Perceptual Consequences of Portfolios: How Allocation Affects Left – Right Placement

Recent research suggests that party leaders can strategically impact the perceived left – right position of their parties by changing their selective emphasis on certain issues. We suggest that a party ’ s ideological image can also be altered by the portfolio allocation of the coalition government in which the party participates. By controlling a portfolio, the party will have a more direct influence on the related issue and will frequently communicate the party ’ s issue position publicly, thereby cultivating a perception of strong emphasis on the related issue. We run a cross-national party-level analysis showing that portfolio allocation matters with regard to the importance of the subdimensions for the general left – right dimension. In particular, the influence of sociocultural stances depends on the share of sociocultural portfolios. In addition, we show that the mechanism does not apply at the beginning of a government ’ s tenure, but only after a year or longer in office.

How does portfolio allocation impact voters' perception of a party? Previous research has mainly been concerned with analysing who gets which portfolio (see e.g. Bäck et al. 2011) or whether obtaining a portfolio has consequences for pledge fulfilment (see e.g. Schermann and Ennser-Jedenastik 2014). What has been neglected in the literature so far, though, is whether and how obtaining a specific portfolio also has consequences for the perceptions of parties in the eyes of the potential voters. Obtaining a particular portfolio could be a liability, especially in situations where parties become responsible for certain policy areas they do not especially care about. This will force them to enact policies and, more importantly, defend and explain these policies even though they are not high on the list of priorities for their potential voters. This, in turn, could have consequences for how these voters perceive the party in general.
With this project, we synthesize two very separate literatures: namely, the literature on issue competition and issue saliency on one hand and the literature on portfolio allocation and coalition bargaining on the other. We show that the portfolios a party obtains do not just serve as the payoffs in coalition bargaining but also have a significant moderating impact on public perceptions of the party. Instead of changing their policy positions or their selective emphasis, party leaders can, intentionally or unintentionally, alter the perceived left-right position of the party through forming coalitions with certain other parties and through the specific portfolio allocation in those coalition governments. This provides another potential mechanism behind the use of 'coalition heuristics' (see e.g. Fortunato and Stevenson 2013). We hypothesize that for each individual coalition member, positions of the party on the subdimension will have a stronger weight on voters' left-right perception if the party controls the relevant ministerial portfolios.
To test this hypothesis, we present a cross-national, party-level analysis of portfolio allocation, issue positions and average ideological party placement. The empirical analysis relies on a combination of manifesto data and ministerial appointment data as well as mass and elite survey data. First, we explore whether the share of portfolios related to the economic and sociocultural issues can explain the relative impact of subdimensions within governments. We find that the more sociocultural portfolios a party controls, the more weight the party position on the secondary dimension will have in the voters' perceptions of the party. Secondly, we take into account that timing might play an important role for the influence of portfolio allocation on the perceived placement of parties on the left-right dimension. We find that only after some time in office does the impact of portfolios on perceptions kick in.
Whether portfolio allocation has an influence on the perceived position on the left-right dimension or not has important consequences for the parties. For instance, a party could lose electoral support if it obtains a portfolio that shifts its perceived policy position too far away from the policy positions of its potential voters. Our results should therefore lead parties to rethink the 'the more, the better' doctrine regarding portfolios as it might sometimes be better to refuse a particular portfolio in order to protect the party's ideological image.
This article will proceed in the following way. In the next section, we offer a brief review of the issue competition literature and the portfolio allocation literature before synthesizing them and presenting two hypotheses. In the third section, we describe our methodology and in the fourth section we present our findings. In the final section, we discuss the implication of our results and position them relative to other studies of voter perceptions of party positions.

How portfolio allocation influences voter perceptions
In this part, we will present our theoretical framework and the hypotheses we deduced from it. Ultimately, the aim of this article is to test whether portfolio allocation has an influence on the placement of parties on the left-right dimension. Accordingly, we will first discuss the left-right dimension as a super-issue in politics before presenting our argument about how portfolio allocation influences the voters' perception of parties on the left-right dimension.

The left-right as a super-issue
The left-right dimension is ubiquitous. Across Western European democracies, the left-right is still the predominant way for voters, politicians, commentators and scholars alike to describe political affiliations and changes. Following the original conceptualization by Anthony Downs (1957: 132), research in this field usually assumes that the left-right dimension simply represents a weighted average of all the specific positions political actors take on a range of issues (see also Inglehart and Klingemann 1976;Sani and Sartori 1983). This is also how the left-right dimension is operationalized in the widely used Manifesto Project (MARPOR) data set (Budge and McDonald 2012).
While we support this general approach, we believe that the existing research overlooks an important aspect: the left-right dimension has a variable structure, meaning that the underlying issues that are salient to the public might vary greatly over time and place. Different weights should be given to different issues at different times. This has led several scholars to describe the left-right as a super-issue that summarizes ideological differences over the most important issues in a given era (Fuchs et al. 1990;Inglehart and Klingemann 1976;Van der Brug 2004: 234) or in a particular party system (Budge and Farlie 1983;Dalton et al. 2011).
However, the weights of various issues do not differ only across time and placethat is, between party systemsbut also within party systemsthat is, between the individual parties. Wouter van der Brug (2004) found that when an individual party changes the relative salience of issues it will cause changes in voters' perceptions of the party's general ideological position. In other words, parties move along the left-right dimension by strategically emphasizing and de-emphasizing certain issues, and not by actually changing issue positions.
Thomas Meyer and Markus Wagner (2020) elaborated further on this theory by showing that voters infer party positions based on two ideological subdimensions: economy and culture. 1 Among scholars of European party competition, there is a widespread consensus on the co-existence of an economic dimension and a sociocultural subdimension (see e.g. Gidron 2022; Lefkofridi et al. 2014). The cultural, or 'new politics', subdimension has been summarized in a number of different ways, but one of the most popular is the GAL-TAN dimension, which separates green-alternative-libertarian parties from traditional-authoritarian-nationalist parties. At one end of the dimension, we find 'libertarian' or 'postmaterialist' parties, which advocate expanded personal freedoms, access to abortion, active euthanasia, same-sex marriage or greater democratic participation, and at the other we find 'traditional' or 'authoritarian' parties, which favour order, tradition and stability, and want the government to be a firm moral authority on social and cultural issues (Bakker et al. 2015: 144). In contrast, the economic subdimension simply captures whether parties want the government to have an active or reduced role in the economy. Meyer and Wagner (2020) suggested that the perceived positions of individual parties reflect the party's positions on the ideological subdimensions that the party itself emphasizes as well as those subdimensions that are salient in the party system. When evaluating positions, signals that are repeated often by the party are more accessible to voters, and party leaders can use this strategically to impact the perceived left-right position of their parties. Meyer and Wagner (2020) found that the impact of economic positions is affected by the individual party's emphasis, whereas the impact of cultural positions is more affected by the systematic salience. Similarly, Heiko Giebler et al. (2021) argued that individuals map a party's left-right position based on the party's positions on underlying ideological subdimensions and the relative importance of these subdimensions in the communications of the party.
This interpretation clearly links left-right positions to the salience theory promoted by Ian Budge and Dennis Farlie (1983), among others. Budge and Farlie (1983) described how parties can signal to voters that the party would prioritize certain issues by continually emphasizing them (see also Budge 2015). Parties might design their public rhetoric strategically, but it is only habitual and continuous emphasis on certain issues that will connect the issue and the party in the mind of voters. Voters do not consciously compare issue emphasis and positions, but they are, continuously and over long time periods, exposed to certain parties advocating their views on specific issues. These impressions accumulate to form the voters' perception of the party's ideology.
In short, if a party emphasizes a subdimension, the perceived left-right position of the party will be more strongly associated with that position. Hence, it is not necessarily about positional shifts and what direction the party moves in, but more about which subdimension is more relevant for the party's overall placement. We add to the previous work by Meyer and Wagner (2020) by suggesting that the importance of the subdimensions for the placement of the parties by the voters is also importantly influenced by portfolio allocation in a way that is not unilaterally in the hands of party leaders.

Portfolio allocation
The literature on portfolio allocation distinguishes between two different approaches: quantitative and qualitative portfolio allocation. Quantitative portfolio allocation is closely linked to Gamson's Law: the share of ministries is almost proportional to the parliamentary seats that a party contributes to the coalition (Gamson 1961). Qualitative portfolio allocation, on the other hand, moves away from only looking at the numbers and takes into account party preferences as well (see e.g. Browne and Feste 1975;Budge and Keman 1990;Laver and Shepsle 1996). The most relevant approach for this article is the so-called salience approach, which claims that parties try to distinguish themselves from other parties not by relying on different positions but by emphasizing different issues (Budge 1982). Hanna Bäck et al. (2011) were among the first to connect portfolio allocation explicitly to issue salience. Their work centres on the idea that obtaining a portfolio should be more likely, the more salient a policy area is in the party's manifesto. By relying on data from 115 government formations, they find support for their hypothesis: higher emphasis on certain topics comes with a higher probability of obtaining the ministry. In this article, we suggest that obtaining the relevant portfolio for a given issue is not only a priority for a party that already emphasizes an issue in its party platforms but that it also independently contributes to how the party is perceived by the voters. In the following, we will explain the mechanisms through which portfolio allocation influences the importance of the subdimensions for the overall party placement.
While the distribution of policy payoffs is difficult to measure (see Klüver and Bäck 2019, for an exception), the allocation of portfolios is rather public and should be visible even to citizens with lower levels of political interest. Obtaining a certain portfolio therefore increases the perceived salience of the issues belonging to the realm of this specific portfolio for the party. This increased salience is the result of at least two different mechanisms.
First, obtaining control over a portfolio gives the party an advantage in implementing policies in that specific issue area given that ministers enjoy a certain degree of ministerial autonomy (Laver and Shepsle 1996). This can be highly beneficial, especially for policy-oriented parties, because they will be able to deliver on their pledges. Previous research has shown that potential voters take into account the pledge fulfilment of parties and reward these parties at the next election (Matthiess 2020). Another study has shown that voters mainly hold parties accountable for outcomes in the policy areas they are responsible for (Angelova et al. 2016). The fact that voters seem to realize what parties are doing during their time in office should also lead to an increase in the perceived salience for the topics belonging to the realm of those portfolios that the parties were able to obtain.
Second, obtaining a portfolio also comes with increased media attention, which in turn should increase the perceived salience of the topic for the parties. Governing parties generally have easier access to the media (Schoenbach et al. 2001), and introducing legislation is a convenient opportunity for parties for policy-signalling because unveiling a legislative proposal generates a lot of media attention (Martin and Vanberg 2011). However, this increased media attention does not automatically lead to a stronger link to just any issue. Stefaan Walgrave and Jonas Lefevere (2017) additionally argue that parties become increasingly associated with a specific issue when they have a minister dealing with itthe responsible minister is not only able, but actually obliged, to talk about the issue she is responsible for. They found that having a cabinet minister dealing with an issue significantly increases the connection between the issue and the minister's party in the minds of voters (Walgrave and Lefevere 2017: 496).

Hypotheses
To summarize, we posit that obtaining a certain portfolio comes with increased power over public policy and media attention. This means that the party will automatically become the face of the government policy. The sheer publicity that naturally follows with cabinet membership will make it impossible for the party not to emphasize the issue. It is important to stress that we do not intend to argue that voters sit down with a list of cabinet members and a stack of party manifestos, and go through it meticulously, painstakingly comparing each portfolio and issue stance, in order to arrive at a weighted average of the party positions. The process of updating party perceptions is perhaps better described as a 'running tally' (Fiorina 1981). Just as voters are not carrying around stable attitudes about every policy issue (Zaller 1992), they are also unlikely to carry around full-fledged ideas about what each of the parties stands for. Similarly, these perceptions are formed based on what is at the 'top of the head' at any given time.
Imagine a typical voter with a moderate interest in politics. This voter's main exposure to parties' policy stances might come from watching the evening news where different ministers present policy drafts and the opposition reacts. However, the ministers are not able to present the full policy programme of their partyonly the fraction that relates to the portfolio they are responsible for. Thus, over time and through repeated exposure, voters will subconsciously start to associate one government member with a certain set of issues and another government member with another.
Of course, sometimes parties get portfolios they do not necessarily want. This can pose a risk and alter the party's policy image in an unintended way. For this article, we remain quite agnostic about whether parties intend to shape the importance of specific subdimensions for their ideological image. The first step is to show that this effect is truly there, meaning that citizens actually realize that parties emphasize more the issues belonging to the realms of a specific portfolio. Accordingly, our hypothesis reads as follows: Hypothesis 1: The greater the share of portfolios a party obtains that belong to either the sociocultural or the economic subdimension, the stronger the weight of this subdimension on the voters' perception of the left-right position of the party.
Our theory rests on the assumption that voters are, at least on average, aware of portfolio allocation. Research has shown that three-quarters of voters can correctly identify the parties in the cabinet and that among this subset of sophisticated voters, the number of portfolios attributed to each coalition member is surprisingly accurate (Lin et al. 2017). In terms of which portfolios a party obtains, we currently only have evidence from Denmark suggesting that knowledge of which party is in charge varies hugely from portfolio to portfolio. However, for nine out of fifteen portfolios more than 60% of voters attributed the portfolio to the correct party (Lin et al. 2017). This gives us a sufficient degree of confidence in our assumption.
Once again, we do not mean to suggest that voters carry around a mental roster of who is who in the cabinet from day one. Instead, by being casually but consistently exposed to a certain minister being the face of a certain government policy over a long period, voters will gradually alter their image of the party. Following this logic, we believe that the timing of the surveys matters. First of all, the longer a party has been in government and has controlled a certain portfolio, the more likely it is that voters made this connection between the party and the portfolio and the stronger should be the impact of portfolio allocation on the relative weights of the subdimensions. Secondly, it will likely take a new government some time to settle into office and draft new legislation, and thus we do not expect it to present its policies in the media until a number of months after the election. Additionally, Lanny Martin and Georg Vanberg (2004) have shown there is also a considerable amount of time between the introduction and the final vote on a bill. In their analysis of legislative delay in Germany and the Netherlands, the mean time until a bill is voted on is approximately 170 days. Since media attention should be strongest in situations where a bill is properly enacted (see e.g. Martin and Vanberg 2011), it is likely that the influence of portfolios on the weight of the subdimensions depends on how long the parties have been in government. Accordingly, our second hypothesis reads as follows: Hypothesis 2: The longer a party has been in government, the stronger the impact of the portfolios on the weight of the relative ideological subdimension on the voters' perception of the left-right position of the party.

Research design
In order to test our hypotheses, we run a cross-country analysis on 20 countries between 1994 and 2019. 23 We divided our analysis into two parts: in a first step, we analyse the influence of portfolio allocation. In our second analysis, we include the time that a party has been in government in order to test Hypothesis 2.
Since we are interested in how portfolio allocation influences the importance of the two subdimensions for the perceived left-right placements of citizens, we only include those parties that were a member of a coalition government. In a singleparty government, the party does not have a choice on the allocation of portfolios, and thus this is an irrelevant case which we do not need to study. For the construction of our data set, we rely on a number of different sources. We combine data from the Chapel Hill Expert Survey (CHES) (Bakker et al. 2020), the European Election Study (EES) (Schmitt et al. 2009(Schmitt et al. , 2015(Schmitt et al. , 2020Van der Eijk et al. 1999;Van Egmond et al. 2013), the MARPOR data (Volkens et al. 2020) and portfolio allocation data provided by Klüver et al. (2023). Since the EES is only available every five years, we only include those parties in our analysis that are in government at the time of the survey. This reduces our sample to 175 parties.

Dependent variable
Our main dependent variable in our analysis is the left-right placement of parties by the respondents of the EES. The EES is a cross-national study conducted in all EU member states at the time of the European Parliament elections. The survey item that we are interested in, left-right party placement, does not specify the European context, so it is fair to assume that voters are evaluating their national parties' ideological positions more generally. The timing of the EES is convenient for our study, because we are more likely to get an accurate idea of how the government's issue emphasis is perceived most of the time and it allows for the necessary variation in how long the government has been in office. In contrast, national election studies would only be capturing perceptions during or immediately after an election campaign or coalition formation. Voter perceptions from the EES are merged to the most recently held national election.
Since our unit of analysis is one government member per survey per country, we aggregate the left-right placements of parties by averaging across the respondents. Hence, the dependent variable is the average left-right placement perception of the respondents of the EES. While this variable theoretically ranges from 0 (left) to 10 (right), in our analysis it varies between 1.36 and 8.42 with a mean of 5.26 and a standard deviation of 1.63. Figure 1 shows the distribution of our dependent variable.

Independent variables
For our analyses, we follow Meyer and Wagner (2020) and rely on two CHES dimensions that are available for all parties and countries for our timeframe: the GAL-TAN and the economic (hereafter) ECON dimension. Both CHES and EES are collected around the time of the elections for the European Parliament, so matching the five waves of the surveys was straightforward. 3 We manually matched each party included in the EES survey to a CHES party identifier as well as to ParlGov (Döring et al. 2022) and MARPOR identifiers. This data are available as part of our replication material.
In order to test our hypotheses empirically, we also need information on government participation and portfolio allocation for the government in office at the time of the European Parliament elections. For our portfolio allocation variables, we draw on data provided by Heike Klüver et al. (2023). Relying on the allocation of portfolios in different countries suggested by the paper by Bäck et al. (2011), they collected data on the allocation of the following portfolios: foreign, finance, budget, interior, defence, economy, industry, justice, social, labour, health, agriculture, environment and education. 4 For our main analysis, we first match portfolios to the GAL-TAN or ECON dimension but not all portfolios can be allocated. The allocation of portfolios looks as follows: 5 • GAL-TAN: environment, justice, social affairs, foreign, interior • ECON: economy, labour, finance, education and health, budget.
We allocate those portfolios with the clearest financial and redistributive implications to the ECON dimension. Meanwhile the GAL-TAN portfolios mainly concern social organization and the cohesiveness of society even though they, of course, also come with financial costs for the government. For instance, we allocate the social affairs portfolio to the GAL-TAN subdimension, while education and health is considered to be ECON. This is due to the fact that education and health has more obvious financial consequences for the government (i.e. providing money to the education and health sector) than social affairs, since this ministry also deals with questions of, for instance, equal treatment and non-discrimination. In addition, the education and health portfolio always has clear consequences for redistribution, while social affairs in some countries has a more diffuse connection to social mobility via cultural assimilation, housing policies, domestic violence and so on.
In a second step, we count the number of portfolios that a party was able to obtain, allocated to the ECON and the GAL-TAN dimension respectively, and divide it by the party's total number of portfolios in the cabinet. Hence, we are left with two variables for each party: the share of the party's portfolios allocated to the GAL-TAN dimension and the share of party's portfolios allocated to the ECON dimension. 6 As a control variable, we also include the RILE scores of the coalition members at the election preceding the EES and CHES. The RILE measure, provided by MARPOR (Volkens et al. 2020), is a somewhat ambiguous concept. On one hand, it is intended to capture the stated left-right positions of parties and as such we expect it to be highly correlated with perceived left-right positions: if parties are portraying themselves as far right, they will be perceived as further right. In this sense, we are not looking at changes in party placements over time as a result of portfolio allocation, but rather systematic deviations from where the party claims to be located.
On the other hand, MARPOR measures issue salience. Though very implicit, the underlying theoretical basis for RILE is that the relative salience of issues determines left/right positions (Budge and McDonald 2012). Bäck et al. (2011) have convincingly shown that salience for specific topics increases the probability that a party obtains the portfolio which, in turn, could be a driver for the results we find in this article. However, by using MARPOR data we can, at least partially, control for this endogeneity: the more sociocultural issues a party emphasizes, the more weight sociocultural positions will have on the overall RILE score, exactly because they will have a large negative impact on the sum of quasi-sentences. In short, the interaction between salience and position is already built into the RILE measure.
Furthermore, we take the potential endogeneity into account by controlling for the relative salience of the economic subdimension as defined by Jelle Koedam (2022). In short, we take 42 MARPOR categories and aggregate them into two subdimensions. We then divide the difference between economic and sociocultural issue salience by the total and subtract the party system average.

Results
In order to test whether obtaining portfolios has an influence on the importance of the subdimensions for the perceived placement of parties by voters, we run multiple ordinary least squares (OLS) regression analyses with country fixed effects to account for the clustering of parties into countries. Table A.1 in the Supplementary Material includes our main analysis with survey-year fixed effects. The results remain substantially the same. Following Meyer and Wagner (2020), we first run a baseline model where we only include the ECON and the GAL-TAN position of the parties to explain the perceived average party position. The results can be found in Table 1. We also include the RILE scores provided by MARPOR as a control variable (Volkens et al. 2020). Similar to the results of Meyer and Wagner (2020), the results are far from a perfect correspondence even though they are both significant and positive. In Model 2, we include the share of ECON and GAL-TAN portfolios and Model 3 additionally includes the interaction effect between the positions and the share of portfolios.
In Model 3, we find a positive and significant interaction effect between the GAL-TAN position and the share of GAL-TAN portfolios. This means that as the share of GAL-TAN portfolios increases, the GAL-TAN dimension becomes more important for the average left-right perception of the voters. This result supports our hypothesis. However, we do not find a similar influence for the interaction effect between the ECON dimension and the share of ECON portfolios. This is rather puzzling in light of the results by Meyer and Wagner (2020), suggesting that the weight of ECON positions is more affected by individual issue emphasis than the weight of GAL-TAN positions. However, we speculate that the lack of interaction effects is perhaps due to a comparatively high general emphasis on economic issues by governing parties.
Since interpreting interaction effects is not very intuitive, we plotted the results of our analysis in Figure 2. 7 Since the share of GAL-TAN portfolios only effectively varies between 0 and 50, we only show this part of the figure to ease interpretation. The figure clearly shows that as the share of GAL-TAN portfolios increases, the average marginal effect of the GAL-TAN position also increases. If a party has no GAL-TAN portfolios, a one-unit move towards a TAN position is only associated with a 0.1-unit perceived shift to the right. For a party with 40% GAL-  TAN portfolios, the same positional change would cause a 0.4-unit perceived shift. Substantially this means that the greater the importance of the GAL-TAN position for the overall left-right perception by the voters increases, the higher the share of GAL-TAN portfolios. Imagine a Christian Democratic party that has rather centrist views on the economy but scores more on the traditional side of the GAL-TAN dimension (i.e. a GAL-TAN position of 9). If this party obtains a large share of GAL-TAN portfolios, this might have important consequences for the overall placement of the party. If this party receives 30% GAL-TAN portfolios, the average perception of the party would be at 6.7. If the party instead receives 40% GAL-TAN portfolios, the average perception would be at 7.2. This is a substantial change in the perception of the party, which could have important implications for, for instance, the electoral success of parties. In a second step, we test Hypothesis 2 by including the time between the cabinet start date and the date of the survey. For our proposed theoretical mechanism to work, it might take some time, meaning that if the survey took place a week after government formation, the citizens might not yet know who is in government and who got which portfolio. But if the government has been in office for over a year, this might be different. Since continuous three-way-interactions are incredibly difficult to visualize, we created a dichotomous variable that is 1 if the government has been in office for more than one year at the time of the survey and 0 otherwise.
Model 4 in Table 2 shows that for a party that joined a coalition government within the past year and does not have any economic portfolios, there is only a moderate impact of GAL-TAN positions on voters' left-right placement of the party. For a new government party without any ECON portfolios the effect of the ECON position is slightly higher. Furthermore, the model shows that there is There is a very strong interaction effect between GAL-TAN positions and the dummy variable indicating whether the party has been in government for more than a year. This effect is strong enough to completely cancel out the baseline effect of GAL-TAN positions, meaning that over time as voters accumulate impressions of the party in government, the position the party takes on the sociocultural subdimension becomes essentially irrelevant for voters' placement of the party on the left-right dimension. However, this is only true for parties with no GAL-TAN portfolios. As is visible in Figure 3, as long as the party has 10% or more of its portfolios dedicated to sociocultural issues the average marginal effect of GAL-TAN position is still significant and positive.
The results in Table 2 lend support to our second hypothesis, at least for the GAL-TAN subdimension. Time matters and importantly moderates the interaction between the GAL-TAN position and the share of GAL-TAN portfolios. Governments need time to draft and enact their bills Vanberg 2004, 2011) and respondents need some time to get informed about government composition before it becomes relevant for their perceptions of political parties. We further added a number of robustness checks in the Supplementary Material. First, we analyse if there are differences depending on the party type. Accordingly, we follow Bonnie Meguid (2005) and define niche parties as nationalist, ethno-regional and ecologist parties and run a three-way-interaction with our main analysis. While the interaction term is not significant at first glance, Thomas Brambor et al. (2006) have demonstrated that a non-significant interaction term is not necessarily a problem. Hence, we plotted the interaction in Figure A.1 in the Supplementary Material. It shows that the interaction effect between the GAL-TAN position and the share of GAL-TAN portfolios is especially pronounced for mainstream parties but not significant for niche parties.
Second, we distinguish between prime minister parties and junior coalition parties. The results can be found in Table A.3 in the Supplementary Material. The three-way interaction between the prime minister's party, GAL-TAN position and the share of GAL-TAN portfolios is positive and significant at the 10% level. This means that the interaction effect between GAL-TAN position and the share of GAL-TAN portfolios is especially pronounced for prime minister parties. This result also lends support to our proposed theoretical mechanism, namely that governing parties have easier access to the media, especially the party of the prime minister (Schoenbach et al. 2001). Figure A.2 also displays the moderating effect of being a prime minister party. While the line for junior parties is flat and always includes 0, the graph for prime minister parties demonstrates that the interaction between the GAL-TAN position and the share of GAL-TAN portfolios is especially pronounced for prime minister parties.
Third, we drop the relative salience measure and instead include the salience for GAL-TAN and ECON separately. The separate salience measures do not reach conventional levels of statistical significance, similar to the relative salience measure. More importantly, we still find a significant positive interaction effect between GAL-TAN position and the share of GAL-TAN portfolios.
Lastly, we include the difference between the survey and the cabinet start date as a continuous measure in Table A.5 in the Supplementary Material. 8 As in Table 2, the three-way interactions are not significant. At the very beginning of a coalition government, controlling a larger share of GAL-TAN portfolios is associated with a high weight on the subdimension on the perceived left-right perception, but counter-intuitively 'authoritarian' parties are automatically perceived as more left-wing the higher the share of GAL-TAN portfolios they have. Over time, this 'unconditional' effect of GAL-TAN portfolios, independent of the actual position, disappears and only the interaction between position and portfolio share remains.

Discussion
How do voters perceive the positions of parties? In this article we have argued that portfolio allocation influences the perception of voters in important ways. We follow previous arguments that the left-right dimension can be seen as a super-issue that consists of multiple policy issues and subdimensions (Downs 1957;Inglehart and Klingemann 1976;Sani and Sartori 1983), but we add to the existing research by suggesting that the individual weight of the multiple issues varies depending on the portfolios that the parties obtained during coalition negotiations. More specifically, we argue that certain issues will be more relevant for the perceived position of a party if the party was successful in obtaining the relevant portfolio. In order to test our hypotheses, we analysed the influence of the share of portfolios on the importance of the ECON and GAL-TAN dimension. We also explored how the impact of portfolios is moderated by the time the government has been in office. Overall, we find that the allocation of portfolios matters for the left-right placement of parties by the voters.
The left-right position is one of the key concepts in the academic field of political behaviour and is included in studies of key topics such as vote choice, representation, polarization and many others. Recent research has suggested that voters rely on coalition heuristics to judge where governing parties are located on the left-right dimension (Fortunato and Adams 2015;Fortunato and Stevenson 2013;Spoon and Klüver 2017). In short, they infer that governing parties are ideologically similar to the other members of the coalition government. David Fortunato and Randolph Stevenson (2013: 463-465) speculated that, on one hand, voters interpret the coalition signal to mean that cabinet formation is more likely if parties were already ideological similar before joining, and on the other, that voters will expect that the process of governing together will inevitably require accommodation and policy compromise between the parties in the future. However, the findings above unveil a third mechanism through which the ideological positions of parties are communicated to voters and this could be driving the results of the coalition heuristic theory.
Parties are more likely to obtain portfolios on the issues where they take a median position relative to the other coalition members (Laver and Shepsle 1996). That implies that the parties will emphasize and will get increasingly associated with the issues where they are already closer to their coalition partners. Once voters update their perceptions accordingly, by weighting the unchanging issue positions of the respective parties differently, the coalition members will automatically converge ideologically. This could in turn lead voters to conclude that governing parties have 'sold out' and to punish the parties in subsequent elections.
Furthermore, we would expect this to matter even more for smaller parties with fewer portfolios. David Fortunato and James Adams (2015) have demonstrated that voters map the prime minister's policy position onto junior coalition members, but not vice versa. Since the prime minister puts together the coalition and describes the guidelines for its political project, researchers have theorized that voters will use the prime minister's party as a focal point when thinking about the ideological position of the entire government. However, this result could also be generated by the prime minister's party obtaining a large share of the other salient portfolios and therefore being less affected by the shift in weights. Although we are not able to test this within the current project, we believe it is a potential consequence worth reflecting on.
Our results also have important implications for our understanding of government participation in general and portfolio allocation more specifically. The findings clearly suggest that the allocation of portfolios could have a strategic component, not only with regard to office and policy goals but also with regard to the perception of voters. Party positions and priorities not only influence which portfolios parties receive, but the allocation of portfolios also influences how voters perceive parties. As long as parties obtain portfolios on the issues where they agree with the other coalition members, voters will often perceive parties governing together as converging ideologically. This influence is rarely beneficial for the party, so a strategic party might consider how to avoid convergence by allocating portfolios differently. Another point that could be relevant for parties is whether an unfortunate portfolio allocation could move the perception of potential voters so far away from their ideal points that they might no longer consider voting for the party. If this is true, then parties should not simply try to get as many portfolios as possible but should also strategically decline certain portfolios if they fear that this might alter their perception too much. Future research should therefore have a closer look at the consequences of portfolio allocation with regard to, for instance, electoral support. We extend this data set with information on portfolio allocation in Italy. 5 We acknowledge the fact that the responsibilities and divisions of ministries are likely to change over time. However, we argue that this is not a major problem for our analysis since our timeframe is rather short (1994-2019) and because we restrict our analysis to the main ministries, which are less likely to change their denomination in such a short timeframe (see also Bäck et al. 2011). 6 The correlation between these two variables is at 0.47, which does not create any problems regarding the use of both of these variables simultaneously. We also multiply the shares by 100 to ease the interpretation. 7 The figure is based on Model 3. 8 Since the coefficients would be extremely small otherwise, we divided the variable by 100.