Moving out of the house: an analysis of congressional quits.

Schansberg, D. Eric

I. INTRODUCTION

Tenure in the House may end by electoral defeat, retirement, the pursuit of higher office, or death. While the literature looking at the possibility of defeat in a congressional election is extensive, studies looking at other types of departures from the House or Senate are few.(1) Schlesinger's "ambition theory" [1966] was the first attempt to explain and document, among other things, House members' moves to the Senate. Others have followed Schlesinger with somewhat more rigorous theoretical frameworks and limited empirical work on the pursuit of certain higher offices.(2) In their work on the congressional labor market, Reed and Schansberg [1990; 1992; 1994] have noted the importance of all types of departures as determinants of congressional tenure.(3) Despite these contributions, little has been done to empirically investigate voluntary departures, and nothing has been done to model the complete range of choices confronting a House member entering an election.

Have voluntary departures been overlooked for the most part because these occurrences are relatively infrequent? It turns out that voluntary departures are more numerous than involuntary departures. Looking at all exiting representatives since 1960, only 39 percent failed to be reelected, while 30 percent retired, 23 percent left to pursue higher office, and 8 percent died. Further, Reed and Schansberg [1992] note that since 1977, defeats in reelection bids account for only 43 percent of all departures by representatives with fewer than five terms in office, although reelection woes are often assumed to be the dominant cause of leaving Congress.

Further evidence to the importance of voluntary departures is seen in Figure 1, which provides a graphical representation of voluntary and involuntary departures since 1960. Voluntary departures are comprised of members of the House who resign their office to work elsewhere or retire, and those who leave to pursue higher office. Involuntary departures are simply those who are defeated in either their party's primary or the general election.(4) In every election cycle since 1969, more members have left the House voluntarily than involuntarily. Because they are so numerous, a comprehensive understanding of the turnover in Congressional membership requires a more thorough investigation of voluntary departures.

There are noteworthy differences in voluntary departure behavior between the parties. Table I presents the number of representatives of each party who have departed since the 87th Congress and their reason for leaving.(5) Note that Republicans have left to pursue higher office more often than their Democratic colleagues; in fact, it is the only type of departure where Republicans leave more frequently than Democrats. This is especially apparent since the 91st Congress: since then 31 percent of all Republican departures have been to pursue higher office while only 22 percent of Democratic stays ended for the same reason. A model that looks at voluntary departures should also address the question of why Republicans have left to pursue higher office more often than their Democratic colleagues.

TABLE I

Number of Departures by Type and Party for the 87th-101st Congresses

 DEFEATED RETIRED HIGHER OFFICE DIED
 D R D R D R D R
87th 26 6 9 14 9 1 8 7
88th 25 13 7 10 10 6 4 1
89th 13 37 12 12 8 5 4 4
90th 46 4 15 5 3 3 3 3
91st 10 3 13 3 4 13 2 1
92nd 11 11 7 6 3 20 6 5
93rd 14 7 14 16 10 5 3 2
94th 8 39 13 20 10 5 1 2
95th 10 5 22 7 17 10 4 0
96th 19 5 23 8 8 10 4 2
97th 31 5 23 8 3 8 3 1
98th 9 30 12 9 8 11 4 1
99th 17 3 3 7 7 6 2 1
100th 2 6 5 8 14 11 2 4
101st 2 6 10 4 7 9 4 1

87-90th 110 60 43 41 30 15 19 15
 (54%) (46%) (21%) (31%) (15%) (11%) (9%) (11%)

91-101st 133 120 145 96 91 108 35 20
 (33%) (35%) (36%) (28%) (23%) (31%) (9%) (6%)

TOTAL 243 180 188 137 121 123 54 35
 (40%) (38%) (31%) (29%) (20%) (26%) (9%) (7%)

Note: Percentages are calculated as a percentage of the party's total for each
type of departure. They may not sum to 100 percent because of rounding.

Source: Congressional Quarterly Almanac.

Moreover, voluntary departures represent the greatest possibility for a change in party control of a given congressional seat and, thus, overall party strength. Since 1960, in races for "open seats"--where the incumbent has voluntarily left--27 percent of the seats switched parties; in races where the seat was occupied by an incumbent, only 5 percent of the seats switched parties. With the oft-cited advantages of incumbency, a race between two non-incumbents represents the best opportunity for one party to wrest control of a seat from the other party.(6) Thus, any patterns in voluntary departure behavior within the parties may influence the extent of the power of the two parties.

In the context of labor economics, we should be concerned with the level of voluntary departures--quit behavior--and how this behavior changes over time. The average quit rate for the House from 1960-1990 was 4.4 percent per year and, referring to Figure 1 again, it is clear that the number of voluntary departures has varied considerably over time. Among other things, voluntary departures are a function of total compensation, including wages and other job characteristics. This is especially relevant in the face of recent congressional complaints about "low pay" and the pay raise that soon followed. Thus a question arises: is congressional quit behavior affected more by pecuniary or non-pecuniary aspects?

Since voluntary departures by representatives are important and have not been adequately investigated, we are provided with the opportunity to more carefully analyze the full range of decisions that are available to members of the House. This paper has three primary purposes: (1) to establish a theoretical framework to model all of the choices a representative faces at the end of any given term, (2) to measure the impact of different variables that affect a representative's decision to voluntarily leave his House seat, and (3) to explain why Republicans have been more likely to pursue higher office than their Democratic colleagues. Section II presents the model, establishing a framework from which we can analyze the impact of variables on a representative's decision. Section III presents the estimation of the model using non-ordered multinomial logit and data from 1960-1990. Section IV analyzes the results and provides an explanation for the observation that House Republicans pursue higher office more frequently than Democrats. The paper concludes with a discussion of some ramifications of these findings.

II. THE MODEL

A member of the House faces three alternatives at the end of his current term: to resign in order to retire or work elsewhere (R), to run for reelection to House (C), or to pursue a higher office (H). Let [X.sub.t] represent the choice made at the end of the period t; let [S.sub.t+1] represent the state occupied by the representative in the next period. A representative's decision is then contingent on the expected value of those alternatives, given by equations (1)-(3):

(1) V([X.sub.t] = C) = [P.sub.cc] [Rho] W([S.sub.t+1] = C) + [p.sub.CR] [Rho] W([S.sub.t+1] = R) - T[C.sub.c],

(2) V([X.sub.t] = H) = [p.sub.HH] [Rho] W([S.sub.t+1] = H) + [p.sub.HR] [Rho] W([S.sub.t+1] = R) - T[C.sub.H],

(3) V([X.sub.t] = R) = [Rho] W([S.sub.t+1] = R),

where

[p.sub.ij] = the probability with choice i of moving to state j,(7)

T[C.sub.i] = the total cost of pursuing choice i,

[Rho] = a constant discount factor.

Given the value of the possible choices, the probabilities of making the respective choices are defined in equations (4)-(6):

(4) Pr([X.sub.t] = C) = Pr{[V([X.sub.t] = C)] [is greater than] max[V([X.sub.t] = H), V([X.sub.t] = R)]},

(5) Pr([X.sub.t] = H) = Pr{[V([X.sub.t] = [is greater than] max[V([X.sub.t] = C), V([X.sub.st] = R)]},

(6) Pr([X.sub.t] = R) = 1 - Pr([X.sub.t] = C) - Pr([X.sub.t] = H).

Thus, a representative's decision is ultimately based on his estimate of the net present value of his options. As the benefits, costs and probabilities of success change, the values of the alternatives are altered and particular choices become more/less likely. If, for example, the benefits of holding higher office were to increase, W([S.sub.t+1] = H) and V([X.sub.t] = H) would increase and moves to higher office would be more likely, ceteris paribus. In addition, continuing in the House and retirement become relatively less attractive and, thus, less probable.(8)

The next step is to look at the variables that affect the V's. Equations (7a)-(7e) give the parameters that are important to the representative's choice, expressed as a function of general variables.

(7a) [p.sub.CC] = f(scandals--brand name quality, strength of challenger).

(7b) [p.sub.HH] = f(scandals--brand name quality, opportunities to achieve higher office ex ante).

(7c) W([S.sub.t+1] = C) = f(power in House seat).

(7d) W(S.sub.t+1] = R) = f(age, opportunities to achieve appointed higher office ex post--in the case of failing in choice H).

(7e) TC = f(scandals--brand name quality, ease of opportunities to achieve higher office ex ante, amount of previously accumulated brand name capital).

Two things warrant some discussion. The cost of pursuing elective office is equivalent to the amount of "brand name" capital to be purchased (the cost of political advertising) in the pursuit of that office. (As I shall discuss in the next section, brand name capital can also be previously acquired from other sources.) If the agent has a large stock of brand name capital entering an election, the costs of continuing in the House and of pursuing higher office are lower. Thus, those options would be more likely while retirement would be less likely.

Finally, while notationally inconvenient, the above model can be amended to allow for finite-lived agents.(9) Accordingly, it can be demonstrated that as a representative's time horizon shrinks, age becomes a factor and retirement becomes more likely. While the costs and benefits of holding office are important, the increasing opportunity costs of growing pension benefits and the increasing marginal utility of having fewer and fewer psychic benefits from the leisure one enjoys in retirement are crucial as well. Note also that to the extent that the initial pursuit of higher office involves a one-time fixed cost, the move to higher office can be regarded as an investment. The shorter time one has to recoup that fixed cost, the less likely one is to make the investment.

III. DATA, VARIABLES AND METHOD OF ESTIMATION

Since the representative evaluates all three alternatives simultaneously, non-ordered multinomial logit is an appropriate empirical framework to determine the impact of the relevant variables on the agent's decision.(10) The probability of alternative j being chosen in period t is

(8) P([X.sub.t] = j) = exp([Z[prime].sub.t][[Beta].sub.j])/[summation of] exp([Z[prime].sub.t][[Beta].sub.j) where j = 1 to 3

where

j = 1,2,3 (the three alternatives: C,H,R),

[Z.sub.t] is a vector of observations of independent variables,

[[Beta].sub.j] is a vector of coefficients.

In order to have an exactly identified system, we need to impose a constraint on the coefficients; one of the three alternatives needs to be expressed as a function of the other two. Thus, the coefficients on the determinants to retire are standardized to zero ([[Beta].sub.R] = 0). Further, let [P.sub.jkt] be the odds in favor of choosing alternative j over alternative k at time t:

(9) [P.sub.jkt] = P[X.sub.jt])/P([X.sub.kt]) = exp[[X[prime].sub.t]([[Beta].sub.j] - [[Beta].sub.k])].

It follows that

(10) [Delta]log([P.sub.jkt])/[Delta][X.sub.t] = [[Beta].sub.j] - [[Beta].sub.k].

Equation (10) can be interpreted as the partial derivatives of the log of the odds of choosing alternative j over alternative k. Thus, if [[Beta].sub.j] [is greater than] [[Beta].sub.k], then for a change in [X.sub.t], alternative j becomes relatively more likely than alternative k. Note that with the imposed constraint of [[Beta].sub.R] = 0, the [[Beta].sub.j] coefficients by themselves may be interpreted as the partial derivative of the log of the odds of choosing alternative j over retirement.

Turning to estimating the model, the general variables from section II will be represented by the specific variables in equations (11a)-(11f):

(11a) Quality of Brand Name Capital = f(SCANDAL, WATERGATE).

(11b) Strength of Challenger = f(LSTVTPCT).

(11c) Ex Ante/Post and Ease of Opportunities for Higher Office = f(PRESIDENT, OPENSEAT).

(11d) Value of a House Seat (Power) = f(PARTYPCT, COMMITTEE, LEADER, RWAGE).

(11e) Quantity of Brand Name Capital = f(TENURE, TENURESQ, DISTRICT, REDIST).

(11f) Time Horizon = f(STARTAGE).

Thus, the complete [Z.sub.t] vector of independent variable is

(12) [Z.sub.t] = (SCANDAL, WATERGATE, LSTVTPCT, PRESIDENT, OPENSEAT, PARTYPCT, COMMITTEE, LEADER, RWAGE, TENURE, TENURESQ, DISTRICT, REDIST, STARTAGE),

where the variables are defined below.

SCANDAL is a binary variable (1=yes, 0=no) that denotes whether a representative was investigated by the House for ethics problems or was involved in some other type of scandal in his private life within the last two election cycles.(11) In addition, Watergate was considered a scandal for Republican representatives entering the 1974 elections. One would expect SCANDAL and WATERGATE to be negatively correlated to continuing or pursuing higher office and positively related to retiring.

LSTVTPCT is the percentage of the vote that the incumbent received in the last election and is a proxy for his expected strength in the next election. If the incumbent had a "tough test" in the last election, he is more likely to have difficulty in the next election as well. Thus, one would expect LSTVTPCT to be positively correlated with the decision to continue and negatively correlated with the decision to retire.

PRESIDENT is a binary variable that signifies whether or not the House member and the president are of the same party in period t. If a party controls the presidency, its representatives have a substantive advantage over members of the other party when trying to gain an appointed higher office within an administration--[P.sub.HH], and thus V([X.sub.t] = H), are higher.(12) Another way the executive branch influences moves to higher office can be seen in the value of alternative options if one loses a race for elective higher office. The party in control of the presidency has more jobs to offer to losing candidates as consolation prizes.(13)

OPENSEAT is a binary variable that signifies whether the representative was in a state that had an open seat for the Senate in the next period.(14) An open seat would represent a relatively low-cost opportunity to pursue a higher office. Thus, an open seat would encourage moves to higher office and discourage continuing in the House or retiring.

PARTYPCT, LEADER, COMMITTEE and RWAGE are components of the value or power of the seat occupied by a House member. PARTYPCT is the ratio of the number of House members in the representative's party to the total number of House members in period t. It represents the relative power of the member's party which, for individual representatives, translates into more prestige and greater ability to get legislation passed. LEADER is a binary variable that measures whether a representative is a member of "the leadership."(15) COMMITTEE is a binary variable that denotes whether or not a representative is a member of the Appropriations, Rules, or Ways and Means committees.(16) One would expect that PARTYPCT, LEADER and COMMITTEE will be positively related to continuing and inversely related to retirements and the pursuit of higher office.(17)

RWAGE is the real wage (in $1000s) that House members were paid in period t. With a real wage increase, one would expect continuing in the House to become more attractive while retiring would become less desirable. Since Senate salaries generally parallel House salaries and since moves to the Senate constitute the majority of attempts at higher office, the effect on pursuing any higher office is not expected to be significant.

The tenure variables (TENURE, TENURESQ) affect the accumulation of brand name capital. Normal exposure through the media will allow a House member increased name recognition over the course of his tenure--brand name capital acquired at no direct cost.(18) Longer tenure enhances an incumbent's name recognition and his stock of brand name capital. And although additions to brand name recognition will have a positive effect, these investments--whether by paid political advertising or by normal media coverage--are expected to exhibit diminishing marginal returns (thus, the use of TENURESQ).(19) Ceteris paribus, since more brand name capital lowers the cost of pursuing another term and higher office, we expect these alternatives to be more likely while diminishing the likelihood of retirements.

The variable DISTRICT is the number of districts in the representative's state. Even though all members of the House represent approximately the same number of people, those in lower population states represent a higher proportion of that state's population and have, therefore, greater statewide name recognition. For example, the lone representative from Alaska would bring much more statewide brand name capital to a Senate campaign than his colleagues from Texas, each of whom is just one of many. One would expect DISTRICT to be inversely related to the probability that a representative will pursue higher office. REDIST is a binary variable which notes whether the representative's decision might have been affected by redistricting.(20)

STARTAGE is a variable which measures the age of the representative when he first served in the House.(21) As discussed above, because of the opportunity costs of retirement, one would expect that representatives would be less likely to continue and pursue higher office as they get older. In addition, the fixed cost component of the initial investment to pursue higher office should further reduce the likelihood of moves to higher office as the agent gets older.

The dependent variable in the model can take three values: one each for the choices to continue, to pursue higher office or to retire.(22) The sample is comprised of observations for all returning House members from 1960-1990. The initial date of the sample period was somewhat arbitrary; consideration was shown to counter concerns about a sufficiently long sample period and getting enough years when a Democrat served as President.

IV. RESULTS

Table II reports the estimates of the coefficients for the variables listed in equation (12). The model as a whole is highly significant, while most of the variables of interest have high degrees of significance and the expected signs. As expected, the coefficients for the quality of an incumbent's brand name, SCANDAL and WATERGATE are negative and highly significant for both continuing and pursuing higher office. Scandals are cited as a frequent cause for reelection woes, and clearly most scandals would preclude ambitious House members from pursuing higher offices. Consequently, scandals are a strong predictor of the choice to retire.

With respect to the expected quality of a challenger, LSTVTPCT has the expected sign and is significant. Thus, as the difficulty of the test that an incumbent faced in the last election cycle increased, he is less likely to continue and more likely to retire in the current election cycle. While the effect on continuing was expected, the impact on higher office was both stronger and more significant than expected. A likely explanation is that strong, albeit secure, candidates have popularity outside their district and are thus also likely to pursue higher office relative to retiring.

The opportunities for a House incumbent to pursue higher office are clearly important; PRESIDENT and OPENSEAT both have the expected signs and are statistically significant. Thus, if the president and the House member are from the same party, the incumbent has more opportunities to achieve higher office ex ante and ex post (in case he fails to be elected to higher office). And if there is an open Senate seat available, the probability of success in that endeavor increases and, thus, a representative is more likely to leave his House seat.

With respect to the value of the seat that the representative holds, only PARTYPCT (for the decision to pursue higher office) and RWAGE are measured with much precision; LEADER and COMMITTEE are both insignificant with respect to continuing and higher office. In some sense, the explanatory strength of RWAGE and weakness of the other variables are surprising since many of the benefits to holding office are generally considered to be non-pecuniary in nature--power, perquisites and the enhanced value of post-House employment opportunities. These results seem to indicate that House members value pay over power.

Turning to the quantity of brand name capital, both tenure variables have the right sign for the decision to pursue higher office, although only TENURESQ is statistically significant. This gives some evidence for diminishing marginal returns to accumulating brand name capital and reputation. Surprisingly, the coefficients for the choice to continue are the wrong sign. One explanation of this lies in the possible collinearity between tenure and other variables. It could be that relatively strong effects from other variables are outweighing the relatively subtle effects of tenure on brand name.

The coefficients for DISTRICT point to representatives from less populous states being more likely to take advantage of their relatively large statewide recognition by pursuing higher office. Another application of this is that a term limit law would be relatively less attractive to representatives of more populous states. Since they tend to stay in the House longer, it is more likely that their career paths would be constrained by term limits. REDIST is negative for the decisions to continue and to pursue higher office. The former was expected since reelection suddenly becomes a more difficult prospect. The latter was unexpected; perhaps one could tell a story that weaker representatives are more likely to be the "victims" of redistricting efforts and so would be the least likely to pursue higher office. Finally, STARTAGE is negative and highly significant as expected. In addition, the coefficient for the decision to pursue higher office is larger than the coefficient for continuing, indicating that starting age affects decisions to pursue higher office more strongly than it affects the decision to continue.

TABLE II

Coefficients from Multinomial Logit Analysis of the Determinants of Voluntary
Departures from the U.S. House

 ([X.sub.t] = H) ([X.sub.t] = C)
 Higher Office Continue

SCANDAL -1.020(**) -1.010(**)
 (3.31) (8.73)
WATERGATE -.980(**) -.810(**)
 (4.39) (6.86)
LSTVTPCT .024(**) .018(**)
 (3.61) (3.94)
PRESIDENT .153(*) -.031
 (1.68) (.51)
OPENSEAT .567(**) -.113
 (5.45) (1.41)
PARTYPCT -1.580(*) .466
 (1.95) (.83)
LEADER .065 -.070
 (.22) (.43)
COMMITTEE .012 .037
 (.11) (.54)
RWAGE .121(**) .128(**)
 (6.21) (9.95)
TENURE .050 -.252(**)
 (.51) (6.27)
TENURESQ -.029(**) .004(*)
 (3.51) (1.97)
DISTRICT -.024(**) .005
 (3.15) (1.08)
REDIST -.232(*) -.154(*)
 (2.10) (2.13)
STARTAGE -.098(**) -.034(**)
 (8.86) (4.96)

Note: t-statistics are in parentheses below the coefficients.
* = significant at the .05 level.
** = significant at the .01 level.
Number of observations = 6410.

In addition to supporting most of the earlier hypotheses, these results also suggest an explanation of why Republicans have been more likely to take or pursue higher office. Since Republican members of the House have been the minority party since 1954, it has been more likely that they would leave since they have had less power and a lower probability of ever seeing majority status relative to their Democratic colleagues (PARTYPCT and thus [V.sub.t] = C have been lower). This may have led many Republican House members to view the House as simply a stepping stone to higher office.(23) In addition, with recent history showing a Republican hold on the presidency, Republicans have been more likely to pursue higher office because the available opportunities were greater (PRESIDENT and thus [V.sub.t] = H have been greater). Recall that the Republican tendency to pursue higher office was stronger after the end of the 90th Congress (1968), the beginning of a period that has seen almost complete Republican domination of the presidency.(24) Further, Zupan's [1991a] hypothesis provides compelling evidence that the past trend of Republican domination in the executive branch is likely to continue well into the future.(25)

What do these results say about the possibility of a future Republican majority in the House? It's even less likely than previously thought. It is widely held that it will be difficult for the Republicans to achieve majority status because of the disadvantages that challengers face in displacing incumbents during election bids. Now, in addition, we have found that Republicans are more likely to voluntarily leave the House to pursue higher office. A prospective Republican majority in the lower chamber would have to overcome the disadvantages of unseating incumbents as well as the tendency for their own members to leave to pursue higher office.

This study fills a void in the existing literature by analyzing the quit behavior of U.S. House members. As expected, agents in this labor market respond to the costs and benefits of the decisions to retire, to pursue higher office and to remain in their House seats.

1. Most of the literature on the congressional labor market has looked at factors that influence the reelection prospects of incumbents in political office. In a vast literature, see Ferejohn's [1977] survey article, Collie [1981], Jacobson [1978; 1985], Lott [1987], Wittman [1989], Hart and Munger [1989], Ornstein et al. [1990], Grossman and Noh [1990] and Zupan [1991a].

2. See Black [1972], Rohde [1979], Copeland [1989] and Kiewiet and Zeng [1994]. In addition, Cooper and West [1981] Hibbing [1982a; 1982b], Gilmour and Rothstein [1993] and Groseclose and Krehbiel [1994] discuss trends and motivations behind retirements.

3. In their theoretical [1990] and empirical work [1992; 1994], Reed and Schansberg also differentiate between the reasons for leaving so that hypotheses concerning trends in congressional tenure can be eliminated as possible explanations.

4. Since it is exogenous and relatively infrequent, death will not be modelled here.

5. Observations were rounded off to the nearest two-year term. For example, if someone retired after the middle of a term, he would be credited with serving the current term and then retirement in the next term.

6. The interested reader is again referred to Ferejohn [1977], Lott [1987] and Ornstein et al. [1990]. Note also that incumbents have an overwhelming edge in acquired name recognition over challengers because of past service, i.e., exposure through the media over the course of their tenure.

7. Note that [p.sub.cc] is the reelection rate and [p.sub.CR] is the probability of defeat in a reelection bid. Similarly, the other p's deal with the probability of success (and failure) when [X.sub.t] = H, that is, the representative decides to pursue a higher office. These p's vary by individual and over time, but time and individual subscripts are eliminated because they don't add measurably to the analysis and would clutter the equations considerably. With [X.sub.t] = H, the value of the alternative W(S.sub.t+1] = R), where the representative has resigned to pursue other goals, is augmented by the expected value (including the probability and valuation) of consolation prizes offered to ambitions but failed candidates. This option is assumed not to be open to individuals who merely pursue another term in the House. Finally, W([S.sub.t+1] = H) may include future aspirations for even higher office.

8. For [p.sub.ij] [is less than] 1, the partial derivatives of equations (1)-(3) with respect to the parameters in the model have the obvious signs. Otherwise, the partial derivatives may equal zero.

9. The use of an infinite-time model is solely for ease of notation. The theoretical results and expectations for specific variables mentioned above remain unchanged.

10. See Maddala [1983]. Another possibility would be to use simple or conditional logits on the decision to retire or not and then--conditional on not retiring--the decision to run for the House seat or for higher office. However, this seems too restrictive. For example, if one considers a representative who is a "victim of redistricting," continuing may not be a viable option--the choice might be between retiring and pursuing higher office.

11. The "scandalous" observations were compiled from mention in Congressional Quarterly Almanac. With respect to the use of "two election cycles," I assume that if the controversy failed to unseat the incumbent within two elections, then it would not affect a representative's future decisions.

12. These non-elective higher offices include cabinet positions under the president, judgeships, diplomatic positions, etc. Crossing party lines to make administrative appointments is very rare.

13. A recent example of this was observed in the 1990 Illinois Senate race between incumbent Paul Simon and Republican challenger Lynn Martin. After Martin failed in her attempt to unseat Simon, she was offered the post of Secretary of Labor.

14. Other open seats, such as for governor, are rarely pursued by House members.

15. The eight offices specially designated by Congressional Quarterly are considered to be leadership posts.

16. Crain, Leavens and Tollison [1986] note that the majority party typically "stacks" the ratio of majority to minority members on the most important committees. Although this puts minority members at a disadvantage, their pursuit of placements on those committees reflects a revealed preference. The author also recognizes that rent-generating activity in particular committees might be just as useful to individual representatives. But, presumably, pork barrel opportunities exist for all committee assignments or are a function of tenure or party.

17. The author recognizes that changes in PARTYPCT along different parts of the range could have disparate impacts on moves to higher office. For instance, movement around a value of .5 is potentially more significant than changes around .6 since majority party status is at stake. However, this variable will to some extent pick up expectations of future party control and, thus, future power for individual members. As a party gains security in majority status, expectations of future power are higher and we would expect its members to leave less frequently. Further, assignments in the important committees are usually made very early in a House member's career. Thus, the problem of measuring expectations of power with respect to committee assignments is mostly avoided.

18. Acquiring brand name capital at "no cost" ignores the psychic costs associating with squelching ideological preferences. See Lott and Davis [1992] for a survey of the relevant literature.

19. One could also tell a sorting story here--that the probability of continuing will increase at a decreasing rate as constituents sort their representatives by ideological preferences over their tenures.

20. REDIST=1 for the years 1962, 1972 and 1982. Ideally, one would like to have more specific data about the consequences of redistricting. What percentage of his constituency is new? What percentage of his new constituents belong to the other party? Unfortunately, such data are not available.

21. Actual age would be preferred but was highly collinear with TENURE.

22. Observations considered to be "higher office" include a few cases where the attempt was not immediate but within two years of leaving the House. Questions about what constitutes "higher office" are answered by assuming that revealed preference theory is operative. Note also that the valuation of an office by a representative is relative.

23. See also: Gilmour and Rothstein [1993]. An observationally equivalent explanation is that Republicans behave differently because they are Republicans and not because they are the minority party per se. It could be that Republicans have a comparative advantage in issue stances over local benefit-seeking, which is more important in the Senate (see Zupan [1991b]).

24. In fact, the House members whose party was in control of the executive office have accounted for 61 percent of the moves to higher office since 1960.

25. Zupan provides a Prisoner's Dilemma explanation for the phenomenon that Democrats dominate legislative bodies (many members) while Republicans dominate executive offices (one member) at all levels of government.

REFERENCES

Black, G. "A Theory of Political Ambition: Career Choices and the Role of Structural Incentives." American Political Science Review, March 1972, 144-59.

Collie, M. "Incumbency, Electoral Safety, and Turnover in the U.S. House of Representatives, 1952-1976." American Political Science Review, March 1981, 119-31.

Congressional Quarterly Inc. Congressional Quarterly Almanac. Washington, D.C., various years.

Cooper, J., and W. West. "Voluntary Retirement, Incumbency, and the Modern House." Political Science Quarterly, Summer 1981, 279-300.

Copeland, G. "Choosing to Run: Why House Members Seek Election to the Senate." Legislative Studies Quarterly, November 1989, 549-65.

Crain, W., D. Leavens, and R. Tollison. "Final Voting in Legislatures." American Economic Review, September 1986, 833-41.

Ferejohn, J. "On the Decline of Competition in Congressional Elections." American Political Science Review, March 1977, 166-76.

Gilmour, J., and P. Rothstein. "Early Republican Retirement: A Cause of Democratic Dominance in the House of Representatives." Legislative Studies Quarterly, August 1993, 345-65.

Groseclose, T., and K. Krehbiel. "Golden Parachutes, Rubber Checks, and Strategic Retirements from the 102nd House." American Journal of Political Science, February 1994, 75-99.

Grossman, H., and S. Noh. "A Theory of Kleptocracy with Probabilistic Survival and Reputation." Economics and Politics, July 1990, 157-72.

Hart, D., and M. Munger. "Declining Electoral Competition in the House of Representatives: The Differential Impact of Improved Transportation Technology." Public Choice, June 1989, 217-28.

Hibbing, J. "Voluntary Retirement from the U.S. House of Representatives: Who Quits?" American Journal of Political Science, February 1982a, 467-84.

-----. Choosing to Leave: Voluntary Retirement from the U.S. House of Representatives. Washington, D.C.: University Press of America, 1982b.

Jacobson, G. "The Effects of Campaign Spending in Congressional Elections." American Political Science Review, June 1978, 469-91.

-----. "Money and Votes Reconsidered: Congressional Elections, 1978-1982." Public Choice, 47(1), January 1985, 7-62.

Kiewiet, D., and L. Zeng. "An Analysis of Congressional Career Decisions 1947-1986." American Political Science Review, forthcoming, 1994.

Lott, J. "The Effect of Nontransferable Property Rights on the Efficiency of Political Markets." Journal of Public Economics, March 1987, 231-46.

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

Maddala, G. Limited Dependent and Qualitative Variables in Econometrics. Cambridge: Cambridge University Press, 1983.

Ornstein, N., T. Mann, and M. Malbin. Vital Statistics on Congress: 1989-1990. Washington, D.C.: Congressional Quarterly Inc., 1990.

Reed, W., and D. Schansberg. "How Long Do Congressmen Stay in Office?" Economics and Politics, July 1990, 173-92.

-----. "The Behavior of Congressional Tenure over Time: 1953-1991." Public Choice, 73(2), March 1992, 183-203.

-----. "An Analysis of Congressional Term Limits." Economic Inquiry, January 1994, 79-91.

Rohde, D. "Risk-Bearing and Progressive Ambition: The Case of Members of the U.S. House of Representatives." American Journal of political Science, February 1979, 1-26.

Schlesinger, J. Ambition and Politics. Chicago: Rand McNally, 1966.

Wittman, D. "Why Democracies Produce Efficient Results." Journal of Political Economy, December 1989, 1395-424.

Zupan, M. "An Economic Explanation for the Existence and Nature of Political Ticket Splitting." Journal of Law and Economics, October 1991a, 343-69.

-----. "Local Benefit-Seeking and National Policymaking: Democrats vs. Republicans in the Legislature." Public Choice, 68(1), January 1991b, 245-58.