Differential Turnout, Quality Challengers, and Party Differences in House Elections in the 1990s Jamie L. Carson Michigan State University Charles J. Finocchiaro* Michigan State University Eduardo L. Leoni Columbia University David W. Rohde Michigan State University Abstract Extant research on congressional elections attributes changes in the partisan composition of the House of Representatives to familiar variables such as the presence of a quality challenger, campaign spending, incumbent performance, and the underlying strength of the partisan base in each district. To date, however, little systematic attention has been given to the effects of district-level turnout on incumbents electoral fortunes. In this paper, we examine the effect of differential turnout on House elections in the 1990s, with particular attention given to the 1994 and 1998 midterm elections. We employ a dataset including traditional variables, along with data tapping population growth and voter turnout throughout the 1990s. We find that not only is differential turnout an important variable to include in such analyses of congressional elections, but in the context of a more fully specified model, traditional variables such as challenger quality decline in explanatory power, with differences across the two major parties. *Corresponding Author: Department of Political Science, 303 S. Kedzie Hall, East Lansing, MI 48824. Phone: 517.432.2622; Fax: 517.432.1091; E-mail: finocchi@msu.edu. An earlier version of this paper was presented at the 2001 Annual Convention of the Southern Political Science Association. We thank Jim Campbell, Bob Erikson, Robert Jackson, Michael Martinez, Mark Souva, and the panel participants for helpful comments, and Gary Jacobson for provid ing us with some of the electoral data used in this analysis.
Differential Turnout, Quality Challengers, and Party Differences in House Elections in the 1990s Introduction The congressional elections of 1994 ushered in an era of narrow Republican majorities in both the House and the Senate and gave rise to an ongoing struggle for control of each chamber. The new electoral context carried with it some unusual events that caught political pundits, journalists, and academics alike by surprise. In 1994, for instance, the Republicans reclaimed control of the House for the first time in 40 years, in the process defeating 34 Democratic incumbents and losing none of their own. The Republicans maintained control of the House two years later (albeit with a slightly smaller margin) in the face of the reelection of a popular Democratic incumbent president. Interestingly, with the president s coattails presumably withdrawn in 1998, the Democrats actually gained 5 seats at the midterm, the first time since 1934 that the party of the president did not lose seats in an off-year election. In 2000, with the election of a Republican president, the Republicans again held control of the House, but with a slight erosion in the size of their majority. There is a long-standing theory of congressional elections that relates aggregate turnout to electoral outcomes. This theory of surge and decline holds that House candidates of the president s party are advantaged in presidential election years, and then face greater electoral risks in the subsequent mid-term elections. In fact, some might suggest that Republicans were advantaged in 1994 by a decline in turnout among Democratic loyalists, while in 1998 a resurgence in the traditional Democratic base aided Democratic candidates and gave rise to the surprising result. While this sort of explanation conforms to aggregate-level expectations of the
surge-and-decline model of congressional elections for 1994 (i.e., the Democrats should have been aided in presidential election years, but hindered in off-year elections), this model does not correspond as closely with the outcome in the 1998 elections. Thus, it seems that a clarification or a more detailed examination of the impact of turnout on each of these elections may shed additional light on this apparent puzzle or paradox in our understanding of electoral patterns in the 1990s. Although the impact of turnout on election outcomes (and Democratic candidates in particular) has received widespread attention in the context of presidential elections (see, e.g., DeNardo 1980, 1986; Tucker and Vedlitz 1986; Radcliff 1994; Nagel and McNulty 1996), we know significantly less about its impact in congressional elections, particularly at the district level. 1 Analyses of congressional elections typically indicate that the alteration in the partisan composition of the House of Representatives can largely be explained by familiar variables such as the strength of the candidate s partisan base, campaign spending, and voters judgments on the performance of incumbents and the president. In our view, however, the election results may also be partially the result of the differential turnout of voters relative to the previous election. While we do not seek to explain completely the electoral outcomes as a result of changes in who chose to vote, we do argue that differential turnout is both a theoretically important aspect for study and potentially a partial explanation of the pattern of outcomes. To address the issue of turnout across congressional elections in the 1990s, we build upon conventional models of congressional elections (see, e.g., Jacobson 1995, 1999a) and incorporate additional variables to account for the role of turnout. In the next section, we briefly review the 1 Some exceptions include Dawson and Zinser (1976) and Cox and Munger (1989). There are, of course, a host of studies that examine the turnout decision at the individual level for presidential, congressional, gubernatorial, and state legislative elections. See Jackson (1996) for a useful review of the general turnout literature. Prominent among the studies of congressional elections are Caldeira, Patterson, and Markko (1985); Gilliam (1985); Jackson (1993); and McDonald and Popkin (2001). 2
surge and decline model of House elections before turning to a more general discussion of the role of turnout in congressional elections. Next, we present findings pertaining to the role of turnout and its effect on the two-party vote share for Democratic candidates in the 1994-2000 House elections, and contrast these results to those for Republicans in the same period. We then shift our attention to explore more generally the impact of a number of related factors on turnout in each of these elections. We conclude the analysis with implications of our results and possible extensions for future work. Midterm House Elections and Turnout Surge and Decline Theory There are few trends in modern American elections as consistent as the decline in turnout between presidential and midterm elections. The theory of surge and decline posits a higher stimulus for turnout in presidential elections resulting from increased interest and information (Campbell 1985, 1997a). The increased attention given to presidential elections is said to spur voter participation, particularly on the part of those more disposed toward the winning candidate. Thus the short-term conditions favor congressional candidates of the party winning the presidency, who benefit from the participation of non-core voters. The theory suggests that in off years the two parties turn out at a more symmetric level, and the independent vote is more evenly divided between Democratic and Republican candidates. These conditions essentially withdraw the advantage enjoyed by the congressional candidates of the president s party in the previous election, setting the stage for increased seat loss. Campbell (1991) encapsulates surge and decline to hold that a party s presidential vote positively affects the change in its share of votes and seats in presidential election years and 3
negatively affects the change in its shares of votes and seats in the following midterm elections (478). Of course, a party s electoral performance is likely conditioned upon how its current standing compares with that of its long-term seat share. For instance, a party that is overexposed, or holds a significantly larger number of seats in contrast to their historical average, may be more likely to suffer at the polls, ceteris paribus. The notion that the relative number of seats held by the two parties in Congress directly influences their electoral fortunes was introduced by Oppenheimer, Stimson, and Waterman (1986) and extended in Waterman, Oppenheimer, and Stimson (1991). The Elections of the 1990s: Declining Turnout and Partisan Advantage The theory of surge and decline detailed by Campbell specifies a general, aggregate-level relationship between presidential and midterm elections, and evidence supports the contention that a party will suffer losses in the midterm in proportion to the influx of partisans in the preceding presidential election. Of course, this effect is not constant across congressional districts. The amount by which turnout declines from a particular presidential election to the successive midterm and the extent to which candidates of each party are affected is likely to vary across districts. For example, there was an obvious shift from 1992 to 1994 in the aggregate division of the vote from favoring the Democrats to favoring the Republicans. While the Democrats in 1992 probably did not receive the degree of on-year surge as might otherwise have been the case, we would argue that there remained a number of reasons to expect a poor electoral showing (and off-year seat loss) for the Democratic party in the House. 2 We suggest 2 Democratic candidates in 1992 likely did not experience a normal gain in votes and seats in fact, Republicans gained ten seats in the House due to the Perot candidacy, pro-gop redistricting, and the anti-incumbency mood (see Campbell 1993: 217-221). However, we argue that it is still reasonable to link the poor Democratic showing in 4
that a component of this aggregate-level change resulted from differential reductions in the voting coalitions of the two parties across the districts held by Democratic incumbents. 3 More specifically, due to these types of factors, and other short-term forces that could be mentioned, we argue that the electoral environment of 1994 was tilted against Democratic candidates. For those who voted Democratic in 1992, became dissatisfied with the party, and viewed the Republican as an acceptable alternative, there was a straightforward option: they could switch their votes to the GOP candidate. However, for those who did not like the Republican candidate, another option was available: they could abstain altogether. Thus, we would expect that the greater the decline in turnout in a district from 1992 to 1994, the greater the damage to the Democratic incumbent in 1994. The case of 1998 presents a different picture from what we observed in 1994. The party of the president actually fared better at the midterm, with the president s party gaining seats for the first time in 64 years. Unlike four years earlier, Clinton s approval ratings remained high in 1998, the economy was strong, and there were few vulnerable Democratic incumbents (Jacobson 1999a). This resulted in a situation where the Republicans were put on the defensive, despite the traditional pattern of seat gain for the opposition party. The narrow balance in the House and the fact that neither political party seemed to have much of an advantage going into the election led to enhanced efforts to stimulate turnout on the part of the parties and allied interest groups. For the Democrats in particular, unions such as the AFL-CIO made a concerted effort to mobilize their members to turn out and vote for Democratic candidates. 4 In contrast to 1994, it appears 1994 to reduced turnout in that the president s approval rating remained low, the public s assessment of Congress was even poorer, and unified government offered a focal point for Republican counterattacks. 3 Some ideas presented in this paper are related to earlier work in Orme and Rohde (1995). 4 The AFL-CIO s change in strategy from a focus on television advertising in 1994 to grass-roots mobilization in 1998 had a significant impact in increasing the union vote as a percentage of the electorate (Abramson, Aldrich, and Rohde 1999: 258-259). 5
that the Democrats (as opposed to Republicans) may have been aided by asymmetric turnout in the 1998 midterm elections. The decline in turnout was likely not uniform across districts and the result was perhaps an aggregate gain for Democrats in general (Abramson, Aldrich, and Rohde 1999: 258). Contextual Factors and Voter Mobilization The general disadvantage to the president s party in the midterm House elections attributed to declining turnout could be affected (either mitigated or worsened) by other factors. For one, turnout itself could be influenced by the electoral context in a particular district. There are three such factors that we deal with here. The first is the impact of a state-wide competitive election. In much the same way that presidential races draw marginal voters with limited political interest into the electorate, major state-wide races could have a similar (albeit less pronounced) impact if they are visible and closely contested. Such contests in midterm years could encourage potential voters who would otherwise abstain to participate, thereby leading to additional votes being cast concurrently in the House races. The effect of this variable is likely to be most pronounced in those districts where the congressional election is not particularly competitive and the incentive to abstain is highest. 5 The second and third contextual factors we seek to account for are related to the electoral impact of two significant segments of the electorate that are influenced by underlying ideological and economic interests: the Christian Right and union members. The level of electoral participation among those associated with the Christian Right has increased substantially since the mid-1980s (Green, Guth, and Wilcox 1995). The electoral effects and partisan impact of the 5 Indeed, Nicholson and Miller (1997) demonstrate that prior beliefs about the competitiveness of an election significantly impact voter turnout. 6
Christian Right took a significant step forward in the 1994 election. The distribution of voter guides and the active recruitment of candidates on the part of the Christian Coalition and other allied organizations were aimed at both stimulating turnout among conservative Christians and affecting their choices of candidates. Thus, we suggest that both turnout effects and direct impact on voters choices should be linked with those locales where the Christian Right had a significant presence. More specifically, we would expect that where the Christian Right was strong and effective, voting among their supporters should have been stimulated. This would likely result both in some new voters that abstained in the preceding presidential election deciding to vote (thereby mitigating the decline in turnout from 1992) and shifting the choices of some former Democratic voters to Republican House candidates. In a similar fashion, organized labor likely had an impact on turnout, particularly as the decade passed. In response to a weak showing on the part of union members in 1994, labor redoubled their efforts in 1996 and 1998. As we mentioned above, they shifted their strategy to one of grass-roots mobilization and voter education campaigns (Jacobson 1999b), and this change in strategy provided a boost to Democrats just as Republicans were advantaged in 1994 (though the magnitude of the effect in 1994 was apparently greater). Understanding the Democratic Vote in the 1994-2000 House Elections The data for this segment of our analysis consists of all Democratic incumbents running for reelection in the 1994, 1996, 1998, and 2000 congressional elections. We exclude representatives from the state of Louisiana because of the unusual mixed primary-general election system that governs this state s congressional races. For the sake of comparison, we 7
also exclude Democratic incumbents who ran unopposed in either the current or previous election since their electoral margins were inflated by the lack of an opposition party candidate. 6 Building upon previous work that examines congressional elections in the 1990s (see, e.g., Jacobson 1995; Jacobson 1999a), our dataset incorporates the usual predictors of congressional election outcomes commonly employed by Jacobson and other congressional election scholars: the Democratic share of the two-party House vote in the previous election, the average Democratic share of the presidential vote in the congressional district in the two previous presidential elections (as a proxy for the political preferences of the constituency), the log of incumbent and challenger campaign spending, 7 and challenger quality (measured dichotomously and coded one if the challenger has previously held elected office and zero otherwise). These variables will establish a baseline from which we can assess whether the turnout-related variables provide additional explanatory power. To this set we add four additional explanatory variables: change in turnout from one election to the next, the presence of a competitive state-wide election, the union membership percentage for each state, and the Christian Right influence within the state Republican Party. Change in turnout is measured as the decline (or increase) in turnout from one election cycle to the next. 8 During midterm elections, we should generally expect turnout across congressional districts to decline whereas we should observe increases during presidential election years. The variable indicating the presence of a competitive state-race is measured dichotomously, coded one if either the gubernatorial or Senate race in the state during the election year was won by 10 6 Note that we do not consider Bernie Sanders, an independent from Vermont, as a Democrat for the purposes of our analysis even though he regularly caucuses with the Democratic Party. 7 The log of incumbent and challenger campaign spending is used in order to transform the data to fit a diminishing marginal returns curve. Following Jacobson (1980), we assume that all challengers and incumbents spent at least 5,000 dollars, as any amount under this figure need not be reported to the Federal Election Commission. 8 See Appendix A for a discussion of the way in which turnout was extrapolated across the 1990s. The appendix also includes sources for each of the variables included in our analysis. 8
percent or less of the two-party vote. The union membership variable captures the percentage of individuals within each state who are members of a union. Finally, the Christian Right variable is dichotomous, coded one if the state Republican Party is viewed as being dominated by the Christian Right and zero otherwise (Persinos 1995). Obviously, a dummy variable is not the most desirable way in which one might capture the impact of such a dynamic. It would be preferable to have membership data similar to that of unions, but since such data is not available, the approach we adopt seems to be the best alternative. We return to the potential limitations of this variable later in our discussion. 9 Table 1 presents the results of the regression analyses examining the impact of the above factors on Democratic incumbent vote share in the 1994 through 2000 House elections. All of the conventional variables included in models of congressional elections are significant and in the predicted direction across all elections except challenger quality (which is negative and insignificant). 10 Additionally, the four turnout-related variables we introduced yield notable and varying effects on Democratic vote share in each of the elections. Following from the theory of surge and decline, which considers aggregate-level changes in turnout, we expect that changes in turnout are also important at the individual or district level. That is, the effect of changes in turnout is likely to differentially impact Democratic incumbents depending on the degree to which turnout declined in their particular districts. In line with our expectations, we observe that the change in turnout from 1992 to 1994 yields a significant effect 9 We were able to obtain from the Christian Coalition the actual number of voter guides distributed in various geographic areas in the 1994 election. While this data was usually reported at the congressional district level, there were some cases in which it was reported only at the state level (e.g., California, Florida, Illinois) and one case where it was not reported at all (New York). While we estimated the models with various operationalizations of the voter guide data, for the reasons stated above, the results were mixed and no one specification was free from unrealistic assumptions. 10 Previous research holds that incumbent spending should either be negative or have an insignificant effect on incumbent vote share (see Jacobson 1980). In this case, it is either negative or insignificant (or both) across the election years. 9
on Democratic incumbents vote shares. A decline in turnout, which occurred in every Democratic district, led to a corresponding decrease in electoral performance. Substantively speaking, the magnitude of the variable is not enormous, however neither is it inconsequential. Consider, for instance, the case of Rep. David Price (D-NC), in whose district turnout declined 25% from 1992 to 1994. The result of this decline was a 3½% decrease in his vote share (ceteris paribus), which in tandem with other factors (such as the strong Christian Right presence in North Carolina and the underlying partisanship of his district) led to his surprising defeat. 11 In addition, the turnout effects in 1994 were likely more pronounced due to the prominent influence of the Christian Right on the Republican Party apparatus in numerous states. On the other hand, in those cases where there was a competitive statewide race, Democratic incumbents seem to have benefited in that such races seem to have tempered the traditional decline in turnout occurring in midterm elections. << Insert Table 1 Here >> In 1996, the change in turnout among voters had an insignificant effect on Democratic incumbents vote share. This is not surprising, however, given that turnout among members of both major parties likely increased significantly in contrast to the preceding midterm election. Interestingly, however, the differential change in turnout from the 1996 to the 1998 election was also insignificant in its effect on Democratic incumbents vote shares. Unlike 1994, where the Democrats had unified control of government (including both chambers of Congress and the Presidency), Republicans held control of Congress in 1998. Democratic voters who were 11 See Price (2000) for his own interpretation of the unforeseen electoral defeat, which he also attributes in large part to the changing electorate. 10
dissatisfied in 1994 were faced with a difficult choice at the polls many probably found voting for the opposition party an unpalatable alternative, so the appeal of abstaining was probably greater than would ordinarily be the case. In 1998, the situation was quite different, in that there was a target for their attention the majority Republicans in Congress. Thus, it may well be the case that the type of effects resulting from differential turnout that we observed in 1994 may be particular to those instances in which one party is the focal point of attention in government. Additionally, it seems that variation in the composition of the electorate also benefited Democrats in 1998. Although turnout was lower this year than in any national election since World War II, it appears that the composition of the electorate was such that traditionally Democratic voters made up a greater proportion of the electorate than was the case in 1994. In fact, some have suggested that Democrats benefited from asymmetric turnout in 1998 in a similar manner as did Republicans in 1994, though not enough for them to recapture control of the House. One gauge of the changing composition of the electorate is the degree to which it was made up of union households, and by this standard most observers note an increase in participation on the part of this segment of the population. 12 The models for 1996 and 1998 do not suggest significant effects for the presence of a dominant Christian Right influence over the state Republican Party or union membership at the state level, although their impact on vote share is in the predicted direction. We did, however, perform joint-f tests to determine whether the variables exhibited a jointly significant impact on the dependent variable in both elections. Interestingly, we find that in both 1996 and 1998, the 12 It is difficult to discern exactly the change in union voting from 1994 to 1998 in that the Voter News Service altered the wording of this question in their exit polling beginning in 1996. While it appears as though there was an increase regardless of the question wording, the extent of the change attributable to this difference is difficult to ascertain (Morin 1998; Abramson, Aldrich, and Rohde 1999: 259). 11
variables together offer a statistically significant improvement in the fit of the model when contrasted with the baseline model that does not include them in the estimation. 13 Once again, turnout is insignificant in the 2000 election, probably for the same reason it was in 1996. As we saw in 1996, the presence of a competitive statewide race increased the vote share for Democratic incumbents in 2000, even in the context of the presidential election. Additionally, both the Christian Right and Union membership variables exert a significant influence on the Democrats vote share. Interestingly, however, the effect of union membership is in a negative direction, perhaps indicating some underlying dynamic about those states where labor union membership was proportionally higher. That is, it is likely not the case that union members defected wholesale from Democratic candidates in 2000, but rather that something else influenced the results in those states with higher levels of union membership. 14 Republican Electoral Fortunes in the 1994-2000 House Elections To examine party-specific differences in electoral fortunes based on the above factors, we run a parallel analysis on Republican incumbents seeking reelection. Table 2 presents the results of the regression analyses examining the effects of these factors on Republican incumbent vote share in the 1994 through 2000 House elections. 15 Most notably, nearly all of the supplementary variables we include in our analysis across elections fail to meet conventional levels of statistical significance. With the exception of the change in turnout 13 The joint F-test allows for a comparison between the F-statistics of a constrained and an unconstrained regression model. The ratio between the two is compared to a critical F-value, which if exceeded indicates a significant loss of explanatory power by excluding the variables of interest. The F-statistic for 1996 is 3.68, which allows us to reject the null of no impact at p = 0.028. Similarly, for 1998 we can say with 95% confidence that the two variables exert a significant effect. 14 In 2000, union households made up 27% of the electorate, splitting 60 38 in favor of the Democratic candidate. In contrast, non-union households (73% of the electorate) split 44 53 in favor of the Democratic candidate (Voter News Service). 15 As before, we exclude Republican incumbents who ran unopposed in either the current or previous election since their electoral margins would be inflated by the lack of an opposition party candidate. 12
variable in the 1994 elections and the Christian Right variable in the 2000 elections (both of which are only marginally significant), none of the other variables we include offer any additional explanatory power in the election models. This is not wholly unexpected, since it is thought that lower turnout disproportionately affected Democratic candidates in the 1990s (see Abramson, Aldrich, and Rohde 2001: 91). << Insert Table 2 Here >> For the most part, the conventional variables included in models of congressional elections are significant and in the predicted direction. For instance, Republican House candidates do worse in districts the higher the Democratic presidential vote share. Additionally, challenger spending is always negative and significant across the 1994-2000 House elections, suggesting (as we would expect) that the more Democratic challengers spend, the smaller the electoral vote share for Republican incumbents (on this point, see especially Jacobson 1980; 2001). Also of interest, we observe that the Republicans vote margin in the previous election is always positive and significant, with the exception of 1994 (in terms of the 1992 elections). One potential explanation for this finding is the fundamental change that occurred in 1994. Campbell (1997b: 850) has argued that Republican presidential coattails were wasted in the 1980s, and the full effect of a realignment was not realized until the 1994 House elections. If this was in fact the case, then we should not be surprised to find that elections prior to 1994 offer little help in explaining results in 1994 and after the two series should possess inherent differences. Perhaps the most interesting aspect of the analysis of Republican incumbents comes in terms of the challenger quality variable. Whereas this variable failed to reach conventional 13
levels of statistical significance for the models of Democratic incumbents, the challenger quality variable is both negative and marginally significant for Republican incumbents in the 1996 (p < 0.06) and 1998 (p < 0.06) elections. In certain respects, this finding is unique in that it suggests that there may be party-specific differences in the impact of this conventional variable on incumbents electoral fortunes. Although the quality of the challenger did not appear to play a significant role in affecting Democratic incumbents electoral margins when controlling for other factors, this is clearly not the case for Republican incumbents running in 1996 and 1998. The fact that the challenger quality variable has apparently declined in significance could indicate a number of possibilities. First, this result may be due in part to its inclusion in a more fully specified model. 16 It may be that the existing and additional variables we include encompass the explanatory power of this dichotomous variable, which Jacobson (1989: 776) acknowledges to be a somewhat crude measure. Second, the pattern we observe could be a sign that the measure is not quite as powerful a proxy for quality in the contemporary electoral context, particularly for Republican challengers. While Jacobson (2000) acknowledges that Republicans have been more successful in recruiting candidates that meet his criterion of quality in the 1990s, it may also be the case that the party also recruited candidates who ran highly effective campaigns yet would not be considered quality candidates according to Jacobson s typology. Predicting Turnout in House Elections, 1994-2000 To this point, we have focused exclusively on understanding whether changes in turnout, along with other related variables, impacted Democratic incumbent vote share. However, an 16 We thank Bob Erikson for drawing our attention to this potential explanation. While a full examination of this possibility, along with the observed differences between the parties, is beyond the scope of this paper, we are exploring this phenomenon in related work. 14
equally important facet of congressional elections involves the factors that influence turnout in House elections. Accordingly, we shift our attention to a related analysis seeking to understand and predict the level of turnout in congressional districts from 1994-2000. Our earlier discussion noted three variables that were hypothesized to influence the level of turnout in House elections: the presence of a competitive state-wide race, the Christian Right s dominance of state Republican Party organizations, and the degree of labor union membership in the state. Additionally, we consider the effects of three more variables. The first examines the degree of competitiveness of the House race. Just as competitive presidential and statewide races stimulate voters to participate in elections, we should expect the degree of competitiveness of a House race to yield an analogous, though likely smaller, effect. This effect should be more pronounced during a midterm election due to the absence of a presidential candidate at the top of the ticket. If voters view the outcome of the House race as a foregone conclusion, as has been the case for many House elections since the 1960s, there is less reason for voters to show up at polling places. Thus, we would expect a House race that is regarded as safe for the incumbent to suppress turnout, all else equal. Our measure for this variable is whether Congressional Quarterly Weekly Report rated the race as safe for the incumbent in its last pre-election analysis. We also include the vote share of third-party presidential candidates at the district level for the 1994-2000 elections to control for the potential impact of third-party movements in stimulating turnout. For instance, there was a five percentage point increase in voter turnout in 1992 compared to 1988, which may be partially attributable to Perot s presence in the 1992 presidential contest. To a certain extent, Perot gave voice to the frustration of a large number of people, and many of his supporters indicated that they would not have voted if he had not been on the ballot. While we do not have strong theoretical expectations concerning the directionality 15
of this variable, it struck us as a potentially important factor to incorporate into the various models of turnout. Thus, we include the vote share received by Perot in the 1992 and 1996 elections for the relevant analyses and the percent of the vote received by Nader in the analysis of turnout in the 2000 elections. Lastly, we control for the level of turnout in the preceding election for each of our four models predicting turnout in the current election. It seems reasonable to expect that higher levels of turnout in one election should be related to higher levels in the subsequent election, after controlling for other factors hypothesized to influence turnout. Turning to the results of our models presented in Table 3, we find that previous turnout is both positive and significant across the four elections (as we would expect). Looking first at 1994, we find that increased vote share for Perot, the presence of a competitive state-wide race, and the degree of union membership all increase turnout at the district level. Conversely, if the race involved an incumbent considered safe, the effect on turnout was negative and significant, just as we would expect. In 1996, once again the vote share for Perot had a positive and significant effect on turnout in the election. On the other hand, we now see that the presence of the Christian Right and the level of union membership statewide yield a negative and significant influence on turnout in 1996. Unlike 1994, the competitive-state wide race and safe seat variables do not significantly influence turnout. This is not surprising to us, however, given the already stimulated turnout resulting from the presidential contest. << Insert Table 3 Here >> 16
The results of 1998 are similar to those of 1994 except for the fact that the district-wide presidential vote share for Perot is no longer statistically significant and the presence of the Christian Right now exerts a positive and significant influence on turnout in this election. By 2000, we find that the only significant variables are the presence of a competitive statewide race (which is positive and significant) and the Christian Right influence (which is negative and significant). In general, and particularly in midterm elections, the results above suggest that turnout is influenced by a number of contextual factors, many of which have little or nothing to do with the House election. For instance, we found that a prominent independent or third-party candidate could have a significant effect on turnout. Indeed, this seems to have been the case for Perot in 1994, and to a lesser degree in 1996. By 1998, however, his effect seems to have nearly dissipated. Similarly, the degree of competitiveness in statewide races also plays a role in influencing turnout beyond what might otherwise occur in House elections. The presence of a competitive statewide campaign reduces information costs for voters since issues come to the forefront of the agenda in campaign discourse (Kahn and Kenney 1999). Thus, we would expect voters to be more likely to get involved in those races where there is a pre-existing incentive for them to participate in the electoral process due to reduced information costs. Finally, there seem to be some mobilization effects on the part of certain socio-economic groups, although they do not appear to be uniform across the elections we consider. In the case of the Christian Right, for example, perhaps this is a function of the operationalization of this variable in our analysis. Future studies may benefit from the inclusion of measures that allow for a closer approximation of the strength of such groups at the state or even the district level. 17
Discussion While students of congressional elections have traditionally focused on the effects of factors such as district presidential vote, challenger quality, and campaign spending when analyzing election outcomes at the district level, we offer a modest addition to existing models. In this paper, we have examined the impact of changes in turnout and a number of related variables purported to impact voting behavior such as the presence of a competitive statewide race and the visibility of organized groups and their efforts to mobilize the electorate. We found that, at times, each of these variables offered some degree of explanatory power beyond the traditional variables included in extant research. In particular, our evidence indicates that differential turnout and related factors did have a consequential independent effect on the vote received by Democratic incumbents in 1994. In contrast, however, differential turnout did not appear to have a similar effect on Democratic electoral fortunes in 1998. This underscores the degree to which 1994 involved a unique set of circumstances that led to an even more unique outcome. Our analysis of Republican incumbents highlighted differences between the parties, both in terms of the turnout-related variables we included and the role of challenger quality in affecting incumbents electoral fortunes. Given the limited scholarly attention to turnout in congressional elections as compared to presidential contests, we also sought to explore the impact of a number of related variables in predicting turnout across congressional districts in the 1994-2000 elections. In particular, we found that certain contextual factors such as the competitiveness of the race, the presence of a competitive state-wide race, the electoral showing of a third-party presidential candidate, and efforts at mobilization by labor unions played a role in affecting voter turnout. While these results are not path breaking, they do provide a more complete picture of this important aspect of 18
congressional elections and they document some factors heretofore not included in existing models of turnout. These results would appear to have some relevance for the 2002 midterm election and the continuing fate of the Republican majority in the House of Representatives. While redistricting will obviously play a role in the outcome of the next election in terms of individual races, we cannot discount the impact of turnout on members electoral fortunes. If we continue to observe an extraordinary decline at the midterm relative to turnout in the presidential election, then efforts at mobilization on the part of the parties and related interest groups will likely make a difference in determining whether the Republicans will maintain their majority status in the House. Of course, should turnout again decline in an asymmetric manner, as was the case for Democrats in 1994, then the party experiencing disproportionate abstention on the part of its core supporters is likely to suffer. 19
References Abramson, Paul R., John H. Aldrich, and David W. Rohde. 1999. Change and Continuity in the 1996 and 1998 Elections. Washington, D.C.: Congressional Quarterly Press. Abramson, Paul R., John H. Aldrich, and David W. Rohde. 2001. Change and Continuity in the 2000 Elections. Washington, D.C.: Congressional Quarterly Press. Caldeira, Gregory A., Samuel C. Patterson, and Gregory A. Markko. 1985. The Mobilization of Voters in Congressional Elections. Journal of Politics 47: 490-509. Campbell, James E. 1985. Explaining Presidential Losses in Midterm Congressional Elections. Journal of Politics 47: 1140-1157. Campbell, James E. 1991. The Presidential Surge and Its Midterm Decline in Congressional Elections, 1868-1988. Journal of Politics 53: 477-487. Campbell, James E. 1993. The Presidential Pulse of Congressional Elections. Lexington: University Press of Kentucky. Campbell, James E. 1997a. The Presidential Pulse of Congressional Elections, 2 nd ed. Lexington: University Press of Kentucky. Campbell, James E. 1997b. The Presidential Pulse and the 1994 Midterm Congressional Election. The Journal of Politics 59: 830-857. Cox, Gary W., and Michael C. Munger. 1989. Closeness, Expenditures, and Turnout in the 1982 U.S. House Elections. American Political Science Review 83: 217-231. Dawson, Paul A., and James E. Zinser. 1976. Political Finance and Participation in Congressional Elections. Annals of the American Academy of Political and Social Science 425: 59-73. DeNardo, James. 1980. Turnout and the Vote: The Joke s on the Democrats. American Political Science Review 74: 406-420. DeNardo, James. 1986. Does Heavy Turnout Help Democrats in Presidential Elections? American Political Science Review 80: 1298-1304. Gilliam, Franklin D., Jr. 1985. Influences on Voter Turnout for U.S. House Elections in Non- Presidential Years. Legislative Studies Quarterly 10: 339-352. 20
Green, John C., James E. Guth, and Clyde Wilcox. 1995. The Christian Right and the Republican Party. Paper presented at the annual meeting of the Midwest Political Science Association, Chicago, IL. Hirsch, Barry T., David A. Macpherson, and Wayne G. Vroman. 2001. Estimates of Union Density by State. Monthly Labor Review 124 (July): 51-55. Jackson, Robert A. 1993. Voter Mobilization in the 1986 Midterm Election. Journal of Politics 55: 1081-1099. Jackson, Robert A. 1996. A Reassessment of Voter Mobilization. Political Research Quarterly 49: 331-349. Jacobson, Gary C. 1980. Money in Congressional Elections. New Haven: Yale University Press. Jacobson, Gary C. 1989. Strategic Politicians and the Dynamics of U.S. House Elections, 1946-86. American Political Science Review 83: 773-793. Jacobson, Gary C. 1995. The 1994 House Elections in Perspective. Political Science Quarterly 111: 203-223. Jacobson, Gary C. 1999a. Impeachment Politics in the 1998 Congressional Elections. Political Science Quarterly 114: 31-51. Jacobson, Gary C. 1999b. The Effects of the AFL-CIO s Voter Education Campaigns on the 1996 House Elections. Journal of Politics 61: 185-194. Jacobson, Gary C. 2000. Reversal of Fortune: The Transformation of U.S. House Elections in the 1990s. In Continuity and Change in House Elections, David W. Brady, John F. Cogan, and Morris P. Fiorina, eds. Stanford: Stanford University Press, 10-38. Jacobson, Gary C. 2001.The Politics of Congressional Elections, 5 th ed. New York: Addison Wesley Longman. Kahn, Kim Fridkin, and Patrick J. Kenney. 1999. The Spectacle of U.S. Senate Campaigns. Princeton, NJ: Princeton University Press. McDonald, Michael P., and Samuel L. Popkin. 2001. The Myth of the Vanishing Voter. American Political Science Review 95: 963-974. Morin, Richard. 1998. Citing Survey Change, Pollsters Retreat on Estimates of Union Vote. Washington Post, November 6, A5. 21
Nagel, Jack H. and John E. McNulty. 1996. Partisan Effects of Voter Turnout in Senatorial and Gubernatorial Elections. American Political Science Review 90: 780-793. Nicholson, Stephen P., and Ross A. Miller. 1997. Prior Beliefs and Voter Turnout in the 1986 and 1988 Congressional Elections. Political Research Quarterly 50: 199-213. Oppenheimer, Bruce I., James A. Stimson, and Richard W. Waterman. 1986. Interpreting United States Congressional Elections: The Exposure Thesis. Legislative Studies Quarterly 11: 227-247. Orme, Robert E. and David W. Rohde. 1995. Presidential Surge and Differential Decline: The Effects of Changing Turnout on the Fortunes of Democratic House Incumbents in 1994. Paper delivered at the Annual Meeting of the American Political Science Association, Chicago, Illinois. Persinos, John F. 1995. Has the Christian Right Taken Over the Republican Party? Campaigns and Elections (September): 21-24. Price, David E. 2000. The Congressional Experience, 2 nd ed. Boulder, CO: Westview Press. Radcliff, Benjamin. 1994. Turnout and the Democratic Vote. American Politics Quarterly 22: 259-276. Tucker, Harvey J. and Arnold Vedlitz. 1986. Does Heavy Turnout Help Democrats in Presidential Elections? American Political Science Review 80: 1291-1298. Waterman, Richard W., Bruce I. Oppenheimer, and James A. Stimson. 1991. Sequence and Equilibrium in Congressional Elections: An Integrated Approach. Journal of Politics 53: 372-393. 22
Table 1 Differential Turnout and the Vote for Democratic Incumbents, 1994-2000 Variable 1994 1996 1998 2000 Democrat s vote in previous election 0.413* (0.061) 0.426* (0.066) 0.344* (0.077) 0.523* (0.057) District presidential partisanship 0.211* (0.050) 0.237* (0.056) 0.210* (0.077) 0.188* (0.038) Challenger has held elected office -0.866 (0.901) -0.990 (1.035) -1.588 (1.026) -0.947 (0.833) Log of challenger s spending -2.340* (0.304) -1.155* (0.193) -1.031* (0.242) -1.499* (0.246) Log of incumbent s spending -0.096 (0.750) -1.355* (0.600) -1.047 (0.765) 0.375 (0.479) Change in turnout 0.144* (0.070) -0.095 (0.081) 0.048 (0.083) -0.057 (0.064) Competitive statewide race present 3.150* (0.716) 2.198* (1.107) -0.541 (0.986) 1.302* (0.567) Christian Right -2.870* (0.704) -0.978 (0.856) -1.018 (0.901) -2.448* (0.883) Union membership 0.015 (0.057) 0.094 (0.065) 0.121 (0.084) -0.158* (0.060) Constant 51.01* (10.85) 55.38* (10.25) 55.70* (12.30) 35.80* (8.14) N F-Value R 2 196 136.45 0.828 143 122.00 0.863 139 144.53 0.843 146 255.49 0.898 Note: Coefficients are OLS estimates, generated in Stata 7.0. Robust standard errors are reported in parentheses. * Significant at p < 0.05
Table 2 Differential Turnout and the Vote for Republican Incumbents, 1994-2000 Variable 1994 1996 1998 2000 Republican s vote in previous election 0.088 (0.099) 0.176* (0.053) 0.306* (0.081) 0.266* (0.073) District presidential partisanship -0.236* (0.081) -0.371* (0.059) -0.285* (0.070) -0.198* (0.063) Challenger has held elected office 0.081 (1.225) -1.627 (0.868) -2.066 (1.085) -0.641 (0.914) Log of challenger s spending -1.876* (0.373) -1.553* (0.260) -0.763* (0.173) -1.781* (0.229) Log of incumbent s spending -1.395 (0.873) -1.144 (0.661) -0.253 (0.212) 0.058 (0.606) Change in turnout -0.255 (0.154) 0.097 (0.085) 0.077 (0.146) 0.004 (0.071) Competitive statewide race present 1.345 (1.049) -0.999 (0.851) 0.942 (0.787) -0.399 (0.705) Christian Right -6.544 (10.463) 0.034 (0.756) -1.428 (1.126) -1.309 (0.756) Union membership -0.119 (0.143) -0.089 (0.067) -0.027 (0.081) 0.076 (0.068) Constant 110.39* (13.12) 101.82* (9.69) 71.66* (9.79) 73.90* (10.96) N F-Value R 2 119 19.86 0.431 172 50.73 0.727 149 26.58 0.661 133 40.61 0.731 Note: Coefficients are OLS estimates, generated in Stata 7.0. Robust standard errors are reported in parentheses. * Significant at p < 0.05
Table 3 Using District Contextual Influences to Predict Turnout, 1994-2000 Variable 1994 1996 1998 2000 Previous turnout 0.594* (0.037) 0.631* (0.083) 0.407* (0.062) 0.562* (0.052) Perot percentage 0.244* (0.058) 0.388* (0.189) 0.300 (0.175) Nader percentage -0.329 (0.354) Competitive statewide race present 2.210* (0.760) -1.091 (1.196) 6.940* (0.874) 5.731* (0.882) Christian Right -0.480 (0.764) -5.824* (0.947) 2.387* (1.000) -4.212* (1.057) Safe seat -2.737* (0.625) 0.301 (0.781) -6.038* (0.841) -1.012 (0.866) Union membership 0.182* (0.051) -0.349* (0.078) 0.607* (0.073) -0.196 (0.110) Constant -0.669 (2.090) 25.492* (3.216) 4.193 (3.074) 29.123* (2.387) N F-Value R 2 435 115.77 0.616 435 47.36 0.501 435 69.11 0.482 435 44.53 0.391 Note: Coefficients are OLS estimates, generated in Stata 7.0. Robust standard errors are reported in parentheses. * Significant at p < 0.05
Appendix A1 Descriptive Statistics of Variables in Table 1 by Year Variable 1994 1996 1998 2000 Mean St. Dev Mean St. Dev Mean St. Dev Mean St. Dev Democrat s vote in current election Democrat s vote in previous election District presidential partisanship Challenger has held elected office 60.17 10.15 67.83 10.40 68.74 11.06 67.65 10.72 64.66 9.38 63.45 9.76 65.88 11.88 67.64 11.66 55.03 10.69 62.62 11.61 62.76 11.74 62.18 13.47 0 = 162 1 = 34 0 = 123 1 = 20 0 = 106 1 = 33 0 = 111 1 = 35 Log of challenger s spending 11.45 1.69 10.69 2.67 9.62 3.33 10.96 1.92 Log of incumbent s spending 13.17 0.64 13.08 0.68 12.86 0.69 13.41 0.77 Change in turnout -12.83 4.84 7.04 4.88-9.48 5.17 8.36 6.82 Competitive state-wide race present 0 = 75 1 = 121 0 = 117 1 = 26 0 = 99 1 = 40 0 = 105 1 = 41 Christian Right 0 = 104 1 = 92 0 = 83 1 = 60 0 = 86 1 = 53 0 = 89 1 = 57 Union membership 16.05 6.66 15.97 6.43 16.14 6.15 16.10 5.96 Note: For dichotomous variables, the frequency of each value is given.