Voting for Votes: Opposition Parties’ Legislative Activity and Electoral Outcomes

Abstract

Scholars frequently expect parties to act strategically in parliament, hoping to affect their electoral fortunes. Voters assumingly assess parties by their activity and vote accordingly. However, the retrospective voting literature looks mostly at the government’s outcomes, leaving the opposition understudied. We argue that, for opposition parties, legislative voting constitutes an effective vote-seeking activity as a signaling tool of their attitude toward the government. We suggest that conflictual voting behavior affects voters through two mechanisms: as a signal of opposition valence and as means of ideological differentiation from the government. We present both aggregate- and individual-level analyses, leveraging a dataset of 169 party observations from 10 democracies and linking it to the CSES survey data of 27,371 respondents. The findings provide support for the existence of both mechanisms. Parliamentary conflict on legislative votes has a general positive effect on opposition parties’ electoral performance, conditional on systemic and party-specific factors.

Introduction

When politicians and parties act in parliament, they frequently intend for their actions not merely to affect the present state of affairs but also to improve their fortunes in elections (Fernandes, Won, and Martins 2020; Müller and Strøm 1999; Proksch and Slapin 2015). This tendency should be even more pronounced for opposition parties, which hold a disadvantageous position compared with the government with respect to their influence on policy in the present. Indeed, recent literature includes evidence that (opposition) parties behave with such a calculus in mind (see, e.g., Williams 2011). This assumption that parties act in parliament with a constant eye to the next election makes sense when considering the vast literature on retrospective voting, the concept according to which voters decide how to cast their vote depending on their evaluation of parties’ performance preceding the elections (see, e.g., Ferejohn 1986; Fiorina 1981; Powell and Whitten 1993).

However, the retrospective voting literature is overwhelmingly focused on outgoing government parties and the extent to which they are rewarded or punished for their performance. ¹ Accordingly, the performance of opposition parties is implicitly dependent on government parties’ performance. This is in line with the stark contrast between the massive body of knowledge regarding governmental characteristics and actions and their electoral effects (e.g., De Vries and Solaz 2017; Ferejohn 1986; Healy and Malhotra 2013; Holbein 2016; Kim 2009; Strøm 1985) and our limited understanding of the consequences of parliamentary opposition behavior.

Therefore, we are led to ask whether opposition parties’ vote-seeking parliamentary behavior has the (purported) intended effect, or indeed—any effect—on electoral outcomes, specifically their own electoral performance in subsequent elections. We posit that voters rely, to some extent, on the visible behavior of parties to formulate their political preferences. Whereas the main object of retrospection for government parties is certain policy outputs, voters must rely on different activities to evaluate opposition parties. We analyze one such activity—namely, legislative voting. Although voters are not expected to observe such legislative behavior directly, we argue that legislative voting constitutes an effective mediatized signaling tool for opposition parties as an explicit and clearly attributable expression of party positions and their attitude toward the government.

As we demonstrate below, opposition parties use their parliamentary voting (also) as a means of communication to voters, often highlighting their differences from the government and their opposition to its agenda. Previous studies investigated the factors that determine the extent to which opposition parties adopt such a conflictual strategy (e.g., Louwerse et al. 2017; Tuttnauer 2018). The central question here is whether this strategy is successful.

We suggest that parliamentary conflict sends two kinds of messages to voters, activating two underlying mechanisms linking opposition parties’ legislative voting activity to their electoral outcomes. First, voters perceive opposition parties that frequently confront the government as doing a better job of being an “opposition,” a mechanism pertaining to what we call opposition valence. Second, voters perceive opposition parties that confront the government frequently as more ideologically distinct from the government, a mechanism we term ideological differentiation. Therefore, we hypothesize that although conflicting with the government will be beneficial to opposition parties in general, greater electoral benefit will be associated with political systems that prefer—institutionally and normatively—competition over cooperation between political sides. We further expect that conflict will mostly benefit larger opposition parties (those often seen as the main government alternatives) and mainstream parties that are more at risk of being perceived as “too similar” to the governing parties.

To test our hypotheses, we present both aggregate- and individual-level analyses. We analyze a cross-sectional longitudinal dataset linking 169 opposition parties’ positions on 389,142 legislative votes in 10 countries with results from 41 national elections. We find that greater conflict is electorally beneficial for opposition parties and that in line with our expectations, this holds for mainstream parties and parties in competition-leaning systems. Linking our data to a CSES dataset of 27,371 respondents, we also find that opposition parties that confront the government more frequently are perceived more favorably and as more ideologically extreme and distinct from the government, lending support to the existence of the two suggested mechanisms.

This article offers several contributions to existing research. First, we add a sorely needed broad, comparative analysis to the field of opposition studies, giving an essential insight into how the opposition influences electoral results and hinting at the kind of behavior for which voters reward parties. Second, we show that opposition parties do affect electoral outcomes through their legislative activity. Thereby, we give credence to a foundational assumption of rationality in the field of legislative studies, meaning that legislative actors behave strategically to achieve goals that they can indeed, at least partially, achieve. Finally, we add to the vast literature on retrospective voting by presenting evidence from the opposition’s perspective, thereby filling a lacuna left by a mostly government-focused field.

Opposition Parties’ Goals and Parliamentary Activity

The existence of political opposition, which criticizes the acts of a sitting government and puts forth its own candidates for election, has been acknowledged as a sine qua non of modern democracy. Furthermore, parliamentary opposition—defined as the sum of all parties represented in parliament and not members of the government—is considered the most advanced and institutionalized form of political conflict (Dahl 1966; Ionescu and de Madariaga 1968; Lipset 1963). Given the importance of oppositions, research into their role and the consequences of their behavior is surprisingly limited (see, e.g., Helms 2008). Most existing work focuses on a small number of cases (e.g., Andeweg, De Winter, and Müller 2008; Kopecky and Spirova 2008; Louwerse et al. 2017; Loxbo and Sjölin 2017; Seeberg 2013), limiting the ability to form and test generalizable theories.

Like all political parties, opposition parties seek policy influence, electoral success, and future office benefits (Müller and Strøm 1999). For opposition parties, gaining office benefits usually (unless a rare opportunity presents itself) entails improving electoral performance and must therefore be relegated to the future. We discuss below how opposition parties use parliamentary tools to achieve policy influence and electoral success, but this article focuses on electorally driven opposition behavior.

We acknowledge that not all opposition parties want to maximize votes to the same degree (see, e.g., West and Spoon 2013) and address two relevant influencing party features: party size and ideological distance from the government. Nevertheless, we argue that in opposition, parties should be particularly mindful of the electoral consequences of their parliamentary activity. This is because although opposition parties may be competent and effective at choosing which issue to politicize and on which to polarize, the policy-making agenda of the parliament is ultimately controlled to a significant degree by the government or its supporting legislative coalition. Our interest is how (successfully) opposition parties take advantage of the resulting legislative agenda to promote their vote-seeking goals, even in the face of limited policy influence. We elaborate on this idea in the following paragraphs.

Plenary time is a scarce resource. Having a say on how the parliament’s agenda is set is therefore of primary interest to all political parties. However, rules that regulate and constrain access to plenary time, directed at avoiding the impassable bottleneck implied by a legislative “state of nature,” are used in all modern parliaments and result in an unequal allocation of agenda-setting power, almost universally to the advantage of the government (Cox 2006).

Thus, it is no surprise that many recent studies focus on the potential influence of opposition parties on agenda-setting and its consequences for policies and electoral success. For example, Green-Pedersen (2019) highlights the opportunity of opposition parties to influence the “party system agenda.” Several scholars have shown that opposition parties use various legislative tools to influence the agenda and, ultimately, also policy outputs. For example, opposition parties use parliamentary questions to draw attention to issues important to them (Borghetto and Russo 2018; Vliegenthart and Walgrave 2011). Opposition parties are also successful in getting the government to react to their initiatives. The latter was found to take up issues emphasized by opposition parties (Green-Pedersen and Mortensen 2010) and react to opposition criticism by changing policies (Seeberg 2013).

However, opposition parties’ path to policy influence is generally limited compared with that of government parties. As discussed in more detail in relation with the opposition valence mechanism below, political systems vary in the advantage they lend to the government in agenda setting (Garritzmann 2017). In many parliamentary systems, the government yields privileged tools to control the parliament’s agenda. Most governments are supported by a legislative majority, which gives them greater control than the opposition even in the face of egalitarian procedural rules. Thus, some opposition parties’ uses of parliamentary tools are frequently explained by alluding to not (only) policy-seeking rationales but also vote-seeking ones. We now turn to discuss such explanations of opposition parties’ behavior.

Legislative Voting as a Vote-Seeking Activity

In distinguishing between policy-making and scrutiny activities, Louwerse and Otjes (2019) ascribe appeals to voters to the scrutiny category, which includes activities such as questions and legislative voting (against the government). Others have identified vote-seeking uses of, for example, no-confidence or impeachment motions (Morgenstern, Negri, and Pérez-Linan 2008; Williams 2011; 2016); parliamentary questions, either as a way to raise constituency concerns (e.g., Kellermann 2016; Russo 2011; Saalfeld 2011; Wegmann and Evequoz 2019) or to criticize the government (Seeberg 2020); and private member bills (Bräuninger and Debus 2009; Bräuninger, Debus, and Wüst 2017; Brunner 2013). Explanations of opposition parties’ use of all these tools have pointed to vote-seeking rationales, whereby these parties seek to communicate to the public the differences between themselves and the government parties, thereby hopefully improving their electoral performance in subsequent elections.

In this study, we focus on a parliamentary activity that has so far received less attention as a vote-seeking strategy of opposition parties: legislative voting. Arguably, it figures among the most important activities of legislatures (Saalfeld 1995). Legislative voting is “the final action taken on controversial issues” (Aydelotte 1977, 13). It is the activity that presents a party’s stand in the most tangible manner and is readily comparable across systems. High cohesion in parliamentary systems makes it more readily instrumental in promoting the party brand rather than personal reputation (compared with, e.g., parliamentary questions). A cohesive party vote is a clear and public expression of its position. And as a party activity, it is easier to distinguish from other parties than are some agenda-setting activities, which may necessitate interparty cooperation. Thus, legislative voting constitutes a particularly useful tool for opposition parties to communicate party positions to the electorate.

Opposition parties indeed allude to their voting behavior on social media and press releases. One example is a tweet by the German Left party: “So far DIE LINKE has voted against every #war deployment” (DIE LINKE 2013, our translation). Another example is a statement from the British Labour: “Labour Votes against the Police, Crime, Sentencing and Courts Bill at Third Reading” (Labour 2021).

Moreover, opposition parties distinguish themselves as a major opposition player by alluding to how they diverge from other parties—for example, the Canadian New Democratic Party (NDP): “New Democrats have been on the front line—leading the charge—opposing Conservative schemes to ram the bill through Parliament at all costs” (NDP 2015a) and “It’s not too late for the Liberals to join the NDP and the millions of Canadians who oppose the Conservatives’ devastating Bill C-51” (NDP 2015b).

Finally, traditional media outlets often cite these messages in their reporting—for example, regarding the then-leading opposition party in Israel: “[Opposition Leader MK Tzipi] Livni added that Kadima party, which she heads, will oppose the bill” (Liberman and Zarhia 2011, our translation). Thus, legislative voting seems to be a prominent tool at the hands of opposition parties to seek votes.

Parliamentary Opposition and Retrospective Voting

Democratic elections can be viewed as sanctioning mechanisms that allow voters to judge the actions of rulers “ex-post by the effects they have” (Cheibub and Przeworski 1999, 225), resulting in retrospective voting. There is a vast literature analyzing the electoral consequences of retrospective voting, focusing on government parties and leaders (see, e.g., Fisher and Hobolt 2010; Maeda 2009; Narud and Valen 2010; Powell and Whitten 1993; Söderlund 2016). The main point of this literature is that voters react to economic conditions, scandals, the fulfillment of pledges, et cetera, by rewarding or punishing ruling parties (see, e.g., Anderson 2000; Dassonneville and Lewis-Beck 2019; Hellwig 2012). However, little is known regarding the determinants of opposition parties’ electoral performance. In other words, when dissatisfied voters decide to punish a ruling party and refrain from voting for it, it is unclear how they choose for which opposition party to vote. Similarly, it is unclear how opposition party voters decide whether they are satisfied with the party’s performance or not.

The information voters can attain—and upon which they base their voting decision—differs between government and opposition parties. Whereas voters can evaluate the government’s current performance in office based on observed or mediatized outputs such as economic indicators, general or personal welfare, et cetera, they tend to have less information on opposition parties and need to rely on other sources to evaluate them (Butt 2006; Lachat and Wagner 2018). However, recent evidence suggests that voters are more adept at finding the political information most useful to them in forming a vote choice than previously thought (Dalton 2021; Fortunato and Stevenson 2021; Seeberg 2020). Therefore, we may expect that voters will be attentive to informative cues from opposition parties. As we explain above, legislative voting is primed to be a viable source of such informative cues, especially considering the government–opposition divide is the main predictor of legislative voting behavior in most parliamentary systems (Hix and Noury 2016). Specifically, we argue that the conflict between an opposition party and the government serves as an informative signal to voters.

Evidence from adjacent literature on personal vote-seeking behavior by individual legislators also shows that information on parliamentary activity does reach voters and influences their decisions. For example, legislative speeches seem to influence voters’ evaluation of MPs as voters reward more active MPs with more preference votes (Marcinkiewicz and Stegmaier 2019). Parliamentary work, specifically bill initiation, positively affects personal votes in general and especially for opposition MPs (Bouteca et al. 2019; Bowler 2010; Däubler, Bräuninger, and Brunner 2016; Williams and Indridason 2018). Similarly, although voters certainly do not observe legislative voting behavior directly or follow it closely, we propose that information on the aggregate behavior of an opposition party is mediated to the voters, through the media as well as by the politicians themselves (see, e.g., Huber et al. 2020). Specifically, voters will take note of an opposition party’s aggregate tendency toward conflict with the government. In the following section, we elaborate on why such conflict should be electorally rewarding.

The Electoral Desirability of Parliamentary Conflict

We conceptualize the effect of legislative voting activity by opposition parties as having an accumulative, aggregate effect of establishing a kind of “performance portfolio” for these parties. Opposition parties emphasize their differences with the government to a lesser or greater extent, on the aggregate, depending on the frequency with which they present opposing stands on the issues voted upon. Admittedly, not all votes are of the same importance. Votes on important or politicized issues are expected to attract more conflict than votes on technical and minor issues and may carry greater importance for voters. Moreover, parliaments vary in the extent to which they record and publish information on less important or less conflictual votes. We acknowledge this challenge for our comparative design and address it using several control variables, as discussed below.

But why would opposition parties conflict with the government in the first place? Downsian spatial theory suggests that vote-seeking parties will converge around a centrist, popular platform, even in multiparty systems (Adams 1999; Downs 1957). However, more recent iterations of spatial models show that when accounting for nonpolicy factors that influence voters’ party identification, parties have incentives to represent diverging preferences to secure the support of their electoral bases (Adams and Merrill 1999; Adams, Merrill, and Grofman 2005; Grofman 2004).

In a similar vein of combining spatial models with nonpolicy considerations, others have argued theoretically and shown empirically that low-valence parties will distance themselves from high-valence parties, as without ideological differences, voters will have no reason to prefer the low- over the high-valence party (Fortunato and Stevenson 2013; Schofield 2003; Spoon 2011). We may similarly expect opposition parties to distance themselves from the more visible and powerful government parties. Importantly, this argument leads us to the expectation that opposition parties will have an electoral incentive to distance themselves not only from an unsuccessful and unpopular government (e.g., Williams 2011) but also from a successful and popular one, with which the opposition simply cannot compete in valence terms.

Although much of the above-mentioned literature deals with convergence or lack thereof in terms of policy promises, similar arguments have been applied to legislation and voting (Ganghof and Bräuninger 2006; Tuttnauer 2018). Together, these works provide us with an argument for why vote-seeking opposition parties should adopt conflictual strategies vis-à-vis the government rather than attempting to converge with it in policy terms. In this paper, we move one step further by asking whether such strategies are successful—that is, whether they influence voters and, thereby, the electoral performance of opposition parties. We suggest two complementary individual-level mechanisms that underlie the party-level relationship between legislative behavior and electoral outcomes.

The Opposition Valence Mechanism

Our notion of opposition valence originates from the fact that a central role of the parliamentary opposition is—as the name suggests—to oppose the government. Thus, all voters may view the frequency in which an opposition party openly opposes the government as it “doing its job,” making such a behavior a valence signal for opposition parties, of which voters can approve regardless of their ideology (Zur 2019). To the extent that prospective supporters value such behavior because it is the expected behavior from an opposition party as such, frequent confrontation may be instrumental in attracting disenchanted government voters and helping the party distinguish itself from other, less confrontational, opposition parties. If the strategies the literature imputes to opposition parties are indeed—at least sometimes—effective, one may expect a more conflictual stance vis-à-vis the government to result in improvement of the opposition party’s electoral performance. Thus, our first hypothesis is a general one:

General hypothesis: A more conflictual stance vis-à-vis the government will have a positive effect on electoral performance for opposition parties.

Beyond its general effect, the importance of opposition valence is likely to vary across parties and systems. First, the importance of conflict in a given political system is likely to condition the importance of legislative conflict as a valence indicator. The degree to which institutional rules give the opposition influence on the policy-making processes is a crucial characteristic of the type of democratic model in a polity and its normative proclivity toward either consensus or confrontation (Powell 2000; Strøm 1990). In some systems, the opposition has little influence on the parliamentary agenda (through tools such as those mentioned above), and therefore confrontation and competition are its raison d’être. In others, cooperation and inclusion of the opposition in the policy-making process is the norm. Such systems lean more toward consensus. In consensus-leaning systems, the opposition is usually less disadvantaged compared to the government in its ability to influence the agenda-setting process and claim credit for policy outcomes, thereby giving opposition parties another path to attract votes by alluding to their policy-related achievements, as mentioned in our discussion of the agenda-setting literature above.

In conclusion, we can expect that voters will not value conflict in polities with strong opposition policy-making powers as much as in polities with weak opposition powers. Consequently, signaling confrontation will be less beneficial to a party in the former type of system compared with a party in the latter type. ² Thus,

Institutions hypothesis: A more conflictual stance vis-à-vis the government will have a positive effect on the electoral performance of opposition parties in polities with weak institutional opposition policy-making power.

Second, the ability and willingness of voters to attribute valence to an opposition party’s conflictual behavior may vary between parties—even in the same political system. To begin with, not all opposition parties (and their voters) give the same priority to competing with the government and replacing it. Some, mostly small and niche parties, may prefer policy gains achieved through a more cooperative attitude toward the parliamentary majority.

Moreover, usually, only some parties are perceived as viable government alternatives. Although several party features may serve as indicators of governmental viability—for example, government experience, mainstream ideology, et cetera—we focus on party size, which has been consistently identified as an important determinant of opposition parties’ competitiveness (see, e.g., Louwerse et al. 2021; Tuttnauer 2018). Voters are also more likely to observe and react to larger opposition parties’ campaigns and behavior (Lachat and Wagner 2018; Plescia and Kritzinger 2017). Consequently, taking a more conflictual stance vis-à-vis the government to highlight policy alternatives might be an exceptionally rewarding vote-seeking strategy for larger parties, meaning we should observe a stronger effect of the level of conflict on electoral performance for these parties. Therefore, we hypothesize the following:

Size hypothesis: A more conflictual stance vis-à-vis the government will have a positive effect on electoral performance for larger opposition parties.

The Ideological Differentiation Mechanism

Our second proposed mechanism pertains to ideological differentiation. Voters may use a general sense of confrontation of opposition parties as shorthand for their ideological placement, perceiving parties that conflict with the government more prominently to be ideologically farther from the government, compared to parties that present a more cooperative demeanor. We already alluded above to the spatial modeling argument supporting the electoral desirability for (some) differentiation vis-à-vis the government for opposition parties (see also Klüver and Spoon 2020). Governing parties usually enjoy nonpolicy advantages over opposition parties in terms of media coverage and brand recognition, reputation for governing ability, et cetera. Therefore, an opposition party that is too ideologically similar to the governing party(ies) risks having little in nonpolicy terms to offer undecided voters to sway them away from voting for a governing party. Thus, some ideological difference is beneficial (and perhaps necessary) for opposition parties to gain votes from—or not lose them to—governing parties.

However, we argue that the perception-skewing effect of conflict will not be uniform across all parties. Voters have more tools and heuristics with which they can place actually extreme parties at the extreme. It is the mainstream parties, the ones closer ideologically to the government, whose perception could be skewed by conflicting with the government, and these are the ones that will benefit from the added perceived distance from the government. Parties that are indeed ideologically distant from the government would be less affected. Thus,

Ideology hypothesis: A more conflictual stance vis-à-vis the government will have a positive effect on electoral performance for opposition parties that are ideologically closer to the government parties.

Here we should reiterate that an opposition party’s decision to vote against the government’s position may be sincere, in the sense that it perfectly reflects the opposition’s preferences, as evident also in their manifesto and other public statements. However, in general, the voting portfolio of an opposition party is the result of numerous strategic decisions—on which votes to actively support the government and on which to absent itself, whether to find the minute details on the grounds of which to oppose a bill or to support it based on a general agreement, et cetera.

Thus, to the extent that no opposition party’s ideal position is truly at a “zero distance” from the government’s position, a party can strategically confront the government more often than its ideology will dictate, skewing its perception among possible voters who value an ideological alternative to the incumbent government. We address this concern empirically in our individual-level analysis below. We now move to describe the methodology and data used to test our hypotheses.

Data and Methodology

Case Selection and Data

Our main analyses are designed to reveal the connection between opposition parties’ behavior in legislative votes and their electoral performance in the following elections. To this end, we collected data from 10 countries, spanning 107,881 legislative votes over 41 terms. All recorded votes are included regardless of the issue, type, initiator, and recording procedure (e.g., roll call or electronic). Importantly, our data do not allow us to differentiate between government and opposition bills, a compromise we had to make in such a large-n comparative design. However, although the dynamics leading to the observed conflict may differ between government- and opposition-initiated votes, we contend that the effect of such conflict on electoral outcomes should be similar.

The underlying data consist of the positions of each opposition party on each vote, resulting in a dataset of 389,142 party-vote observations (see Table 1). Our unit of analysis is the individual party over a cabinet term (see below) because our ideological distance hypothesis and some of our controls relate to features of the cabinet facing the opposition. ³ Therefore, we aggregated the data to the level of party-in-cabinet-term, resulting in 169 observations. We included only the last noncaretaker cabinet preceding elections to ensure that the parties analyzed reached elections directly following their tenure in the opposition.

Table 1. Cases and Time Frame

As in many studies of legislative voting, data availability has been the primary criterion for case selection. However, the included countries are diverse in their institutional and cultural settings including Westminster systems (the United Kingdom and Canada), established proportional systems (Finland, Germany, Israel, and Sweden), and new democracies (Czech Republic, Lithuania, Poland, and Spain). As we describe below and in greater detail in Table A.1 in the online appendix, these countries also display cross-country variation in all system-level variables and within-country variation in the party-level dependent and independent variables. We are therefore confident that this study’s findings should be generalizable to most cases of parliamentary democracy.

Dependent and Independent Variables

The dependent variable in our aggregate-level analyses is the party’s vote share in the elections after said term. ⁴ Because the lagged vote share is a strong predictor of the vote share in the next elections, we follow the convention in the literature (e.g., Pacek and Radcliff 1995) by including the party’s vote share in the previous elections (ranging between 0.3% and 36.4% in our data) as a control. We also use the lagged vote share as a constitutive term for the interaction of interest in testing the size hypothesis.

Our main independent variable of interest is the party–government conflict rate. First used by Tuttnauer (2018), the conflict rate is measured as the share of votes over the cabinet’s term in which an opposition party voted against the government’s stance. The original, legislator-level data on individual parliamentary votes, coded as For (+1), Against (-1), or Abstain (0), was aggregated to a party-vote-level position by averaging across all voting party MPs, resulting in a party position for each vote, ranging between -1 and +1. The per-vote party positions were then used to calculate the government’s average positions for each vote (weighted by seat share). A party vote is coded as conflictual when the government’s rounded position is for (+1) and the opposition party is against (-1) or vice versa. Government coalitions, and individual parties, vote in an overwhelmingly cohesive manner in parliamentary systems, especially if one is not interested in cohesion per se but only in ascertaining their position on a vote. ⁵ Thus, a simple rounding of each position to one of for, against, or abstain was sufficient for our needs. ⁶ Finally, the per-vote measure was aggregated across all votes in the last cabinet term before each election to create the share of votes in which each opposition party conflicted with the government.

Data on voting behavior were obtained by scraping the official parliamentary websites (Canada, Czech Republic, Israel, Lithuania, Poland, Spain, and Sweden), using nongovernmental sources compiling such data (the United Kingdom: www.publicwhip.org.uk), or relying on previous publications (Germany: Sieberer et al. 2020; Finland: Pajala and Jakulin 2012). The parties included in our study display a wide range of parliamentary behavior, with variation both between and within countries. Thus, country means range between 0.19 (Israel and Lithuania) and 0.73 (Canada). Behavior also varies within countries. In some, the range between the most cooperative and most confrontational parties is relatively small (e.g., 0.27 in Poland and Spain), whereas in others the range is vast (e.g., 0.61 in Finland and 0.83 in the United Kingdom). Overall, conflict rate values cover nearly the whole theoretically possible range of the scale, ranging between less than 0.01 and 0.91 (see Table A.1 in the online appendix for more detail).

We also include several other independent variables. ⁷ To test the institutions hypothesis, we use the Policy-Making Power of Opposition Players index (PPOP; Wegmann 2022). In contrast with previous research that relied mainly on the committee system’s structure to assess the parliamentary opposition’s influence in the policy-making process (Maeda 2015; Powell 2000; Strøm 1990; Tuttnauer 2018), PPOP is a multifaceted index that measures the influence given to the opposition by various parliamentary rules. ⁸ Our cases range on this measure between 0.43 (Spain) and 0.85 (Germany).

To test our two party-level hypotheses, we include the above-mentioned lagged vote share as well as the party’s ideological distance from the government. We use Parlgov.org’s (Döring and Manow 2019) placement of each party on an 11-point left–right scale, rescale it to range between 0 and 1, and calculate for every opposition-party-in-cabinet term its distance from the government’s size-weighted mean ideological placement. The observed values for this measure are between 0.007 and 0.68.

Control Variables

We include several control variables to be as confident as possible that the effect captured is indeed related to the opposition parties’ behavior and not merely a by-product of the government parties’ changing fortunes. First, we include a dummy for early elections, those that happen within the first three years of the legislative term and may indicate some government scandal, crisis, or otherwise external shock that may result in the government suffering significant losses. Second, we include cabinet-level controls to capture the cabinet’s clarity of responsibility (Powell and Whitten 1993), including the seat share of the governing part(ies) and the fragmentation of the governing coalition in terms of the effective number of parties in it (Laakso and Taagepera 1979). In further specifications (see Table A.3 in the online appendix), we included dummy variables for oversized and minority cabinets and for government alternation patterns (Otjes and Willumsen 2019). These did not change any of our results, so due to high collinearity between the control variables themselves, they were subsequently dropped. ⁹

A third battery of controls includes parliamentary fragmentation (again in terms of the effective number of parties) and dummies for proportional systems (either Proportional Representation or Multimember Proportional) and federations. These seek to capture the ease of access for parties to the parliament and the government as well as the degree to which the most fundamental political institutions lend themselves to zero-sum competition in the (lower) house of parliament. ¹⁰

Fourth, we include the GDP growth for either the preceding year (if elections were held on or up to June 30) or the same year as the elections (if they were held on or after July 1) from the World Bank’s World Development Indicators dataset (2020). The economic control variable captures the fact that the success of opposition parties at elections is also contingent on the government’s popularity and macroeconomic performance measures.

Finally, as legislative votes in general, and particularly roll-call votes, are a product of various strategic political decisions (Hug 2010; Saalfeld 1995), we follow Tuttnauer (2018) in including both a control for the frequency of votes in the plenum and a dummy distinguishing between systems for which the whole record of votes was available and those for which only a subset (usually roll-call votes) was available.

Model Estimation Strategy

Because our dependent variable is a fraction (vote share), we use fractional logit regression to estimate our models. Like other extensions of logistic regression, and unlike ordinary least squares (OLS) regression, the coefficients of fractional logit regression are not straightforward to interpret. However, we are not interested here in the point estimates of the coefficients per se but, rather, in the interaction between coefficients, which we go on to analyze graphically to determine the ranges for each independent variable in which conflict rate has a significant effect on vote share. Therefore, the disadvantage of using a fractional logit regression is negligible, and its advantage over OLS, especially in avoiding predicted outcomes outside of possible values, easily outweighs it. Nonetheless, using either OLS or Tobit regression instead does not change our results (see Tables A.5 and A.6 in the online appendix).

Our observations are nested—parties in elections in countries—but the number of observations is insufficient for a multilevel modeling strategy. Furthermore, we decide against including country and election fixed effects but instead control for the aforementioned relevant systemic factors. However, including such fixed effects (see Table A.6 and Table A.7 in the online appendix) does not change the party-level results reported below.

Results: The Influence of Opposition Behavior on Electoral Outcomes

The results are shown in Table 2. All models include the individual opposition party’s vote share in election t + 1 as the dependent variable. We further include all the above mentioned independent and control variables and use clustering by country-election. ¹¹ Whereas we assess the nonconditional effect of conflict on a party’s electoral result in model 1, we test our interactive hypotheses in models 2–4. Therefore, models 2–4 differ in the variable with which we interact our main independent variable of interest, the party–government conflict rate. Following Brambor, Clark, and Golder (2006) and Clark and Golder (2006), we analyze each interaction graphically.

Table 2. The Conditional Effect of Party–Government Conflict on Electoral Performance

Note: Dependent variable is the individual opposition party’s electoral performance (vote share) in election t + 1. Robust standard errors in parentheses, clustered by country-election. *p < 0.10, ^** p < 0.05, ^*** p < 0.01.

Model 1 shows a strongly positive and significant coefficient for our conflict variable. The only other variable with a significant effect is the party’s vote share in the previous elections. Figure 1 shows the predicted values of vote share for a party with all dummy variables set to zero and all continuous variables set to their medians, with varying values of its conflict with the government. The figure shows that such a party, having agreed with the government on all votes (conflict = 0), is predicted to receive just below 5% of the votes in the next elections. This represents a significant drop from the median in our data, which is 7.5%. It also represents a predicted loss of votes, as the median vote share for the previous election is 7.9% in our data. In contrast, a similar party that conflicted with the government on 90% of the votes is predicted to receive almost 11% of the votes in the next elections, representing a considerable increase from the median in our data of both the previous and the next elections. Confronting the government on legislative votes seems to be electorally beneficial for opposition parties. These findings affirm our general hypothesis.

Figure 1. Predicted Vote Share (t + 1) by Conflict Rate

Note: Based on model 1 in Table 1. Gray bars represent density of the x-axis variable; 95% confidence intervals used.

We now turn to the interactive models, starting with models 2 and 3 that address the opposition valence mechanism. In model 2, we interact conflict with the opposition powers index. Both constitutive terms, the interaction term, and the lagged vote share are significant. Figure 2 shows conflict has a significant positive effect in the range between 0.426 (the lowest value of PPOP in our data) and 0.603, a range including 52.6% of our cases and half of the countries studied. In countries with values above 0.603, which lend the opposition many opportunities and capabilities to influence parliamentary decision making, conflict has no significant effect. Overall, the findings affirm our institutions hypothesis and therefore give support to the proposed opposition valence mechanisms.

Figure 2. Conditional Effect of Conflict on Vote Share in t + 1 by Opposition Powers

Note: Based on model 3 in Table 1. Gray bars represent density of the x-axis variable; 95% confidence intervals used.

Model 3 includes the interaction with the party’s vote share in elections t, acting as a proxy for party size. ¹² The coefficients of the constitutive terms are both statistically significant, but the interaction term is not. Indeed, Figure 3 shows that the effect of conflict on the next vote share does not significantly change with different values of the previous vote share. The fact that in high previous vote shares the effect of conflict loses statistical significance is probably due to the scarcity of observations in this range. Therefore, we have to reject our size hypothesis, implying opposition parties of all sizes would benefit electorally from confronting the government more often.

Figure 3. Conditional Effect of Conflict on Vote Share in t + 1 by Lagged Vote Share

Note: Based on model 3 in Table 1. Gray bars represent density of the x-axis variable; 95% confidence intervals used.

We now move to test our ideology hypothesis, which addresses the ideological differentiation mechanism. Model 4 includes the interaction with party–government ideological distance. The conflict coefficient is statistically significant, as are the interaction term and the lagged vote share. As Figure 4 shows, conflict rate has a positive effect on electoral performance for parties ideologically close to the government, and this effect declines until it becomes indistinguishable from zero at a difference of 0.42 (on a scale between 0 and 1). In total, 130 of the parties in our data (77%) are in the range of a significant effect. These findings affirm our ideology hypothesis and therefore give support to the proposed ideological differentiation mechanism.

Figure 4. Conditional Effect of Conflict on Vote Share in t + 1 by Ideological Distance

Note: Based on model 4 in Table 1. Gray bars represent density of the x-axis variable; 95% confidence intervals used.

In summary, our findings support all our hypotheses except for the size hypothesis. Confronting the government more frequently has a general beneficial effect on opposition parties’ electoral performance, though this effect is mitigated by the institutional policy-making power of the opposition. Both findings are in line with our proposed opposition valence mechanism. Furthermore, the effect of conflict with the government on electoral performance is more pronounced for parties that are ideologically closer to the government, a finding compatible with our ideological differentiation mechanism.

Robustness Checks and Competing Explanations

As mentioned above, we estimated our models using additional control variables. These include cabinet-level factors, such as oversized coalitions and minority cabinets, and system-level factors such as the age of democracy, a dummy for Central and Eastern Europe, and the structure of competition and government alternation (Table A.3 in the online appendix). We also ran the same specifications using different statistical models—OLS and Tobit regression (Tables A.4 and A.5 in the online appendix). All such alternative specifications and estimations yielded similar results. We also reran the models dropping a specific country from the analysis each time. The results (Tables A.11–A.14 in the online appendix) show that no one country drives our results, and the coefficients never change substantially in their direction or size.

To further test the notion that some parties are expected to compete for office more than others, and thus be rewarded more for conflict with the government, we tested a model including interaction with past government experience, either as the time the party has been in government divided by the time it has been in parliament or as a binary variable for existence or absence of such experience. As with our size hypothesis test, we found no significant interaction (see Table A.9 in the online appendix).

We also test the notion that the effect of conflict might be contingent on the government´s popularity and macroeconomic performance. Confronting a well-liked and well-operating government may involve more electoral risk than opportunity. Therefore, we test several models (see Table A.10 in the online appendix). First, we exclude conflict to test whether the economic conditions themselves affect the opposition’s electoral performance, finding no such effect. Second, we test for an interaction between conflict rate and GDP growth. Although the interaction term is significant, a graphic analysis shows there is, in fact, little to suggest that the effect of conflict is substantially conditional on economic growth (Figure A.2 in the online appendix). We further split the population between left and right opposition parties and interacted conflict with unemployment and inflation, respectively. The interaction is insignificant in the former and in the opposite direction to the expectation for the latter. We address the meaning of these null findings in the conclusion.

Furthermore, if the government’s performance were indeed the only driver of individual opposition parties’ electoral results, we would expect to find a strong (positive) correlation between each party’s vote-share change and the other opposition parties’ vote-share change. However, in our data, this correlation is medium in strength and negative (r = -0.29, p < 0.001), suggesting that if at all, opposition parties gain (lose) votes from (to) one another rather than from or to the government.

We now turn to investigate further the micromechanisms we suggest are at the foundation of the macro-level findings described above.

Individual-Level Analysis: Opposition Behavior and Voters’ Assessment

Although our central phenomenon of interest—electoral results—is aggregate in nature, our theoretical argument is based on two microlevel mechanisms by which parties’ legislative behavior affects individual voters. In other words, for opposition behavior to affect electoral results, information of this behavior must reach the voters and voters must care about it.

In the following analysis, we present evidence that is compatible with the two mechanisms proposed above—namely, that higher conflict rates by a given opposition party make voters (a) give the party more favorable ratings and (b) perceive it as more extreme or distant from the government. Although we rely on existing survey data and therefore cannot observe a causal effect directly, we show that the statistical effect of conflict holds when accounting for possible contravening factors.

Data and Variables

We link our data on opposition parties’ behavior to existing electoral survey data. We use the CSES IMD unified dataset, which includes electoral surveys between 1996 and 2016 (modules 1–4). The modules in CSES all ask respondents to place themselves and the major parties in the polity on a single left–right scale as well as to rate their attitude toward each of these major parties on a like–dislike scale. Because our data on Spain and Lithuania lie outside of the CSES IMD time frame, we drop them from this test. ¹³ We also include only opposition parties that were mentioned in the CSES election surveys. After dropping cases with missing values for our dependent variables and various control variables, we are left with 84 party observations. In total, 27,371 respondents replied to questions about these parties, resulting in 81,839 respondent-party observations.

To validate our opposition valence mechanism, we test whether the rate of conflict of an opposition party correlates with the respondents’ attitude toward it. We expect respondents to give higher ratings to opposition parties that confront the government more frequently. This effect should also be more pronounced for larger parties. Our dependent variable here is each respondent’s likability score for an opposition party. Our independent variables are each party’s conflict rate, lagged vote share, and their interaction. We control for the perceived (absolute) difference between the respondent’s placement of the party on the left–right scale and her own placement, CSES’s expert placement of the party on the same left–right scale, recalculated to depict its distance from the center, and a dummy indicating whether the respondent voted for the party in previous elections. We further control for the respondent’s age and age-squared, sex, and level of education. We also include country and party fixed effects and cluster by respondent. ¹⁴

To validate our ideological differentiation mechanism, we analyze how conflict affects the respondents’ placement of an opposition party on the ideological scale. We expect that the more conflictual an opposition party is, the farther respondents will tend to place it from the government and therefore closer to the extremes. Furthermore, drawing on our ideology hypothesis and its affirmation by the aggregate-level analysis, we expect conflict to have a stronger effect on respondents’ placement of parties that are closer to the government.

We prefer to use perceived extremism rather than the respondent’s perceived distance between each opposition party and the average placement of the governing parties, a measure that is more directly linked to our theoretical argument but that we believe requires too many assumptions on respondents’ cognitive process. However, running the same analysis with such a measure yields consistent results (see Table A.16 in the online appendix). Perceived extremism is calculated by taking each respondent’s placement of each opposition party on the left–right scale and computing its absolute distance from the middle of the scale. We regress this variable on the party’s conflict rate, expert-based ideological distance from the government, and their interaction. Additional controls are the same as in the likability analysis.

Results

The results of both analyses are shown in Figure 5 (with full results in Table A.15 in the online appendix), with coefficients of the basic models (without interaction) in circles and those of the models interacting conflict with ideological distance or vote share in squares. Higher conflict rates correlate with higher scores of both perceived extremism and likability. Furthermore, the interaction terms for both the extremism and likability analyses are strong and significant.

Figure 5. Likability, Perceived Extremism, and Conflict with the Government

Note: Black circles correspond to models without interaction with party-level data. Gray squares correspond to models with interaction with party-level data. Lines are 95% confidence intervals.

As shown in Figure 6, both interactive effects are in the predicted directions. A higher conflict rate has a stronger effect on larger parties than on small ones (left panel) and on perceived extremism for parties closer to the government than on ones farther from it (right panel). The effects of conflict are not only statistically significant, which is to be expected with such a large population, but also substantively strong. A 50-percentage-point (roughly two standard deviations in the measure) increase in conflict for a party with a 10% vote share, which is around the median in our sample, is correlated with a 5%, or 0.5-point, increase in likability. Regarding perceived extremism, for parties with a distance of 0.3 points from the government, which is roughly the median in our data, the coefficient for conflict is around 0.7. Thus, an increase of 50 percentage points in conflict is correlated with an increase of approximately 7%, or 0.35 points, in perceived extremism. Importantly, although this is a strong effect, and the effect is stronger for parties closer to the government, it does not exceed a full point. Considering the interval nature of the measure, this means that respondents are not likely to perceive a conflictual party as more extreme than other parties. Their overall perception of the party system is therefore not likely to be skewed. Thus, according to our findings, conflictual parties do not face a substantial risk of losing their voters to other parties by seeming too extreme.

Figure 6. Conditional Effect of Conflict on Perceived Extremism and Likability

Note: Average conditional effects of conflict rate. Gray bars represent density of the x-axis variable; 95% confidence intervals used.

Overall, the individual-level results support both of the mechanisms we suggest, although we do not go as far as claiming they show how voters’ assessments of likability and ideology are later translated to voting decisions. As we control for the respondents’ own perception of how ideologically different the party is from themselves and because the conflict rate measure is not strongly correlated with the ideological distance from the government (r = 0.13, p = 0.087), we believe the effects of conflict do not arise from an unidentified ideological factor but from the parliamentary behavior being successfully mediated to the voters.

Conclusion and Implications

In this study, we analyzed the effect of legislative voting on the electoral fortunes of opposition parties, asking whether a confrontational voting behavior could be an effective vote-seeking strategy. We argued that because voters rely on visible party behavior to formulate their political preferences, legislative voting constitutes an important means for opposition parties to communicate their differences with the government to the voters. We suggested two individual-level mechanisms by which conflictual opposition behavior can affect voters’ perception of it: as a signal of opposition valence and as means of ideological differentiation from the government.

Our results lend support to both mechanisms. We find that a more conflictual stance of opposition parties vis-à-vis the government has a general positive influence on their electoral performance. However, conflict is less beneficial in systems that grant opposition parties influence in the policy-making process and therefore prefer cooperation over conflict. Both findings are in line with our opposition valence mechanism. We also find that ideological distance from the government mitigates the effect of conflict on electoral performance. As ideological differences between the opposition party and the government become more pronounced and thus clearer to the public without the added value of observing parliamentary behavior, the benefit of conflict in legislative voting is diminished. This finding is in line with the ideological differentiation mechanism.

Following our party-level analyses, we present individual-level analyses showing that respondents perceive more conflictual opposition parties as more likable (supporting the opposition valence mechanism) and more extreme (supporting the ideological differentiation mechanism). Furthermore, the effect on likability is stronger as the party is larger, and the effect on perceived extremism is stronger as the party is closer to the government.

This article makes three important contributions. First, it contributes to the legislative studies literature by highlighting that opposition parties can successfully use the tools at their hands to their electoral benefit. This confirms an underlying assumption that has to this point been mostly left untested: that politicians and political parties are rational actors, inasmuch as they can take account of their goals and apply the means at their disposal to reach those goals. Second, our finding that the behavior of opposition parties affects electoral outcomes calls for the research on retrospective voting to further include the opposition in its models and analyses to seriously take into account opposition parties as targets of the voters’ ex post judgment. Third, this study adds to an accumulated body of work showing that actions in parliament seem to affect electoral outcomes, which ultimately suggests that voters are informed about these actions. That such information reaches voters, and that they care about it, is not trivial. This is highlighted by various studies looking into political knowledge and competence (Achen and Bartels 2016; Ashworth, Bueno de Mesquita, and Friedenberg 2018; Lau and Redlawsk 1997).

Testing and finding a correlation between parliamentary behavior and electoral performance is not to be taken lightly, given the difficulties in assembling a comparative dataset of the former. We have controlled for as many contravening factors as possible. Furthermore, the difference in time between our independent and dependent variables ensures the direction of causation—if there is such a relationship—is not in the opposite direction. Moreover, some obvious confounding factors such as the government’s behavior or fortunes do not lend themselves to explaining the observed variation within each country and electoral cycle. By being able to explain this variation between individual opposition parties in the dependent variable by relating to their own activity, we believe we provide substantial evidence that this activity may indeed be partially driving the results.

We came closer to assessing the underlying mechanisms by showing that the correlative effect of conflict is also found with respect to respondents’ attitudes toward parties and perceptions of them. More research is warranted to identify the exact role these attitudes and perceptions play in the voting decision making of citizens. Experimental designs may be specifically promising, although we should stress that using hypothetical scenarios rather than relying on citizens’ reactions to actual parties in the world presents its own pitfalls and challenges. We hope this study serves as an impetus for consideration of such future endeavors.

One interesting clash in our findings concerns our size hypothesis: although we find respondents are more affected by conflict in their approval of large parties compared with small ones, this does not translate to a similar conditioning effect of size when it comes to electoral results. Two questions arise. First, why do voters seem to respond differently to conflictual behavior of parties of different sizes but not vote differently? This question calls for further behavioral research. Second, if all opposition parties stand to gain electorally from more conflict, why do smaller ones not confront the government as frequently as larger ones (Tuttnauer 2018)? We surmise that small parties are more policy driven, whereas large parties are primarily vote driven (West and Spoon 2013). Getting the right policy enacted might be as—or more—important for certain parties than receiving more votes (see Strøm 1990). Therefore, future research should look more closely at whether voters of such parties also have varying expectations from their parties.

Furthermore, we do not find any effect of economic growth, as a proxy of government performance, on individual opposition parties’ performance. This holds even if we exclude our main explanatory factor—the opposition party’s behavior—from the analysis altogether. Although this null finding may seem to clash with the vast literature on retrospective economic voting, we argue it actually highlights the fact that merely knowing what affects the government’s electoral fortunes cannot tell us much about how each individual opposition party benefits from the government’s losses or is hurt from the government’s gains. How electoral gains or benefits are distributed between the opposition parties, our analyses show, also depends on what these parties do. Future research should analyze the relationship between economic performance and electoral success of opposition parties more closely.

The general electoral benefit in conflictual behavior raises further empirical and normative questions. Excessive conflict may lead to deadlocks and reduced governability, something the opposition may find desirable even if detrimental to the “general good.” But it may also backfire, pushing the government to “play hardball” and ultimately lead to a threat to the democratic regime (see Levitsky and Ziblatt 2018). Therefore, what makes opposition parties resist the electoral temptation of excessive conflict is a question of utmost importance. A second and related question is what effect the resulting government–opposition relations have on the public’s satisfaction with democracy and support for it. Some scholars have suggested that too much consensus between government and opposition leads to dissatisfaction among citizens with the choices presented to them by the parties and, as a corollary, with the democratic institutions in general (e.g., Mair 2013). On the other hand, elite behavior has been shown to affect public attitudes such as affective polarization (e.g., Banda and Cluverius 2018), which may lead us to expect that too much conflict would increase such attitudes that are possibly detrimental to democratic well-being. Therefore, we present these questions as a research agenda for future studies.