Abstract
We undertake an examination of southern state legislators who changed their party affiliation between 1992 and 2012. Not surprisingly, the vast majority went from being Democrats to Republicans, and this change has fuelled the dynamics at the national level by pulling the Republican Party further to the right and Democrats to the left. Hence, these are the incumbent party switchers we evaluate. With the use of constituency, electoral and contextual data on both switchers and non-switchers, we assess which factors influence the likelihood that a Democratic office-holder will switch to the GOP (Grand Old Party). We find that the typical demographic correlates of southern Republicanism, such as race and education, do indeed have a significant effect on the probability of the GOP label being adopted. In addition, the disruptive effect of redistricting induces switches, as does a change in the party of the elected governor. However, electoral pressure does not have an independent effect on party switching. Our results suggest that party switching is a response to several district-level factors as well as the broader political context within the state. We conclude with a discussion of why increasing partisan polarization and the maturation of the southern GOP is likely to forestall future party switching.
Introduction
The story of contemporary southern politics is one of Republican rule. Once considered a one-party Democratic stronghold, the South can now be properly characterized as Republican-dominant (Hayes and McKee, 2008; Lublin, 2004; McKee, 2012); talk of a competitive two-party region appears antiquated. More than 30 years after Ronald Reagan’s election, Republican hegemony in the South is firmly established, and this region anchors the party in national politics (Black and Black, 2007), a trend unlikely to abate in the foreseeable future.
The Grand Old Party’s (GOP’s) control of southern politics has national implications that go beyond the confines of the region. The expansion of the Republicans’ southern wing has coincided with polarization at the national level (Bartels, 2000; Hood et al., 2012; Levendusky, 2009), helped by the conservative turn of an increasingly southern-dominated GOP (Theriault, 2008). Recent trends in national politics are, to some degree, a product of this massive shift. Understanding the partisan change in one American region sheds light on larger national trends that we have witnessed over the past several decades.
Along with changes in the voting behaviour of the southern white electorate and the election of Republican candidates to office, one critical aspect of the contemporary ‘rise of southern Republicans’ (Black and Black, 2002) has been the decision by elected office-holders to shed their long-standing Democratic affiliation in favour of the Republican label. This behaviour among elites has been a catalyst for the rise of the GOP, yet we know little about the factors that foster it. In this article we provide a long overdue systematic analysis of southern Democratic legislators who defect to the Republican Party while in office, an important but understated mechanism driving the modern partisan realignment of the South, with significant ramifications for national politics (Black and Black, 2002, 2007).
To fully comprehend the southern realignment, we contend that it is important to understand the decision by lower-office incumbents to switch to the GOP. We systematically model this behaviour among southern Democratic state legislators between 1992 and 2012. Specifically, we incorporate constituency-level factors, elections-related variables, and contextual forces, and assess their relative importance in explaining party switching. We show that party switching is a response to both bottom-up constituency factors as well as top-down forces.
But this article is more than a simple exploration of party switching. Another theoretical contribution is to test various explanations regarding partisan change. Our theoretical contribution, then, is to model a specific career decision systematically in a way that enhances our general understanding of partisan change. Our study also examines the role of race in southern politics, which is especially relevant in light of the rise of the Latino population as well as recent work showing that race still plays a prominent role in electoral politics and representation (Grose, 2011; Hood et al., 2012). If race plays a systematic role in fostering party switching, it would suggest that it is still inextricably linked to partisan politics in a region where it has been the dominant factor shaping public affairs both past and present (Glaser, 1996, 2005; Grose, 2011; Key, 1949; Kousser, 1974; McKee, 2012).
Finally, our research contributes to the emerging literature on party switching in comparative perspective (e.g. Desposato, 2006; Heller and Mershon, 2005, 2009; O’Brien and Shomer, 2013). Much of that literature examines aggregate, institutional factors or electoral factors, which is consistent with comparative scholars’ emphasis on institutional features such as the electoral system, the nature of executive–legislative relations, the party system, and the like. Likewise, we add to this literature by testing for a variety of constituency effects (in addition to electoral and contextual factors), which are a dominant feature of the American electoral system and, hence, of the American scholarly literature.
Explaining partisan change
A large literature on the southern partisan transformation examines the contributors to Republican advancement in the erstwhile cradle of American Democracy. Chief among these are race (Carmines and Stimson, 1989; Hood et al., 2012), a rising middle class (Shafer and Johnston, 2006), urbanization (Bartley and Graham, 1975), in-migration (Beck, 1977; Scher, 1997), generational change (Green et al., 2002), ideological and partisan sorting (Abramowitz and Saunders, 1998; Levendusky, 2009), redistricting (Hill, 1995; Petrocik and Desposato, 1998), and religious realignment (Green et al., 2010). Moreover, the general pattern of Republican advancement has been ‘top-down’ (Aistrup, 1996), starting with presidential elections and moving down to lower offices. In the most salient of elections, presidential contests, southern Republicans have been dominant since 1972 (Black and Black, 1987). As Black and Black (2002: 4) argue, the racially polarizing 1964 election triggered the first ‘great white switch’; with southern whites casting a majority of their (two-party) presidential votes for Republicans ever since. The second ‘great white switch’ came some two decades later during the Reagan presidency, when southern white identification with the GOP finally outstripped Democratic affiliation.
To grasp this remarkable Republican turn, consider Table 1, which displays the massive increase in the share of GOP seats for various offices between 1990 and 2012. After the 2012 elections, the GOP controlled 63 percent of these 1,953 southern elective offices, more than twice as many as they did two decades previously. Top-down advancement has reached all the way to state legislative elections, which until recently were the last bastion of Democratic control.
Republican electoral gains in the South between 1990 and 2012.
Note: Data compiled by the authors. The second number displayed in parentheses (for the U.S. House and Total columns) accounts for the number of U.S. House districts in 1990.
Historically, because of being low profile, localized and insulated from national politics, southern GOP gains were the lowest and slowest in down-ballot state legislative races. Tables 2 and 3 show that this is no longer the case. The growth in GOP state legislative office-holding is perhaps the most telling statement of a modern Republican South. In 1990, Democrats were the majority in all state legislative chambers. Twenty-two years later, Republicans have majorities in all but one chamber (the Virginia senate being evenly split), and hold over 61 percent of the South’s 1,782 state legislative seats.
The increase in Republican state senators, 1990–2012.
Note: Data compiled by authors from various issues of The Almanac of American Politics (1992–2012 editions). The 2012 data are from the National Conference of State Legislatures (NCSL), valid for 16 March 2013: www.ncsl.org/legislatures-elections/elections/statevote-charts.aspx. Numbers in boldface indicate a Republican majority and bracketed numbers a split chamber.
The increase in Republican state representatives, 1990–2012.
Note: Data compiled by authors from various issues of The Almanac of American Politics (1992–2012 editions). The 2012 data are from the National Conference of State Legislatures (NCSL), valid for 16 March 2013: www.ncsl.org/legislatures-elections/elections/statevote-charts.aspx. Numbers in boldface indicate a Republican majority and bracketed numbers a split chamber.
Much of the literature examining this change focuses on vote choice and partisan identification of the southern electorate. Few, if any, offer a theoretically and empirically rigorous account of another factor fuelling partisan change: party switching by incumbent politicians. This is unfortunate since many scholars argue that elite behaviour is the main driver of mass preferences (see Zaller, 1992). Indeed, a high profile switch, like Strom Thurmond’s in 1964, sends a strong signal to the electorate regarding where the parties stand on highly salient issues like civil rights. In this respect, career politicians are betting on which party will sustain them in office, and when these individual-level choices disproportionately occur in favour of one party, it enhances its electoral viability (Aldrich, 1995). This is an understated component of the rise of southern Republicans. As Canon and Sousa (1992) and Yoshinaka (2012) contend, incumbent office-holders who convert from the Democratic Party to the GOP constitute a fundamental feature of Republican party-building in the contemporary American South.
Party switching
While this phenomenon is not limited to a single region, contemporary party switching among state legislators has occurred mostly in the South (Shor and Tomkowiak, 2010) and the lion’s share comprises defections to the GOP. Both trends have contributed to the rise and expansion of the GOP in the region and to the changes in the ideological and geographical make-up of both parties at the national level. Yet, we know little in terms of systematic explanations that cover the entire region over the course of several decades; scholarship on party switching at the state level generally, and in the South specifically, is thin, with only a handful of studies examining this phenomenon systematically.
Legislative party switching in the US South
In an unpublished study, Shor and Tomkowiak (2010) investigate the effect of legislative polarization on the incidence of legislative party switching. Yoshinaka (2012) looks at some of the correlates of party switching in five southern states over a 12-year period. This general lack of scholarly attention is even more surprising given recent high-profile instances of party switching. For instance, several cases of party switching ‘flipped’ control of a legislative chamber (e.g. Georgia in 2002, Mississippi in 2007). Many high-ranking officials, from former governors Sonny Perdue (GA) and David Beasley (SC) to Texas Governor Rick Perry, were Democratic law-makers who switched parties before becoming powerful Republican politicians. The importance of party switching in the South, then, is a question of real import ‘on the ground’, with partisan control of legislatures and the governor’s office hanging in the balance.
Still lacking is a comprehensive understanding of the conditions under which party switching occurs in all states of the former Confederacy during the most important period of Republican ascendancy in state legislatures – the post-1990 era. The literature on party switching offers some insights into the decision-making process, and a common theme is that party switching is costly (Evans et al., 2012; Grose, 2004; Grose and Yoshinaka, 2003; Nicholson, 2005). Party-switching politicians are often depicted as untrustworthy, opportunistic individuals driven by crass political ambition. For instance, this charge was evident in the case of the most recent congressional party switcher, former Alabama Rep. Parker Griffith. After losing his primary, Griffith acknowledged that switching parties ‘may have been a mistake, politically’. 1 This begs the question of why someone would choose to switch parties. In this study we test three sets of explanations centred upon constituency, elections and context.
Constituency demographic effects
Given the palpable and sustained shift in favour of the GOP in the South, we can expect that Democratic legislators – especially those from districts whose demographics portend a Republican breakthrough – will look to jump on the GOP bandwagon. In the context of a realigning southern electorate, politicians whose district demographics are more receptive to the GOP should be prone to switch parties. This is because most politicians are forward-looking and may wish to stem the incoming tide by pre-emptively switching parties. This is a plausible expectation in the context of contemporary southern politics, especially for Democratic legislators. If it is true that party switching exacts some costs, then we surmise that politicians are willing to trade off some of these short-term costs for the long-term benefits of joining a party on the rise.
Our theoretical framework is similar to David Canon’s work on amateurs, where he uncovers a similar pattern with regard to elite behaviour (1992; Canon and Sousa, 1992). When the minority party rises to majority status at a rapid clip, a disproportionate number of political amateurs will win elections due to the lack of experienced politicians in the ascending party (see also Canon, 1990; McKee, 2010). The rising minority party has a ‘short bench’ (Gaddie and Bullock, 2000) and amateurs step up to fill the void. Likewise, this time of electoral upheaval may provide the necessary impetus for switching parties. This leads us to our constituency-based hypothesis.
Constituency demographic hypothesis: Democratic incumbents representing districts whose demographic characteristics are favourable to the GOP are more likely to switch parties.
From 1992 to 2012, 196 southern state legislators switched parties and 85 percent of them to the GOP, and we assume that most of these switchers represent districts whose demographics favour Republicans. Rather than being sitting ducks, these Democrats choose to be proactive, switching to the GOP with the expectation that if they can weather the short-term electoral storm, winning under their new party label will ensure a more promising electoral future. And for those politicians who switch to the ascending party, it may enable them to gain more legislative influence (Yoshinaka, 2005) or attain higher office. 2
Electoral pressure effects
Constituency factors, however, may not tell the whole story. It is possible that what drives elite behaviour is not an underlying basis of support among the electorate, but rather a more immediate electoral threat. Similar to the ‘marginality’ hypothesis, whereby marginal incumbents behave differently from safe incumbents (e.g. Griffin, 2006), we surmise that electoral pressure, in either the general election or the primary, may ‘push’ Democratic incumbents toward the Republican Party.
Electoral pressure hypothesis: Democratic incumbents are more likely to switch parties in the face of significant electoral pressure.
During a realigning period like the one present in the South, the palpable advantages to the ascending Republican Party provide a clear signal with respect to electoral competition. It amounts to a bandwagon effect and many Democratic legislators view the benefits of switching as outweighing the costs of remaining loyal (Aldrich, 1995; Aldrich and Bianco, 1992). The changing political opportunity structure shifts decidedly in favour of one party and political ambition adjusts accordingly (Schlesinger, 1966). Specifically, the act of switching can be viewed as a pre-emptive move if a southern Democrat perceives a difficult election contest is imminent or is ‘pushed out’ by a primary threat. 3
Contextual effects
We posit that a third set of factors can lead politicians to reconsider their party affiliation. When there is instability caused by institutional and contextual factors, the decision calculus changes: a risk-averse incumbent may no longer put a premium on maintaining the status quo, since other forces affect the political equilibrium. For example, it is possible that, when redistricting is taking place, incumbents may no longer feel the need to keep the status quo intact, since boundary changes are on the horizon anyway. Or, perhaps the election of a new governor changes the dynamics in the state capital such that incumbent legislators revisit their party affiliation; this happened in 2002 when four state senators switched parties in Georgia right after the election of Republican Sonny Perdue to the governorship. This leads us to our third hypothesis.
Contextual hypothesis: Democratic incumbents are more likely to switch parties when the context is more unstable and uncertain.
In contrast to the previous two hypotheses that are district-centred, we are interested in capturing whether party switching also responds to broader factors (instability and uncertainty) whose reach goes beyond the confines of individual districts. Juxtaposed with the other hypotheses, this allows us to test for the effect of both district-level and state-level factors on the decision to switch parties. This adds to the theoretical import of our analysis by testing both types of effects within a unified model of party switching.
Data and methods
We use a multi-prong approach to identify party switchers. By combining existing lists 4 with a canvassing of yearly almanacs (e.g. the Council of State Governments’ State Directory), election data (Klarner et al., 2013), and various media and government sources (e.g. websites of individual Secretaries of State), we are confident that we were able to identify the universe of party switchers in southern state legislatures.
Estimation method
We pit our three hypotheses against each other in two unified regression models: one for 1992–2002 and another for 2002–2012. With each ‘inter-election period’ 5 of each legislator from the 1990s to the present as the unit of analysis, our universe contains almost 15,000 cases of potential party switchers. With the number of party switchers just shy of 200, this means that more than 98 percent of cases did not switch parties. When faced with such data, King and Zeng (2001) argue that it is appropriate to gather data on a sample of non-switchers and all cases of switchers. They show that a logistic regression correcting for this oversampling of 1’s and under-sampling of 0’s will produce consistent parameter estimates as well as a more accurate estimate of the probability that switching occurs.
In all our analyses the dependent variable equals 1 when a legislator switched parties and 0 otherwise. The decision to conduct separate analyses on 2002–2012 and 1992–2002 was driven mainly by the fact that demographic data available for the 2000s (from Lilley et al., 2008) differ from the data available for the 1990s (from Barone et al., 1998). Thus, a pooled model would come at the cost of either including fewer covariates or including covariates measured differently. Allowing for some missing cases this gives us 7,404 cases prior to 2002 and 7,300 cases since 2002. Since we are interested in the motivations of Democratic law-makers, the relevant populations are the 4,454 Democratic lawmakers/terms prior to 2002 and the 3,482 legislator/term dyads since 2002. In 1992–2002, we find 93 Democratic law-makers joining the Republicans, whereas in the post-2002 era we find 73 such instances. We follow the recommendation of King and Zeng (2001) and sample into the data about four times as many instances of non-switchers.
Constituency demographic variables
We include demographic independent variables known to affect the partisan leanings of legislative districts (Gelman et al., 2008; Hamm and Hogan, 2008; Hogan, 2004). For 1992–2002, we include each district’s percentage of African Americans and percentage of Hispanics, as well as average household income. We also include the percentage with at least a two-year associate’s degree; the percentage of households receiving Social Security; and the percentage of the district that is rural and the percentage that is suburban (percentage urban is the omitted category). All these demographic variables come from Barone et al. (1998). 6
To capture constituency effects in 2002–2012, we include each district’s percentage of African American constituents and Hispanic constituents. We also include the average household income, the percentage of the population with some college experience, and a variable measuring population growth. These variables are from Lilley et al. (2008). 7
We expect the coefficients for race to be negative, since districts with large minority populations should be more receptive to Democratic incumbents, who will in turn be less likely to switch to the GOP. We expect income to exert a positive effect, while education should exert a negative effect. While the effect of Social Security income is not as clearly directional, it is reasonable to expect that districts with a high proportion of Social Security recipients would be more inclined to support the party that is viewed as favouring government support for the elderly, in this case the Democratic Party. The coefficients for the two variables measuring the size of the rural and suburban constituencies should be positively signed as the GOP has made much deeper inroads in rural and suburban communities than in urban areas. Finally, we expect population growth to exert a positive effect on party switching since rapid growth can often present difficulties – and new opportunities – for incumbents (see Fenno, 2000).
Electoral pressure variables
To test our second hypothesis we include an incumbent’s general election vote percentage in the previous election. If incumbents switch parties when facing electoral pressure in the form of a strong Republican challenge, we would expect the coefficient to be negative and significant, implying that, all else equal, safe Democrats are less likely to switch parties than marginal incumbents. However, perhaps party switching is a reaction to intense pressure at the primary stage (Aldrich, 1995; Aldrich and Bianco, 1992). This would be the case if a Democratic incumbent viewed the Republican primary as an easier road to the nomination. We test this conjecture in a few ways in 2002–2012.
8
In one specification, we include an incumbent’s primary margin, in percentage points, with the closet opponent. In another specification, we include the competitiveness of the incumbent party’s primary and that of the opposing party’s primary. Each competitiveness score is computed:
where p i is the proportion of the primary vote obtained by candidate i. An incumbent who faces no opposition in the primary will have a competitiveness score of 0, whereas an incumbent who barely ekes out a win in a three-way race will have a competitiveness score near 0.67. The theoretical maximum for this variable is 1, though in the data the maximum observed is 0.79. If, as expected, incumbents prefer a less competitive primary environment, they will be more likely to switch parties when their own party’s competitiveness is high or when the opposing party’s competitiveness is low. 9
Contextual factors
Finally, party switching may also be a response to a changing institutional or elite-driven environment. Specifically, Yoshinaka (2012) shows that redistricting (including mid-decade redistricting) and a change in the partisan control of the governorship affects the incidence of party switching at the aggregate level. In the case of redistricting, it certainly would make sense for a legislator to re-think past career decisions (such as party affiliation) in light of imminent changes to the electoral map.
As for changes in the partisan control of the governor’s mansion, we expect that a new governor would create the sort of changing conditions that are propitious to party switching, especially when the GOP first wins the governorship after more than a century of Democratic governors. For instance, after Sonny Perdue became in 2002 the first Republican governor elected in Georgia since Reconstruction, four Democratic senators changed parties, which gave Republicans control of that chamber. This sort of upheaval of the executive – which triggered these party switches in Georgia – could therefore have trickle-down effects. Moreover, it is possible that a new executive will trigger a wave of party switching away from his party, especially when the incoming governor’s policy positions are at odds with many co-partisan legislators.
In order to estimate the effects of redistricting we include a dummy variable that equals 1 when the legislative term coincides with the implementation of a new redistricting plan (including mid-decade plans). 10 To capture the effects of a change in the party affiliation of the governor, we include a variable that equals 1 when the governorship changes partisan hands during the legislative term. We also include a dummy variable indicating whether the legislator belongs to the majority and the size of the legislative majority at the beginning of the term (measured as a percentage of all seats). We also include control variables that may be linked to party switching: seniority measured in years and the length of the term (4 years versus 2 years) since we should expect a higher probability of switching during longer terms.
Results
For each decade we estimate a rare-events logistic regression model with the variables listed above and a second model that adds state dummy variables. We also test for primary effects in 2002–2012. We begin by discussing the results in the most recent period when Republican ascendancy was in full force, but it will also be evident that many of the same factors were important in influencing the decision to switch parties in the 1990s.
2002–2012
Table 4 presents the results of our analyses for 2002–2012. In the first model, three constituency variables and three contextual variables affect the likelihood of party switching among southern Democratic state legislators: a district’s percentage of African American constituents, percentage of Hispanic constituents, and percentage with some college education; the occurrence of redistricting, a change in the party of the governor, and majority status in the chamber.
Rare-events logistic regression analysis of party switching among Democratic southern state legislators, 2002–2012.
Note: In each model we show the factor change in the odds that Y = 1 when varying statistically significant continuous variables between the 10th percentile and the 90th percentile and dichotomous variables between 0 and 1.
***sig. at the 0.01 level; **sig. at the 0.05 level; *sig. at the 0.10 level (all two-tailed); †sig at the 0.10 level (one-tailed).
The substantive effects of these variables are shown in the second column, which displays the factor change in the odds that a Democratic southern state legislator will switch parties when varying each statistically significant independent variable between its 10th and 90th percentile, or 0 and 1 in the case of dichotomous variables. 11 We find that going from the 90th percentile to the 10th percentile in the size of the black constituency increases the odds of a party switch by a factor of about 6.4, while going from the 90th percentile to the 10th percentile in the size of the Hispanic constituency increases the odds of switching parties by about 1.5. 12 These two results indicate that party switching among southern Democratic state legislators is more likely to occur for those who represent mostly white districts. And, unsurprisingly, party-switching behaviour is much more responsive to variation in the size of the African American constituency than to variation in the size of the Hispanic constituency. Also, party switching is negatively related to a district’s level of education. Going from the 90th percentile to the 10th percentile on that variable increases the odds of a party switch by a factor of about 4.
With respect to contextual factors, party switching is more likely to occur when districts are being redrawn, with the odds of a party switch about four and a half times higher when a state reconfigures its legislative boundaries. This shows that the uncertainty induced by redistricting can provide the necessary push for a momentous career change. Also, a new party in the governor’s mansion increases the odds of a party switch by a factor of about 2, and being in the majority decreases the odds of switching parties by a factor of nearly 3.
What these results tell us is that party switching is a response to a changing constituency as well as larger contextual forces within the state. Also telling are some of the non-significant coefficients as they speak to the various hypotheses presented earlier. Specifically, general election vote-share shows no independent impact on party switching, and neither does the size of the legislative majority. This suggests not only that legislators are responding to their constituency independent of electoral pressures, but also that electoral pressure does not induce party switching independent of the constituency factors that underlie the electoral environment.
In the second model we see that the results generally hold even after the inclusion of state dummy variables. 13 With respect to constituency factors, race is again significant and in the expected direction and a district’s education level exerts a negative impact. In addition, the district income variable is now significant and in the expected direction. Going from the 10th to the 90th percentile in district income increases the odds of a party switch by a factor of about 2. Also, population change now exerts a negative effect on party switching.
In terms of contextual effects, once again redistricting and being in the majority party have large and statistically significant effects on the incidence of party switching. On the other hand, the only result that ‘goes away’ is that a change in the party of the governor is no longer a statistically significant predictor of party switching (though the coefficient is in the right direction). In terms of elections, again we find that vote-shares have no effect on party switching. 14
Does primary competition induce party switching?
In order to explore the effects of electoral pressure more fully, we estimated a series of models in which we included additional variables on the right-hand side that capture primary competition. The results of these additional models are shown in the Appendix. In one model, we included a variable that measures the margin, in percentage points, between the legislator and her closest primary opponent. In a second model, we included the competitiveness score in the legislator’s party’s primary and a competitiveness score in the opposing party’s primary. And in a third model, we included the difference in competitiveness between the legislator’s own primary and that of the opposing party.
In all three models, the coefficients for the primary variables fail to reach statistical significance. Thus, it appears that as was the case with general elections, competitiveness in the primary does little to induce party switching. Moreover, the inclusion of those variables does not affect our inferences with regard to other constituency and contextual variables. Thus, our earlier findings were not an artifact of having excluded primary variables.
1992–2002
Turning to 1992–2002, the first model in Table 5 shows that the size of the black constituency and the education level of a district are again negatively related to party switching, a change in party control of the governorship increases the odds of a party switch, and majority status is negatively correlated with the likelihood of switching parties. These results mirror those from 2002–2012. The substantive effects are shown in the second column. Going from the 90th percentile to the 10th percentile in the size of the black constituency changes the odds of a party switch by a factor of about 5.9. Going from the 90th to the 10th percentile in district education level changes the odds by a factor of about 2. At the contextual level, a new party in the governor’s mansion increases the odds of a party switch by a factor of just about 4, while being in the majority decreases the odds by a factor of just over 3.
Rare-events logistic regression analysis of party switching among Democratic southern state legislators, 1992–2002.
Note: In each model we show the factor change in the odds that Y = 1 when varying statistically significant continuous variables between the 10th percentile and the 90th percentile and dichotomous variables between 0 and 1.
***sig. at the 0.01 level; **sig. at the 0.05 level; *sig. at the 0.10 level (all two-tailed); †sig at the 0.10 level (one-tailed).
The second model includes dummy variables for each state. The results show that even after including these state dummies, the coefficients for race, education, a change in the governor’s party, and majority status are still in the expected direction and statistically significant. Their substantive effects are also quite large. In addition, the coefficient for the size of the rural constituency is now negative and statistically significant (at the 0.10 level, one-tailed test). Similarly, the effect of the size of the suburban constituency is negative and statistically significant. Also, the contextual factors of change in the party of the governor and majority status affect the likelihood of switching parties. 15
Discussion
In both models we find consistent support for the constituency demographic and the contextual hypotheses, but no support for the electoral pressure hypothesis. By and large, the effects that are significant in one decade are also significant in the other (e.g. percentage black, education, governor, majority status). However, two variables are only significant in 2002–2012: percentage Hispanic and redistricting. The significant effect of percentage Hispanic is likely due to an increase in the importance of the Hispanic vote in the last decade in both the South and across the country, as well as an increased movement toward the Democratic Party (Leal et al., 2005).
As for redistricting, we surmise that the slew of court-ordered redistricting plans in the 1990s may have had a different impact from that of the largely legislatively drawn plans of the 2000s. The latter perhaps gave incumbents a better picture of the changes brought about by redistricting, which they then took into consideration when deciding whether to switch parties. It is possible that due to the plethora of court cases in the 1990s and the various instances of plans being rejected either by the Department of Justice or the courts, the instability we posited may have been present throughout the decade, even in years when redistricting did not occur.
Finally, similar to 2002–2012, we fail to uncover any significant effect of electoral pressure independent of constituency factors. While perhaps surprising, it must be noted that the vast majority of elections lack any meaningful competition (Forgette et al., 2009), especially in the South (Squire, 2000). Perhaps as a result, general election competition – or lack thereof – provides a very weak signal as to the underlying level of opposition. Or perhaps incumbents are insulated in these down-ballot, low-information elections, such that they do not feel the need to revisit their party affiliation simply because they face a meaningful challenge. We note that our results do not change if we substitute unopposed re-election for vote-share in all our models (results available upon request). Party switching simply does not respond to immediate past election results.
Conclusion
This study presents the first systematic analysis of all Democratic southern state legislative switches to the Republican Party in the post-1990 era. As such, we contribute to the growing literature on party switching in the United States (Castle and Fett, 2000; Evans et al., 2012; Glaser, 2001; Grose, 2004; Grose and Yoshinaka, 2003; Nokken, 2009; Nokken and Poole, 2004; Yoshinaka, 2005, 2012). Second, we discriminate between various explanations by showing that constituency and contextual factors affect party switching, but that electoral pressure alone does not induce party switching. Third, we add to the literature on realignment and partisan change by focusing on elected officials rather than voters.
What does the future hold in terms of party switching generally and in the South in particular? First, we note that the 2010 elections were followed by a wave of party switchers. While such a result is not that surprising considering the extent of the Republican sweep in 2010, given that the Republican Party has reached a mature stage in its control over the southern electoral system it is surprising to see so many latecomers to the Republican Party. However, because of the overriding pattern of top-down advancement, Republican inroads in southern state legislative contests have historically been the most difficult to establish. Indeed, it was not until after the 2010 mid-term that southern Republicans occupied a majority of state legislative seats for the first time since Reconstruction. So GOP control in these lower level offices is a very recent development.
If Republicans hold on to their current state legislative majorities and possibly expand them, we expect to see fewer switches in the foreseeable future as opportunities for ambitious Democrats diminish and those left behind are unlikely to be conservative enough for the GOP. The main reason we expect this to be true is because of increasing partisan polarization at a time when the southern GOP has attained its dominant status, which will make it much more difficult for a party switcher to credibly defend a conversion. Not only might accusations of placing political ambition over political principles stick, but the likely outcome is electoral defeat because loyal partisans may be advantaged in a primary contest.
This kind of scenario is now likely to hold in state legislative contests because of the growing ideological distance between the parties in these races and the greater prevalence of loyal partisans who have the qualifications to mount strong challenges against legislators bold enough to switch. Party switching always comes with considerable costs and they have only increased as the parties have become more ideologically distinct. We are also seeing, at the local level, potential party switchers who are rebuffed by their prospective new party. 16 We may be getting to a point when the Republican Party is much less receptive of former Democrats.
A natural next step, then, is to examine more systematically what happens to party switchers after they change sides. Do they face significant electoral costs or are they better able to win re-election? For those who switch parties in preparation for a run for higher office (Yoshinaka, 2012), does the switch help or hinder their prospects? Answering these questions would allow for a more complete assessment of party affiliation decisions among ambitious law-makers than what is found in the extant literature, and it would go a long way in establishing the conditions under which elite decision-making produces the intended outcomes and, conversely, the conditions under which it can backfire.
Footnotes
Appendix
Acknowledgements
We thank Pam Bachilla and Dominique Green for providing research assistance and Jonathan Knuckey, Dave Rohde, Joe Stewart, Arnold Fleischmann and Kathleen Burke for their feedback and suggestions. An earlier version of this article was presented in March 2012 at the biennial Citadel Symposium on Southern Politics, Charleston, SC; and in January 2013 at the annual meeting of the Southern Political Science Association, New Orleans, LA.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.