Abstract: We use our "PM and the Pendulum" Model to forecast the outcome of the 2011 General election. The vote function of the model, aside from a cyclical dynamic, relies of approval of the prime mister as the sole predictor. We find that PM Approval predicts the vote (and vote intention between elections) more accurately than does Government Approval. Turning to the forecasting of seats, we examing the accuracy of the autoregressive model of the vote-seat translation against the uniform-swing model, which is widely used by pollsters and the media. Testing the alternatives on election data since 1910, our autoregressive vote-seat translation model proves superior to the uniform-swing model.
Abstract: Models of presidential success in the legislature have sometimes focused on the importance of political capital and presidential approval and at other times looked at the partisan environment within which the president acts. We develop time series models of success that refine and integrate these two perspectives while attempting to reframe the matter in terms of the literature on congressional parties. Measures of the ideological and partisan makeup of the House of Representatives, including a measure of conditional party government (CPG), are used to explain how the president becomes more or less successful from 1953-2006. We fine that partisanship is an important lens through which to judge success but that the approval of the president's base is important as well. We also demonstrate the electoral consequences to congressional parties of presidential success. While work in the Congress literature has discussed the importance of collective reputations as a link between legislative success and electoral success, our findings indicate that congressional parties gain and lose seats based on the battles won and lost by the president. This gives legislators (not) of his party an incentive to see his agenda implemented (defeated). Looking at both the causes and consequences of presidential success is meant to integrate theories of the two institutions along with extant theories of party behavior. Presidential Data (xls) House Data (xls) RATS program
The Electoral Costs of Party Loyalty in Congress (With Jamie L. Carson, Gregory Koger, and Everett Young, American Journal of Political Science, 2010, 54(3): 598-616) Abstract: To what extent is party loyalty a liability for incumbent legislators? Past research on legislative voting and elections suggests that voters punish members who are ideologically “out of step” with their districts. In seeking to move beyond the emphasis in the literature on the effects of ideological extremity on legislative vote share, we examine how partisan loyalty can adversely affect legislator’s electoral fortunes. Specifically, we estimate the effects of each legislator’s party unity—the tendency of a member to vote with his or her party on issues that divide the two major parties—on vote margin when running for reelection. Our results suggest that party loyalty can indeed be a liability for incumbent House members. In fact, we find that voters are not punishing elected representatives for being too ideological, they are punishing them for being too partisan. Online Appendix
(With Everett Young, The Journal of Elections, Public Opinion & Parties, 19(1), 117-124 2009)
Abstract: Modeling electoral support time series as fractionally integrated has been criticized on grounds that electoral support could instead be an autoregressive (AR) process with equilibrium shifts where major events occur—and common statistical tests for fractional integration (FI) cannot distinguish true FI from such a process. It has been argued additionally that even where FI is present, autoregressive moving-average models are adequate approximations of FI and are desirably less complicated, and in any event that estimates of d, the amount of FI, should account for suspected equilibrium shifts. We argue that strong theory predicts electoral support would be fractionally integrated, not autoregressive; that fractional differencing represents a more, not less, elegant solution to the need to filter electoral support series; that modeling FI processes as AR invites a host of undesirable statistical properties; and that theory demands major events be modeled as long-memoried shocks rather than equilibrium shifts. We show that a simulated autoregressive process with deliberately added equilibrium shifts behaves measurably differently in Box-Jenkins models (in being mean-reverting) than electoral support (which is not). We also argue that estimating d under the false assumption of equilibrium shifts invites bias. Tangentially, we are unconvinced by arguments that public opinion is asymptotically integrated of order 1. In sum, electoral support series are likely fractionally integrated, lack equilibrium shifts, and should be modeled that way.
(With Everett Young, The Journal of Elections, Public Opinion & Parties, 19(1), 73-103 2009)
Abstract: Political leadership has long been established as a key determinant of party support. Yet, the extent to which this may vary across parties in and out of government is less well understood. In fact, there are good reasons to expect leadership to matter less for “protest” parties. We demonstrate that the relationship between satisfaction with the Liberal Democratic leader and support for the British Liberal Democratic Party is fundamentally different than those between the Prime Minister and the Leader of the Opposition and their respective parties. Using 28 years of monthly public opinion data, we develop models of the political and economic factors that determine support for each of the three major parties in Britain in each of the 1979-1997 and 1997-2006 periods. We find that vote intentions for the two major parties are closely related with the satisfaction levels of their respective leaders but that this relationship is substantially weaker for the Liberal Democrats under the period of Tory rule and non-existent for the years of Labour rule. These findings are in contrast to individual-level studies that have shown the importance of leadership approval to vote choice for all British parties. More generally, we show the need to refine theories about the relationships between parties and their leaders.
(With Janet M. Box Steffensmeier, American Journal of Political Science, 52(3), 688-704 2008)
Abstract: Time-varying relationships and volatility are two methodological challenges that are particular to the field of time series. In the case of the former, more comprehensive understanding can emerge when we ask under what circumstances relationships may change. The impact of context—such as the political environment, the state of the economy, the international situation, etc.—is often missing in dynamic analyses that estimate time invariant parameters. In addition, time-varying volatility presents a number of challenges including threats to inference if left unchecked. Among time-varying parameter models, the Dynamic Conditional Correlation (DCC) model is a creative and useful approach that deals effectively with over-time variation in both the mean and variance of time series. The DCC model allows us to study the evolution of relationships over time in a multivariate setting by relaxing model assumptions and offers researchers a chance to reinvigorate understandings that are tested using time series data. We demonstrate the method’s potential in the first example by showing how the importance of subjective evaluations of the economy are not constant, but vary considerably over time as predictors of presidential approval. A second example using international dyadic time series data shows that the story of movement and comovement is incomplete without an understandin of the dynamics of their variance as well as their means. Approval Data (xls) Levant Data (xls) RATS program
(With Andrew Sidman and Maxwell Mak, Forthcoming at the International Journal of Forecasting) (With Daniel Cassino, Political Psychology, 28(6), 719-746 2007) Abstract: Research in political psychology has shown the importance of motivated reasoning as a prism through which individuals view the political world. From this we develop the hypothesis that, with strong positive beliefs firmly in place, partisan groups ignore or discount information about the performance of political figures they like. We then speculate about how this tendency should manifest itself in presidential approval ratings and test our hypotheses using monthly presidential approval data disaggregated by party identification for the 1955–2005 period. Our results show that partisan groups generally do reward and punish presidents for economic performance, but only those presidents of the opposite party. We also develop a model of presidential approval for self-identified Independents and, finally, a model of the partisan gap, the difference in approval between Democrat and Republican identifiers. *Data File (.xls) Data Dictionary RATS program
Abstract: Scholars differ regarding the reason for the institutionalization of a large, functionally specialized White House-centered presidential staff system during the last six decades. Among the explanations cited are a general growth in government size and complexity, an increase in the presidential workload, and the institutional rivalry between the president and Congress. However, using new advances in time series analysis based on fractional integration, we show that these models of staff growth are plagued by conceptual and methodological shortcomings that render their substantive conclusions misleading. In response, we develop and test a more comprehensive model of presidential staff institutionalization that takes into account the different ways in which this process takes place. Presidents institutionalize staff support, we conclude, to better manage relations with Congress, the media and the public. Government growth, policy complexity and presidential workload, on the other hand, play a much smaller role in this process than previous research indicates. *Appendix to paper
Abstract: We apply a dynamic perspective to forecasting votes and seats in British elections. Our vote model captures the swing of the electoral pendulum between the two major parties while using prime ministerial approval as the (sole) short-run predictor of vote choice. The seat model incorporates the inertia of the previous seat distribution while translating votes into seats. The models forecast the lead of one major party over the other (percentage for votes and number for seats). The statistical estimation includes data on British elections since 1945, although the test for cycles (swing of the electoral pendulum) goes as far back as 1832. The vote model picks the winner of every one of the 1945-2005 elections (out-of-sample forecasts) and is rarely off by more than 2 percentage points. The seat model does almost as well, rarely missing the seat lead by more than 25.
Abstract: Many studies have sought to clarify how voters’ opinions of the economy predict evaluations of leaders and parties. Following Kramer’s (1983) work on the problems of studying individual-level data, many authors have employed aggregate data in dynamic analyses to estimate rival models and choose favored variables. A restriction with such analyses is that they are unable to look closely at different periods within the overall study and thus may miss the importance of contextual factors. Our study complements existing aggregate-level inferences by analyzing repeated cross-sections of opinion polls in Britain over several years. We estimate the effects of subjective economic variables on vote intention in monthly public opinion surveys and examine how the parameters vary across individuals and over time. We suggest that the choices of whether voters are forward-looking, backward looking, egocentric or sociotropic are overly restrictive. We find that the sociotropic dimension dominates the egocentric dimension in evaluations of the government and that the relative importance of prospective and retrospective evaluations vary in predictable patterns over the election cycle. (With Benjamin G. Bishin and Robert Barr, Comparative Political Studies, 2006, 39(10): 1194-1219) Abstract: Elites’ support for democracy and their satisfaction with political leadership are important factors in evaluating Latin American leaders’ progress towards consolidating their democracies. However, we know surprisingly little about how elites understand or define democracy and thereby evaluate leaders in terms of their progress towards democracy. Much of the literature on the opinions of elites is focused on their relative interest in democratic values and formal institutions. But is progress in these two areas really of utmost importance to elites? In order to better understand elite views of democracy, we use new survey data in which elites assess current politicians’ progress towards democracy. We find that the importance of perceived progress in democratic values – rights and liberties – and formal institutions is minor compared to the impact of perceptions of economic progress. That is, elite evaluations of democratic progress depend primarily on their perceptions of economic success and only secondarily on their perceptions of achievement of democratic values like respect for civil rights and civil liberties.
Abstract: Recent developments in the analysis of long-memoried processes provide important leverage for analysing time-series variables of interest to political scientists. This article provides an accessible exposition of these methods and illustrates their utility for addressing protracted controversies regarding the political economy of party support in Britain. Estimates of the fractionally differencing parameter, d, reveal that governing party support, prime ministerial approval and economic evaluations are long-memoried and non-stationary, and that governing party support and prime ministerial approval are fractionally cointegrated. Pace conventional wisdom that party leader images matter little, if at all, analyses of multivariate fractional error correction models show that prime ministerial approval has important short-run and long-run effects on party support. Prospective and retrospective personal economic evaluations are influential but, contrary to a longstanding claim, national economic evaluations are not significant. The article concludes by suggesting that individual-level heterogeneity is a likely source of the observed aggregate-level fractional integration in governing party support and its determinants. Specifying parsimonious models that incorporate theoretically meaningful heterogeneity is a challenging topic for future research.
Abstract: Recent research by Box-Steffensmeier and Smith (Box-Steffensmeier, J.M., Smith, R.M., 1996. The dynamics of aggregate partisanship. American Political Science Review 90, 567–580; Box-Steffensmeier, J.M., Smith, R.M., 1998. Investigating political dynamics using fractional integration methods. American Journal of Political Science 42, 661–689) has alerted scholars to the problems involved in the analysis of fractionally integrated time series. This paper pursues this line of inquiry by compiling evidence on the time series properties of a number of common variables used in political research, including macropartisanship, presidential approval, the monthly and quarterly index of consumer sentiment, percentage liberalism in Supreme Court decision making, and others. In applying a variety of formal tests to these series, we fail to reject hypotheses of random walk or fractionally integrated processes while commonly rejecting hypotheses of stationary behavior. Evidence obtained from point estimates of the fractional differencing parameter, d, supports these findings while providing a glimpse into the long-memory characteristics of many political time series. Finally, Monte Carlo studies are performed that demonstrate the likelihood of spurious regressions when researchers fail to account for the fractional dynamics of time series. (Forthcoming at Congress and the Presidency. Accepted 11/07.) Abstract: A theory of partisan control might expect that during times of divided government, approval of one Congress and the Presidency would move in opposite directions. Yet, the Congressional approval question is an ambiguous one in the minds of voters and makes understanding the movement of the aggregated series much more difficult. Here, monthly Congressional and Presidential approval data for the 1995-2005 period are studied and found to move in tandem, even during periods of divided government. Multivariate ARFIMA Models show a strong and positive relationship between Congressional approval and lagged Presidential approval, even during the period of fiercely divided government under President Clinton. This means that, rather than paying attention to partisan control, the electorate transfers feelings about the president to the institution of Congress. |
     |
