An analysis of the impact of congressional term limits.

Reed, W. Robert ; Schansberg, D. Eric

We examine the impact of a six-term tenure limit on members of the U.S. House of Representatives. The first part of this study develops a methodological framework for simulating the quantitative impact of term limits. The second part of this study uses this framework to estimate the effects of term limits on (i) the value of holding office, (ii) turnover, and (iii) the size of the Democratic majority in the House. Both steady-state and transition period effects are explored.

I. INTRODUCTION

The widespread support for term limitations seems to have caught both popular commentators and academic researchers by surprise.[1] Why do voters want them? and What is the likely impact of such legislation? A small but growing literature concentrates on the first of these two questions (Adams and Kenny [1986], Dick and Loft [1993]). In contrast, this paper addresses the second. In particular, it investigates the empirical consequences of imposing a twelve-year (six-term) tenure limit on the members of the U.S. House of Representatives.

Both the degree and timing of the impact of the alternative regimes are investigated here. The issue of timing arises because term limits can have very different effects depending upon how they treat currently serving congressmen. We focus our investigation on the two legislative regimes that we believe have the greatest chance of being implemented: limited grandfathering and complete grandfathering.[2] Assuming term-limit legislation is passed in 1992, we trace out the time paths of the effects associated with each of these regimes. As we shall see, there are important transitional differences between the two.

The paper proceeds as follows. Section II addresses the methodology used in this analysis. Section III considers the impact of term limits on (i) the value of holding office, (ii) turnover, and (iii) the Democratic majority in the House. Section IV concludes.

II. METHODOLOGY

The first concern of this paper is to model the seniority structure of the House in the absence of congressional term limits. Let N,(t) identify the number of congressmen in their nth term at time t. Let r, define the proportion of congressmen that successfully move from term n to term n+l. Note that there are a number of reasons for not continuing to an additional term. Besides

electoral defeats, there are voluntary departures to seek other elective office or private-sector employment, retirements, and deaths? Given any set of continuation rates, [r.sub.1], [r.sub.2], .... and an initial distribution of representatives across terms, [N.sub.1](t), [N.sub.2](t), .... one can construct what the seniority structure of Congress will look like any number of periods into the future. For instance, the number of congressmen in their jth term at time t+m is given by

[MATHEMATICAL EXPRESSION OMITTED]

where [N.sub.1](t+1+m-j) identifies the number of representatives who enter Congress in time period (t+1+m-j)> (t+1), and j and m > O. Ifj-m > 1, one simply takes the number of congressmen currently serving their (j-m)th term, and multiplies this by the appropriate product of continuation rates.

If j-m (1, one needs to identify the number of congressmen who will enter Congress in time period (t+l+m-!). Since Congress always has 435 members, the number of arrivals must always equal the number of departures. The number of first-term congressmen who leave office at the end of time period t is given by (1- [r.sub.1])[N.sub.1](t). Similarly, the number of second-term congressmen who leave office at the end of time period t is given by (1-[r.sub.2])[N.sub.2](t). Thus, the number of representatives who will enter Congress in time period t+l is given by

[MATHEMATICAL EXPRESSION OMITTED]

Equations (1) and (2) allow one to recursively solve for [N.sub.1](t+1+m-j) for all m-j > -1 (i.e., j-m (1). They provide the flamework for addressing a host of questions concerning the impact of term-limit legislation.

The choice of an appropriate set of continuation rates, [r.sub.1] ..... is crucial for the validity of the analysis. Obviously, one would like to use the actual continuation rates that will prevail in the future. Since these are not known, we proceed by using two continuation rate patterns from recent history that might reasonably be supposed to approximate the continuation behavior of future Congresses: (i) those from the years 1985-1991 (corresponding to the continuation experiences of the last four Congresses at the time of this writing), and (ii) those from the years 1977-1991 (corresponding to the last eight Congresses).[4] Having two sets of continuation rates provides a check of sorts on the sensitivity of our results to different continuation rate regimes.

It is straightforward to modify the preceding methodology to incorporate the effects of congressional term limits. In the case of a six-term limit, all continuation rates after [r.sub.5] are zero: [r.sub.n] = 0 for n > 5. Two complications arise. The first issue concerns the degree of grandfathering. In our analysis, we will assume that the term limit legislation takes effect beginning with the 102nd Congress. Accordingly, under limited grandfathering, rn = 0, n > 5, for all incoming representatives beginning with the 103rd Congress. In contrast, for each member of the 102nd Congress currently serving his ith term, [r.sub.n] =0 for n > i+5. Under complete grandfathering, incoming representatives are treated as described above, but currently serving congressmen are unaffected by the legislation (i.e., the r,'s remain unchanged).

The second issue concerns changes in the continuation rates after term limits are implemented. Term limit legislation will likely change the incentives facing both politicians and voters in their electoral decisions. Since continuation rates reflect these decisions, they may change as well. Some commentators believe that term limits would serve to discourage challengers from running against incumbents before they completed their maximum service. They argue that term limits would function as term guarantees (Benjamin and Malbin [1992]). Others believe that term limits would encourage greater competition. Since incumbents would not stay in office as long, special interests would have less incentive to financially support them. This would reduce disparities between challenger and incumbent campaign spending, thus making races more competitive. In lieu of a better alternative, our working assumption is that there is no change in continuation rates under term limits.s As discussed above, the consideration of two sets of continuation rates provides some measure of the sensitivity of our results to changes in the continuation rate regime.[6]

III. RESULTS

The Value of Holding Office

Figure 1 plots the mean completed spell length for every entering cohort of representatives from 1953 to 1989. For those House members who left office prior to 1991, completed tenure is measured by the actual length of time between their entry and departure dates. For those House members still serving in 1991, the remaining spell length is estimated using the methodology described above? These values were then added together and divided by the number of representatives in the respective cohort to obtain the associated mean values of completed tenure. Two estimates of mean completed tenure are used for each cohort, as indicated by the dark solid lines in Figure 1. One line is based on computations of remaining spell lengths of currently serving House members using the continuation rates from 1985-1991. The other uses the continuation rate pattern from 1977-1991. The recent trend towards increased congressional tenure is clearly evident, and is discussed in detail elsewhere (Reed and Schansberg [1992]).

Superimposed on this historical trend are three horizontal lines, identifying the expected values for mean completed tenure both in the absence and presence of a twelve-year (six-term) restriction on tenure. In each case, the values are constants, since all members entering future Congresses are assumed to experience identical survival probabilities. The two upper lines identify future expected lengths of stay for House members assuming no term restrictions exist, each corresponding to a given continuation rate pattern. Using the 1985-1991 continuation rates, the average length of a completed spell is estimated to be 17.8 years for a new member of the House. Using the 1977-1991 continuation rates, the corresponding value is 13.5 years.

The bottom dotted line identifies the expected length of a completed stay in the House assuming that term limits are imposed. In general, one would expect different estimates for different sets of continuation rates. In fact, they diverge only slightly. Using continuation rates from 1985-1991, it is estimated that the average length of a completed House stay will be 6.32 years. If one

uses the continuation rates from 1977-1991, the corresponding value is 6.24 years. The difference is not distinguishable on the graph.

Figure 1 illustrates that a six-term limit would dramatically reduce mean completed tenure for future cohorts of U.S. House members. We estimate that the average stay in the House would be cut by seven to twelve years, corresponding to a decrease of 50 to 65 percent. This would cause average lengths of stay in the House to fall to levels far below anything that has been observed in recent times. Indeed, one has to go as far back as the turn of the century to find average length of stay values that approximate those that would arise under a six-term limit (Polsby [1968]).

While expected length of time in office is an important determinant of the overall value of holding office, there are other factors. One such factor concerns the impact of term limits on the opportunity that House members have to achieve positions of leadership. In general, leadership can be expected to enter into the value of holding office via three dimensions: the amount of tenure one needs to accumulate before becoming eligible for a leadership position, the probability of being appointed to a leadership position once one becomes eligible, and the amount of time one can expect to occupy a leadership position.

Note, however, that the latter two dimensions are inversely related. Holding constant the number of members and the number of leadership positions, increases in the probability of attaining a leadership position are necessarily accompanied by decreases in the mean length of tenure in a leadership position. As a result, it is difficult to assess how term limits would affect the overall value of holding office through these two dimensions.[8] For that reason, the subsequent analysis focuses on the effect of term limits on the expected waiting time before appointment to a House leadership position.

This study defines the leadership of the House as the forty to fifty members who comprise the ranking membership of the House's most important committees and top party positions.[9] Seniority still plays a key role in these appointments. As a rough approximation, we shall assume that an individual becomes eligible for a party leadership position upon having attained a seniority rank equal to the 80th percentile.[10]

Figure 2 demonstrates the impact of term limits on the amount of time required to become eligible for leadership. It plots the projected 80th percentile seniority rank beginning with the 102nd Congress, under three different regimes: (i) no term limits, (ii) term limits with limited grandfathering, and (iii) term limits with complete grandfathering. In order to place in the 80th percentile in terms of seniority, a member of the 102nd Congress needed to have sixteen years of service in the House. If future congressional continuation behavior follows the pattern established from 1985-1991, this figure can be expected to rise to twenty-two years. Alternatively, if the years 1977-1991 are representative, the 80th percentile mark for future Congresses can be expected to remain relatively constant at sixteen years.

In contrast, if term limits were imposed, at first six, and eventually eight, years of service would be sufficient for one's length of service to place in the 80th percentile. Note that this result is robust with respect to both sets of continuation rates and both grandfathering regimes, though complete grandfathering would delay the onset of these changes. With respect to the value of holding office, this reduction in the time required to achieve a position of leadership within the House has an unambiguous impact: ceteris paribus, it will increase the attractiveness of obtaining and holding office for incoming congressmen.[11]

This positive impact would be vitiated during the first few years under term limits, though the extent would differ between limited and complete grandfathering regimes. Under complete grandfathering, those representatives already serving in office at the time term limits were implemented would be allowed to stay in Congress until they were defeated or voluntarily left. Newly entering representatives could serve their six terms and be forced to leave before they had acquired sufficient seniority to be eligible for leadership. We estimate that under complete grandfathering a period of six to twelve years would arise during which no entering congressman could expect to obtain a leadership rank during his career in the House. Under limited grandfathering, this effect would not be so severe. Only the first entering class subsequent to the adoption of term limits would be denied leadership positions.

In summary, an evaluation of the effect of term limits on the value of holding office depends crucially on the degree to which representatives value achieving positions of House leadership. On the one hand, shorter terms in office unambiguously reduce the value of holding office. On the other hand, if the opportunity to become a leader in the House is highly valued, then term limits could potentially increase the value of holding office. The effect of term limits on political performance will largely be determined by which of these two effects dominates.[12] This is a major reason why it is difficult to predict the future behavior of continuation rates under term limits. Without knowing the effect of term limits on the value of holding office and, a fortiori, political performance, we are unable to forecast the effect of term

limits on future continuation rates. In lieu of a better alternative, we assume the continuation rates from the recent past will continue to prevail into the future.

First- and Last-Termers

This section focuses on the degree and timing of the impact term limits are expected to have on congressional turnover. It is an ineluctable consequence of congressional turnover that for every newcomer to the current Congress, there is a representative who served his last term in the previous Congress. Thus, potential last-period control problems must be kept in mind when evaluating changes in congressional turnover.[13]

Figure 3 plots the expected percentages of first-termers for three different regimes: (i) no term limits, (ii) limited grandfathering, and (iii) complete grandfathering. The top panel reports results based on average continuation rates from 1985-1991. The bottom panel does the same using average continuation rates from 1977-1991.

Consider first the percentage of first-termers in the absence of term limits. If current trends are maintained into the future, we estimate the steady-state turnover rates to be 11.3 and 14.7 percent, using the 1985-1991 and 1977-1991 continuation rates, respectively. Looking now at the effect of term limits, it is clearly seen that the percentage of first-termers in Congress will increase relative to the amount that would arise in the absence of term limits. The steady-state values under term limits are just over 20 percent. While these percentages of first-termers are substantially larger than those that would arise in the absence of term limits, they are not without recent historical precedent. Since 1953, the percentage of first-termers has ranged between 10 and 23 percent.

Thus, while term limits are predicted to somewhat increase congressional turnover in steady state, these increases would not likely yield dramatic differences in the numbers of first- and last-termers represented in Congress, at least when compared to recent historical experience. However, during the transition to steady state, both limited grandfathering and no grandfathering (not shown) term-limit regimes would create enormous twelve-year waves of incoming congressmen. For example, using continuation rates from 1985-1991, a limited grandfathering regime would generate a "super class" of 251 freshman congressmen (57.7 percent) that would enter Congress twelve years after enactment of term-limit legislation. Twelve years later, another super class of 180 freshman congressmen (41.4 percent) would enter Congress. If continuation rates from 1977-1991 are used, the corresponding numbers are 204 (46.9 percent) and 159 (36.6 percent), respectively.[14] To provide some historical perspective, the last time Congress had a turnover rate over 30 percent was 1933. One has to go back over a hundred years to 1883 to find a Congress with a turnover rate over 50 percent (Polsby [1968]). However, as Figure 3 also demonstrates, these huge swings in the sizes of entering classes of congressmen would be avoided if a regime of complete grandfathering were adopted.

Political Party Representation

One of the major hindrances of implementing term limits is the belief that term limits would greatly increase representation of Republicans at the expense of the Democratic party. To the extent that this is true, partisan political interests are likely to derail reform efforts. In this section we investigate the impact of term limits on political party representation.[15] Figure 4 plots the predicted numerical advantage in representation enjoyed by the Democratic party in future Congresses.[16] As before, the top panel identifies the expected composition of Congress assuming that future behavior is based on the average continuation rates from 19851991. The lower panel does the same, using average continuation rates from 1977-1991.

The standard for comparison is the behavior of the House Democratic advantage in the absence of term limits. In 1991, Democrats held a 101 seat advantage. The analysis here suggests that, even in the absence of term limits, that advantage will gradually decline to steady-state values of either seventy-nine or fifty-nine seats. The primary reason for this decline is that a disproportionate number of Democrats entered the House in the mid- to late 1970s, due to Watergate and its aftermath? As these cohorts leave Congress, the Democratic advantage will decrease, ceteris paribus.

Now consider the impact of term limits. As Figure 4 clearly shows, term limits would further reduce the Democratic party's advantage in Congress, since Democrats are represented in greater number in the more advanced terms. Under limited grandfathering, the Democrat's advantage would immediately decline once the term limit became binding on the original class of congressmen. Mild cycling behavior would also occur. After the first "super class" of freshman congressmen entered, the Democratic advantage in the House would gradually increase because Republicans leave office at a higher rate than Democrats. Twelve years later, a disproportionate number of Democrats would reach their sixth term and leave Congress, and the cycle begins again. Similar behavior arises under a no grandathering regime.

Thus, while term limits would decrease the Democratic advantage in the House, the numerical size of the impact would not be dramatic. The steady-state estimates suggest that term limits would reduce the Democratic advantage by ten to forty seats. Changes of this magnitude have been common in the post-1953 period.

We offer one last observation with regard to the implications of these results for the likelihood of a term-limit law being passed. While the House majority party would be disadvantaged by the adoption of term limits, this is not necessarily the case for its representatives in Congress. A term-limit law with complete grandfathering would distort power in the future towards those currently serving since they would be unaffected by the law. Their individual power would be enhanced by virtue of their gain in relative seniority with respect to new members. To the extent that currently serving representatives value their individual welfare above their party's, the likelihood of a term-limit law being passed is enhanced.

IV. CONCLUSION

First, it is not clear how term limits would affect the value of holding office, and thus political performance. On the one hand, a six-term limit would dramatically cut mean lengths of stay of U.S. representatives. On the other hand, it would also reduce the time that representatives would have to acquire before becoming eligible for leadership positions in the House. Our estimates of this latter reduction are sufficiently large that it not unreasonable to suppose that many will find service in the U.S. House to be more attractive under term limits. Many commentators emphasize the negative implications of term limits for political performance. For example, Gary Becker [1990, 18] writes, "Only an unrealistic view of human nature could presume that taking away the right to continue at a job will improve performance. More likely, congressmen will take less interest in their work and will spend their time arranging future careers." Our findings suggest that term limits might actually improve performance.

Second, while a six-term limit would increase steady-state turnover in the House, we estimate that it would still remain within the range of values realized during the post-1953 period. During the transition period, however, assuming a regime of limited grandfathering, it would create enormous "super classes" that would enter in twelve-year cycles. For example, assuming average continuation rates from the years 1977-1991 are maintained during future years, we estimate that turnover would equal approximately 47 percent of total House membership in the first year that term limits became binding. It would then fall below 20 percent until twelve years later, when it would rise to over 35 percent, and then fall again. While complete grandfathering would avoid the creation of cycles, it has its own deficiencies.

Finally, we find that term limits would have only a modest effect on the Democratic advantage in the House. We estimate that term limits would cause a net total of five to twenty seats to change from Democratic to Republican control. Changes of this magnitude have been frequent in the past, and so would not be expected to substantially alter the strengths of the respective political parties.

The Author

W. Robert Reed and D. Eric Schansberg*

* Assistant Professor, University of Oklahoma, and Assistant Professor, Indiana University-Southeast. We acknowledge helpful comments from John Lott, Jonathan Nagler, Dan Wood, and seminar participants at George Mason University and the 1992 Public Choice Society meetings. The constructive suggestions of three anonymous referees and an editor served to significantly improve the manuscript. The authors retain property rights to all remaining errors

1. There is historical precedent for term limits. Restrictions on the number of terms representatives could serve were part of the Articles of Confederation. After great debate, they were not included in the Constitution because they were considered to be "too much detail" {Fund [1992]). Despite the lack of a binding legal constraint, there were self-imposed limits for a long time after the establishment of the nation. Included in those self- imposed limits, House seats were frequently rotated between parties and competitors--a tradition that was practiced well into the nineteenth century (Kessler [1992]). Rotation began to decrease as the South realized the effectiveness of senior legislators in protecting the institution of slavery. The pursuit of additional terms became more attractive with the advent of the standing committee and a House leadership driven by seniority.

2. At the time of this writing, fifteen states had passed some kind of term limits for federal legislators. All of them call for limited grandfathering. This study includes the case of complete grandfathering for two reasons. First, were Congress to pass its own term-limit legislation, it would not be restricted to proposals which had been passed at the state level. For obvious reasons, Congress might find complete grandfathering more palatable than limited grandfathering. Second, as shall be discussed below, complete grandfathering addresses some potential problems associated with limited grandfathering, and is thus of analytical interest. The case of no grandfathering, where all incumbents having more than six terms experience are forced to leave office upon passage of the legislation, is discussed only briefly.

3. Non-electoral reasons for leaving office are important even during the early terms of a congressman's career, Schansberg [forthcoming 1994] reports that since 1960, only 39 percent of congressional spells have ended because of defeat, compared to 30 percent from retirement, 23 percent from pursuit of higher office, and 8 percent from death.

4. The mid-1970's represent a natural cut-off point for a number of reasons. Bullock and Loomis [1985] note that "the nature of congressional committees and subcommittees changed drastically during the early and mid-1970's." Congressional reforms included election of chairmen by secret ballot and a limit of one subcommittee chair per member. Further, Reed and Schansberg [1990] find that the continuation rates from elections after 1976 are significantly different from those of a preceding time period (1964-1976). The last four Congresses have been characterized by unusually high survival rates, and relatively low rates of entry, implying particularly long stays in office. To calculate r, based on the last four Congresses, we look at all the representatives in their nth term office during the years 1983 to 1989; and then calculate the percentage who actually succeeded to their n+lth term. The calculation of rn based on the last eight Congresses follows analogously, except that it expands to include the experiences of all House members in their nth term of office during the years 1975 to 1989. For a more detailed discussion of the data, including the table of term-specific continuation rates used for this study, see Reed and Schansberg [1992].

5. As a check on the assumption of invariance of continuation rates under term bruits, this study investigated all U.S. governors who served in office between 1968 and 1990. At question was whether governors from states with term-limits experienced markedly different continuation rates than governors from states without term limits. Three general term limit regimes were observed

at the state level: (i) one-term limits, (ii) two-term limits, and (iii) no term limits (terms are either two or four years). The first to second term continuation rate (rt) for governors serving under two-term limits during this period was 47/74 = 0.635. The first to second term continuation rate for governors serving without term restrictions was 52/72 = 0.722. Given differences in the length of terms across states (continuation rates for two-year terms are higher than continuation rates for four-year terms), among other things, we conclude that this evidence is not inconsistent with the invariance assumption.

6. There is a third issue. Our analysis assumes that politicians forced out of office by term limits will not return to serve additional terms (no reentry). Without this assumption, the problem of calculating the future seniority structure of the House becomes unmanageable. We checked this assumption by examining reentry behavior under term limits at the state level. Of all U.S. governors who served in office between 1960 and 1978 and who left office after meeting their respective term limits, only 8.43 percent (seven out of eighty three) had returned to office by 1992. This suggests that the assumption of no teentry may not be overly severe.

7. A more detailed discussion of the methodology employed in calculating expected lengths of completed tenure can be found in Reed and Schansberg [1992].

8. Not surprisingly, term limits would both increase the probability of eventually being appointed to a leadership position and reduce the length of time that a congressman could expect to spend in a leadership position. We estimate the expected probability of eventually being appointed to a leadership position would more than double under term limits. In contrast, we estimate that term limits could be expected to reduce by three to seven years the amount of time representatives spent in leadership positions.

9. This number is arrived at by adding (i) chairmen of non- "select" committees; (ii) ranking members of important non- "select" committees, (iii) party leaders, such as Speaker of the House, majority and minority whips, etc.; and (iv) chairmen of important House subcommittees.

10. We arrived at this number after surveying recent congressional experience. Of course, not all representatives eligible for leadership positions are actually appointed to them. While reasonable individuals can differ about the number of leadership positions and the amount of seniority required to reach these positions, the important thing is that seniority is still the key to real influence in the House. In this context, Hibbing [1991, 405] writes, "...pure legislative involvement..., despite widely held beliefs about the death of apprenticeship, is now more than ever dependent on tenure. In relative terms, junior members are less [HibbIng's emphasis] legislatively involved than they were a few decades ago."

11. Perhaps seniority would become less important in a Congress under term limits. Of course, if that were the case, it would strengthen the argument that term limits could increase the value of holding office by making leadership more accessible.

12. Note that there are both moral hazard and adverse selection consequences at play here. When the value of holding office increases, we would expect representatives to work harder to win voter support (moral hazard). However, candidates will not identically value the decrease in waiting time to leadership. Those candidates who more highly value a reduced waiting time to leadership will be more likely to run for office under term limits. Those who find the prospect of shortened careers in Congress most costly will be less likely to run for office. This adverse selection effect in favor of "non-professional" politicians is cited by many proponents of term limits.

13. For theoretical analyses of the last-period control problem, see Barro [1973] and Ferejohn [1986]. Little evidence of last- period shirking behavior exists. Recent papers reporting little or no evidence of shirking are Vanbeek [1991], Lott and Davis [1992] and Lott and Bronars [1993]. An exception is Zupan [1990]. For a theoretical discussion of why empirical studies may have difficulty identifying last-period shirking behavior, see Lott and Reed [1989].

14. Under a no-grandfathering term-limit regime, and using continuation rates from 1985-1991, 229 congressmen (52.6 percent) would enter the first Congress after term-limit legislation was passed. Twelve years later, 168 congressmen (38.6 percent) would enter. The numbers based on the 1977-1991 continuation rates are similar.

15. For a general theoretical treatment of the effect of term limits on political party strength, see Gilmour and Rothstein [1992].

16. The methodology used to generate these results follows very closely that presented in section II. There are two modifications. The first is that party-specific values are used for the continuation rates and the initial seniority structure (the rn's and Nn(t)'s, respectively). The second is that the number of new arrivals is divided between Republicans and Democrats. Once again, recent historical experience is used as a guide for future behavior. For the time period 19851991, 51.0 percent of freshman representatives were affiliated with the Democratic party. The corresponding value for the time period 1977-1991 is 53.5 percent.

17. Entering classes for the 94th to the 98th Congress 0975-1983) were 61.2 percent Democratic, with an average class size of 83.4. Entering classes for the 99th to the 102nd Congress (1985-1991) were 51.0 percent Democratic with an average class size of 45.7.

REFERENCES

Adams, J., and L. Kenny. "Optimal Tenure of Elected Public Officials." Journal of Law and Economics, October 1986, 303-28.

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

Becker, G. "Reforming Congress: Why Limiting Terms Won't Work." Business Week, 8 August 1990, p. 18.

Benjamin, G., and M. Malbin. eds. Limiting Legislative Terms. Washington, D.C.: Congressional Quarterly, Inc., 1992.

Bullock, C., and B. Loomis. "The Changing Congressional Career," in Congress Reconsidered 3rd ed., edited by L. Dodd and B. Oppenheimer. Washington, D.C.: Congressional Quarterly, Inc., 1985.

Dick, A., and J. Lott. "Reconciling Voters' Behavior and Legislative Term Limits." Journal of Public Economics, January 1993, 1-14.

Ferejohn, J. "Incumbent Performance and Electoral Control." Public Choice 50(1), 1986, 5-25.

Fund, J. "Term Limitation: An Idea Whose Time Has Come," in Limiting Legislative Terms, edited by G. Benjamin and M. Malbin. Washington, D.C.: Congressional Quarterly, Inc., 1992.

Gilmour, J., and E Rothstein. "The Impact of Term Limitation on Relative Party Strength in Legislatures." Working Paper, Department of Economics, Washington University, 1992.

Hibbing, J. "Contours of the Modem Congressional Career." American Political Science Review, June 1991, 405-28.

Kessler, C. "Bad Housekeeping: The Case Against Congressional Term Limits," in Limiting Legislative Terms, edited by G. Benjamin and M. Malbin. Washington, D.C.: Congressional Quarterly, Inc, 1992.

Lott, J., and S. Bronars. "Time Series Evidence on Shirking in the U.S. House of Representatives." Public Choice, June 1993, 125-49.

Lott, J., and M. Davis. "A Critical Review and an Extension of the Political Shirking Literature." Public Choice, December 1992, 461-84.

Loft, J., and W. Reed. "Shirking and Sorting in a Political Market with Finite-lived Politicians." Public Choice, April 1989, 75-96.

Polsby, N. "The Institutionalization of the U.S. House of Representatives." American Political Science Review, March 1968, 144-68.

Reed, W., and E. Schansberg. "How Long do Congressmen Stay in Office?" Economics and Politics, July 1990, 173-91.

______ "The Behavior of Congressional Tenure Over Time: 1953-1991." Public Choice, March 1992, 183203.

Schansberg, D. "Moving Out of the House: An Analysis of Congressional Quits." Economic Inquiry, forthcoming 1994.

Vanbeek, J. "Does the Decision to Retire Increase the Amount of Political Shirking?" Public Finance Quarterly, October 1991, 444- 56.

Zupan, M. "The Last-Period Problem in Politics: Do Congressional

Representatives Not Subject to a Reelection Constraint Alter Their Voting Behavior? Public Choice, May 1990, 167--80.