The Jeffords Switch and Public Support for Divided Government
by STEPHEN P. NICHOLSON
British Journal of Political Science, April 2005, volume 35, issue 2, pages 343-356
On 24 May 2001, US Senator James M. Jeffords of Vermont formally announced that he intended to leave the Republican party to become an Independent. Although congressional patty switches in the United States are noteworthy because of their infrequency, the Jeffords switch was especially significant because he would caucus with the Democratic party, handing the Democratic party control of the Senate by a one-vote margin. When Jeffords formally left the Republican party on 5 June 2001, it was the first time that partisan control of the US Senate changed hands mid-session. Perhaps most importantly, with the House and presidency under Republican control, the Jeffords move ushered in the return of divided government--split party control of the executive and legislative branches. Indeed, Jeffords' switch received vast amounts of media attention precisely for this reason. The scant attention given to other recent party defectors in the US Senate such as Ben Nighthorse Campbell of Colorado or Robert (Bob) Smith of New Hampshire suggests that the Jeffords switch was much larger than a senator simply changing his party affiliation.
Over the last twenty years, split partisan control of the executive and legislative branches occurred in American government roughly two-thirds of the time. Given the frequency of divided government, especially in the post-war era, scholars have written much about its causes. (1) The focus of much of this research concerns voting behaviour, especially ticket-splitting and voting against the president's party at the midterm (midterm voting). Some scholars advocate 'cognitive madisonianism' or 'policy-balancing' explanations, which posit that a subset of voters with moderate policy preferences intentionally, but not necessarily consciously, engage in ticket-splitting or midterm voting to bring about policy moderation or 'balance' among the parties. (2) Others contend that ticket splitting and midterm voting is unintentional, a by-product of factors that shape voting behaviour such as candidates, partisanship and campaigns. (3)
Though ticket splitting and midterm voting are the primary causes of divided government, they are behaviours distinct from attitudes. In other words, research on ticket splitting or midterm voting informs us of the origins of divided government, but it does not directly speak to whether citizens actually prefer divided government. For example, Petrocik and Doherty (4) find that 'support for divided government does not explain much of the actual behavior [ticket-splitting].' Finding no relationship between opinions about divided government and voting behaviour, Sigelman, Wahlbeck and Buell conclude that a stated preference for divided government is not associated with ticket-splitting. (5) Fiorina, one of the most ardent supporters of intentional ticket-splitting, would be likely to agree that a stated preference for divided government is not associated with ticket-splitting because of the 'as if' nature of voting for divided government. (6) In sum, scholars have largely neglected public opinion about divided government as a dependent variable because it is a poor predictor of behaviour.
In contrast, the Jeffords switch provides an intriguing case for studying public attitudes towards divided government. Similar to voting behaviour research on divided government, the Jeffords switch takes place in the 'real world' and thus avoids the 'non-attitudes' problem inherent in abstract questions about divided government. (7) Yet, unlike voting behaviour research where split-ticket or midterm voting may (and often does) mean something entirely different from a preference for divided government, the Jeffords switch provides a relatively straightforward opportunity to measure public attitudes about divided government.
To anticipate the argument and conclusion, my findings suggest that unintentional explanations, notably partisan considerations and attitudes towards political figures, shaped public approval of the Jeffords switch. The available evidence provides little support for intentional or policy-balancing explanations. Although I am cautious about generalizing these findings to studies of voting behaviour and divided government, they are suggestive of what citizens think about divided government away from the voting booth.
EXPLAINING SUPPORT FOR DIVIDED GOVERNMENT
Studies of ticket-splitting provide the theoretical foundation for examining public attitudes towards divided government. As mentioned, some scholars argue that divided government is the product of intentional but not necessarily conscious, voting behaviour, while others argue that divided government is unintentional or the by-product of other factors such as candidate-centred politics. Although these approaches seek to explain voting behaviour, the logic behind them works the same for understanding attitudes about divided government.
Fiorina was one of the first scholars to propose an intentional ticket-splitting model of divided government voting. (8) In Fiorina's model, ideologically moderate voters choose candidates from different political parties in an attempt to 'balance' them through divided control of government, thus producing middle-of-the-road policy outcomes. Faced with parties on the left and right of the ideological spectrum, voters located between the two parties choose a candidate from each side. To avoid the extremes of either party, then, middle-of-the-road voters choose a Democratic president and a Republican House member, or vice versa. This behaviour is especially strong under conditions of certainty such as those found at the midterm election. (9)
Crucial to policy-balancing models is the idea that voters see meaningful differences between the parties. (10) Thus, party polarization will increase split-ticket voting and party convergence will diminish its frequency. Despite the many critics of policy balancing, (11) evidence supportive of, or consistent with, its propositions represents a growing area of scholarship. (12) One burgeoning area of inquiry about policy balancing comes from research on non-separable voting preferences--whether a voter's preference for partisan control of Congress depends on the outcome of the presidential race or vice versa. Here, researchers have found evidence that voters take into account both branches of government--the larger policy-making picture--when expressing voting preferences. (13) In the context of the Jeffords switch, purposeful explanations of divided government would predict that ideological moderates who see a difference between the political parties will support Jeffords' party switch.
Yet, while an overall appreciation of the larger policy-making process lingers in the background of citizen consciousness, the primary actors involved in the Jeffords switch make up citizens' conscious thinking on the matter and, I contend, constitute the most accessible or important considerations on the matter. Democrats taking control of the Senate away from Republicans, obstructing President Bush's agenda and ultimately restoring divided government by the action of a single senator should increase the salience and importance of these actors. Unintentional or non-purposive theories of divided government emphasize such considerations. Non-purposive explanations propose that divided government is the product of well-recognized factors that shape voting decisions, namely candidate attributes, issues and voter predispositions, such as party identification. None of these explanations posits purposive behaviour in the sense that voters intentionally choose candidates from different parties for Congress and the presidency to produce 'balanced' policy outcomes.
Voting for a congressional incumbent or quality challenger, for instance, may produce a split-ticket vote without any desire for divided government. (14) Based on this reasoning, I expect Jeffords' constituents to evaluate the switch through the lens of incumbency and the 'personal vote', (15) the ingredients that routinely help congressional incumbents win re-election. Congress and the president, of course, also lie at the centre of Jeffords' decision and approval (or disapproval) of these actors will affect attitudes about the switch for both Jeffords' constituents, and the public at large. Supporters of President George W. Bush, not wanting to hinder his administration' s agenda, will oppose the switch. Since at the time of the survey (late August) the Senate was firmly under the control of the Democrats and the House under the control of the Republicans, my expectations about congressional approval are less clear.
The weight voters assign to candidate characteristics depends on how strongly they identify with a political party. Voters with strong party attachments are less susceptible to candidate-driven factors whereas voters with weak party attachments are more likely to consider them. Not surprisingly, weak partisans or Independents are more likely to split their tickets than strong partisans are. (16) I expect partisanship to act as a lens through which citizens interpret politics, in this case to guide a citizen's response to Jeffords' decision. (17) Simply put, Democrats will approve of the switch and Republicans will oppose it.
Issues may also cause voters to split their tickets in a given contest by giving one party an agenda advantage in a presidential or congressional race (18) or by matching the qualities voters seek in congressional and presidential candidates better, (19) In Petrocik and Doherty's theory of issue ownership, parties try to set the agenda with issues they 'own' or on which they have an advantage. (20) Republicans, therefore, stress taxes and crime and Democrats emphasize the environment and health care. In the end, the electoral advantage goes to the party that succeeds in this endeavour. The argument is easily extended to opinions towards the parties between elections. If the major issues of the day belong to the Republicans, they should benefit in the arena of public opinion and the same should hold true for the Democrats. Thus, citizens who believe the Democrats better able to handle the major issues of the day are more likely to approve of the Jeffords switch.
MEASURING ATTITUDES ABOUT DIVIDED GOVERNMENT
The data for this analysis come from the CBS News Vermont State Poll (ICPSR 3345) and a CBS News Monthly Poll (ICPSR 3346), both taken at the end of August, several weeks after Jeffords' defection from the Republican party. (21) CBS News asked respondents in both polls a nearly identical question about Jeffords' party change--the differences attributable to the question being asked of a national sample and the senator's home state:
As you may know, in May Senator Jim Jeffords of Vermont left the Republican party and became an Independent, giving control of the Senate to the Democrats. Do you approve or disapprove of Jeffords' decision'?
The question provides most of the essential information without explicitly mentioning, 'divided government'. As discussed below, it probably represents an improvement over standard survey questions that ask about divided government in the abstract. Of course, it also has its limitations. Given that understanding public attitudes about divided government necessarily lies in the questions asked, I expound on the strengths and weaknesses of this question, especially as it relates to prior work on measuring voter attitudes about divided government.
Ideally, the question would have mentioned the president as well as his party affiliation. Since it does not, one could argue that it simply taps attitudes about Jeffords switching parties and handing control of the Senate over to the Democrats--and no more. (22) Yet media accounts of the switch make this interpretation unlikely. News stories focused on what the Jeffords switch meant for partisan control of government and how a Democratic Senate would frustrate the efforts of the Bush administration. Interviews with Senators Joseph Lieberman, Tom Daschle, Trent Lott and White House spokesman Ari Fleisher, to name a few, focused on what Jeffords' decision meant for partisan control of the Senate, and national policy making. In short, reporting on the switch highlighted the implications of Jeffords' decision for national governance, not that an obscure Senator changed parties. As noted, recent party switches by other US senators received far less news coverage probably because the implications of their party switches for national governance were minimal.
In addition, the fact that the question mentions control of the Senate switching to Democratic party control fits nicely with policy-balancing explanations that 'presume that some citizens have a general appreciation of the institutional structure of American government and that such institutional considerations enter into their voting decisions'. (23) I too assume that citizens have an overall sense of government by understanding that handing control of the Senate to the Democrats means divided government. This is a plausible interpretation of citizens' reactions to this question given that it highlights the fact that Jeffords' decision gave control of the Senate to the Democrats rather than other equally plausible considerations (such as admiration of politicians who exit the two-party system). Coupled with media coverage that highlighted the partisan implications of the switch, it is likely that respondents interpreted the question mostly in terms of the balance of power between the parties across the branches of government. Of course, many citizens may not think about 'divided government' or split-party control of government in explicit terms. Rather, they are likely to consider how a Republican president and Democratic Senate will offer alternative programmes, producing 'gridlock' and/or moderate policy outcomes.
Another possible limitation of the question is that it concerns specific personalities and circumstances. Many studies on divided government have traditionally examined the House of Representatives given that it was (and is) more ideological and more party polarized than the Senate. Candidate-centred elements have traditionally played a larger role in Senate contests, so it is less likely that voters would easily discard these considerations. Despite this limitation, many of the idiosyncrasies of the Jeffords switch should not hinder or benefit either explanation for divided government. For example, if partisan control of government had been reversed, the predictions for intentional and unintentional explanations would not change. Ideological moderates who perceive differences between the parties, for instance, would be equally as inclined towards Jeffords' creation of divided government as they would be towards voting for it--the desire for split-party control is the same. (24)
Some may also argue that pre-existing partisan arrangements are troublesome for assessing unbiased attitudes towards divided government. For example, Lacy and Paolino criticize the NES question for containing 'a strong partisan component'. They conclude, 'The NES question ... does not separate a sincere preference for divided government from a conditional preference that depends on which party is already in power'. (25) The idea is that a Democrat and Republican party identifier will have a different opinion about divided government depending on the existing partisan control of government. However, the group posited as most desiring of divided government, ideological moderates, should not care about pre-existing partisan control of government. Indeed, theories of voting for divided government at the midterm presume knowledge of partisan control of the presidency and argue that because of such knowledge voting for divided government is more pronounced at the midterm. (26) Under these conditions ideologically moderate voters--including voters previously unwilling to split their tickets under uncertainty during the presidential election year because they do not know the outcome of the presidential contest--choose candidates from the opposite party to 'balance' the president. (27) In this way, the Jeffords switch 'mimics' the political environment of the midterm election. Reasoning under complete information, the Jeffords switch enables ideological moderates to express something akin to a 'real attitude' about divided government.
Despite limitations, the Jeffords' question has a few notable advantages over previous questions on divided government. A distinct advantage of the Jeffords' question is that it makes specific reference to all the actors involved and avoids abstraction. In contrast, survey questions about divided government typically inquire about partisan control of government in the abstract. The standard National Election Study (NES) question, for example, asks respondents whether they like it 'better when one party controls both the presidency and Congress; better when control is split between the Democrats and Republicans; or doesn't it matter?' Using this question from the 1992 NES, Sigelman, Wahlbeck and Buell tested whether voters who prefer divided government are more likely to split their tickets and found no relationship. (28) They concluded that this question is a 'non-attitude', 'a complicated matter to which most people probably have given little thought'.
Lacy and Paolino offer this same criticism of the standard NES question in arguing that it makes little sense to ask questions about political parties per se since voters evaluate candidates, not parties. (29) While a voter may not care about unified or divided government, they argue that a 'voter might vote for Clinton for president and Republicans for Congress if she thinks Clinton and a Democratic Congress would be too far left and Dole and a Republican Congress too far right'. (30) In a candidate-centred era of elections, this point makes sense: questions about divided government should be specific about political actors. Although Lacy and Paolino offer the names of presidential candidates, they do not present the names of congressional candidates. Indeed, no studies of non-separable voting preferences make specific mention of congressional candidates. (31) Consequently, these studies neglect the effects of congressional candidates on voter preferences, a leading cause behind unintentional theories of divided government. When possible, then, the proper names of congressional candidates should be included in survey questions about divided government. Assuming partisan knowledge of the president, the Jeffords question pits a Republican president versus a Democratic congress, and for Vermont respondents, a noted elected representative--all familiar ingredients in the choices voters face.
CITIZENS' ATTITUDES ABOUT THE JEFFORDS SWITCH
The Jeffords switch was popular among citizens. A majority of respondents in both the Vermont and national samples approved of Jeffords party switch. (32) In Vermont, 65 per cent of citizens approved, 28 per cent disapproved, and 7 per cent fell into the don't know/no answer category. In the national sample, 49 per cent approved, 30 per cent disapproved, and 21 per cent were coded don't know/no answer. Obviously, Jeffords' move was more popular in Vermont than the nation.
To get a better sense of which citizens approved, Table 1 reports approval and disapproval of Jeffords' decision for each sample by various political characteristics and attitudes. The greatest support for the Jeffords switch comes from Democrats. Indeed, of all the categories, Democratic identifiers in both samples were the most likely to approve of Jeffords' party change. In the Vermont sample, 94 per cent of Democrats approved and in the national sample 87 per cent of Democrats approved. Interestingly, although the majority of Republicans in both samples opposed the switch (about three-quarters), Republican disapproval is not as high as Democratic approval. The largest partisan difference between the two samples is among Independents: 81 per cent of Independents in Vermont approved of Jeffords' switch, whereas 61 per cent of Independents in the national sample did so. This difference is consistent with an incumbency effect, whereby non-partisans rely on name recognition and constituency service more than party labels. Of course, like many Senators, it may also reflect Jeffords' success at cultivating support among Independents, a core electoral constituency.
Approval of the Jeffords switch also differs by ideology. Liberals in Vermont and nation-wide approved of Jeffords' decision, 91 per cent and 86 per cent respectively. Majorities of conservatives in both samples disapproved of Jeffords' switch, although these percentages, like the party differences between Republicans and Democrats, were not as high as liberals' approval. Of Vermont conservatives 52 per cent disapproved of the switch and 56 per cent of conservatives from the national sample did so. The potency of partisan and ideological considerations, for whatever reason, appears to be greater for Democrats and liberals than for Republicans and conservatives. Consistent with policy-balancing explanations, ideological moderates also showed support for the Jeffords switch. In the Vermont and national samples approximately two-thirds of moderates approved of Jeffords' switch. There is not much of a difference between Vermont respondents when asked whether they saw a difference between the political parties, the other essential ingredient of the policy-balancing model. Although 67 per cent of respondents who saw differences between the parties approved of the Jeffords switch, 75 per cent of those who saw no difference also approved.
Moderates who saw differences between the two parties make up the crucial group that policy-balancing models propose as the citizens that desire divided government. Contrary to the expectations of the balancing model, this group does not stand out in its approval of the Jeffords switch (not shown). Of moderates who saw a difference between the parties 68 per cent approved of Jeffords' switch, but this percentage is a bit smaller than the 71 per cent of respondents overall who approved of the Jeffords switch. Looking just at moderates, seeing a difference between the parties did not appear to make any difference towards approval of the Jeffords switch either. Of moderates who saw a difference between the parties 68 per cent approved of the Jeffords switch, whereas 67 per cent of moderates who did not see a difference between the parties approved. Furthermore, neither of these relationships is statistically significant. (33)
Citizens' attitudes towards political figures also figure prominently in approval/disapproval of the Jeffords switch. Although the percentages differ somewhat between the national and Vermont samples, respondents who approved of President Bush generally did not approve of the Jeffords switch. Among those who disapproved of Bush, 86 per cent approved and 14 per cent disapproved of the switch in both the Vermont and national samples. Those who approved of Bush's job performance, however, were less uniform in their disapproval of Jeffords' switching parties. In Vermont, 46 per cent of respondents who approved of the president approved of the switch while nationally 38 per cent did so. Among Jeffords' constituents who approved of his job performance, 90 per cent approved of his party switch. By contrast, of those respondents who disapproved of Jeffords' job performance, 23 per cent approved of his party switch.
Though it appears that approval of the Jeffords switch lies firmly in partisanship, ideology and attitudes towards political figures, other factors, when taken into account, may obscure these relationships. For these reasons, I formulate and test multivariate models of approval of the Jeffords switch.
MULTIVARIATE ANALYSES
To examine the impacts of the various factors on approval of Jeffords' party switch, I estimate a series of probit models. The dependent variable consists of two outcomes, approval (coded 1) and disapproval (coded 0). The independent variables include the partisan, ideological and politician-centred variables from Table 1, as well as control variables (see the Appendix for variable coding).
Where possible, I include all variables in each model. However, as mentioned, the Vermont poll includes items specific to constituents' opinions about the Jeffords switch. This discrepancy is not surprising given that Jeffords represents Vermont and many questions tapping the constituent--representative linkage are not relevant to citizens in other states. Not surprisingly, then, the Vermont poll includes more questions specific to the switch and generally allows a more thorough investigation than the national poll. Nevertheless, the national poll includes a few items not found in the Vermont data for testing theories of divided government.
Table 2 presents the results of the probit analyses for Vermont respondents. The goodness of fit statistics suggest that the model does a good job explaining attitudes towards the Jeffords switch. The proportional reduction in error statistic ([[lambda].sub.p]) suggests that the model produces over a 54 per cent reduction in error over the modal category. Using a two-tailed test of statistical significance, many of the predictors in the model are statistically significant at p < 0.05 or better.
Partisan considerations and politician-centred variables appear to be the predominant influences on approval of the Jeffords switch. Taking control of the Senate from Republicans, Democrats approved of the Jeffords switch; Democrat is statistically significant (p < 0.001) and in the predicted direction. For ease of interpretation, I calculated the change in predicted probability of approval of Jeffords' party switch. This procedure involved looking at a change in an independent variable from its minimum to maximum value while holding all other variables at their mean. Using this method, if a person identifies as a Democrat his or her probability of approving the Jeffords switch is 0.27 greater than a Republican, holding other variables at their mean. Independents were also likely to approve of Jeffords' decision. The coefficient is positive and statistically significant, p < 0.001. Substantively, identifying as an Independent increased the probability of approval by 0.18 compared to Republicans.
The coefficient for Jeffords' Approval is statistically significant at p < 0.001 and in the predicted direction. Substantively, this variable appears to have the largest effect on approval of the Jeffords switch. Vermonters that approved of Jeffords had a probability of approving of his party switch 0.58 greater than citizens who disapproved. This finding is consistent with candidate-centred explanations of ticket splitting that show the causes of divided government flow from well-funded and recognized, high quality candidates. (34) Opinion about President Bush also figured prominently in Jeffords' defection as well. The coefficient for Presidential Approval is statistically significant (p < 0.001) and negative, which suggests that respondents who approved of President Bush's job performance disapproved of Jeffords' party switch. Substantively, approving of President Bush's job performance decreases the probability of approving of the Jeffords switch by 0.26, holding other variables to their means. It is likely that respondents who approved of President Bush's job performance did not want to see an obstructionist Democratic Senate obstructing or modifying Bush's policy agenda. Although I did not have any prior expectations about congressional approval, approval of congress' job performance had a positive effect on approval of the Jeffords switch.
Ideological leanings, however, did not appear to have had an effect on approval of the Jeffords switch for Vermont citizens. Although significantly different from zero, the sign for Moderate is negative, contrary to the policy-balancing model. (35) The crucial test of the balancing model, however, involves moderate voters who saw a difference between the parties. The coefficient for the interaction effect between Moderate and Party Difference is in the predicted direction, but it is not statistically distinguishable from zero. Contrary to policy-balancing models, then, there appears to be no relationship between moderate citizens who saw a difference between the parties and approval of the Jeffords switch. (36) Although included as a control variable, Age is the only demographic variable that is statistically distinguishable from zero. I will discuss this finding below.
These results suggest that the preference for divided government, at least as measured by approval of Jeffords' party switch, is a by-product of factors devoid of any policy-balancing intention on the part of citizens. The most important predictor appears to be attitudes towards Jeffords, namely his constituent's approval of his job as Senator. In the final analysis, the constituent-representative relationship is the most appropriate place to look for the causes of divided government. When voters produce divided government, it happens congressional district by congressional district or state by state. In other words, it is the local choices of candidates, and all that they bring to the table, that bring about divided government. (37)
Table 3 presents the results of a probit analysis for the national sample. Although the fit of the national sample model is not as impressive as its Vermont counterpart, it is nevertheless respectable. The proportional reduction in error statistic is 0.55 and the pseudo [R.sup.2] is 0.37.
The results, in many respects, look similar to the Vermont model. For example, partisan identification plays an important role in both models. Democrats and Independents both approved of the Jeffords switch, the coefficients for each variable are statistically significant and in the predicted direction. Respondents identifying as Democrats and Independents have a probability of approving of the Jeffords switch 0.29 and 0.19 greater than Republican party identifiers, respectively. Presidential and congressional approval evaluations also play a similar role in shaping national opinion about the Jeffords switch. Each of these variables is statistically significant, with approval of President Bush negatively associated with approval of the Jeffords switch and approval of Congress positively associated. Disapproving of the president decreases the probability of approving the Jeffords switch by 0.24 compared to those who approve of the president. Congressional approval, by contrast, has a substantive effect of 0.15, almost identical to the Vermont analysis.
As mentioned, additional questions on the national poll permit further examination of unintentionalist claims. Specifically, the national poll included items about the party 'most likely to make the right decisions' about the federal budget, Social Security, keeping the economy strong and spending taxpayers' money. By combining these items into a single additive index, I am able to test the theory of issue ownership--the notion that the party best able to handle the important issues of the day will garner advantage in the court of public opinion. For each 'Democratic party' response to a question, the respondent receives a score of one. Thus, Democratic Issue Ownership ranges in value from 0 to 4, lower values representing Republican ownership and higher values representing Democratic ownership. Of course, I expect this variable to be associated with an increase in the probability that a respondent will approve of the Jeffords switch. As seen, this variable is statistically significant (p > 0.001) and in the predicted direction. It also appears to be the strongest predictor in the model, increasing the probability of approval of the Jeffords switch by 0.32. Taken together, partisan considerations--party identification and beliefs about the party's capabilities of handling important issues--constitute the major ingredients of approval of the Jeffords switch for the national sample. Thus, in the absence of constituency-representative factors, partisan considerations and thus unintentional approaches were central to evaluating Jeffords' decision.
In contrast, the policy-balancing model does not fare well. Although in the predicted direction, Moderate is not statistically distinguishable from zero. Unfortunately, the national poll did not include a question about whether the parties differed, so I cannot address a nuanced policy-balancing argument.
Although the control variables do not appear to have much effect in explaining attitudes towards the Jeffords switch, Age merits discussion. The coefficient for Age is statistically significant with a negative sign, suggesting that older citizens were less likely to approve of the Jeffords switch. Curiously, this effect is quite strong. Holding all other variables at their mean values, a change from the lowest to highest age category decreases the probability of approval of Jeffords' switch by 0.17. This finding is consistent with DeVries and Tarrance's research that shows ticket-splitting is less common among older voters. (38)
CONCLUSION
The Jeffords switch provided a unique opportunity to investigate questions about divided government. In particular, it allowed an investigation of public attitudes about divided government apart from ticket-splitting, a behaviour that ultimately causes divided government, but may have little to say about whether citizens actually desire it. By eschewing an abstract question about divided government, an approach that scholars find deficient for studying mass opinion on divided government, this study is grounded in 'real' American politics.
The results of this study provide support for unintentional explanations of divided government. Indeed, partisan and politician-centred cues made up the major ingredients of public attitudes towards the Jeffords switch. Although partisan considerations were strong, the constituent-representative linkage in the Vermont analysis was especially potent as constituents' approval of Jeffords' job as senator appeared to have the greatest impact on approval of his switch. Without the dominance of incumbency-centred factors, partisan considerations and issues made up the primary ingredients of public attitudes towards the Jeffords switch in the national data. Thus, outside the constituency-representative link, it appears that party best explains whether citizens want a unified or divided government. Democrats want Democratic party control of government and Republicans want Republican party control of government. In addition, the content of the agenda and its partisan tilt had a substantial effect on public attitudes towards the Jeffords switch as well. Citizens who believed the Democrats (Republicans) were best able to handle important issues facing government were more (less) likely to approve of the Jeffords switch.
Policy-balancing explanations, however, did not fare well. Awash in considerations of party and political figures, the desire for balanced government among ideological moderates' did not matter much. In particular, the Vermont analyses provided an in-depth examination of the policy-balancing model by looking at moderates who saw a difference between the two major parties. As with looking simply at ideological moderates, I did not find support for this more nuanced analysis of the policy-balancing argument. Given the circumstances of the Jeffords switch it is surprising that ideological moderates, especially those who saw a difference between the parties, did not approve. Jeffords handed these citizens divided government, an outcome they should have cheered. Furthermore, he did so under conditions of certainty, a situation not unlike that of the midterm election where even the most cautious ideological balancer votes for divided government.
The Jeffords switch, however, takes place within a specific historical context, thus limiting broad generalizations. In contrast to earlier studies of divided government where there was greater certainty (or less confusion) about the ideological status of the branches of government, the fact that control of the Senate was nearly evenly split may have clouded voters' perceptions of its ideological or policy orientations. In addition, the controversy surrounding the 2000 presidential contest had not entirely subsided creating the possibility that opinion on the Jeffords switch was anchored to a president with a feeble mandate and questionable legitimacy.
Although I am reluctant to generalize these findings beyond this time period, they suggest that moderate citizens do not endorse divided government any more than they endorse unified government. The supporters of divided government included those citizens uneasy or upset by unified Republican government: Democrats, those who disapproved of President Bush's job performance, and backers of Senator Jeffords. Those opposed to divided government, not surprisingly, had the opposite profile: Republicans, citizens who approved of President Bush's job performance, and Jeffords detractors back home. In short, support for divided government, at least as measured by approval of the Jeffords switch, is a matter of sincere preferences for parties and politicians and has little, if anything, to do with intentionally balancing the branches of government.
APPENDIX: VARIABLE CODING
1. Partisan Identification. Unfortunately, the survey did not ask respondents the items required to construct the traditional seven-point scale. The measures of party identification, Democrat and Independent, are based on the question, 'Generally speaking, do you usually consider yourself a Republican, a Democrat, an Independent, or what?' Democrat = 1, Otherwise = 0; Independent = 1, Otherwise = 0. Republican identification is the unexpressed category.
2. Ideology. The measures of ideology, Liberal and Moderate, are based on a question asking respondents, 'How would you describe your views on most political matters? Generally, do you think of yourself as liberal, moderate, or conservative?' Liberal = 1, Otherwise = 0; Moderate = 1, Otherwise = 0. Conservative is the unexpressed category.
3. Presidential Approval. This variable is based on the question, 'Do you approve or disapprove of the way George W. Bush is handling his job as President?' Approval = 1; Disapproval = 0.
4. Congressional Approval. This variable is based on the question, 'Do you approve or disapprove of the way Congress is handling its job?' Approval = 1; Disapproval = 0.
5. Economy. This variable is based on the question, 'Do you think the economy is getting better, getting worse, or staying about the same?' Getting Worse = 1; Otherwise = 0. This item is available only in the national poll.
6. Democratic Issue Ownership. The variable is based on four questions asking respondents their opinions about the party more likely to keep the economy strong, make the fight decisions about the federal budget, Social Security, and how government should spend tax payers' money. The questions, with slight variations, 'Regardless of how you usually vote, do you think George W. Bush or the Democrats in Congress are more likely to--?' Each Democratic response is coded l, 0 otherwise. The variable, then, ranges from 0 (all Republican responses) to 4 (all Democratic responses). This item is available only in the national poll.
7. Jeffords' Approval. This variable is based on the question, 'Do you approve or disapprove of the way Jim Jeffords is handling his job as Senator?' Approval = 1; Disapproval = 0. This item is only available in the Vermont poll.
8. Party Difference. This variable is based on the question, 'Would you agree or disagree with the following statement: it makes no real difference which party controls Congress--things go on just as they did before.' Disagree = 1; Agree = 0. This item is only available in the Vermont poll.
9. Female. Female = 1; Male = 0.
10. Age. This variable consists of four categories: 18 and 29 = 0; 30 and 44 = 1; 45 and 64 = 2; Over 64 = 3.
11. Education. This variable consists of five categories: did not graduate from High School = 0; High School graduate = 1; some college = 2; college graduate = 3; post graduate work or degree = 4.
12. Income. This variable consists of five categories: under $15,000 = 0; $15,000 to $30,000 = 1; $30,000 to $50,000 = 2; $50,000 to $75,000 = 3; Over $75,000 = 4. Available only in national poll.
13. African-American. African-American = 1; Otherwise = 0. This item is only available in the national poll.
TABLE 1 Approval of the Jeffords Switch by Party Identification,
Ideology and Political Attitudes
Vermont
Approve Disapprove
Variable % %
Party Identification
Democrat 94 6
Republican 24 76
Independent 81 19
N 593
Ideology
Liberal 91 9
Conservative 48 52
Moderate 68 32
N 606
Presidential Approval
Approve 46 54
Disapprove 86 14
N 636
Jeffords' Approval
Approve 90 10
Disapprove 23 77
N 636
Party Difference
Difference 67 33
No Difference 75 25
N 608
Nation
Approve Disapprove
Variable % %
Party Identification
Democrat 87 13
Republican 26 74
Independent 61 39
N 645
Ideology
Liberal 86 14
Conservative 44 56
Moderate 67 33
N 673
Presidential Approval
Approve 38 62
Disapprove 86 14
N 684
Jeffords' Approval
Approve Not included
Disapprove
N
Party Difference
Difference Not included
No Difference
N
Sources: CBS News Vermont State Poll (ICPSR 3345), August 2001, and
CBS News Monthly Poll (ICPSR 3346), August 2001.
TABLE 2 Probit Analysis of Vermont Citizens'
Approval of Senator Jeffords' Party Switch
Independent Standard Change in
variables Coefficient error probability
Democrat 1.34 *** (0.251) 0.27
Independent 0.73 *** (0.168) 0.18
Liberal 0.23 (0.232) 0.06
Moderate -0.55 * (0.259) -0.15
Presidential
Approval -0.89 *** (0.169) -0.26
Congressional
Approval 0.32 (0.171) 0.08
Jeffords'
Approval 1.83 *** (0.171) 0.58
Female 0.29 (0.156) 0.07
Age -0.19 * (0.086) -0.15
Education 0.04 (0.068) 0.04
Party
Difference -0.48 (0.227) -0.12
Moderate x
Party
Difference 0.40 (0.315) 0.10
Constant -0.45 (0.296) --
N 601
[chi square] 394.05 ***
Pseudo [R.sup.2] 0.54
Predicted correctly 87%
PRE ([[lambda].sub.p]) 0.55
*** p < 0.001 (two-tailed test); ** p < 0.01 (two-tailed test);
* p < 0.05 (two-tailed test).
Note: Change-in-probability values are calculated by looking at changes
in an independent variable from the minimum to maximum value holding
all other variables at their mean values.
TABLE 3 Probit Analysis of Approval of Senator Jeffords' Party Switch,
National Poll
Independent Standard Change in
variables Coefficient error probability
Democrat 0.82 *** (0.230) 0.29
Independent 0.54 ** (0.186) 0.19
Liberal 0.55 * (0.235) 0.19
Moderate 0.07 (0.171) 0.03
Presidential
Approval -0.65 ** (0.215) -0.24
Congressional
Approval 0.40 * (0.159) 0.15
Democratic
Issue Ownership 0.22 *** (0.066) 0.32
Economy 0.30 (0.166) 0.11
Female 0.24 0.09
Age -0.15 * (0.076) -0.17
Education 0.05 (0.073) 0.08
Income -0.01 (0.065) 0.01
African-American 0.08 (0.290) 0.03
Constant -0.65 * (0.317) --
N 442
[chi square] 219.47 ***
Pseudo [R.sup.2] 0.37
Predicted Correctly 82%
PRE ([[lambda].sub.p]) 0.55
*** p < 0.001 (two-tailed test); ** p < 0.01 (two-tailed test);
* p < 0.05 (two-tailed test).
Note: Change-in-probability values are calculated by looking at
changes in an independent variable from the minimum to maximum
value holding all other variables at their mean values.
(1) For background studies, see Alberto Alesina and Howard Rosenthal, Partisan Politics, Divided Government, and the Economy (New York: Cambridge University Press, 1995); R. Michael Alvarez and Matthew M. Schousen, 'Policy Moderation or Conflicting Expectations? Testing the Intentional Models of Split-Ticket Voting', American Politics Quarterly, 21 (1993), 410-38; Richard Born, 'Split-Ticket Voters, Divided Government, and Fiorina's Policy-Balancing Model', Legislative Studies Quarterly, 19 (1994), 95-115; Barry C. Burden and David C. Kimball, 'A New Approach to the Study of Ticket Splitting', American Political Science Review, 92 (1998), 533-44, and Why Americans Split Their Tickets: Campaigns, Competition, and Divided Government (Ann Arbor: University of Michigan Press, 2002); Morris Fiorina, Divided Government, 2nd edn (Needham Heights, Mass.: Allyn and Bacon, 1996); Paul Frymer, Thomas Paul Kim and Terri S. Bimes, 'Party Elites, Ideological Voters, and Divided Party Government', Legislative Studies Quarterly, 22 (1997), 195-216; James C. Garand and Marci Glascock Lichtl, 'Explaining Divided Government in the United States: Testing an Intentional Model of Split-Ticket Voting', British Journal of Political Science, 30 (2000), 173-91 ; Bernard Grofman, William Koetzle, Michael P. McDonald and Thomas L. Brunell, 'A New Look at Split Ticket Outcomes for House and President: The Comparative Midpoints Model', Journal of Politics, 62 (2000), 34-50; Gary C. Jacobson, The Electoral Origins of Divided Government (Boulder, Colo.: Westview Press, 1990); Dean Lacy, 'Back from Intermission: The 1994 Elections and the Return to Divided Government', in Herbert F. Weisberg and Samuel C. Patterson, eds, Great Theatre: The American Congress in the 1990s (New York: Cambridge University Press, 1998); Dean Lacy and Philip Paolino, 'Downsian Voting and the Separation of Powers', American Journal of Political Science, 421 (1998), 180-99; Walter R. Mebane Jr, 'Coordination, Moderation, and Institutional Balancing in American Presidential and House Elections', American Political Science Review, 94 (2000), 37-58; Bruce I. Oppenheimer, 'Split Party Control of Congress, 1981-86: Exploring Electoral and Apportionment Explanations', American Journal of Political Science, 33 (1989), 653-69; Charles E. Smith, Robert D. Brown, John M. Bruce and L. Marvin Overby, 'Party Balancing and Voting for Congress in the 1996 National Election', American Journal of Political Science, 43(1999), 737-64.
(2) Alesina and Rosenthal, Partisan Politics, Divided Government, and the Economy; Robert S. Erikson, 'The Puzzle of Midterm Loss', Journal of Politics, 50 (1988), 1011-29; Fiorina, Divided Government; Lacy and Paolino, 'Downsian Voting and the Separation of Powers'; Everett C. Ladd, 'Public Opinion and the "Congress Problem"', The Public Interest, 100 (Summer), 57-67; Mebane, 'Coordination, Moderation, and Institutional Balancing in American Presidential and House Elections'; Smith et al., 'Party Balancing and Voting for Congress in the 1996 National Election'.
(3) Alvarez and Shousen, 'Policy Moderation or Conflicting Expectations?'; Paul Allen Beck, Lawrence Baum, Aage R. Clausen and Charles E. Smith Jr, 'Patterns and Sources of Ticket Splitting in Subpresidential Voting', American Political Science Review, 86 (1992), 916-28; Born, 'Split-Ticket Voters, Divided Government, and Fiorina's Policy-Balancing Model'; Burden and Kimball, 'A New Approach to the Study of Ticket Splitting' and Why Americans Split Their Tickets: Campaigns, Competition, and Divided Government; Frymer et al., 'Party Elites, Ideological Voters, and Divided Party Government'; Grofman et al., 'A New Look at Split Ticket Outcomes for House and President'; Jacobson, The Electoral Origins of Divided Government; John R. Petrocik, 'Divided Government: Is It All in the Campaigns?', in Gary W. Cox and Samuel Kernell, eds, The Politics of Divided Government (Boulder, Colo.: Westview Press, 1991); John R. Petrocik and Joseph Doherty, 'The Road to Divided Government: Paved without Intention' in Peter F. Galderisi, ed., Divided Government: Change, Uncertainty, and the Constitutional Order (Lanham, Md.: Rowman and Littlefield Publishers, 1996).
(4) Petrocik and Doherty, 'The Road to Divided Government'.
(5) Lee Sigelman, Paul J. Wahlbeck and Emmett H. Buell Jr, 'Vote Choice and the Preference for Divided Government: Lessons of 1992', American Journal of Political Science, 41 (1997), 879-94.
(6) Fiorina, Divided Government.
(7) For more information, see Sigelman et al., 'Vote Choice and Preference for Divided Government'.
(8) Fiorina, Divided Government.
(9) Alesina and Rosenthal, Partisan Politics, Divided Government, and the Economy.
(10) Fiorina, Divided Government, p.81.
(11) Alvarez and Shousen, 'Policy Moderation or Conflicting Expectations?'; Beck et al., 'Patterns and Sources of Ticket Splitting in Subpresidential Voting'; Born, 'Split-Ticket Voters, Divided Government, and Fiorina's Policy-Balancing Model'; Burden and Kimball, 'A New Approach to the Study of Ticket Splitting'; Frymer et al., 'Party Elites, Ideological Voters, and Divided Party Government'; Grofman et al., 'A New Look at Split Ticket Outcomes for House and President'; Petrocik and Doherty, 'The Road to Divided Government'; Gary M. Segura and Stephen P. Nicholson, 'Sequential Choices and Partisan Transitions in U.S. Senate Delegations', Journal of Politics, 57 (1995), 86-100; Sigelman et al., 'Vote Choice and Preference for Divided Government'.
(12) Alesina and Rosenthal, Partisan Politics, Divided Government, and the Economy; Fiorina, Divided Government; Dean Lacy, 'A Theory of Nonseparable Preferences in Survey Responses', American Journal of Political Science, 45 (2001), 239-258; Lacy and Paolino, 'Downsian Voting and the Separation of Powers'; Mebane, 'Coordination, Moderation, and Institutional Balancing in American Presidential and House Elections'; Smith et al., 'Party Balancing and Voting for Congress in the 1996 National Election'.
(13) Lacy, 'A Theory of Nonseparable Preferences in Survey Responses'; Lacy and Paolino, 'Downsian Voting and the Separation of Powers'; Smith et al., 'Party Balancing and Voting for Congress in the 1996 National Election'.
(14) Alvarez and Shousen, 'Policy Moderation or Conflicting Expectations?'; Born, ,Split-Ticket Voters, Divided Government, and Fiorina's Policy-Balancing Model'; Burden and Kimball, 'A New Approach to the Study of Ticket Splitting'; Jacobson, The Electoral Origins of Divided Government.
(15) Bruce Cain, John A. Ferejohn and Morris P. Fiorina, The Personal Vote: Constituency Service and Electoral Independence (Cambridge, Mass.: Harvard University Press, 1987).
(16) Beck et al., 'Patterns and Sources of Ticket Splitting in Subpresidential Voting'; Garand and Lichtl, 'Explaining Divided Government in the United States'; Angus Campbell and Warren E. Miller, 'The Motivational Basis of Straight and Split Ticket Voting', American Political Science Review, 51 (1957), 293-312; Sigelman et al., 'Vote Choice and Preference for Divided Government'.
(17) Angus Campbell, Philip E. Converse, Warren E. Miller and Donald E. Stokes, The American Voter (New York: John Wiley, 1960).
(18) For presidential, see Petrocik, 'Divided Government', for congressional, see Petrocik and Doherty, 'The Road to Divided Government'.
(19) I would have liked to test Jacobson's thesis, but neither poll contained appropriate items.
(20) For presidential, see Petrocik, 'Divided Government', for congressional, see Petrocik and Doherty, 'The Road to Divided Government'.
(21) CBS News, 2002; CBS News Monthly Poll, August 2001 [Computer File]. ICPSR version. New York, NY: CBS News [producer], 2001. Ann Arbor, Mich.: Inter-university Consortium for Political and Social Research [distributor] and CBS News, 2002. CBS News Vermont State Poll, August 2001 [Computer File]. ICPSR version. New York, NY: CBS News [producer], 2001. Ann Arbor, Mich.: Inter-university Consortium for Political and Social Research [distributor].
(22) The question also does not inform respondents about party control of the House of Representatives. Yet, this should not be a problem because policy-balancing takes place across branches, not within (e.g., House and Senate). Ideologically moderate balancers who thought the Democrats controlled the House would still prefer the Senate be in the hands of the opposite party of the president. Thus, because these models feature a president of one party balanced against a Congress controlled by the other party, not knowing a respondent's knowledge of control of the House should not stack the deck against intentionalist approaches. Furthermore, this should not be troublesome for testing unintentional explanations since they do not depend on this distinction. Partisans, for instance, will always want their party in control, regardless of who is currently in control.
(23) Fiorina, Divided Government, p. 64.
(24) Nor should it matter if control switched from divided to unified--cognitive Madisonians would disapprove just as they would approve party control going from unified to divided. Similarly, unintentional explanations should not be at a comparative advantage or disadvantage. If Al Gore had won the 2000 presidential election with a Democratic Congress, Republican citizens would surely have approved a Democratic Senator such as Zell Miller of Georgia switching sides and creating divided government. Furthermore, given Miller's popularity in Georgia, many of his constituents would likely have approved (in much the same way as Jeffords' constituents approved). In short, although more cases with different politicians and circumstances would be helpful, the specifics of the Jeffords case should not help or hinder any of the competing explanations for citizens' opinions of divided government.
(25) Lacy and Paolino, 'Downsian Voting and the Separation of Powers', p. 1196.
(26) Alesina and Rosenthal, Partisan Politics, Divided Government, and the Economy; Erikson, 'The Puzzle of Midterm Loss'.
(27) Alesina and Rosenthal, Partisan Politics, Divided Government, and the Economy.
(28) Sigelman et al., 'Vote Choice and Preference for Divided Government', p. 891.
(29) Lacy and Paolino, 'Downsian Voting and the Separation of Powers'.
(30) Lacy and Paolino, 'Downsian Voting and the Separation of Powers', p. 1196.
(31) Lacy, 'Back from Intermission'; Lacy, 'A Theory of Nonseparable Preferences in Survey Responses'; Lacy and Paolino, 'Downsian Voting and the Separation of Powers'; Smith et al., 'Party Balancing and Voting for Congress in the 1996 National Election'.
(32) All analyses are weighted using CBS News' 'final' weight.
(33) z = 0.569 (p = 0.5693) comparing moderates who see a difference between the parties to all citizens and z = -0.821 (p = 0.4116) when comparing moderates who see a difference between the parties to moderates who see no difference between the parties.
(34) It is likely that approval of Jeffords' job as senator is an endogenous variable. In other words, not only does approval of Jeffords' job as senator affect approval of his party switch but also the reverse should be true. Unfortunately, the limitations of the data preclude a two-stage estimation of this potential endogeneity. Although the CBS News poll includes additional items concerning opinions of Jeffords (e.g., vote to re-elect) that might serve as instrumental variables, these variables correlate highly with both the endogenous regressor and the dependent variable, approval of the party switch. Accordingly, using these instruments would produce estimates that are biased. Given that any potential instruments in the data are themselves endogenous to the dependent variable and thus the error term, it appears unwise to proceed with an instrumental variables approach. For a discussion of this point, see Larry M. Barrels, 'Instrumental and "Quasi-Instrumental" Variables', American Journal of Political Science, 35 (1991), 777-800.
Despite this limitation of the data, it seems unlikely that the effect of approval of the Jeffords switch on approval of Jeffords' job as senator is a more powerful cause than feelings about Jeffords' job as senator on opinions of Jeffords' party switch. The general orientation towards Jeffords, approval of his job performance, is likely to affect opinions of an action, his party switch, since the former is a general disposition and the latter is a specific action.
Furthermore, in the context of the survey, the question on approval of Senator Jeffords job as senator was asked third following standard questions about presidential and congressional approval, respectively. The question about approval of Jeffords' party switch, the dependent variable, was Question 7. Thus, there were no cues in the survey about Jeffords' party switch to affect opinion of his job as senator. Coupled with the fact that the poll was conducted more than three months after Jeffords' party switch and media attention had shifted elsewhere, it is unlikely that respondents were thinking about it at the beginning of the survey when asked about his job approval.
(35) To evaluate the robustness of this finding, I dropped the variable Independent from the model to evaluate whether its presence affected the results for Moderate. This variable still did not obtain statistical significance nor did its sign change.
(36) To evaluate the robustness of this finding, I also ran this model excluding Independents. The coefficients for Party Difference and the interaction term remain insignificant. However, Moderate obtains statistical significance (p < 0.05) but its sign remains negative, contrary to the expectations of the policy-balancing model.
(37) Burden and Kimball, 'A New Approach to the Study of Ticket Splitting' and Why Americans Split Their Tickets; Jacobson, The Electoral Origins of Divided Government.
(38) Walter De Vries and Lance Tarrance, The Ticket-Splitter: A New Force in American Politics (Grand Rapids, Mich.: Eerdmans, 1972).