The importance of attitudes and the processes by which they are formed and changed is ubiquitous throughout political science. Perhaps the most obvious example is research exploring citizens' attitudes toward candidates, how these attitudes are influenced by political advertising and other persuasive messages, and how these attitudes influence decisions and behavior (see McGraw's chapter in this volume). Attitudes toward candidates are fundamental to the democratic process because they help voters make vote choices, perhaps the most basic way in which citizens can express their opinions and influence government (e.g., Rosenstone and Hansen 1993). Other key attitudes in the political domain include attitudes toward specific policies that also help voters make important decisions about voting, vote choice, and activism (e.g., Rosenstone and Hansen 1993). Attitudes toward institutions such as political parties and government entities also influence people's view of government. Finally, attitudes toward other groups in society (e.g., African Americans, women) may help determine support for specific policies (e.g., Transue 2007). Thus, attitudes play a central role in many of the democratic processes studied by political scientists reviewed in this volume (e.g., Gadarian and Lau; Hutchings; Lodge and Tabor; McGraw; Nelson; Wilson and Eckel).

Collective action problems are difficult problems that pervade all forms of social organization from within the family to the organization of production activities within a firm and to the provision of public goods (PGs) and the management of common pool resources (CPRs) at local, regional, national, and global scales. Collective action problems occur when a group of individuals could achieve a common benefit if most contribute needed resources. Those who would benefit the most, however, are individuals who do not contribute to the provision of the joint benefit and free ride on the efforts of others. If all free ride, then no benefits are provided.

Political scientists trying to analyze collective action problems have been influenced by a narrow, short-term view of human rationality combining an all-powerful computational capacity, on the one hand, with no capability to adapt or acquire norms of trustworthiness and fair contributions to the provision of collective benefits, on the other. To provide PGs, it is believed that governments must devise policies that change incentives to coerce citizens to contribute to collective action.

Formal analysis of collective action problems has been strongly affected by the pathbreaking work of Olson (1965) on The Logic of Collective Action and the use of game theory (e.g., Hardin 1982; Taylor 1987), which improved the analytic approach to these problems.

Within the prevailing Fisher-Neyman-Rubin framework of causal inference, causal effects are defined as comparisons of potential outcomes under different treatments. In most contexts, it is impossible or impractical to observe multiple outcomes (realizations of the variable of interest) for any given unit. Given this fundamental problem of causality (Holland 1986), experimentalists approximate the hypothetical treatment effect by comparing averages of groups or, sometimes, averages of differences of matched cases. Hence, they often use (Ȳ|t = 1) − (Ȳ|t = 0) to estimate E[(Yi|t = 1) − (Yi|t = 0)], labeling the former quantity the treatment effect or, more accurately, the average treatment effect.

The rationale for substituting group averages originates in the logic of the random assignment experiment: each unit has different potential outcomes; units are randomly assigned to one treatment or another; and, in expectation, control and treatment groups should be identically distributed. To make causal inferences in this manner requires that one unit's outcomes not be affected by another unit's treatment assignment. This requirement has come to be known as the stable unit treatment value assumption.

Until recently, experimenters have reported average treatment effects as a matter of routine. Unfortunately, this difference of averages often masks as much as it reveals. Most crucially, it ignores heterogeneity in treatment effects, whereby the treatment affects (or would affect if it were actually experienced) some units differently from others.

In his 1909 American Political Science Association presidential address, A. Lawrence Lowell (1910) advised the fledgling discipline against following the model of the natural sciences: “We are limited by the impossibility of experiment. Politics is an observational, not an experimental science…” (7). The lopsided ratio of observational to experimental studies in political science, over the one hundred years since Lowell's statement, arguably affirms his assessment. The next hundred years are likely to be different. The number and influence of experimental studies are growing rapidly as political scientists discover ways of using experimental techniques to illuminate political phenomena.

The growing interest in experimentation reflects the increasing value that the discipline places on causal inference and empirically guided theoretical refinement. Experiments facilitate causal inference through the transparency and content of their procedures, most notably the random assignment of observations (a.k.a. subjects or experimental participants) to treatment and control groups. Experiments also guide theoretical development by providing a means for pinpointing the effects of institutional rules, preference configurations, and other contextual factors that might be difficult to assess using other forms of inference. Most of all, experiments guide theory by providing stubborn facts – that is, reliable information about cause and effect that inspires and constrains theory.

In political surveys, many citizens fail to answer, or provide incorrect answers to, fact-based questions about political figures and institutions. A common inference drawn from such failures is that citizens' poor performance on surveys reflects their incompetence in democratically meaningful contexts such as voting booths.

The scholarly home for such findings is the academic literature on political knowledge. A common analytic definition of political knowledge is that it is a measure of a citizen's ability to provide correct answers to a specific set of fact-based questions. Typical political knowledge questions include “What is the political office held by [name of current vice president, British prime minister, or chief justice of the United States]?” and “Which political party has the most seats in the U.S. House of Representatives?” Many people have used responses to survey-based political knowledge questions to criticize the public for its general incompetence.

In recent years, these criticisms have come under increasing scrutiny (e.g., Graber 1984; Popkin 1994). Some scholars raised questions about the practice of basing broad generalizations of citizen competence or knowledge on a relatively small set of idiosyncratic, fact-based survey questions (e.g., Lupia 2006). Others uncovered logical and factual errors in claims about the kinds of political knowledge that are needed to make important political choices competently (e.g., Lupia and McCubbins 1998; Gibson and Caldeira 2009).

Knowledge about negotiation and mediation comes primarily from laboratory experiments. The question asked in this chapter is what value is added by experiments for understanding processes of elite bargaining? This question is addressed in the following sections. After describing the international negotiation context, I provide a brief overview of the experimental approach. Then, key studies on distributive and integrative bargaining are reviewed, as well as examples of experiments that capture complexity without forfeiting the advantages of experimental control. The chapter concludes with a discussion of the value added by experiments.

The Context

Negotiating in the international context takes several forms. It occurs from a distance and face to face; deals with multiple complex issues; and includes bilateral, multilateral, and global participation. National leaders often make demands or exchange proposals from a distance. Well-known examples include the bilateral exchanges between the United States and the Soviet Union over the 1948–49 blockade of Berlin, between Kennedy and Khrushchev in 1962 over Soviet missile bases in Cuba, and between Carter and Khomeini concerning American hostages in Iran in 1979–80. Leaders and their representatives also confront each other face to face to discuss their interests over security, monetary and trade, or environmental issues.

In this chapter, I follow other authors (e.g., Kagel and Roth 1995; McDermott 2002; Camerer 2003; Morton and Williams 2010) in focusing on a few key dimensions of difference between experiments in the economic and psychological traditions.

Section 1 considers the level of stylization typical in economics and psychology experimentation. Although research in the political psychology tradition tends to place an emphasis on the descriptive realism of laboratory scenarios, work in experimental economics tends to proceed within a purposefully abstract, “context free” environment.

Section 2 considers the kinds of incentives offered to subjects by experimentalists from these two schools of thought. Experimental economists generally offer subjects monetary incentives that depend on subjects' choices in the laboratory – and, in game-theoretic experiments, the choices of other subjects as well. In contrast, psychology research tends not to offer inducements that are conditional on subjects' actions, instead giving subjects fixed cash payments or fixed amounts of course credit.

Section 3 considers the use of deception. The psychological school tends to see deception as a useful tool in experimentation, and, at times, a necessary one. In contrast, the economic school by and large considers deception to be taboo.

Nowhere can democracy be better seen “in action” than during political campaigns. As Lau and Pomper (2002) argue, “Democracy is a dialogue between putative leaders and citizens. Campaigns provide the most obvious and the loudest forums for this dialogue. Candidates try to persuade voters to cast a ballot and to support their cause. Voters respond by coming to the polls and selecting their preferred candidates” (47). Candidates make their arguments in speeches at campaign rallies and on their Web sites, but those venues are primarily experienced by the most committed of partisans. It is only through their campaign advertisements that candidates have any chance of reaching uncommitted voters. And in times of even very approximate party balance, it is uncommitted voters who usually determine election outcomes. Hence, political ads are arguably the vehicle through which democracy operates.

Consider the choices facing a political candidate at the outset of a democratic election campaign. To simplify, let us assume a candidate's goal is to win the election, but he or she is facing one or more opponents who have the same goal and therefore want to defeat him or her. All candidates must decide what strategy to follow in order to maximize their chances of winning. But all candidates face resource limits that put very real constraints on what it is possible to do. Therein lies the rub. Candidates must decide how they can get the biggest bang for their buck.

Civic participation is an essential component of a healthy democracy. Voting allows citizens to communicate preferences to elected officials and influence who holds public office. At the same time, deficiencies and asymmetries of participation in the United States call into question the representativeness of elected officials and public policies. Yet, although political activity is crucial for the equal protection of interests, participation is often seen by individuals as irrational or excessively costly, and it is well known that turnout in the United States lags well behind that of other democracies. Scholars have consistently found that participation is linked to socioeconomic variables, psychological orientations, and recruitment. Candidates, parties, and organizations thus spend considerable effort mobilizing electoral activity. This chapter highlights contributions made by field experiments to the study of voter mobilization, the problems faced by such work, and opportunities for future study.

Observational Studies

Nonexperimental studies have primarily relied on survey research to demonstrate correlations between self-reported mobilization and civic-minded behaviors, while also controlling for demographic characteristics (e.g., age, education, income) that are known to be significant predictors of turnout. The conclusion usually reached is that mobilization efforts are generally effective (e.g., Rosenstone and Hansen 1993; Verba, Schlozman, and Brady 1995). However, four major empirical hurdles render this conclusion suspect.

For citizens to responsibly exercise one of their primary democratic duties, namely, voting, two preliminary psychological processes must occur. First, citizens must learn something about the candidates, that is, come to some understanding, even if amorphous, about the candidates' characteristics and priorities. Second, citizens must reach a summary judgment about the candidates. There is a long and distinguished history of scholarly studies of the linked processes of voting, perceptions of candidates, and evaluations of them. In his recent review of the voting behavior literature, Bartels (2010) notes, “The apparent failure of causal modeling [of observational data] to answer fundamental questions about voting behavior produced a variety of disparate reactions” (240), including scholars turning to experimentation to better understand these basic processes. My goal in this chapter is to outline experimental work that has contributed to our understanding of citizens' impressions and evaluations of political candidates, as well as to identify questions that future experiments might answer.

Clarification of Basic Concepts

Two sets of conceptual distinctions should be made clear at the outset. First, the chapter title distinguishes between candidate impressions and evaluations. By “impressions,” I mean an individual's mental representation – the cognitive structure stored in memory – consisting of knowledge and beliefs about another person. “Evaluation,” in contrast, refers to a summary global judgment ranging from very negative to very positive.

The literature on coalition experiments is blessed by a close connection between theory and empirical work, using both experimental and field data. In political science, the main research goal has been to understand the formation of coalition governments in multiparty democracies. Although other aspects of coalitions could be considered (e.g., coalition stability or the creation and maintenance of military alliances), they have not been the focus of much existing research. We therefore focus on experiments in coalition formation.

The goal of this chapter is not to provide an accurate description of the complete history of coalition experiments, but rather to discuss some of the key questions that continue to occupy researchers today. For example, much of the recent experimental literature has tested noncooperative models of coalition government. Although there has been a substantial amount of research on solution concepts from cooperative game theory, its performance to explain experimental data has long been considered unsatisfactory, leading some researchers such as Gamson (1964) to propose an “utter confusion” theory.

In addition to testing of theoretical models, experiments can also be used to supplement field studies. For example, field studies (Martin and Stevenson 2001; Diermeier, Eraslan, and Merlo 2003) have indicated the importance of constitutional features on government formation, specifically the presence of an investiture vote. Yet, the equilibria predicted by noncooperative models in general depend on institutional detail, such as the protocol in which offers can be made and so on.

Discoveries regarding the scope and meaning of majority rule instability have informed debate about the most fundamental questions concerning the viability of democracy. Are popular majorities the means of serving the public interest or a manifestation of the absence of equilibrium (Riker 1982)? Should majority rule legislatures be suspect or even avoided in favor of court decisions and bureaucratic delegation? Are the machinations of agenda setters the true source of what we take to be the legislative expression of majority rule? Are rules themselves subject to the vagaries of shifting majority coalitions (McKelvey and Ordeshook 1984)?

These and other questions were raised as a result of explorations in Arrovian social choice theory, which visualized group decisions as being the product of individual preferences and group decision rules such as majority rule. The biggest challenge to the research agenda was majority rule instability. In general, majority rule may not be able to produce a majority rule winner (an outcome that beats every other in a two-way vote). Rather, every possible outcome could lose to something preferred by some majority coalition. McKelvey (1976) showed that the potential for instability was profound, not epiphenomenal. A population of voters with known preferences might easily choose any outcome, or different outcomes at different times. If this were true, then how could scholars predict the outcome of a seemingly arbitrary and unconstrained majority rule institution, even with perfect knowledge of preferences? Was literally anything possible?

The largest literature on gender and experimentation in political science concerns voter reaction to candidate gender. One of the earliest and most enduring questions in the study of gender and politics concerns women's election to office. Because the number of women candidates and officeholders has increased in the United States over the past several decades, there are more cases of women candidates and officeholders available for empirical analysis. Today, women are a majority of the electorate, and women candidates tend to win their races at rates similar to those of men. Yet, the gender gap in candidacy and office holding remains large and stable. Understanding how voter beliefs about candidate gender shape attitudes and political behavior remains an important area for research.

Experimentation has helped scholars overcome some of the limitations of using observational studies to investigate candidate gender. As Sapiro (1981) observed, public opinion surveys may not be able to detect prejudice against women candidates if voters provide socially desirable responses. And if prejudice against women is subconscious, then voters may not even be aware of their attitudes. Observational studies are also limited in helping us understand what we cannot observe, namely, why far fewer women than men seek office. If women fail to run because they fear a gendered backlash from voters, then we are unable to evaluate the experiences of those women.

Randomized experiments, the most rigorous methodology for testing causal explanations for phenomena in the social sciences, are experiencing a resurgence in political science. The classic experimental design randomly assigns the population of interest into two groups, treatment and control. Ex ante these two groups should have identical distributions in terms of their observed and unobserved characteristics. Treatment is administered based on assignment, and by the assumptions of the Rubin causal model, the average effect of the treatment is calculated as the difference between the average outcome in the group assigned to treatment and the average outcome in the group assigned to control.

Randomized experiments are often conducted within a multilevel setting. These settings can be defined at the level at which the randomization occurs, as well as at the level at which the treatment is both directly and indirectly administered. These indirect effects most often occur as a result of social transmission of the treatment, which is particularly likely when the treatment consists of information. This chapter explores the implications of these multilevel settings to highlight the importance of careful experimental design with respect to random assignment of the population and the implementation of the treatment. There are potential problems with analysis in this context, and this chapter suggests strategies to accommodate multilevel settings. These problems are most likely to occur in field settings where control is lacking, although they can sometimes occur in other settings as well.