Robust Estimation and Outlier Detection for Overdispersed Multinomial Models of Count Data

We develop a robust estimator—the hyperbolic tangent (tanh) estimator—for overdispersed multinomial regression models of count data. The tanh estimator provides accurate estimates and reliable inferences even when the specified model is not good for as much as half of the data. Seriously ill-fitted counts—outliers—are identified as part of the estimation. A Monte Carlo sampling experiment shows that the tanh estimator produces good results at practical sample sizes even when ten percent of the data are generated by a significantly different process. The experiment shows that, with contaminated data, estimation fails using four other estimators: the nonrobust maximum likelihood estimator, the additive logistic model and two SUR models. Using the tanh estimator to analyze data from Florida for the 2000 presidential election matches well-known features of the election that the other four estimators fail to capture. In an analysis of data from the 1993 Polish parliamentary election, the tanh estimator gives sharper inferences than does a previously proposed heteroskedastic SUR model .     

An Estimator for Some Binary-Outcome Selection Models Without Exclusion Restrictions

This article provides a new maximum-likelihood estimator forselection models with dichotomous dependent variables when identicalfactors affect the selection equation and the equation of interest.Such situations arise naturally in game-theoretic models whereselection is typically nonrandom and identical explanatory variablesinfluence all decisions under investigation. When identicalexplanatory variables influence selection and a subsequent outcomeof interest, the commonly used Heckman-type estimators identifyfrom distributional assumptions about the residuals alone. Whenits own identifying assumption is reasonable, the new estimatorallows the researcher to avoid the painful choice between identifyingfrom distributional assumptions alone and adding a theoreticallyunjustified variable to the selection equation in a mistakenattempt to "boost" identification. The article uses Monte Carlomethods to compare the small-sample properties of the estimatorwith those of the Heckman-type estimator and ordinary probit.     

Parliamentary procedure: principal forms and political effects

A standard conclusion of theorists who model bargaining as a non-cooperative game is that the party designated to make the first move—the formateur party—will determine the bargaining outcome. Most empirical studies of parliamentary coalition formation have paid surprisingly little attention to the formation process. In this paper we model government formation as a two-stage unordered discrete choice problem that better reflects this process. The first step involves the selection of a formateur party, and the second involves the choice of partners by the predicted formateur. We evaluate several hypotheses for the two stages, using a data set of all cabinets formed in the Western European countries from 1970 to 2006. In our analyses of formateur selection, we find that party size is clearly the dominant feature. In the second stage, we show that when predicting government composition it is fruitful to add information drawn from a first stage analysis.     

Central bank independence and inflation revisited

We re-examine the relationship between central bank independence (CBI), proxied by the central bank governor’s turnover rate and an indicator based on central bank laws in place, and inflation using a random coefficient model with the Hildreth-Houck estimator for more than 100 countries in the period 1980 to 2005. We conclude that there exists no general significant negative relation between our indicators of central bank independence and inflation. Central bank independence has a significant effect only in a minority of the countries in our sample.     

Ecological inference under unfavorable conditions: Straight and split-ticket voting in diverse settings and small samples

Problems of ecological inference have long troubled political scientists. Thomsen's (1987) estimator for ecological inference has been shown to produce estimates close to the individual level estimates for transitions across elections, but it is unknown how well it performs under unfavorable conditions. We fill this void by testing the estimator as the across-unit variance increases and introduce a new procedure to examine the bias of the estimates as the number of aggregate units decreases. Looking at partisan voting patterns across races within the 2000 general election in Florida counties and taking advantage of ballot image data to study straight-ticket voting we demonstrate that the estimator performs well in both heterogeneous societies and when the number of aggregate units is limited.     

Efficient Estimation of Time-Invariant and Rarely Changing Variables in Finite Sample Panel Analyses with Unit Fixed Effects

e-mail: tpluem{at}essex.ac.uk (corresponding author) This paper suggests a three-stage procedure for the estimationof time-invariant and rarely changing variables in panel datamodels with unit effects. The first stage of the proposed estimatorruns a fixed-effects model to obtain the unit effects, the secondstage breaks down the unit effects into a part explained bythe time-invariant and/or rarely changing variables and an errorterm, and the third stage reestimates the first stage by pooledOLS (with or without autocorrelation correction and with orwithout panel-corrected SEs) including the time-invariant variablesplus the error term of stage 2, which then accounts for theunexplained part of the unit effects. We use Monte Carlo simulationsto compare the finite sample properties of our estimator tothe finite sample properties of competing estimators. In doingso, we demonstrate that our proposed technique provides themost reliable estimates under a wide variety of specificationscommon to real world data. Authors' note: Earlier versions of this paper have been presentedat the 21st Polmeth Conference at Stanford University, PaloAlto, July 29–31, 2004, the 2005 MPSA Conference in Chicago,April 7–10, and the APSA Annual Conference 2005 in Washington,September 1–4 2005. We thank the referees of PoliticalAnalysis and Neal Beck, Greg Wawro, Donald Green, Jay Goodliffe,Rodrigo Alfaro, Rob Franzese, Jörg Breitung, and PatrickBrandt for helpful comments on previous drafts. Any remainingdeficiencies are our responsibility.     

Patterns of democracy reconsidered: The ambiguous relationship between corporatism and consensualism

Arend Lijphart's Patterns of Democracy, similar to most of his work, elicited fierce scientific debate. This article replicates some of the analyses proposed in its second edition (published in 2012) in the light of the critiques received by the first edition (published in 2009). It primarily examines the relationship between institutional setup and interest group representation, disentangling the effect of consensualism from that of corporatism on issues such as macroeconomic performance and governance capabilities. The article further deepens our understanding of the complex causal mechanisms connecting these variables, proposing a more sophisticated empirical investigation that emphasises selection effects and conjunctural causation.     

Membership vote for party leadership changes: Electoral effects and the causal mechanisms behind

Political parties in parliamentary democracies have increasingly democratized their leadership selection processes, incorporating the votes of party members. Despite generating numerous headlines, there has been a relative dearth of cross-national scholarly work on the electoral effects of selectorate expansion and the causal mechanisms behind them. This study fills this gap in the literature. Using observational data from eleven parliamentary democracies, we show that parties using membership selection can expect a polling boost when compared to those using more exclusive mechanisms. However, membership selection does not affect electoral performance. Nevertheless, our crossnational analyses and results from a survey experiment from Australia suggest that incorporating members generates excitement, demonstrates an openness to new ideas, and can be a signal of leader work ethic and a commitment to the democratic process, increasing leader legitimacy. We discuss the disconnect between these positive evaluations and the lack of electoral effects, and suggest possible strategies for parties to improve their electoral standing.     

Bias in Conditional and Unconditional Fixed Effects Logit Estimation

Fixed-effects logit models can be useful in panel data analysis,when N units have been observed for T time periods. There aretwo main estimators for such models: unconditional maximum likelihoodand conditional maximum likelihood. Judged on asymptotic properties,the conditional estimator is superior. However, the unconditionalestimator holds several practical advantages, and thereforeI sought to determine whether its use could be justified onthe basis of finite-sample properties. In a series of MonteCarlo experiments for T < 20, I found a negligible amountof bias in both estimators when T 16, suggesting that a researchercan safely use either estimator under such conditions. WhenT < 16, the conditional estimator continued to have a verysmall amount of bias, but the unconditional estimator developedmore bias as T decreased.     

Sex isn’t Gender: Reforming Concepts and Measurements in the Study of Public Opinion

The importance of sex and gender to political behavior is reflected in the volume of work examining gender gaps in public opinion and partisan choice. Despite their centrality, sex and gender are poorly measured in survey research. The principal problem is the conflation of gender with sex in survey research. Consequently, gender is typically treated as a dichotomy, with no response options for androgynous gender identities, or indeed degrees of identification with masculine or feminine identities. We compare a new measure of genuine gender identification with a conventional measure of biological sex to determine whether the practice of using sex as a proxy for gender is sound. Sex is a fair proxy for gender, but for about a quarter of our sample, it is not. Moreover, greater nuance is gained when analyses incorporate a finer-grained measure of gender than is possible by using biological sex as a substitute. We argue that this is simply the start to an important conversation and that more research is needed to ascertain how we might best measure “gender” in the future.     

Analyzing Censored and Sample-Selected Data with Tobit and Heckit Models

Political scientists are making increasing use of the Tobitand Heckit models. This paper addresses some common problemsin the application and interpretation of these models. Throughnumerical experiments and reanalysis of data from a study byRomer and Snyder (1994), we illustrate the consequences of usingthe standard Tobit model, which assumes a censoring point atzero, when the zeros are not due to censoring mechanisms orwhen actual censoring is not at zero. In the latter case, wealso show that Greene's (1981) well-known results on the directionand size of the bias of the OLS estimator in the standard Tobitmodel do not necessarily hold. Because the Heckit model is oftenused as an alternative to Tobit, we examine its assumptionsand discuss the proper interpretation of the Heckit/Tobit estimationresults using Grier and co-workers' (1994) Heckit model of campaigncontribution data. Sensitivity analyses of the Heckit estimationresults suggest some conclusions rather different from thosereached by Grier et al.     

Random Coefficient Models for Time-Series-Cross-Section Data: Monte Carlo Experiments

Jonathan N. Katz Division of the Humanities and Social Sciences, California Institute of Technology, Pasadena, CA 91125 e-mail: jkatz{at}caltech.edu e-mail: nathaniel.beck{at}nyu.edu (corresponding author) This article considers random coefficient models (RCMs) fortime-series–cross-section data. These models allow forunit to unit variation in the model parameters. The heart ofthe article compares the finite sample properties of the fullypooled estimator, the unit by unit (unpooled) estimator, andthe (maximum likelihood) RCM estimator. The maximum likelihoodestimator RCM performs well, even where the data were generatedso that the RCM would be problematic. In an appendix, we showthat the most common feasible generalized least squares estimatorof the RCM models is always inferior to the maximum likelihoodestimator, and in smaller samples dramatically so. Authors' note: We gratefully acknowledge the financial supportof the National Science Foundation. Katz also acknowledges thesupport of the Center for Advanced Study in the Behavioral Sciences.We are thankful to Jake Bowers, Rob Franzese, Andy Gelman, SandyGordon, Bill Greene, and Luke Keele for comments; to Larry Bartelsfor always reminding us that our judgment may outperform thedata; as well as to the anonymous reviewers of Political Analysis.     

Individual and contextual characteristics of the German extreme right-wing vote in the 1990s. A test of complementary theories

Abstract In this article we set out to improve our knowledge on voting for extreme right–wing parties, i.e. the Republikaner, by taking into account social, political and contextual characteristics. We test four theories that provide explanations as to why certain social categories are more likely to vote for the Republikaner. The hypotheses are tested with multilevel analyses, with data from a national sample (N = 4688). Multinomial analyses provide additional information on how theoretically derived political attitudes effect voting behaviour.     

Individual and contextual characteristics of the German extreme Right–Wing vote in the 1990s. A test of complementary theories

In this article we set out to improve our knowledge on voting for extreme right–wing parties, i.e. the Republikaner, by taking into account social, political and contextual characteristics. We test four theories that provide explanations as to why certain social categories are more likely to vote for the Republikaner. The hypotheses are tested with multilevel analyses, with data from a national sample (N = 4688). Multinomial analyses provide additional information on how theoretically derived political attitudes effect voting behaviour.     

External Validity in Policy Evaluations That Choose Sites Purposively

Evaluations of the impact of social programs are often carried out in multiple sites, such as school districts, housing authorities, local TANF offices, or One‐Stop Career Centers. Most evaluations select sites purposively following a process that is nonrandom. Unfortunately, purposive site selection can produce a sample of sites that is not representative of the population of interest for the program. In this paper, we propose a conceptual model of purposive site selection. We begin with the proposition that a purposive sample of sites can usefully be conceptualized as a random sample of sites from some well‐defined population, for which the sampling probabilities are unknown and vary across sites. This proposition allows us to derive a formal, yet intuitive, mathematical expression for the bias in the pooled impact estimate when sites are selected purposively. This formula helps us to better understand the consequences of selecting sites purposively, and the factors that contribute to the bias. Additional research is needed to obtain evidence on how large the bias tends to be in actual studies that select sites purposively, and to develop methods to increase the external validity of these studies. © 2012 by the Association for Public Policy Analysis and Management.     

The limits and rewards of political opportunism: How electoral timing affects the outcome of currency crises

Politicians are often assumed to be opportunistic. This article examines both whether there is a limit to this opportunism and whether voters reward policy makers for opportunistic behaviour. By looking at currency crisis situations, the article presents a graphic rational opportunistic political business cycle model in which incumbents face a tradeoff between their wish to signal competence and the economic constraints imposed by the crisis. It analyses how electoral incentives affect policy makers' management of currency crises and how this management in turn affects the subsequent election outcome. The empirical results of probit models with selection using a sample of 122 crises in 48 industrial and developing countries between 1983 and 2003 confirm the model's prediction that under certain circumstances some types of policy makers do indeed have incentives to deviate from optimal policy in the run-up to elections – and that voters reward this behaviour by re-electing policy makers who follow such strategies. However, there is a limit to the readiness to manipulate: when speculative pressure is too severe, incumbents no longer manipulate policy but implement the least painful policy option instead.     

Stability through control? The influence of coalition agreements on the stability of coalition cabinets

Why do some coalition cabinets terminate early and others run until the end of the legislative term? This article analyses whether coalition agreements lower the risk of early government termination. The main argument is that coalition agreements can increase the stability of coalition cabinets as they lower the probability of intra-cabinet conflict. The theoretical expectations are empirically evaluated on the basis of a newly compiled comprehensive dataset on cabinet duration and control mechanisms in coalition governments. Drawing on event history analysis, the effect of coalition agreements on cabinet duration is tested for 420 coalition cabinets from 1945 until 2015 in 23 Western and Eastern European countries. The results show that the existence and duration of a coalition agreement lower the risk of early government termination. These findings have important implications for our understanding of the interaction between coalition governance and coalition termination.     

Locating power in party leader selection

In this article, the aim is to enhance our understanding of who has power over leader selection in political parties. To this end, we apply an analytical framework in which the selection process is divided into three phases: gatekeeping, preparation and decision. The focus is on determining the extent to which each of these phases is influential for the outcome and thereby locating the distribution of intra-party power. Underpinning the analysis is the conviction that the comparison of leader selection is too limited if it relies solely on information about formal procedures, including the composition of the selectorate. We should also take the preselection phase of leader selection into account. Empirically, we examine a sample of recent selection processes in European parliamentary democracies. In contrast to previous research on intra-party politics, which has suggested an ascendancy of the party in public office, our results suggest an enduring strength of the party on the ground and the party in the central office.     

Robust voting

The formal equivalence between social choice and statistical estimation means that criteria used to evaluate estimators can be interpreted as features of voting rules. The robustness of an estimator means, in the context of social choice, insensitivity to departures from majority opinion. In this paper we consider the implications of substituting the median, a robust, high breakdown estimator, for Borda's mean. The robustness of the median makes the ranking method insensitive to outliers and reflect majority opinion. Among all methods that satisfy a majority condition, median ranks is the unique one that is monotonic. It is an attractive voting method when the goal is the collective assessment of the merits of alternatives.     

Cherry Picking with Synthetic Controls

We evaluate whether a lack of guidance on how to choose the matching variables used in the Synthetic Control (SC) estimator creates specification-searching opportunities. We provide theoretical results showing that specification-searching opportunities are asymptotically irrelevant if we restrict to a subset of SC specifications. However, based on Monte Carlo simulations and simulations with real datasets, we show significant room for specification searching when the number of pre-treatment periods is in line with common SC applications, and when alternative specifications commonly used in SC applications are also considered. This suggests that such lack of guidance generates a substantial level of discretion in the choice of the comparison units in SC applications, undermining one of the advantages of the method. We provide recommendations to limit the possibilities for specification searching in the SC method. Finally, we analyze the possibilities for specification searching and provide our recommendations in a series of empirical applications.