Congressmen in Exile: The Politics andConsequences of Involuntary Committee Removal

Justin Grimmer Stanford University

Eleanor Neff Powell Yale University

We show how preferred committee assignments act as an electoral subsidy for members of Congress—empoweringrepresentatives’ legislative careers. When holding preferred assignments, legislators are free to focus on legislativeactivity in Washington, DC. But when the subsidy is removed, legislators are forced to direct attention to thedistrict. To test our theory of legislative subsidy, we exploit committee exile—the involuntary removal of committeemembers after a party loses a sizable number of seats. Legislators are selected for exile using members’ rank onthe committee, causing exiled and remaining legislators to appear strikingly similar. Using exile, we show that ithas only limited electoral consequences, but this is partly due to compensatory efforts. Exiled legislators shift attentionaway from Washington and towards the district: they raise and spend more money for reelection, author lesslegislation, are absent for more days of voting, and vote with their party less often.

Congressional committees, as Fenno (1973)famously contends, have diverse purposesand allow members to pursue a variety of

career goals. We show how congressional committees,across these varying types, act as an electoral subsidythereby empowering legislators to pursue policy-focused careers in Washington, DC, in place ofelectorally focused careers in the district. With prestigeor preferential committee assignments, legislators focusmore of their attention on politics in Washington—both in their participation in the institution and theirsupport for the party. In the absence of those com-mittee assignments, legislators engage in a compensa-tory effort, returning their focus to the district andcultivating electoral support.

We demonstrate how committees subsidize policywork using new estimates of the effect of committeeson legislative behavior. To construct the estimates, weexploit a regularly exercised—though rarely studied—Congressional institution: committee exile. When newmajorities arrive in Washington, they reapportion seatson a committee to favor the new party. Becausecommittee sizes are often fixed and losses are unevenlydistributed across committees, this forces the minorityto remove legislators from committees. We exploitcommittee exile to create a robust panel research design

to measure the effects of congressional committees thatmitigates both measured and unmeasured confoundingfound in cross-sectional designs.1

To explain why committee exile is a useful toolfor estimating committee effects, we first study theincidence and characteristics of committee exile. Weshow that committee exile is a consequence of unevenlosses across congressional committees and not dueto strategic targeting by the new majority. Bothcongressional parties follow strict norms of seniorityto determine who to exile from the committee, thus thelowest-ranked legislators are the most likely to beremoved. The result is that exiled legislators have servedless time in Congress than those who remain on thecommittees, but they are otherwise strikingly similar tothe remaining legislators. After exile, representatives aresometimes deposited on lower prestige committeesbut more often are given no compensatory committeeassignment.

Building a robust panel research design aroundexile, we show that committee exile has few electoralconsequences but substantial legislative consequences.After exile, a legislator’s party receives the same level ofsupport in the subsequent election, and the exiledlegislator is no more likely to retire. But this lack ofelectoral consequences is due, in part, to legislators

The Journal of Politics, Vol. 0, No. 0, XXX 2013, Pp. 1–14 doi:10.1017/S0022381613000704

� Southern Political Science Association, 2013 ISSN 0022-3816

1Replication data is available at justingrimmer.org and eleanorneffpowell.com.

1

shifting their priorities away from legislative work inCongress and towards electoral considerations in thedistrict. Exiled legislators have greater campaign ex-penditures for their reelection efforts and, in turn,raise more money to support those reelection efforts.Perhaps due to their increased fundraising activities,exiled representatives participate less in Congress. Theexiled legislators author fewer pieces of legislation andmiss more days when Congress is in session, andmarginal legislators vote less often with their party.Further, the effects endure: they are felt both imme-diately and in the subsequent Congress.

We caution that preferential committee assign-ments’ lack of an electoral effect does not imply thatcommittees are not beneficial electorally. Rather, ourresults show how committee assignments allow stra-tegic legislators to invest their effort across theirdiverse goals. With preferred assignments, legislatorsare free to invest time in cultivating a Washingtoncareer. But in the absence of those preferred assign-ments, legislators compensate with increased attentionin the district. The result is that strategic legislators areable to maintain their electoral support—their primarygoal—but at a cost to their broader goals of policyinfluence in Washington. While evidence of this com-pensatory effort to overcome electoral effects is new tothe literature on committees, it is consistent with recentstudies that have exploited institutionalized random-izations to study legislative committees (Broockmanand Butler 2011; Kellermann and Shepsle 2009).

To demonstrate the subsidizing effect of com-mittees, we introduce a new approach to studying theeffects of committees—an approach that we showprovides unusually good leverage on the value ofcommittee assignments. Our approach removes con-cerns that limit inferences that can be made fromcross-sectional studies of committee effects—in par-ticular matching estimators. And the usefulness ofexile as an identification strategy will only increase inthe coming Congresses. With large-scale swings inHouse membership over the most recent congressio-nal elections, committee exile is an increasinglycommon phenomenon. To that end, we provideguidance on how to use exile data, the trade-offsthat are made in using this identification strategy,and useful specification strategies. Together, thismethodological contribution provides a strategy tobegin addressing long-standing questions on howcommittee memberships affect representation inCongress.

We begin with an explanation of why exile is sucha useful tool for limiting confounding in the study ofcommittee assignment effects.

Institutional Design: CommitteeAssignment Process

After a defeat in a congressional election, the outgoingmajority party is forced to relinquish power in a varietyof ways. One of the most frustrating for returningincumbents is the loss of a valued committee assign-ment, which can occur when electoral losses are spreadunevenly across committees. In addition to losing itsmajority status, a party loses at least a proportionalnumber of seats on every committee as committeeratios are adjusted to reflect the new majority’s seatadvantage. This causes some legislators to lose theircommittee seats—or to be exiled—because electorallosses are unevenly distributed across committees.

To better understand how legislators are exiledfrom committees, we first review how seats oncommittees are determined.2 The committee assign-ment process for any new congress begins with thecommittee assignments and party ratios of the pre-vious congress. Following the election, both new andreturning members submit committee (and transfer)requests. Before any assignments can be made, however,the majority and minority party leaders must negotiatethe committee sizes and party ratios for each commit-tee.3 Once the committee sizes and ratios are set, theassignment process by each party’s Steering Committeetakes place, and finally the slates are approved by theparty caucus and eventually the full House. (Wesummarize the process and provide additional detailsin the online supplemental information).

While technically each party’s committee assign-ment slates are subject to votes at both the caucusand chamber level, they are almost always upheld.Therefore, the decisions party leaders face in thecommittee-assignment process are of the greatestinterest to us here. Both congressional parties employa deference or seniority system approach to commit-tee assignments. Once a member has received anassignment on a committee, it is assumed that shewill have the option to continue serving on it.4 Thereis, however, an important and greatly understudied

2The committee assignment process itself has been the subject ofconsiderable academic study; see Frisch and Kelly (2006) for anoverview of the literature. See the supplemental information forfurther discussion.

3The House Committee on Standard of Official Conduct is thelone exception: House Rules guarantee both parties an equalnumber of seats.

4This is a strong norm in both parties but not a rule formalized inthe Republican Party Conference Rules in the 112th Congress(Conference 2010).

2 justin grimmer and eleanor neff powell

exception to that rule: when there are an insufficientnumber of slots to accommodate the returning mem-bers.5 We call this process—the removal of a committeeassignment due to electoral losses—exile.6

Once electoral losses have occurred and the needfor exile arises, the party must select whom to exile fromthe committees. In deciding whom to exile, both partiesemploy a seniority system removing the least seniorcommittee members (those with the fewest terms servedon the committee). Contrary to other areas of in-stitutional and party politics where seniority normshave eroded, this strict seniority system continuestoday.7 While this aberration in the trend towarda weakened seniority system is worthy of study in itsown right, we can take advantage of this application ofa seniority rule in the process of committee exile to gaincausal leverage on how committees affect representation.

Our strategy for leveraging exile relies heavily onthis seniority rule. Given our reliance upon it, it isnatural to ask if the seniority rule reflects some otherprocess that might undermine the usefulness ofcommittee exile—such as compensation with othercommittee assignments or consideration of pet legis-lation. But a more benign logic explains the persis-tence of the seniority norm (at least for ourpurposes): maintenance of comity within the partisancaucus. For example, after one of the largest exilewaves following the 1994 Republican Revolution,Carney summarized the problem facing Democraticleaders: ‘‘Whatever they decide, incoming Demo-cratic leaders will be hard-pressed not to alienatefellow Democrats fighting over the shrinking com-mittee pie . . . Any attempt to consider factors otherthan seniority on Appropriations could prove partic-ularly explosive. ‘It would be very, very ugly,’ a Dem-ocratic House aide said. ‘It would be a real bloodbath

to start throwing people off the committee who aremore senior in favor of people who are more junior’’’(1994, 2916). In an interview we detail in the supple-mental information, Rep. Rosa DeLauro (D-CT)—co-chair of the Democratic committee that determinescommittee assignments—stated that the seniority rulehelped Minority Leader Nancy Pelosi (D-CA) explainthe loss of seats and reassure exiles. DeLauro said that‘‘Pelosi could hold out is that we have precedent. Shecould say ‘look this is what we did in the past, peoplecame off in order of seniority’.’’ Thus for the minority,the persistence of the seniority rule may largely beabout preserving comity within the party duringa particularly challenging period for party leaders.

We identified all instances of involuntary com-mittee reassignment from the 80th Congress to thepresent, which resulted in 230 cases. We identifiedcases of exile by beginning with Nelson (2011) andStewart and Woon’s (2011) databases of congressionalcommittee assignments. We first identified every casein which a member left a previously held committeeassignment. We then examined each of these casesindividually to determine whether the member left dueto a shortage of party seats created by electoral losses.The exiles are exclusively the result of reduced com-mittee sizes or surplus members of a party sitting ona committee after substantial electoral losses, poten-tially compounded by a reduction in committee size.8

The observed pattern of committee exile alignswith the stated rule. If we rank copartisan legislators ona committee from 0 (the most senior) to 1 (the mostjunior), we find that exiled legislators have substantiallyless tenure on the committee (0.24 units higher),a difference extremely unlikely under the null ofrandomly selecting legislators for exile (p-value � 0).Given the strong evidence of selection, it also is notsurprising that exiled legislators have a much shortertenure in the institution (3.7 fewer years, 95% confi-dence interval [-5.04, -2.36]). In the supplementaryinformation, we provide detailed examples of how exileoperates in practice.

Patterns of Exile

To better characterize the aggregate patterns of exile,in this section we examine more systematically when

5As Stewart explains, ‘‘Thus, over the past century, a type ofproperty right in committee assignments has emerged in bothchambers–members are allowed to hold onto their committeeseats from one Congress to the next and may not be removedunless the party ratios change so dramatically between Congressesthat junior members of the minority party find their seatsabolished altogether’’ (2001, 299–300). While the term ‘‘propertyright’’ is frequently used, exceptions and violations includingexile suggest that it is more of a behavioral regularity.

6While there are a handful of rare cases in which a member maylose his or her seat due to, say, a felony conviction, we confineour attention here to those following an exogenous (most likelyelectoral) induced change in committee ratios and when there isa surplus of return legislators to a particular committee.

7It is worth noting that in our data the seniority exile norms arenever violated for what have traditionally been viewed as the twomost powerful and desirable congressional committees: Appro-priations, and Ways and Means.

8Inevitably, this process required making some assumptions,which probably are most strongly supported for exclusive andprestige committees that members never voluntarily leave. Anymisclassified cases (voluntary departures) bias our results againstour findings.

congressmen in exile 3

and how exile occurs in congressional committees.Committee exile occurs under both Democratic- andRepublican-controlled congresses and, as we expectgiven the electoral circumstances that lead to exile,members of the minority party make up the vastmajority of cases. While Democrats controlled thechamber, 79% of exile cases were Republicans, andwhile Republicans controlled the chamber, all exileswere Democrats.

The temporal patterns in committee exile furtherdemonstrates that majority parties rarely exile theirown members. Figure 1 below shows the partisan-exile breakdown over time beginning with the 81stCongress at the far left and moving forward towardthe 112th Congress at the far right, omittingCongresses where no exile occurred. The bulk ofexile cases occur after a major wave election thatchanged control of the chamber—such as the104th, 110th, and 112th Congresses. The few casesof majority parties exiling their own membersoccurs during the long period of Democrat dom-inance in the House: during the 90th, 97th, and99th Congresses.9 The exile of majority-party legis-lators from committees has ceased: the most recentof these majority exile cases occurred in the 99thCongress immediately following the 1984 Congres-sional elections.

Committee exiles, therefore, are primarily con-centrated among the new minority and create vari-ation in who belongs to committees. But to use exileas a strategy to identify the effect of committees, weneed to demonstrate that leaders of the new majorityare not strategically targeting committees to removeminority members. The possibility for manipulationarises if the new majority manipulates the party ratiosor committee size to force minority members to beremoved. Evidence against this strategic manipula-tion would be party ratios that remain relatively fixedfrom Congress to Congress. But if there is strategicmanipulation, then we would expect large biases infavor of the new majority.

Figure 2 presents the majority-party bias on eachcommittee from the 80th to the 112th Congress. Amajority bias of zero (along the solid horizontal line)indicates that the partisan balance of the committeeexactly reflected the partisan balance of the chamber,while a positive majority bias indicates a majority-partycushion above the chamber balance, and a negative biasindicates a rare minority-party cushion. The dashed

vertical lines indicate a party takeover year, in whichthere was a change in control of the chamber. Thebottom row of graphs represent the prestige commit-tees, which are those that have been traditionallyconsidered by scholars to be the most desirablecommittee assignments. Consistent with the committeeliterature, we can see that committees that areessential to the majority’s control of the chamber,such as the Rules Committee, maintain a largepositive majority-party bias over time, while others,such as the Science, Space and Technology Com-mittee are consistently more faithful to the partisanbalance in the chamber.

Most relevant for our understanding of commit-tee exile is the stability of the majority-party biasduring the wave election years (alternations incontrol of the chamber), which generate most ofour exile cases. If we look at the majority-party biasaround the party takeover years (dashed verticallines), we see that the majority-party bias is quitestable before and after takeover. This stability isreassuring for the purposes of research design: themajority party is not engaging in deliberate manip-ulations of committees to target individual mem-bers. Rather, the stability of the majority bias showsthat, contingent on the election results, committeeexile is largely exogenously determined by pastparty shares on committees.

While minority-party members do not appear tobe strategically targeted for exile, another possibilitythat could limit the usefulness of committee exile isthat exiled legislators could receive compensatory com-mittee assignments from the new minority. Figure 3shows there is little compensation offered: most exiledlegislators receive few additional, new committeeassignments. To demonstrate this, we present inthe bottom left-hand figure the exile patterns—where the exiles from a committee (rows) are sent(columns). For example, the shaded box in the lowerright-hand corner of the figure represents exiledmembers of the Agriculture Committee who re-ceived no compensating assignment. The darker thecell, the larger the share of exiles. To help in theinterpretation of this distribution of committeeassignments, the bottom right-hand plot shows thetotal number of exiles from each committee, whilethe top-most plot sums over committees and showsthat the modal outcome is no compensation.

Across committee assignments, the bottom left-hand plot in Figure 3 shows that legislators receiveonly marginal compensation after leaving a com-mittee. This is even true for the most prestigiouscommittee assignments. For example, the modal

9These elections involved majority party (Democratic) losses of48, 33, and 15 seats.

4 justin grimmer and eleanor neff powell

FIGURE 2 Exile Is Not the Result of Strategic Committee Manipulation

Note: This figure shows that parties rarely manipulate party ratios on committees to induce committee exiles. Each cell in the figure shows the proportion of majority party members on a committee, less the proportion of majority party members in the House. Positive numbers are indicative of a promajority bias, negative numbers are evidence of an antimajority bias, and zero indicates no bias. On the whole, there is not a deviation in the majority party’s share on a committee—there is little evidence new majorities are intentionally creating exiles.

FIGURE 1 Exile Is Most Prevalent after a Wave Election

Note: This figure shows the number and the party of exiled legislators, and the Congress number of the legislators that were exiled. Members of the new minority are the most likely to be exiled.

congressmen in exile 5

post-exile committee assignment for a representa-tive from Ways & Means was a seat on Oversight.Similar compensation patterns in which membersreceive what might be perceived as a less desirablecommittee assignment can be seen on most of the

remaining committees. In the supplemental infor-mation, we directly measure the lack of compen-sation. We use measures of committee transfersfrom Stewart (2012), based on the method intro-duced in Groseclose and Stewart (1998), to show

FIGURE 3 Exiles Receive Little Compensation after They Are Removed

Note: This figure shows where representatives go after their exile from committees. The right-hand figure is a histogram that counts the number of exiles from committees—it shows that with a handful of exceptions, exile is roughly evenly distributed across committees. The bottom left-hand figure shows where a committee’s exiles (rows) are sent (columns). The darker the cell, the more legislators sent there. Notice, most exiles from most committees receive no compensation. The top-most plot sums over committees and shows that the modal outcome is no compensation. On the whole, this shows that exiled legislators receive little, and often times no, compensation.

6 justin grimmer and eleanor neff powell

that exiled legislators are transferred to lower pres-tige committees.10 The top plot in Figure 3 aggregatesacross committees to show the dramatic lack of com-pensation: more often than not, exiles fail to receive anew committee assignment.

Not only do members receive relatively little tocompensate them for their exile, but the exile endures.Only about 11% of members ever return to serve ona committee from which they were exiled. In the onlinesupplemental information, we show that there is higherrate of return on prestige committees such as Appro-priations (9 out of 17 return) and Ways and Means(5 out of 13). But, using two case studies, we show inthe supplemental information that there is substantialuncertainty about whether an exiled legislator will beable to return to a committee, even among those bestplaced to return. Rosa DeLauro (D-CT), who wasexiled from Appropriations in the 104th Congress andreturned in the 105th Congress, told us in an interviewthat even with the assurances of the leadership, as anexiled member ‘‘you don’t know’’ if you’ll be able toreturn to the committee assignment.

Exiles’ uncertainty is about if and when they willreturn to their original committee. Exiles are unsure ifcurrent (and future) party leaders will honor promisesof return based on seniority. Even if legislators trustparty leaders about the order of return, exiles are stilluncertain about when a vacancy on the committee willoccur. With prestige and preferred committees popu-lated with well-entrenched copartisan incumbents, it ishard for exiles to forecast when a vacancy will occur.

Committees as Electoral Subsidy

Instrumental legislators have diverse and multifacetedgoals when deciding what to do when in Washington.While the primary goal of legislators is reelection(Mayhew 1974), legislators are also interested insecondary goals of developing good policy, advancingtheir careers, and even developing a reputation ofpolicy effectiveness (Fenno 1973). Committee assign-ments can simultaneously help members accomplishboth these primary and secondary goals, effectivelyserving as an electoral subsidy for campaign activity.

Scholars of congressional politics have often lookedto the ability of members to use their congressional

committee assignments to their electoral advantage.Committee work provides this boost in electoralprospects by providing legislators the opportunity todeliver policy and particularistic goods to their district.When legislators work on committees, they developexpertise in the area (Clapp 1963), which is a credibleand valuable signal to constituents that their represen-tative exerts influence on policy (Fenno 1973; Katz andSala 1996). Some committee assignments such asAgriculture or Armed Services allow members to signalexpertise and take public positions on issues of greatsalience to their constituents. While other committeeassignments, such as Ways and Means provide repre-sentatives with the capacity to develop nonpartisanbases of support through the delivery of particularisticgoods to the district.

We argue that these electoral effects of congres-sional committees effectively serve as an electoralsubsidy for campaign activity. When members reapelectoral benefits from congressional committee as-signments, those benefits free members to spend lesstime campaigning and instead more time legislating.In addition to the time and energy members savefrom campaigning, which can be spent legislatively onmember’s secondary goals, committees further facili-tate the accomplishment of legislative goals by servingas the primary venue for policymaking activity. Much,if not most, of that legislative action happens in con-gressional committees. Committees, then, are a naturalvenue where instrumental legislators will use their in-stitutional positions to pursue their diverse goals.

Other committee assignments allow legislators toexercise substantial policy influence—even if it isdifficult to use positions on the committees to bolsterelectoral support. While Appropriations, Agriculture,and Armed Services are helpful in a member’s questfor reelection, other committees might be more usefulfor members seeking influence within the chamber ora lobbying career after they retire. For example, Fennoargues that, ‘‘[t]he opportunity to achieve the threegoals [reelection, influence within the House, andgood public policy] varies widely among committees.House members, therefore, match their individualpatterns of aspiration to the diverse patterns of oppor-tunity presented by House committees’’ (1973, 1). Forthis reason, we might expect that once members losethe opportunities provided by a given committeeassignment they will alter their behavior.

When legislators sit on committees that are eitherprestigious or desired, it allows them to invest in theircareer in Washington. As former Rep. Tom Delay(R-TX) observed, Congress is filled with two types oflegislators: policy-focused legislators—members of

10A notable exception is on the Science, Space and TechnologyCommittee in which members receive surprisingly desirablecompensating committee assignments, though the small numberof exiles from the Space, Science and Technology Committeelimits the finding.

congressmen in exile 7

Congress who focus on their work in the institution—and district legislators—those legislators who focus onreelection and the concerns of the district (Draper2012). While legislators sit on prestigious com-mittees, they are able to pursue careers as policylegislators. But when legislators lose their committeeassignments, they lose the opportunity to pursuetheir career through their committee assignment.They also lose clout—the exiled legislators are nolonger able to exercise direct influence over thecontent of bills before they reach the floor. Theresult is that the legislators shift towards a districtand reelection focus.

Part of this district focus will manifest in howlegislators campaign. If more focused on reelection,we expect that legislators will raise and spend moremoney on their reelection effort. The district focuswill also limit their work in Washington. So, exiledlegislators should author fewer pieces of legislationand miss more days when Congress is in session. Afinal implication is that committee exile will affecthow legislators vote in Washington. If exiled legis-lators adopt a greater focus on the district, theyshould defer less to their party when casting roll-callvotes. This is particularly true when the district andparty preferences clash, implying marginal represen-tatives should be less likely to vote with their party.

An implication of a greater district focus after exileis that involuntary committee removal will have a limitedeffect on legislators’ electoral support. When legislatorsadopt a greater focus on their district after exile, theycompensate for losing their seat on the committee. Theresult is that exiled legislators will maintain levels ofsupport to those legislators who remain on the presti-gious committees. But it is worth emphasizing that weexpect that this sustained electoral support comes at thecost of legislators shifting their career goals.

Deterministic AssignmentBased on Observables

We test our expectations by building a research designaround committee exile. The goal of our inference is tomeasure the effect of a committee assignment for thelegislators who are assigned to the committee: theaverage treatment effect on the treated or ATT.Adopting the now familiar potential outcome notation(Holland 1986), for each legislator i, treatment statusTi, and outcome Yi, we will suppose that a legislator hasa response under treatment, YiðTi ¼ 1Þ—assigned tothe committee—and control, YiðTi ¼ 0Þ—not assigned

to the committee. Our effect of interest is the averagetreatment effect among those who receive a committeeassignment, E½Yð1Þ � Yð0ÞjT ¼ 1�.

The primary challenge in estimating this effect isthat it is impossible to measure all the characteristicsthat affect committee assignments. Some characteristicsare easy to identify—legislators more senior, more loyalto the party, and those with greater fundraising prowessare more likely to obtain seats on preferred committees(Cox and McCubbins 1993; Currinder 2008). Butobtaining desired committee assignments also dependson political skill, the persuasion of party leaders, andthe demonstration of legislative capacity in a particulararea (Shepsle 1978). These characteristics are difficult tomeasure in observational studies. This is a particularproblem for matching estimators: as the balance im-proves between the treatment group—legislators whoobtain desirable committee assignments—and the con-trol group—legislators who fail to obtain the desirableassignments–the unobservable characteristics are likelyto remain and confound estimates of the treatmenteffects.

Committee exile provides a deterministic assign-ment mechanism for determining who remains onparticular committees. This deterministic selectionon observables is useful because it mitigates theunmeasured confounding. This mitigation occurs,in part, because all legislators—both those who areexiled and those that remain on committees—hadsufficient political acumen to obtain a seat on theprestige committee. The hope is that legislators whoremain on the committee and those that are removedfrom the committee are similar on unmeasuredcharacteristics, limiting their potential influence onthe estimation of the effect of committee assignmentson legislative behavior.

We build our modeling strategy around exile,comparing the change in exiles’ behavior before andafter committee removal to the change in behavior ofcopartisan members of the same committee over thesame time period. That is, we make our design morerobust by comparing exiled legislators’ behavior toexiled legislators’ copartisans who remain on thecommittee in the new Congress. We use only copar-tisans to avoid party-specific swings in support thatcan occur after a new majority arrives in Congress.11

And we restrict our sample to those who remain on

11By limiting analysis to copartisans and comparing, for example,Republicans exiled from a given committee to returning Repub-licans from that Committee, party-specific swings should impactboth groups (treated and untreated) and thus not affect ourresults.

8 justin grimmer and eleanor neff powell

committees to ensure that we are comparing legis-lators with similar political skills and interests—thosewith sufficient political acumen to obtain a seat onthe committee—limiting potential confounding fromdifficult-to-measure characteristics. To further miti-gate biases, all of our models include lagged measuresof exiled and nonexiled legislators’ behavior. Thisensures that we are examining how exile changeslegislators. We further mitigate this bias with addi-tional covariates and fixed effects for committee andyear.

Demonstrating that any research design providesbalance on unmeasured characteristics is, by definition,impossible. But an implication of balance on unmea-sured variables is that exiled and remaining legislatorsare similar on measured characteristics. Figure 4 showsthat this is the case—presenting the standardized differ-ences between exiled legislators and nonexiled legisla-tors of the same party as the exiled legislators (on thehorizontal axis) across several pre-exile characteristics.For characteristics not explicitly used to select legislatorsfor involuntary removal, both exiled and nonexiledlegislators are quite similar—this includes prior cam-paign spending, nominate scores, and support for thesame party presidential candidate in the district. The

similarity also extends to the pre-exile values of de-pendent variables that we will use in our analysis—exiled and nonexiled legislators had similar levels ofprior vote share, days missed in Washington, numberof bills sponsored, money raised, and party-unity score.

But exiled and remaining legislators are quitedifferent on characteristics that determine who isselected for exile. Consistent with the rules used todetermine exile, Figure 4 shows that legislators whohave a lower rank are much more likely to be exiled.This closely covaries with tenure in the institution, sonot surprisingly, exiled legislators have spent feweryears in Congress. This demonstrates the trade-offsthat must be made when using exile to study theeffects of committees. Because we include both tenureand relative rank on committees in our analyses, weare able to mitigate the bias that this induces.

Committee exile is useful, therefore, because itprovides a clear mechanism to identify who is allowedto remain on committees. Exile provides a determinis-tic assignment mechanism, but Figure 4 shows thatsome differences remain across exiled and nonexiledlegislators. To address this remaining imbalance, weuse lagged values of our dependent variable of interest,covariates that are potential confounders, and fixedeffects for committee and years. Specifically, for eachlegislator i, we model her response after exile Yi;1as,

Yi;1 ¼ b0 þ tExilei þ gYi;0 þ b0Xi

þ +j¼1

C

ajCommij þ +t¼1

T

htYearit þ eið1Þ

where Yi;0 is the lagged dependent variable, Xi is a vectorof covariates, Commij is an indicator of whetherlegislator i serves (or was exiled from) committee j,Yearit is an indicator of the year the legislator wasincluded, and ei is an error term. To reiterate, we applyequation (1) only to exiled legislators’ copartisans whoserved on the same committee in the prior Congress.Further, we include baseline characteristics Yi;0 to limitconfounding due to potential differences in legislators’baseline rates of performing particular activities. Wealso include in Xi a set of covariates that are potentialconfounders—such as the same-party presidential can-didate’s vote share in the district (District Partisanship)and spending in the prior election (Prev. CampaignExp.). We also include variables directly related to theselection of exiled legislators–including a legislator’srelative rank on the committee (Rel. Rank) and thenumber of years in the institution (Tenure). Note thatthis specification strategy makes use of all legislatorswho remain on the committee. In the supplemental

FIGURE 4 Exiled Legislators Are Similar to theCopartisans Who Remain onCommittee

Note: This figure shows the standardized difference in means across the pre-exile covariates used in this study. For most covariates, the exiled legislators are strikingly similar to the legislators who remain on the committees. But this is not true for the variables used to select the exiles: tenure and relative rank.

congressmen in exile 9

information, we show that a matched sample that usesonly the exiled and nonexiled legislators most similaron tenure and relative rank provides the same findings.

Using this robust research design and the specifi-cation in equation (1), we will use our estimate of t asthe effect of exile on the behavior of legislators—theeffect of losing a committee assignment. To summa-rize, our design has three steps to mitigate potentialbiases: (1) we use legislators removed due to exile,providing a clear and deterministic assignment basedon observables; (2) we restrict our sample to coparti-san members of the same committee prior to exile,who remain on the committee, ensuring both treat-ment and control group experience the same nationalpartisan swings; and (3) we include lagged values ofall our dependent variables to examine how exiledlegislators’ behavior changes relative to change innonexiled legislators’ behavior and regression toinclude additional covariates.12

Committee Exile and IncreasedDistrict Focus

Using the research design described in the previoussection, we examine the far-reaching effects of com-mittee exile on how members of Congress approachtheir job. First, we consider the electoral effects ofcommittee exile for the minority party. The left-handpanel of Figure 5 demonstrates the average effect ofinvoluntary removal on the exiled legislator’s party’selectoral support in the district. To measure thiseffect, we use equation (1) with the percentage pointsupport for legislators in the election after switchingto the minority as the dependent variable. Thebottom line shows the average effect of exile acrossall committees, while the remaining lines show thecommittee-specific effect, estimated using a multilevelmodel, as detailed in Gelman and Hill (2007). In thisplot and the subsequent plots, the points in each plot

represent the average effect of committee exile, whilethe thick and thin bars are 80% and 95% confidenceintervals, respectively.

The bottom line in the left-hand plot of Figure 5shows that committee exile has no real electoral effect.For districts with a representative exiled from a com-mittee, the minority party experiences a minute de-crease in vote share of about 0.55 percentage points,and this confidence interval overlaps substantially withzero (95% confidence interval, [-1.46, 2.54]).13 Theremaining lines show the committee-specific effects ofexile—these too exhibit substantively small and statis-tically insignificant differences from zero. Perhaps themost notable exception is the Agriculture committee—which is traditionally viewed as a reelection committeethat farming district representatives use to build sup-port with constituents (Fenno 1973).

The right-hand plot in Figure 5 shows the effectof committee exile on retirement rates overall andfor each of the committees. To compute these effects,we use equation (1) with an indicator of whether alegislator leaves the House as the dependent variableand estimate the overall effect with probit regressionand the committee-specific effects with a multilevelprobit regression. Similar to vote share, there is littleeffect of exile on retirement. Exiled representativesare only 0.1 percentage points more likely to retire,with a 95% confidence interval nearly centered onzero ([-0.08, 0.08]). The committee-specific effectsshow that there is a nearly uniform null effect acrosscommittees. And exile does little to affect whenlegislators decide to leave Congress. Exiled legislatorstend to leave only slightly earlier (hazard ratio, 1.14)and the 95% confidence intervals imply we cannotreject the null that exile has no effect on whenlegislators leave Congress ([0.76, 1.72]).

Committee membership not only affects legisla-tors support among voters, it also enables legislatorsto pursue legislative careers that may advance theirnonelectoral goals. The implication is that exile haslimited electoral consequences because it changeshow legislators invest their time in Washington. Afterexile, legislators become more focused on the districtand therefore have less time available to participate inthe policymaking process in Washington. One impli-cation of legislators focusing more on the district andreelection is that they should become more prolificfundraisers and, in turn, spend more on reelection.Figure 6 shows that this is the case. The topline in thisfigure shows the effect of exile on the amount

12As this article was written before the 2012 congressionalelections, we drop all exiles from the 112th Congress. As analternative design strategy, we could use legislators who arrive oncommittees, but this will induce severe biases. Legislators who arestriving to obtain a committee assignment are altering theirbehavior to obtain the assignment. Our case studies show thatthere is substantial uncertainty in who will return, therefore wedo not expect that this will pollute our results. For all analysesthat measure the effect of exile after the first Congress, we removeall legislators who return to the committee. As with all suchdesigns, we are making a trade-off in favor of internal validity byrestricting the population of legislators for whom our inferencesapply (Sekhon and Titiunik 2012).

13Tables that summarize the models used to produce the figuresin this section are found in the supplemental information.

10 justin grimmer and eleanor neff powell

nonretiring legislators spend on reelection. We useequation (1) with the total dollar amount spent onthe election as the dependent variable. On thehorizontal axis is the effect of exile on the amountspent (measured in thousands of dollars).

The top line of Figure 6 shows that exiled andnonretiring legislators spend about $111,750 more ontheir reelection effort (95% confidence interval[-8433, 221309]). This is a meaningful increase inexpenditures: the average candidate spent about$945,000 on their campaign. After exile, legislatorssubstantially increase the money spent to win theirelections.

To support this increased campaign spending, leg-islators increase the amount of money they fundraise—drawing on both individuals and political action com-mittees (PACs) to bolster their campaign funds. Wemeasure the effect of exile on fundraising usingequation (1) with the dollar amount raised as thedependent variable and estimate the model with leastsquares. The second line from the top of Figure 6shows that exiled legislators substantially increase

the amount of money they raise. Exile causes leg-islators to raise $153,822 more for their reelectionefforts (95% confidence interval [$27360, $293803]).

FIGURE 5 The Minimal Electoral Consequences of Committee Exile

Note: This figure shows the effect of exile on the minority party’s vote share in the subsequent election (left-hand plot) and retirement (right-hand plot). Exile has largely a minimal effect on vote share or retirement.

FIGURE 6 The Campaign FinancingConsequences of Committee Exile forNonretiring Legislators

Note: This figure shows that after exile legislators spend much more money (top line) and raise much more money to be spent in total (second line) from both individual (third line) and PAC (bottom line) donors.

congressmen in exile 11

The next two lines in Figure 6 show that the increasein fundraising comes from donations from bothindividuals and political action committees (PACs).This effect endures beyond the first Congress aftera legislator is exiled. In the second Congress afterbeing removed from the committee, exiled legislatorsraise an additional $110,790 in total (95% confidenceinterval [-27782, 249362]), $47,000 from individuals(95% confidence interval, [-34580, 128744]), and$53640 from PACs (95% confidence interval [6424,100850]).

Not only are legislators raising and spendingmore money, they are participating less in Washingtonpolitics. One manifestation of the decreased partici-pation is that exiled legislators author fewer pieces oflegislation. To assess the effect of exile on billintroduction, we use equation (1), with the numberof bills introduced in the Congress after exile as thedependent variable. To calculate the number of billsintroduced, we use the collection of bill introduc-tions from the Congressional Bills Project (Adlerand Wilkerson 2012). We estimate equation (1) witha poisson regression.

The top line in the left-hand plot of Figure 7 showsthat exile reduces the number of bills that legislatorsproduce. After exile, legislators author 2.3 fewer piecesof legislation (95% confidence interval, [-4.15, -1.05]).This reduction of two bills authored is large, relative tothe authorship rates in our sample, where legislatorsaverage 9.7 bill introductions each session. The im-mediate drop in the number of bills authored mayoccur as legislators adjust to their new committee

assignments. But the decrease in sponsorship contin-ues in the second Congress after exile. There, exiledlegislators author 1.4 fewer pieces of legislation (95%confidence interval, [-2.78, -0.01]).

Exiled legislators are also absent from Congressduring more days when it is in session. To show thatexiled legislators are absent more often, we use theroll-call voting record to measure absence fromWashington. Specifically, using the roll-call votingdata from www.voteview.com (Poole and Rosenthal1997), we calculate the number of days a legislator isabsent. If a legislator misses all roll-call votes on a day,we record her as absent. For each legislator we thencalculated the total number of days that she missed.We estimate the effect of exile on the total number ofdays missed using equation (1), which we estimatewith a poisson regression.

The bottom line in the left-hand plot of Figure 7shows that exiled legislators spend less time inWashington. After exile, legislators are absent anadditional 3.3 days of voting (95% confidence in-terval, [1.42, 7.03]). As with the number of bills in-troduced, this seemingly small difference is actuallyquite large when compared to baseline rates of ab-sence. On average, legislators only miss about 10 daysof Congress in a given session—so a three-day shiftconstitutes a large increase in the amount of absences.And as with the number of bills introduced, this is anenduring effect on legislators’ behavior in Washington.In the second Congress after exile, exiled legislatorsmiss an additional 1.8 days (95% confidence interval,[0.13, 4.37]). The combination of a decline in the

FIGURE 7 The Participation and Voting Consequences of Exile

Note: This figure shows that legislators author fewer pieces of legislation, are absent for more days of voting, and vote less often with their party.

12 justin grimmer and eleanor neff powell

number of bills sponsored and the number of daysabsent provides strong evidence that exiled legislatorsturn their attention to the district. After all, the declinein the number of bills sponsored could be a conse-quence of the lost policy influence from losing a com-mittee assignment. But the increase in number of daysabsent shows that legislators are spending less time inWashington (or at least less time casting roll-call votes).

Exiled legislators also change their voting behav-ior to be more in line with the district. If committeesact as an electoral subsidy, then legislators who areexiled have increased incentive to align themselveswith their constituent views—no longer able tocultivate leeway through their policy influence inWashington. Not all legislators have equal incentiveto deviate from the party once pressure from theirparty leaders is removed. Those representatives frommarginal districts—those composed of a large shareof the other party’s partisans—should have muchgreater incentive to deviate than representatives frommore aligned districts. To assess whether this is true,we use measures of party unity posted on the www.voteview.com website. Because we expect that effectof exile on party unity scores will depend upona legislator’s constituency, we modify equation (1)to include a term that interacts exile with the partisancomposition of a legislator’s district.

The right-hand plot in Figure 7 demonstratesthat exiled legislators from marginal districts deviatesubstantially away from the party. The horizontal axisin Figure 7 presents the vote share for the same partypresidential candidate as the representative in thedistrict (our measure of the district’s partisanship),and the small bars along the horizontal axis is a rugplot that demonstrates where the observed districtpartisanships occur. The vertical axis presents theeffect of exile on a legislator’s party-unity score. Theblack line is the average effect of exile, conditional onthe partisan composition of a legislators district, andthe gray area is a 95% confidence envelope.

This plot demonstrates that the legislators whodeviate from their party the most after exile are thoselegislators who reside in the most marginal districts. Forexample, exile causes legislators in relatively marginaldistricts (40% of the vote for the same party presiden-tial candidate, 10th percentile of districts in our sample)to decrease their party-unity scores 4.7 percentagepoints (95% confidence interval, [-8.15, -1.38]). Andin the second Congress after being removed, exiledlegislators maintain their decreased party unity, -2.8percentage points (95% confidence interval, [-6.69,0.97]). But for legislators from well-aligned districts,exile has little effect on their party unity score. For

example, legislators who represent districts where 68%of voters voted for the same party presidential candi-date (80th percentile of districts in our sample) decreasetheir party-unity score only 0.3 percentage points—adecrease that is neither statistically nor substantivelysignificant (95% confidence interval [-2.70, 1.94]).14

Conclusion

Together, our results show the wide-reaching effectsof congressional committees as electoral subsidy.Leveraging a new strategy for estimating the effectof congressional committees, we find that committeesempower legislators to maintain high levels of sup-port while focusing less on the district. Once theelectoral subsidy of the preferred committee assign-ment is removed, legislators are forced to turn theirattention to the district. The result is that we findcommittee exile has only a conditional effect ona party’s vote share.

Legislators’ shift in attention from policymakingin Washington to campaigning back home manifestsitself across a variety of activities. After losing a com-mittee assignment, exiled legislators spend moremoney on their reelection effort. To support thisspending, they also raise substantially more money.There is a also a clear shift away from policy work inWashington. Exiled legislators author fewer pieces oflegislation and miss more days when Congress is insession. Exiled legislators also align more closely withtheir district preferences when casting roll-call votes.Legislators who represent districts composed of legis-lators from the other party vote with their party lessoften, while legislators from districts composed ofcopartisans change their voting behavior little or votewith their party more often.

Our research contributes a new view of the effectof committees on legislative behavior and a new re-search design to identify these effects (e.g., Broockmanand Butler 2011; Kellermann and Shepsle 2009). Thisresearch shows that legislators are able to compensatefor the loss of a preferred committee assignment. Theresult is that there appears to be few electoral benefits toobtaining a committee assignment, but this is onlybecause legislators are able to compensate for their lackof a preferred assignment. This compensation, though,

14An alternative explanation is that exiles pursue higher office. Inthe supplemental information, we show this is unlikely: ourresults are inconsistent with running for higher office, and exiledlegislators are no more likely to run for Senate than comparablelegislators.

congressmen in exile 13

restricts the amount of policy work legislators canperform in Washington. Committee exile, therefore,can substantially alter who is contributing new policyproposals to Congress and who is evaluating thoseproposals.

Our research design is quite general: committeeexile can be useful in tackling other substantivelyinteresting congressional questions that are usuallyconfounded by selection and identification problems.This future agenda will expand upon what we havedemonstrated here: that the involuntary loss of com-mittee assignments has far-reaching consequences—both for the policy creation process in Washington andthe representational process in congressional districts.

Acknowledgments

We thank Josh Revesz and Tom Dec for researchassistance. We thank Gary Jacobson for sharing hiscandidate-quality data, Arjun Wilkins for sharing hiselection data, and Adam Bonica for his campaignfinance data. We thank Scott Ainsworth, Peter Aronow,David Broockman, Daniel Butler, Gary Cox, JimFearon, Emily Hickey, Dan Lee, Keith Krehbiel, DavidMayhew, Eric Schickler, Heather Stoll, and BarryWeingast for helpful comments. We also thank confer-ence and seminar participants at APSA 2011, Visions inMethodology 2012, Congress & History 2012, StanfordUniversity, and Yale University for helpful comments.

References

Adler, E. Scott, and John Wilkerson. 2012. ‘‘Congressional BillsProject: 1947–2008.’’ Accessed August 11, 2011. http://www.congressional.org.

Broockman, David E., and Daniel M. Butler. 2011. ‘‘Is Member-ship on more Desirable Committees Value for Legislators?Evidence from the Seniority Lottery in Arkansas’ StateLegislature.’’ Yale University Mimeo.

Carney, Eliza Newlin. 1994. ‘‘Bickering Over Committee Seats.’’The National Journal 26 (50): 2916.

Clapp, Charles L. 1963. The Congressman: His Work As He Sees It.Washington, DC: Brookings Institution.

Conference, House Republican. 2010. ‘‘Rules of the HouseRepublican Conference for the 112th Conference as Adoptedon December 8, 2010.’’ Accessed on August 11, 2011. www.gap.gov/resources/library/documents/needed/112th.Conference-Ruels-As-adapted.pdf.

Cox, Gary, and Mathew McCubbins. 1993. Legislative Leviathan.Berkeley: University of California Press.

Currinder, Marian. 2008. Money in the House: Campaign Fundsand Congressional Party Politics. Boulder, CO: Westview Press.

Draper, Robert. 2012. Do Not Ask What Good We Do. New York:Free Press.

Fenno, Richard. 1973. Congressmen in Committees. New York:Little Brown and Company.

Frisch, Scott A., and Sean Q. Kelly. 2006. Committee AssignmentPolitics in The U.S. House of Representatives. Norman:University of Oklahoma Press.

Gelman, Andrew, and Jennifer Hill. 2007. Data Analysis UsingRegression and Multilevel/Hierarchical Models. Cambridge:Cambridge University Press.

Groseclose, Tim, and Charles Stewart, III. 1998. ‘‘The Value ofCommittee Seats in the House, 1947–91.’’ American Journal ofPolitical Science 42 (2): 453–74.

Holland, Paul. 1986. ‘‘Statistics and Causal Inference.’’ Journal ofthe American Statistical Association 81 (396): 945–60.

Katz, Jonathan N., and Brian R. Sala. 1996. ‘‘Careerism, Com-mittee Assignments and the Electoral Connection.’’ TheAmerican Political Science Review 90 (1): 21–33.

Kellermann, Michael, and Kenneth A. Shepsle. 2009. ‘‘Congres-sional Careers, Committee Assignments, and Seniority Ran-domization in The U.S. House of Representatives.’’ QuarterlyJournal of Political Science 4 (2): 87–101.

Mayhew, David. 1974. Congress: The Electoral Connection.New Haven, CT: Yale University Press.

Nelson, Garrison. 2011. ‘‘Committees in the U.S. Congress,1947–92, 80th-102nd Congresses.’’ Accessed August 11,2011. http://web.mit.edu/17.251/www/data.pege.html

Poole, Keith, and Howard Rosenthal. 1997. Congress: A Political-Economic History of Roll Call Voting. Oxford: Oxford Univer-sity Press.

Sekhon, Jasjeet, and Rocio Titiunik. 2012. ‘‘When NaturalExperiments are Neither Natural nor Experiments.’’ AmericanPolitical Science Review 106 (1): 35–57.

Shepsle, Kenneth A. 1978. The Giant Jigsaw Puzzle: DemocraticCommittee Assignments in the Modern House. Chicago: Uni-versity of Chicago Press.

Stewart, Charles, III. 2001. Analyzing Congress. New York:Norton.

Stewart, Charles, III. 2012. ‘‘The Value of Committee Assign-ments in Congress Since 1994.’’ Presented at the annualmeeting the Midwest Political Science Association.

Stewart, Charles, III and Jonathan Woon. 2011. ‘‘CongressionalCommittee Assignments, 103rd to 112th Congresses,1993–2011: House of Representatives.’’ August 11, 2011.http://web.mit.edu/17.251/www/data-page.html

Justin Grimmer is an Assistant Professor in theDepartment of Political Science at Stanford Univer-sity, Stanford, CA, 94305.

Eleanor Neff Powell is an Assistant Professor inthe Department of Political Science and Institutionfor Social and Policy Studies at Yale University, NewHaven, CT, 06511.

14 justin grimmer and eleanor neff powell