I. INTRODUCTION

Despite general and particular economic considerations being critical influences on Argentine voting ( ^{Ratto 2013} ; ^{Gervasoni & Tagina 2019} ), previous research has revealed the importance of both socioeconomic and sociocultural preferences in structuring the political competence and divisions between supporters of the main political forces in Argentina ( ^{Catterberg & Braun 1989} ; ^{Alessandro 2009} ; ^{Ostiguy 1997, 2009} , ²⁰⁰⁹ ; ^{Mora y Araujo 2011} ; ^{Sendra 2022)} . Indeed, other authors have shown 'elite-mass congruence', meaning a strong correlation between the average policy opinions of mass partisans and those of partisan elites in Argentina, Chile, and Uruguay ( ^{Luna and Zechmeister 2005)} , and that left-right orientations may specifically condition the policy expectations of the voters of conservative parties in Argentina ( ^{Calvo & Murillo 2013)} .

Argentina is an interesting case study of programmatic voting. Despite the severe process of disarticulation of its party system from the 1990s to 2001 ( ^{Roberts 2013} ), linkages between voters and political forces do not seem to have completely unravelled. Especially from 2015 onwards, they seem to increasingly crystallise around two major systems of social and political representation, expressed into two polarized electoral coalitions, one centre-left and one centre-right, Frente de Todos (FdT) and Juntos por el Cambio (JxC). At the same time, both coalitions are made up of political forces with different views and voices that do not always coincide on the central issues of the public agenda.

In this sense, the aim of this research is to examine to what extent the electoral support for the main Argentine parties is driven by policy preferences. Using data from an original online survey, conducted in Argentina during September 2020, we focus on the so-called 'propensities to vote' and policy preferences of Argentine voters. We perform a restructuration of the original data matrix into a 'stacked' data matrix, according the method proposed by ^{van der Brug, van der Eijk and Franklin (2007)} . This methodological approach has several benefits and allows for systematic comparisons. Additionally, we measure programmatic and ideological distance between mass-based issue preferences and parties positions, relying on the Political Representation, Executives, and Political Parties Survey (PREPPS) ( ^{Wiesehomeier et al. 2021)} .

Our results highlight the role of programmatic voting in the Argentine political space. In general, the party-voter distance is relevant in almost all the dimensions considered, with the exception of social policy and transfers. Notwithstanding, notable differences emerge when looking specifically at parties and coalitions. We found that economic spheres (deregulation and redistribution) are crucial for the orientations towards FdT, while support for the centre-right coalition, JxC, seems to be influenced by a more diverse nature of policy dimensions. Lastly, the intra-coalition analysis reveals important differences regarding the intensity of the policy dimensions among the political forces. Inside the Frente de Todos , individual liberties and transfers are more important for Frente Renovador than for Partido Justicialista . On the other hand, in Juntos por el Cambio , economic dimensions are less important for UCR voters than for the rest of the coalition.

This article is structured as follows. After this introduction, we present and develop the theoretical review and argument that underpin the reasons why the last realignment of the party system may have a correlate in programmatic structuration party linkages. Hypotheses are also formulated in this section. In the third section, we provide the case selection, variables, and models to test these hypotheses. We then present the multivariate analyses, and, finally, we conclude by discussing the implications of our findings.

II. WHY ARGENTINE PARTIES' SUPPORTERS CAN BE PROGRAMMATIC VOTERS

Within a rational choice paradigm, programmatic voting assumes that on the basis of a supply of parties' policy packages, citizens reward them with their vote. According to ^{Kitschelt et al. (2010)} , there will be programmatic voting if the voters' policy preferences are similar to those advocated by the parties. A broad literature explains voting by considering parties' positions and individuals' preferences ( ^{Downs 1957} ; ^{Rabinowitz & Macdonald 1989} ; ^{Merrill and Grofman 1999)} . Most authors follow, explicitly or implicitly, ^{Downs' (1957)} spatial voting theory and his so-called 'proximity model', according to which individuals prefer parties closer to their own political preferences. This is usually studied by identifying, at the individual level on a left-right dimension, the ideological distance between voters' self-positioning on an issue (or issues) relevant to the electoral campaign and the position that voters impute to the candidates on the same issue ( ^{Green & Hobolt 2008)} . The theoretical expectation is that the likelihood that a voter will support a candidate naturally decreases with the ideological distance between them. Thus, programmatic voting occurs when a voter chooses the candidate whose publicly announced platform best approximates the voter's own policy preferences.

Although a rich body of literature has considered policy preferences voting since the seminal study of ^{Downs (1957)} , research on Argentina has not being sufficiently sensitive to these approaches, except for a few notable cross-country works mainly focused on Latin American political congruence that include the Argentine case ( ^{Luna & Zechmeister 2005} ; ^{Herrera & Morales 2018)} or studies that analyse the dimensional nature of citizen attitudes and party choice for particular elections ( ^{Alessandro 2009} ; ^{Cataife 2011} ; ^{Schiumerini 2019} ; ^{Sendra 2022)} . ^{Schiumerini (2019)} pointed out that in cases where the electorate largely agrees on the parameters of the policy regime, voting based on performance assessments indicates lower reliance on valence issues and higher reliance on positional issues.

Within political congruence studies, previous research has revealed a strong level of 'elite-mass policy congruence' in Argentina ( ^{Lupu and Warner 2017} , ^{Herrera and Morales 2018)} . ^{Herrera and Morales (2018)} found that, despite having a less institutionalized party system, Argentina has higher levels of programmatic congruence than Chile and Uruguay, in at least two of the policy dimensions that structure political competence. This could be explained by the fact that Argentina has a proportional electoral system which, from an institutional perspective, generates incentives for fragmentation and for searching for programmatic spaces by leaders and parties ( ^{Herrera & Morales 2018)} .

The path dependence of partisan linkages in Argentina can also be explained by the long trajectory of the party system. According to ^{Bornschier (2019)} , party systems in Latin America that exhibited congruence in the 1990s were those in which a challenging party had started polarizing the party system several decades earlier. Following a trajectory of inclusion and sustained polarization, conflict along the state-market dimension was channelled into party competition in countries such as Uruguay, Chile and Argentina, and became embedded in lasting political identities and partisan loyalties. In Argentina, the Peronist/ anti-Peronist antagonism in fact represents a sectoral divide: Peronism allied with the urban working class and other social groups, which pulled protectionist segments of the rural elites toward anti-Peronism ( ^{Bornschier 2019} ).

Despite the Peronist move to the right in pursuit of market reform under Menem in the 1990s in a move of what ^{Stokes (2001)} calls 'policy switches', this did not exceedingly damage its representation in the short run ( ^{Bornschier 2019} ). There is widespread consensus that the Peronists' (increasing) reliance on clientelistic mobilization strategies has hampered representation in the longer run. As ^{Calvo and Murillo showed (2014} , ²⁰¹⁹ ), political networks play a large role in defining the distributive expectations for Argentine voters regarding the PJ and UCR. Furthermore, the existing studies in Argentina on the relationship between political clientelism and voting suggest that the link is more complex than expected. Using ecological data at the municipal level, ^{Nazareno, Stokes, and Brusco (2006)} find that clientelistic distribution of goods reaffirms the loyalty of loyal voters; yet, it does not affect the vote of opponents, and it induces marginal voters to vote against. Using individual level data, ^{Lodola & Seligson (2013)} found that there was no significant effect of the supply of goods and the receipt of the universal child allowance on the probability of voting for Peronism in the 2011 presidential elections.

Moreover, from 2003 when the Peronism moved back to the left under Néstor and Cristina Kirchner's leadership, ideological differences increased in the following elections, while until that moment, Argentina had had low levels of polarization ( ^{Singer 2016} ). As pointed out by ^{D'Alessandro (2013)} , in its beginnings Peronism had a very regulatory stance on the economy, which changed in 1989 for almost two decades, to return to a more interventionist position in the campaigns of Cristina Fernández.

The emergence in the same period of the party Propuesta Republicana (PRO) achieved, in alliance with UCR and the Coalición Cívica (CC), a centre-right alternative with electoral returns ( ^{Alessandro 2009} ). According to ^{Anria & Roberts (2019)} , the triumph of the coalition led by Mauricio Macri in the 2015 presidential elections meant a reinforcement of the post-neoliberal left-right programmatic structuring of political competition. Indeed, it has been found that identification on the left-right spectrum has strengthened in Argentina since 2015, due to the decreasing trend in the number of non-responses to the self-left-right location (Lupu, Oliveros, & Schiumerini 2021). This was even more evident in the 2019 elections, when the political space aligned around an axis of competition from left to right, with two antagonistic coalitions ( Frente de Todos and Juntos por el Cambio ) that structured the electoral supply in a much clearer way than under the classic two-party system of Peronism/ Unión Cívica Radical .

This structuration is based on both ideological factors, the existence of a realignment of the political system along the left/right axis, and sociological factors, the existence of differentiated social bases in terms of socioeconomic sectors and collective identities ( ^{Anria and Vommaro 2020} ).

This shape of the party system, we argue, may have a correlate in programmatic party linkages. Based on the assumption that Argentine political parties' supporters can be rational consumers in the electoral market, who guide their voting decisions based on policy preferences and without denying the importance of alternative explanations, we expect that 'policy voting' can explain, to a large extent, the electoral preferences for Argentine political parties (H1), in line with previous findings ( ^{Alessandro 2009} ; ^{Cataife 2011} ; ^{D'Alessandro 2013} ; ^{Sendra 2022)} .

However, it is necessary to consider the main policy dimensions that structure Argentine party systems and society. Studies of economic-distributive conflict in Latin America have shown that, in the late 20th century, in countries such as Argentina, which had adopted relatively comprehensive social welfare policies during the ISI (Import Substitution Industrialization) period, the electorate was able to crystallize its preferences around specific political-economic achievements; hence, there was programmatic competition largely but not exclusively structured by rival preferences towards economic and redistributive policies ( ^{Kitschelt et al. 2010} ). ^{D'Alessandro (2013)} points out that there is a high level of consensus among the Argentine political class on the importance of economic issues, as well as those linked to the provision of social services by the state. Although in the past decade, a new interest of political discourse in line with moral and law and order debates was confirmed ( ^{Cataife 2011} ; ^{Tagina & Varetto 2013} ), we expect that these policy issues will be subordinated to economic dimensions. In summary, we want to test the expectation that economic policy dimensions (redistribution, deregulation, and transfers, which will be treated separately) have greater explanatory power than the rest of the policy dimensions in explaining electoral support for the party supply in Argentina (H2) .

Moving deeper into an inter-party analysis, we argue that there might be some dimensions that are more salient for a limited-rationality party voter than others. Thus, we also expect some differences between the two main political coalitions of parties regarding the impacts of particular policy dimensions. As we mentioned, the shift to the left of Peronism with more statist positions carried a left-wing agenda in sociocultural terms as well. New moral issues emerged in the public debate as a result of a subset of policies promoted by Peronism to expand minority rights ¹ . Additionally, from 2003 onwards, another non-economic issues with electoral returns began to take centre stage: crime or insecurity ( ^{Tagina & Varetto 2013} ). ^{Cataife (2011)} finds robust evidence that crime policy is a statistically significant factor in vote choice across voter groups in Argentina. This topic was brought to the agenda mainly by the PRO, who alike other current Latin American centre-right parties (e.g. RN and UDI in Chile), went through a programmatic moderation in socio-economic terms, accepting part of the distributive policies displayed during the past years, but at the same time, found grounds to install a more favourably conservative agenda in topics abandoned by the left, especially in matters of insecurity ( ^{Monestier & Vommaro 2021} ; ^{Luna & Rovira-Kaltwasser 2021} ). In this sense, the PRO took the discussion on the means by which the government fights crime closer to a position of mano dura even when this could mean affecting individual liberties. Different views on migration issues are also connected to crime as well. The centre-right has argued that the arrival of illegal immigrants is one of the factors contributing to increase crime or conflict social situations, such as the occupation of land to demand housing ( ^{Vommaro et al. 2015} ; ^{Vommaro 2019)} .

Considering the above, that is, the issue membership that each political force emphasizes, we expect that the individual liberties dimension has greater explanatory power for the centre-right parties (PRO, UCR, and CC) than for the centre-left parties (PJ and FR) (H3). On the other hand, the preferences for centre-left parties (PJ and FR) are expected to be more strongly affected by the social policy dimension in comparison with centre-right parties (PRO, UCR, and CC) (H4).

Lastly, we must take into account that, given the fragmentation of the party system and the territorialisation of voting ( ^{Calvo 2005}: ^{Calvo & Escolar 2005} ), coalition governments became an inherent feature of the Argentine political system from 2001 onwards ( ^{Tcach 2011} ; ^{Clerici 2018)} . In the latest 2019 elections, it became clear that political competition continues to be structured in terms of electoral coalitions, through a kind of 'bipartisanship of coalitions' or 'bicoalitionism' ( ^{Sendra 2020} ; ^{Cruz 2021)} . Therefore, it is of interest to distinguish the relative importance of policy dimensions in an intra-coalition analysis as well. Both coalitions that currently lead the political scene are quite heterogeneous and represent different political voices. While Frente de Todos is formed by the Peronism movement, which includes the Partido Justicialista , Kirchnerism, the Frente Renovador , and other progressive minority sectors, Juntos por el Cambio is integrated by a relative newer party, PRO, the centenary party UCR, and CC ( Coalición Cívica ), a political split of the UCR created in 2002. Whether we understand coalitions not only as electoral and government instruments but also, from a sociological perspective, as convergences or mergers between social actors ( ^{Di Tella 1999} ) , we can expect coalitions reflect the existence of an alternative set of preferences to those of the parties individually that make them up. Thus, we want to put to test the general expectation that the role and the intensity of the policy dimensions varies among parties within each coalition (H5), although we cannot predict in which direction this variation in intensity would occur.

III. DESIGN AND METHODOLOGY

This study aims to examine the Argentine political space from the point of view of voters' electoral orientations. We focus on voters' support to the main Argentine political parties. Table 1 shows the six selected parties, all of which are currently represented in the national parliament. All of them are established parties that play a key role in the political system and reflect the main ideological sensitivities in contemporary Argentina. Moreover, this table shows the specific electoral coalitions, which are necessary to understand the Argentine political map and the findings of our research. In addition, the shares of votes obtained for each coalition in the last 2019 elections is added, showing that together the Frente de Todos and Juntos por el Cambio gather almost the 90% of the electorate. Lastly, it is worth to mention that these are pre-agreed coalitions, for which the citizens vote for in the elections. However, parties often keep their own parliamentary group. That is the reason why it is important to analyse not only the coalitions, but the parties as well.

Our data mainly come from an original survey of the University of Salamanca ( ^{Plaza Colodro et al. 2021} ) that was implemented online by the company Netquest. The survey was conducted during September 2020 in Argentina and covered a wide range of sociodemographic and attitudinal variables, typically used to study electoral behaviour. It had N = 1003. To avoid bias in the distribution of age, gender, education, and territorial distribution of respondents, the 1003 respondents were selected using quota sampling (see details in Online Appendix) ² .

We also used the Political Representation, Parties, and Presidents Survey (PREPPS) ( ^{Wiesehomeier et al. 2021} ). PREPPS is an expert survey that provides information about the policy positions of 156 political parties in 19 Latin American countries for the period 2018-2019.

The dependent variable of the research is operationalized through the so-called 'propensity to vote' (PTV). In PTV questions, respondents are asked to indicate how likely they would be to vote for each political party in a scale of 0-10 (0 represents 'not at all probable' and 10 'very probable'). PTV is gaining increasing popularity in Political Science, being accepted as an excellent indicator of electoral preferences because of its benefits ( ^{Paparo, De Sio, & Brady 2020} ). Basically, PTV allows maximizing the available N, since almost all respondents express their opinion for almost all parties. Thus, the same voter can simultaneously have a high probability of voting for two different parties. This is a great advantage compared with the traditional qualitative 'voting recall' variable, in which minor parties have a very small N. In this sense, PTV measures is not so much electoral choice, but rather preferences or orientations towards political parties, that is, the electoral attractiveness for each party. Moreover, PTV makes it possible to perform linear regressions models that have analytical advantages in comparison with discrete-choice models, most notably, logit and multinomial logit models, traditionally associated with the 'voting recall' question ( ^{van der Eijk et al. 2006)} .

However, the raw PTV question (scale 0-10) and the data matrix itself are subjected to transformation. Specifically, we carried out a method that has been subsequently applied by several studies ( ^{van der Eijk & Franklin 1996} ; ^{van Der Brug et al. 2000} ; ^{van der Brug & Fennema 2003, 20} , ²⁰⁰⁹ ; ^{van der Brug et al. 2007} ; ^{Ortiz et al. 2021)} which consists in stacking the data matrix. In this new 'stacked' matrix, each case is represented by the combination of each respondent and the answer to the PTV question for each party (respondent x PTV _n ...). When the data matrix is stacked so that each voter appears as many times as there are parties for which her probability has been measured, the question can be posed 'what is it that makes a vote for a party attractive to voters? ( ^{Van der Brug & Fennema 2009)} . To perform this transformation, we used the Stata package PTVTOOLS ( ^{De Sio & Franklin 2011)} (See more details about the stacked data matrix in Online Appendix).

Then, having the new stacked matrix, we focused on several item statements from our survey. In these questions, respondents should indicate their agreement or disagreement toward particular policy issues (scale 1-5, where 1 = totally agree; 5 = totally disagree). In particular, we select six items regarding different policy dimensions. Three of them related to economic-distributive and state intervention issues that are particularly important in the Argentine context: redistribution (attitudes towards policies to reduce the income gap by redistributing wealth), deregulation (attitudes towards the state's intervention by protecting domestic industry and charging taxes on exports to the agricultural sector), and transfers (attitudes towards the state assistance of socio-labour vulnerability groups). Then, we included a social policy dimension, with attitudes towards LGTB rights, abortion, and drugs legalization), immigration (attitudes towards the impact of immigration in society); individual liberties, which taps attitudes towards tough measures, including an authoritarian regime, to fight crime. To build the six policy dimensions, whose importance has been reviewed in section 2, we ran exploratory factorial analyses using the sets of statements. We found that all the items loaded in the same factor, with Cronbach's alpha of >0.4.

In parallel, we extracted the positions of the main six Argentine political parties (PJ, FR, UCR, PRO, CC, and FIT) in the same six policy dimensions described above according to data from PREPPS (See the wording and matching of questions in Table OA.2). After standardizing the scale of responses for each dimension in both datasets (voters and parties), we calculated the distance between the positions of respondents and political parties on each dimension ³ .

According to the policy voting explanation, the smaller the distance is (or, in other words, the greater the agreement between voters and parties), the higher the probability of supporting a party. It is true that past research focused on the ideological and thematic congruence between voters and parties ( ^{Luna & Zechmeister 2005} ; ^{Otero Felipe 2017} ; ^{Herrera & Morales 2018)4} . However, offering a rich picture of programmatic voting by distinguishing specific policy dimensions, beyond analysing whether parties and voters' positions are congruent, is an added value of the article. In conclusion, the distances between respondent's position and parties' position on the policy dimensions were the main independent variables of the research.

Apart from the abovementioned distances on six policy dimensions, other independent variables were added as controls. On the one hand, we used the question on the general assessment of economic situation (1-Very good to 5-Very bad), to operationalize the 'economic voting' explanation. As argued before, we expected that economic voting may have strong effect on electoral preferences in Argentina in line with previous research. On the other hand, we operationalized the 'class voting' explanation through three questions: net monthly income (1-No income to 11-More than $62.000), self-perceived social class (ranging from 1-upper class to 5-lower class) and education (1-primary and high school studies/2-university studies). We further controlled for other kinds of party mass linkages by considering the distance between voters and parties on a people-centrist dimension. This variable measures attitudes referring to perceptions of people as a virtuous entity that share key characteristics and have a clearly defined popular will. We needed to include this indicator since populists discourse and ideas have played a crucial role in the formation of Argentine mass politics, and for the fact that populists ideas are successful in a context where ideological divides are not present and parties do not have strong links to like-minded constituencies ( ^{Bornschier 2018} ). Finally, usual sociodemographic variables on sex and age were added.

Regarding the independent variables for which distances could not be calculated (i.e., all with the exception of the six which referred to policy dimensions), we follow the procedure suggested by ^{van der Brug et al. (2007}: 43-45). We performed a linear transformation of the original variables by running a series of bivariate regressions to predict the PTV question for each independent variable. The predicted values of these regressions, known as y-hats, were used as the new independent variables.

IV. RESULTS

Table 2 presents different OLS regression models ⁵ in which the PTV is predicted by the independent variables explained above. Dummies for the 24 Argentine provinces have been added in order to control for regional differences and potentially omitted variables, thereby removing idiosyncratic explanations from the models. We also estimated these models without province dummies, and the findings were essentially the same (see models with dummies coefficients in Table OA.3). Standardized beta coefficients are used to enable a direct comparison of the strength of the independent variables. These coefficients range to -1 to 1 and show the effect of a particular variable when the rest are hold constant.

In particular, model 1 estimates the PTV for all six Argentine political parties combined, while models 2 to 7 estimate the same for the parties separately (See additional models for the coalitions separated in Table OA.4). As explained before, several independent variables were created inductively through a linear transformation. These variables are those for which it is not possible or does not make sense to calculate the distance between respondents and parties: economic evaluation, income, education, sex, and age. As a result of this transformation, the standardized coefficients are generally positive. In sum, no conclusions can be extracted about directionality but only about the intensity of the effects. While some may argue that this is a limitation, this allows us to properly test the proposed hypotheses from the point of view of our research design. Furthermore, the fit of the models is in general quite acceptable, explaining between the 10% and 20% of the variance of the dependent variable.

First, examining model 1, we observe that policy voting effectively structures the party preferences for the six main Argentine parties. In other words, policy dimensions such as immigration, redistribution, individual liberties, and deregulation explain the electoral orientations in the Argentine party system (when controlling for other variables of sociodemographic factors, economic evaluation, social class, and people-centrism). As expected, the shorter the distance was between a respondent's position and the party's position, the more attractive the party would be. For this reason, the standardized coefficients are negative. Thus, model 1 is in line with H1, which stated the explanatory power of policy voting. In addition, age and economic evaluation were the factors with the best power prediction of party preferences in the Argentine party system (coefficients of 0.125 and 0.164 respectively, significant at a <0.001 level).

In addition to the above, H1 is also supported by the fact that in the separated models for each party (models 2 to 7), policy voting seems to be important. For all the parties with the exception of FR, at least one policy dimension is statistically significant. In sum, the results generally point to the explanatory power of policy voting models in order to explain party preferences in Argentina. Other variables (sociodemographic or economic voting) are also significant in several cases. Obviously, policy voting does not play a role alone, but alongside with other determinants. What we aim to illustrate here is the explanatory power of the policy voting approach to understand the complex sphere of voting behaviour (and considering that this approach has not been given sufficient attention in the Argentine and Latin-American cases and, specifically, using the PTV as dependent variable).

Now, we turn to review the rest of the hypotheses regarding the specific role of policy dimensions. On the one hand, H2 stated the pre-eminence of economic dimensions in comparison with the rest of policy dimensions. First, it should be noted that positions on the transfers dimension do not explain voting in any of the models, while redistribution and deregulation are clearly important for the whole party system and for the parties separately. While this hypothesis has empirical support for redistribution and deregulation in the case of PJ, PRO, CC, and FIT, it does not have support for UCR and FR. Looking separately at the different political parties, the results point to partial support for this hypothesis.

On the other hand, the individual liberties dimension is statistically significant for PRO and CC, but not for UCR. Meanwhile, as expected, it did not have significance neither for PJ nor FR. This is consistent with the initial expectation made in H3 regarding the greater relevance of this policy dimension for predicting preferences for centre-right than for centre-left. As explained before in the theoretical background, the previous literature has pointed out the importance of the individual liberties dimension for the ideological right on both the supply and demand sides. Moreover, the individual liberties dimension seems to be significant in model 7, that of the radical left FIT, but with a lower magnitude and significance. All this suggests that this dimension is important at orienting the party preferences for the centre-right and the radical left. The ideological position of potential voters for these parties are in the ideological antipodes. However, the focus of our research is not on the 'raw' positions, but on the distances between voters and parties. Hence, we are not examining the orientation but the intensity of party preferences.

Moreover, the expectation made in H4 does not have empirical support, since the social policy dimension are not more important for centre-left parties than for centre-right parties. Indeed, this dimension is not significant either for PJ, FR, or the leftist coalition FdT. Surprisingly, the social policy dimension predicts party preferences for CC (model 5).

As shown previously, the models in Table 2 present results that allow ad oculum comparisons of differences in the effect coefficients. The stacked nature of our data matrix allows us to go one step further and to perform more systematic comparisons to identify intra-coalition similarities and differences. For this purpose, we ran four regression models including several interaction terms between each party and the policy dimensions, comparing it with the main effects of the coalition ( Table 3 ). That is, if a policy dimension has a different effect for one party as opposed to all the other parties that integrate the coalition, the regression model will contain an interaction term between the respective party, on the one hand, and this variable, on the other ( ^{van der Brug & Fennema 2003} ). In other words, in Table 2 we can only know if "X policy dimension is significant for X party", while the interactions ( Table 3 ) indicate us whether differences among parties (a higher or lesser effect of a policy dimension for a party regarding the whole coalition) are significant. These models also contain the other independent variables presented in the previous analyses (see Table 2 ), but they are not shown here, since we are interested only in the interaction terms.

Model 8 refers to the coalition FdT (integrated by PJ and FR) and shows the interaction terms between FR and the different policy dimensions, as well as the main effects. To interpret the meaning of the interaction terms, the main effects should be considered. For example, the main effect for the coalition regarding the redistribution dimension is -0.635 (significant at a <0.001 level), which indicates that, as expected, the less distance in this dimension, the more attractive these parties. The positive interaction effect shows that the policy voting is somewhat weaker in this dimension for FR than for PJ: the unstandardized effect for FR is 0.027 (-0.608 +0.635). Moreover, the negative signs of the interaction terms regarding individual liberties and transfers indicates the stronger effect of policy voting for FR in comparison with PJ. In conclusion, model 8 illustrates the existing intra-coalition dissimilarities of the party preferences of voters.

In addition, the rest of models in Table 3 focus on the intra-differences of the coalition JxC (integrated by UCR, PRO and CC). The negative significant coefficient of the interaction term of UCR and immigration is noticeable (Model 9), since points to the greater importance of policy voting in this dimension for this party, than for the rest of the coalition. In the rest of dimensions (social policy, redistribution, and deregulation), policy voting seems to be weaker for UCR than for the rest of the coalition partners (that is, the coefficient is positive). Moreover, models 10 and 11 allow us to observe the differences between PRO and CC in comparison with the rest of the coalition. In this sense, policy voting is weaker for PRO regarding social policy and immigration, while it is stronger for CC regarding social policy. There results point to the fact that party preferences for UCR seem to be more idiosyncratic regarding its electoral orientations, as it is the party for which more significant differences can be found.

In sum, the systematic examination of the interaction terms has shown the heterogeneous nature of the party preferences for the different parties within each coalition. As we expected in H5, some relevant similarities and differences can be found at the intra-level coalition. The two main Argentine coalitions capture different sensitivities in terms of the nature and strength of policy voting.

V. CONCLUSIONS

This article addresses to what extent party-choice might be driven by 'policy preferences' in the Argentine party system. The prior literature provided evidence supporting the importance of policy dimensions for party choice in Argentina but mainly focused on the super-dimensions of Peronism/anti-Peronism ( ^{Ostiguy 2009} ) or pointed out the relevant dimensions in party choice and programmatic structuration of political competence in the last years ( ^{Alessandro 2009} ; ^{Cataife 2011} ; ^{Luna & Rovira Kaltwasser 2021} ; ^{Monestier & Vommaro 2021} ; ^{Sendra 2022)} . However, these studies do not unpack the programmatic voting effect, as they did not look at political party positions along the same dimensions where voter preferences are measured. Instead, in this research we considered the stances of voters and the political party they support on a number of dimensions and tested whether the distance between them affects their electoral attractiveness. In this way, we are truly assessing whether there are programmatic linkages in Argentina between voters and parties for which those voters would potentially vote for.

In sum, within the constraints posed by our available survey data, the results generally point to the explanatory power of policy voting models in order to explain party preferences in Argentina, along and compatible with other deter- minants of vote. Although we cannot deny the importance of economic performance perception and social class on voting, it could be possible to fundamentally consider Argentine voters as rational consumers in the electoral market who guide their voting decisions through rational cognitive mechanisms.

Focusing on the overall party model, we see that the party-voter distance was relevant in almost all the dimensions considered, with the exception of social policy and transfers, which indicates that perhaps the political supply does not produce differentiating messages on these themes and, therefore, do not generate a programmatic effect that would link them to voters' preferences.

We also showed that the party system structuration has a correlate in programmatic terms. The two main coalitions, the left-centre (FdT) and right-centre (JxC), present different patterns of representative linkages. We found that although Peronism (FdT) promotes LGTB or abortion policies, it does not activate a programmatic vote in these issues, but its links with voters are focused exclusively on economic dimensions (deregulation and redistribution). This can be explained by the fact that the move that Peronism made to the left on the socio-cultural dimension may have been a move from the elites, which was not necessarily accompanied by a shift of voters towards an emphasis on these issues. From the agency, the party decided to enhance the salience on these issues, but at the mass level, this does not seem to have been replicated. Support for the centre-right coalition, Juntos por el Cambio , seems to be influenced by a more diverse nature of policy dimensions. In this sense, JxC would be more akin to the new Latin American right that seeks to reach broader electorates, by activating not only distributional conflicts but also aiming toward mobilization on sociocultural and law and order issues ( ^{Roberts 2014} ; ^{Luna and Rovira Kaltawasser 2021} ). For a long time, interpretations of political competition in Argentina identified Peronism as a hegemonic force adaptable to different ideological platforms that eclipsed the possibility of partisan-programmatic structuration among parties. Nevertheless, the emergence of the left-wing Peronist trend, Kirchnerism and, afterwards, the formation of a centre-right force (PRO) with strong electoral mass linkages is something novel in the Argentinian case, having strengthening the programmatic structuration of party-mass programmatic linkages between two main coalitions that separate the political field between a centre-left and a centre-right.

Lastly, the intra-coalition analysis revealed important differences regarding the role and the intensity of the policy dimensions among the political partners of each coalition. An insight of the Frente de Todos showed that Frente Renovador , a moderate split space from Peronism in 2013, seemed to active a programmatic vote in similar terms as the centre-right. Individual liberties and transfers were more important for Frente Renovador than for PJ, whereas redistribution was more important for PJ than for FR. On the other hand, in Juntos por el Cambio , we found that the historical party, Unión Cívica Radical , was the one that presented more differences in terms of programmatic vote regarding the rest of the coalition. UCR voters do not support the party because of its programmatic position in the most important conflict of the political competition, that is, the economic-distributive one. This role seems to be occupied by the PRO, the partner with the greatest weight of leadership within the coalition. Likewise, the finding that the UCR is articulating a programmatic vote on immigration is absolutely novel and may indicate that the party positioned itself in a dimension whose ownership was vacant and where its potential voters are receptive to them.

In a nutshell, these intra-coalition differences could imply that whether parties take their voters' preferences seriously, consensus-building may become an arduous task and, thus, jeopardize the coalitions' cohesion. In other words, the intra-coalition differences bring up to discussion the question whether the difference on the electoral bases of the coalitional partners have consequences in the political relationship among them. Mainly, whether despite their differences, the parties decide to stick together in order to be able to win electorally against the other coalition, even if they have to face the internal dilemma of offer different policy packages to its supporters.

To sum up, we proved that electoral support in Argentina could be guided by policy preferences, with economic dimensions playing a major role for the whole party supply as the previous literature has already pointed out. However, within the two main coalitions, PJ and PRO's voters were mainly driven by programmatic motivations in the economic-distributive issues, while their minority partners were more peripheral in this policy dimension. Additionally, we added that the sociocultural programmatic dimension was important for voters of the new centre-right political space in Argentina, but not for Peronist voters.

As we have mentioned, these conclusions have limitations. Given the stacked nature of the data and the linear transformation of the independent variables, directional effects cannot be inferred. However, this is an implicit deficit of the research design and does not preclude achievement of the initial objectives. Our findings have several implications and suggest new questions for future research. It would be especially relevant to include cases that allow to shed light on whether 'policy voting' is also preeminent in other Latin-American cases. Overall, our findings suggest some methodological and theoretical keys in order to improve their examination, like the use of the propensity to vote. This approach allows to know how probable respondents consider to ever vote for a given party, regardless of their discrete vote intention. This comparative focus will allow discernment of cross-country similarities and differences regarding the role and intensity of policy voting in Latin America.