 7], while Cain et al. [1987, 176-89] provide direct empirical evidence of the role that such services play in determining the probability by which a congressman and a member of Britain's parliament will be reelected.

(2) Lott [1987b] is the only attempt I know of that tries to explain how attendance rates for politicians change when they no longer face reelection. While in general, a similar specification will be used here, Lott [1987b] does not address the question of how post-elective office employment of politicians and/or their children affect the cost of shirking through attendance rates during their last term. While that paper does investigate the effect of post-elective office employment on how a politician votes when he does vote, it is not a "fair" test of the usefulness of future employment in constraining shirking. This is because if the ideology of politicians coincides sufficiently closely to that of their constituents, no cheating will exist in terms of how a politician votes and so there is nothing to constrain. As section II points out, the question of how often a politician votes can be fundamentally different from how he votes when he does vote (see also Lott [1987a]).

(3) Becker and Stigler [1974, 11-12] and Barro [1973] were the first to recognize explicitly that a premium may exist for holding office, the threatened loss of which makes shirking by politicians costly. Lott [1987a; 1987b] argues that opportunistic behavior by politicians in terms of how they vote, when they do vote, can be solved though ideology. Benson and Baden [1985] discuss the problems of legislators monitoring the shirking or cheating behavior of bureaucra Landes and Posner [1975] and Crain and Tollison [1979] discuss various mechanisms used to guarantee "contracts" made with special-interest groups. (See also Coats and Dalton [1989] and Van Beek [1989] for recent work supporting my earlier findings.) The discussion here focuses on how an alternative mechanism--post-elective office employment--enforces implicit contracts made by legislators.

(4) As Eckert notes [1981, 120] his evidence is equally consistent with future employment acting either as a bribe "for votes on the bench that were favorable to the industry or a particular firm" or as "the return on the investment in human capital that the commissioner made by learning the details and politics of regulation...."

(5) As the cost of shirking in terms of forgone future support is reduced, politicians are expected also to purchase more leisure by producing less output in the other dimensions mentioned earlier (e.g., case work). While it is hyothetically possible that a retiring congressman votes less congressmen in their last term suddenly begin to produce a different mix of output, this seems doubtful. First, it is not clear why changes in voter preferences would be related to their congressman's term in office. Second, given that retiring incumbents do not have to run for reelection, it seems unlikely that the increase in the amount of time necessary to produce these other outputs would be so great that it not only absorbs the time savings from not having to campaign but also requires a reduction in the time devoted to voting on legislation.

(6) The appendix, based on questioning of the retired politicians used in this sample, shows that 58 percent of the sample made their "final" decision to retire at least by November 1976, and that 80 percent had made it by the end of 1977. The Congressional Quarterly [1978, 33] reports that by January 7, 1978, seventeen congressmen had announced their retirement. It is assumed here that congressmen who died during the 95th Congress did not anticipate their death and thus they were not included as retiring. Those who died early were not included in the sample due to a lack of data on their voting patterns.

(7) Their telephone numbers were obtained from either (i) the Association of Former Members of Congress, (2) old telephone directories and directory assistance, or (3) other members of Congress for whom I had telephone numbers. (See also Lott [1987b, 177-79].)

(8) Both the VD7876 and PD7876 variables assume a linear relationship between changes in the level of shirking and changes in the relative difficulty of reelection. Peltzman [1976, 214, equation 1] makes a similar assumption when he assumes that politicians weight each additional vote equally in their objective functions.

(9) There is no significant difference in the attendance rate level between Republicans and Democrats in either the 94th or 95th Congresses.

(10) One question arises from the observation that since politicians occasionally announce that an upcoming election will be their last, as well as the existence of tenure limits for certain offices, voters will at times know when the last period will occur (see Fremling and Lott [1988] and Lott [1988]). If voters realize that political lame ducks will represent their interests less energetically, why would voters ever knowingly place politicians in this last period? One possible explanation is provided in the discussion on sorting of politicians by their ideological beliefs found in Lott [1987b]. There I argued that a politician can produce a signal to convince constituencies that he is an "ideologue" by engaging in activities the constituencies desire to have continued once the last period arrives nsee also section II of this paper). Since ideological politicians can produce this investment (e.g., past voting records and/or "good deeds") at a lower cost than nonideologues since they intrinsically value the signal, competition to produce this signal will produce too low a return to mimicking nonideologues. The longer this signal must be produced, the greater the difficulty in mimicking. Thus, voters may be more willing to put up with a politician that they are certain about with respect to how he will vote when he does vote versus a less certain prospect who will vote more often. After all, since retiring politicians have been in office for appreciable periods of time, the nonideologues who would cheat in terms of how they will vote in the last period may have already already been weeded out.

(11) The weakly significant coefficients for TENURE also provide some evidence consistent with the barrier to entry arguments discussed in Lott [1987a]. However, as noted in those discussions, other interpretations are also possible.

REFERENCES

Amacher, Ryan C, and William J. Boyes. "Cycles in Senatorial Voting Behavior: Implications for the Optimal Frequency of Elections." Public Choice, 33, 1978, 5-13.

Barone, Michael and Grant Ujifusa, The Almanac of American Politics, Washington, D.C.: National Journal, 1976, 1978, and 1980.

Barro, Robert. "The Control of Politicians: An Economic Model." Public Choice, 14, 1973, 19-42.

Becker, Gary and George Stigler. "Law Enforcement, Malfeasance, and the Compensation of Enforcers." Journal of Legal Studies, January 1974, 1-18.

Benson, Bruce and John Baden. "The Political Economy of Government Corruption: The Logic of Underground Government." Journal of Legal Studies, June 1985, 391-410.

Cain, Bruce, John Ferejohn, and Morris Fiorina. The Personal Vote: Constituency Service and Electoral Independence. Cambridge: Harvard University Press, 1987.

Crain, W. Mark and Robert Tollison. "The Executive Branch in the Interest-Group Theory of Government." Journal of Legal Studies, June, 1979, 555-67.

Coats, R. Morris and Thomas R. Dalton. "Entry Barriers in Politicis and Uncontested Elections." Nicholis State University Working Paper, 1989.

Congressional Quarterly Weekly Report, Washington, D.C.: Congressional Quarterly, Inc., 1976, 1977, 1978, 1979.

Davis, Michael and Philip Porter. "Identifying Pure and Apparent Ideology in Congressional Voting." Public Choice, February 1989, 101-11.

Dougan, William and Michael Munger. "The Rationality of Ideology." Journal of Law and Economics, april 1989, 119-42.

Eckert, Ross. "The Life Cycle of Regulatory Commissioners." Journal of Law and Economics, April 1981, 113-20.

Fenno, richard F., Jr. Home Style: House Members in Their Districts. Boston: Little Brown, 1978.

Fiorina, Morris P. Congress--Keystone to the Washington Establishment. New Haven: Yale University Press, 1977.

Feregohn, John. "Incumbent Performance and Electoral Control." Public Choice, 50, 1986, 5-25.

Fremling, Gertrud and John R. Lott, Jr. "Televising Legislatures: Some Thoughts on Whether Politicians Are Search Goods." Public Choice, 58, 1988, 73-78.

Jacobson, Gary and Samuel Kernell. Strategy and Choice in Congressional elections. New Haven: Yale University Press, 1981.

Johnston, R. J. "Campaign Spending and Votes: A Reconsideration." Public Choice, 1978, 83-92.

Kalt, Joseph and Mark Zupan. "Capture and Ideology in the Economic Theory of Politics." American Economic Review, June 1984, 279-300.

Kau, James and Paul H. Rubin. "Self-Interest, Ideology, and Logrolling in Congressional Voting." Journal of Law and Economics, October 1979, 365-84.

Laband, David and Bernard F. Lentz. "Favorite Sons: Intergenerational Wealth Transfers Among Politicians." Economic Inquiry, July 1985, 395-414.

Landes, William M. and richard A. Posner. "The Independent Judiciary in an Interest-Group Perspective." Journal of Law and Economics, December 1975, 875-901.

Lott, John R., Jr. "Brand Names and Barriers to Entry in Political Markets." Public Choice, 51, 1986, 87-92.

Lott, john R., Jr. "The Effect of Nontransferable Property Rights on the Efficiency of Political Markets: Some Evidence." Journal of Public Economics, march 1987a, 231-46.

Lott, John R., Jr. "Political Cheating." Public Choice, 52, 1987b, 169-87.

Lott, John R., Jr. "Explaining Challengers' Campaign Expenditures: The Importance of Sunk Nontransferable Brand Name." Public Finance Quarterly, January 1989, 108-18.

Maddalla, G. S. Econometrics. New York: McGraw-Hill, 1977.

Nelson, Douglas and Eugene Silberberg. "Ideology and Legislator Shirking." Economic Inquiry, January 1987, 15-25.

Nelson, Phillip. "Political Information." Journal of Law and Economics, August 1976, 315-36.

Peltzman, Sam. "Constituent Interest and Congressional Voting." Journal of Law and Economics, January 1984, 181-210.

Telser, Lester. "Comment on Political Information." Journal of Law and Economics, August 1976, 337-40.

Van Beek, James. "Does the Decision to Retire Increase the Amount of Political Shirking?" Texas A&M University Working Paper, 1989.

Welch, William P. "The Economics of Campaign Funds." Public Choice, 27, 1974, 83-97.

Wright, Matthew. "Shirking, Political Support, and Agency Costs in Political Markets." University of California at Los Angeles mimeo, 1988.

Anderson Graduate School of Management. University of California at Los Angeles. I would like to thank Mark Crain, Gertrud Fremling, Russell Roberts, an anonymous referee from this journal, and participants at the 1987 Western Economic Association meetings for their helpful comments and Jim Van Beek for his research assistance. This paper originally appeared as Hoover Institution working paper #E-86-60/10. Any remaining errors are mine.