Partisanship, Ideology, and Selective Exposure: A Longitudinal Analysis of Media Consumption in Spain (2008–2019)

The literature on selective exposure has shown that ideology and voting govern media consumption decisions, but longi‐ tudinal studies are still scarce. To fill this gap, this work analyzes: (a) whether selective exposure guided by ideology and partisanship to three media types (television, radio, and newspapers) is a stable phenomenon in Spain (2008–2019) as expected from a polarized pluralistic media system; and (b) which variable has the strongest effects on news choices (ide‐ ology or partisanship). Multinomial logistic regressions are carried out using data sources from six post‐electoral surveys conducted by the Center for Sociological Research in the last six Spanish general elections (2008–2019). As dependent vari‐ ables, we consider the four most important preferred media outlets for each media type, and as independent variables, we include the vote in general election and ideology. Moreover, interest in politics, gender, age, education, and social class are included as control variables, too. Findings show strong evidence that selective exposure based on ideology and partisanship is a well‐established phenomenon in Spain for the three media types during the 11 years. Results also show that the effect of partisanship is always stronger than that of ideology for the three news media types. We conclude that selective exposure in Spain needs therefore to be conceived as partisan selective exposure.


Introduction
The classic hypothesis of selective exposure suggests that individuals expose themselves to the media in search of messages that confirm their attitudes and avoid those that contradict them (Steppat et al., 2022;Stroud, 2008).Although this argument has been discussed by experts such as those who claim that there is only de facto selectivity (Freedman & Sears, 1965), experimental research has provided a more nuanced and complex picture of political selective exposure, showing that it only occurs under certain conditions (Meffert et al., 2006).In this sense, the multiplication and fragmentation of media systems over the past few decades have led the classic hypothesis of selective exposure to be adapted to new forms of media consumption (Bennett & Iyengar, 2008;Iyengar & Hahn, 2009;Stroud, 2010).
One of the topics regarding which selective exposure has been corroborated most strongly has been its application to the consumption of political information.The growth of less objective and more politically biased journalism has generated more evidence that shows that party-specific selective exposure is increasing (Iyengar & Hahn, 2009), particularly in more polarized media contexts (Bos et al., 2016;Fletcher et al., 2020;Skovsgaard et al., 2016).Most studies have used cross-sector data, but very few analyses consider how selective exposure evolves over time through longitudinal approaches.
Over the past two decades, the number of media outlets through which Spaniards receive political information has increased, giving each person more power when it comes to choosing the sources of information that are available to them, and creating niches around ideological and political predispositions (Humanes, 2014;Martín-Llaguno & Berganza-Conde, 2001;Valera-Ordaz & Humanes, 2022).However, scholars have not considered whether ideology or party affinity is the variable that best explains media selection from a longitudinal perspective.This long-term vision is especially crucial given the major transformation of the Spanish party system-which shifted from a well-established two-party system to a multi-party system in 2015-and the effects of digitalization on the media system: the emergence of digital newspapers, the dramatic increase in supply, and the transformation of media consumption habits.Thus, analyzing selective media exposure using a broad time frame allows us to verify whether media consumption driven by political attitudes is a structural feature of Spanish audiences' behavior regardless of the political context and changes in opportunity structures derived from information environments (Skovsgaard et al., 2016).Based on this evidence, this study focuses on selective exposure to political information through three types of media (radio, television, and the print press) in Spain over a period of 11 years, addressing two objectives: (a) whether selective exposure guided by ideology and party affiliation is a stable phenomenon in Spain, as would be expected in a polarized pluralist media system; and (b) which variableideology or party affinity-has the strongest impact on news choices.

Longitudinal Perspective on Selective Exposure
One of the limitations of the research on selective exposure is the lack of longitudinal studies that address continuities or changes in news consumption in different media (Karlsen et al., 2020).While some authors have shown that ideologically guided selective news consumption has increased over the past few decades, few studies introduce time as an explanatory variable, as most are based on cross-sector data.
As Strömbäck et al. (2013, p. 416) have stated, "people's news consumption cannot be explained by individual-level factors only…because changes in the media environment affect individual-level factors."Furthermore, it is important not to assume that relationships between media consumption and other variables are stable, as they change over time (Strömbäck et al., 2016).In this sense, Dahlgren et al. (2019) state that one must address how changes in media environments have made it easier for people to seek out and selectively expose themselves to information that aligns with their attitudes and how this influences the political polarization of attitudes (Levendusky, 2013a).
The study of Garrett et al. (2013) showed that the use of political news sources aligned with one's own attitudes correlated to the use of discrepant media, and this pattern remains stable over time and across different online media outlets.Strömbäck et al. (2013), using surveys from 1986 and 2010 with data on media use on different platforms, found that news consumption in Sweden has become more polarized among both individuals who seek out news and those who avoid them over time.
In the US, Rodriguez et al. (2017) corroborated the increase in party-aligned and ideological selective exposure between 2000-2012.The most ambitious longitudinal study was conducted by Dahlgren (2019), who analyzed which political preferences (ideology, party affinity, or political interest) drove selective exposure to public media between 1986-2015 in Sweden.He showed that ideology and interest in politics are stable variables over time for explaining selective exposure, while party support for extra-parliamentary organizations explains a major drop in the use of public service news during the period.
In the case of Spain, no studies have been conducted that would introduce the longitudinal perspective.To contribute to this lack of longitudinal studies, we propose the following hypothesis: H1: Ideology and party identification are stable predictors of selective exposure to the media (print press, radio, and television) over time.

Partisan Selective Exposure to Political Information
Selective exposure to political information can be guided by ideological approaches or by interest in politics (Skovsgaard et al., 2016).However, most empirical evidence refers to the consumption of politically oriented media (Camaj, 2019;Iyengar & Hahn, 2009;Messing & Westwood, 2012;Rodriguez et al., 2017;Stroud, 2010), especially in polarized political and media contexts (Bos et al., 2016;Fletcher et al., 2020;Skovsgaard et al., 2016), where individuals tend to select the media outlets that align with their ideology, leading to what is known as partisan selective exposure (Garrett & Stroud, 2014).
In Spain, various studies conducted over the past two-plus decades have shown that media use is politically oriented (Martín-Llaguno & Berganza-Conde, 2001).More recently, Humanes (2014) found a clear connection between political positions and selective media consumption.The print press is the political and electoral news media outlet that is most conditioned by the effect of selective exposure due to political orientation.
The relationship is stronger with voting recall than ideology.The news use of broadcast media outlets, though also coherent with the process of selective exposure, shows less consistent results than those obtained based on newspaper consumption.For their part, radio stations align more with the trend toward selective exposure to the print press than television does.
The study conducted by Valera-Ordaz (2018, 2022) explored the importance of positions on the centerperiphery axis as predictors of media preferences in the case of the media in Catalonia.Using data from the Center for Sociological Research (Centro de Investigaciones Sociológicas, hereafter CIS) post-electoral study for the 2015 autonomous elections, the results showed that the national identification variables, support for independence, and nationalism are the most important factors for explaining the preference for regional TV and radio stations.However, the pattern did not repeat in the case of the print press.While the indicators related to nationalist sentiment also have a strong explanatory capacity, the left-right ideological axis determines the selection of the newspaper that one reads for informational purposes.
Valera-Ordaz and Humanes (2022) found that ideological-party orientations are clearly the most important variables that govern selective exposure, especially for the digital press and radio, in addition to signaling an important trend toward selective avoidance of news media perceived to be ideologically incongruent.
However, none of these studies have tested which of the two variables-ideology or party affinity-has more weight when it comes to explaining the selection of political information.The Spanish media system fits the polarized pluralist model (Hallin & Mancini, 2004), which is characterized by a high level of political parallelism, external pluralism, and an underdeveloped press market.Furthermore, the audience is ideologically polarized, and the journalistic profession combines opinion reporting style with weak professional autonomy due to a dependence on political and economic systems (Humanes et al., 2013;Humanes & Roses, 2018;López-Rabadán & Casero-Ripollés, 2015).
Based on this background we propose the following hypothesis: H2: Party identification has more important effects than ideology as a predictor of selective exposure for all media types (print press, radio, and television).

Data Sources and Temporal Framework
This study is based on a classic source of data for analyzing selective exposure: post-electoral surveys.While post-electoral surveys are susceptible to post-hoc rationalizations on the part of survey respondents, their benefits are worthy of highlighting.These include evaluating the phenomenon of selective exposure in an environment that is not simulated or manipulated by researchers and using representative samples that allow the results to be generalized to the entire population.
Specifically, we used post-electoral studies conducted by the CIS after the last six general elections in Spain, which were held between 2008-2019 (Table 1).
This 11-year period covers six elections and includes the process by which Spain's two-party system became a multi-party context.While the 2008 and 2011 elections took place in the context of a consolidated two-party system-with the Partido Popular (PP) and Partido Socialista Obrero Español (PSOE) as the only forces with any real chance of forming a government, the other four elections (December 2015, June 2016, April 2019, and November 2019) were characterized by the emergence of new political groups (Unidas Podemos, hereafter UP, and Ciudadanos after 2015; and VOX after 2019) and by a much more politically fragmented context.This is therefore an appropriate period for observing whether selective exposure to different types of media is a stable behavior that has taken root among Spanish audiences beyond the political context.

Research Technique and Variables
We proposed conducting a multivariate analysis based on multinomial logistic regression given the categorical and multiple nature of our dependent variable: the preferred media outlet for consuming political information.
We used the three media consumption questions that post-electoral surveys contain as the dependent variable.They are: Which newspaper/radio station/television channel do you prefer for following political and electoral information during the campaign?Unfortunately, the surveys did not account for the preferred digital outlets, so we could not include this element in the analyses.These are open questions, and Table 1.Data sources.

General elections
Number of CIS study Sample the respondent is allowed to respond spontaneously, providing their favorite source in each of the three formats.To limit the multinomial regression analysiswhich requires a limited number of categories for the dependent variable and high sample sizes (Cea d'Ancona, 2005)-we have limited ourselves to the four most populated response categories for each type of media outlet (Table 2).
In the case of newspapers, El País, El Mundo, La Vanguardia, and ABC are the four national publications with the highest rates of penetration among gen-eral information newspapers.They dominate the written press market in Spain (Asociación para la Investigación de Medios de Comunicación, 2020).For their part, the four radio stations mentioned most frequently by survey respondents during the 11-year period were Cadena SER, COPE, Onda Cero, and RNE.Together, they represented 75.3% of the market share in 2019 (Asociación para la Investigación de Medios de Comunicación, 2020).The four TV stations selected for the analyses were TVE1, Antena3, Telecinco, and La Sexta.The first three were always among the four most frequently mentioned by given that it occupies the same political space and shows a significant association with the more central ideological positions on the scale (5-6).This is reflected in the calculation of the standardized residuals in 2008 and 2011.Reference categories for each variable are indicated by in bold.
survey respondents during the 11-year period.This is not the case for La Sexta, which had a smaller showing in 2008 (placing fifth behind Cuatro) because it was founded in 2005 and took a few years to outperform Cuatro in terms of audience size.However, we decided to include La Sexta in the analyses to be able to compare the results over time.The four channels chosen had 42.9% of the market share in 2019 (Barlovento Comunicación, 2019).We included ideology in the left-right cleavage and party identification as independent variables.The positions on the left-right axis are operationalized using a 10-point scale that is recorded as three variables: left (1-4), center (5-6), and right (7-10).Party identification is operationalized as vote recall in the last general elections.The regression models only include national parties with the highest parliamentary representation during the period analyzed (2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019).These are the PP and PSOE based on all elections, Izquierda Unida and Unión Progreso y Democracia in 2008 and 2011, and Ciudadanos and UP in 2015 and thereafter.VOX was added for the two elections held in 2019.
Furthermore, we added time-operationalized as continuous-as an independent variable in order to contrast H2, offering a single regression model for each media type.This allows us to observe which variable has the most important effects on media preferencesideology vs. party identification-comparing the size of the coefficients.
Finally, we introduced control variables that are related to socio-demographic attributes, which are also associated with media consumption (Cardenal et al., 2019).These include age, education level, gender, social class, and interest in politics, which also have a notable influence on news consumption decisions (Skovsgaard et al., 2016;Valera-Ordaz & Humanes, 2022).
Multinomial logistic regression requires setting up a dependent variable category outside of the model that acts as a "reference group" regarding which one can compare the effects of the independent variables that the model specifies (Cea d'Ancona, 2005).Given that our purpose is to analyze the scope of selective exposure guided by ideology and vote, we chose to select as a reference a media outlet for each format that is relatively uninfluenced by our variables of interest or, to put it differently, influenced by the center of the ideological spectrum: Telecinco in the case of television, RNE for radio, and El Mundo for the print press.The latter was the most difficult to choose given that the print press is the media form that is most significantly segmented by ideology and vote (Humanes, 2014;Valera-Ordaz & Humanes, 2022).The same logic applies to the rest of the independent and control variables, and the most central categories are selected when possible as reference groups (in bold in Table 2).
In total, 18 multinomial logistic regression models are specified (three for each election, and one for each type of media outlet).These include anti-logarithmic coefficients that should be interpreted as follows: when the value is greater than 1, it indicates a positive association (the interest variable increases the likelihood of the event) and when its value is less than 1, we are looking at a negative association (the likelihood of the event decreases).The results should always be interpreted in relation to the reference category.Three more models are added to these (one per media type) that include time as an independent variable.

Results
In the case of the print press, the results show that both ideology and vote are significant predictors of exposure to newspapers (Table 3).For example, those who have a right-wing ideology were more likely to prefer ABC (compared to El Mundo) in 2011 and 2019.Furthermore, those who voted PP were three to five times as likely to read ABC (compared to El Mundo) in 2015 and April 2019, respectively, and that same year voting VOX tripled the likelihood that one would choose ABC over El Mundo.
The differences are also notable regarding other newspapers.Those who have more left-leaning positions are much more likely to prefer El País (compared to El Mundo) in 2011 and April 2019, and voting PSOE is a significant predictor of exposure to El País throughout the entire time analyzed.In fact, voting PSOE increased the likelihood of preferring El País (to El Mundo) three to 36 times depending on the year.In 2015 and 2016, voting UP also increased the likelihood of reading El País instead of El Mundo significantly, though that was not the case in 2019.Furthermore, voting PP significantly reduced the likelihood of reading La Vanguardia vs. El Mundo in 2011 (the 2011 regression model had to be completed with Vote PSOE as the reference category for statistical reasons) and 2016, while voting PSOE increased it in 2008, 2016 and April 2019, and voting UP did so in 2016 and April 2019.
In summary, exposure to Spanish newspapers is clearly influenced by ideology and vote, which emerge as significant predictors of newspaper consumption, controlling for age, gender, education, social class, and interest in politics.
Furthermore, the specified models explain between 21%-42% of the variance.
Table 4 shows that ideology and voting also have a significant influence on television preferences.Respondents who identify as leftists were less likely to watch Antena3 (vs.Telecinco) in 2008 and 2015.The same is true of those who voted Izquierda Unida in 2008, UP in November 2019, and PSOE voters in practically all the temporal points analyzed (except for 2011 and 2016).All of them are less likely to choose Antena3 (vs.Telecinco) to get information about the campaign.
For their part, those who identify as leftists were more likely to consume La Sexta (vs.Telecinco) in 2019.This was also true for UP voters beginning in 2015.Both variables increase the likelihood of consuming this television station.Furthermore, PP voters (in December 2015  and June 2016) and VOX voters (in November 2019) were less likely to choose this station.As such, we are looking at both selective media exposure based on political reasons and the phenomenon of selective avoidance through which some ideological groups avoid consuming media that they perceive to be ideologically distant (Garrett et al., 2013).
In the case of TVE1, however, there are less significant differences compared to Telecinco.Voting PP increased the likelihood of preferring TVE1 over Telecinco only in 2015.Voting UP decreased that likelihood.However, there are no differences based on the positions on the left-right axis.This is because the ideological composition of Telecinco and TVE1 audiences are similar.
But the political variables are not the only ones that explain the population's television preferences.Sociodemographic attributes do so as well.Older adults prefer to watch TVE1 and La Sexta over Telecinco.Women prefer Telecinco a great deal and are less likely to consume La Sexta, Antena3, and TVE1.Those with higher levels of education tend to consume La Sexta, Antena3, and TVE1 more intensely (compared to Telecinco) and those who state that they are interested in politics are much more inclined to watch La Sexta.
These results are partly due to the type of content that television stations offer.While Telecinco focuses on entertainment, La Sexta offers many more formats related to politics as both information and politainment (Berrocal-Gonzalo et al., 2014).In this sense, programs like "El Intermedio" could generate selective exposure with their satirical news format, as studies conducted in other contexts have found (Stroud & Muddiman, 2013).Furthermore, it is necessary to note that the goodnessof-fit of the regression model is lower in the case of television than in the print press.The model explains only between 8.6-15.5% of the variance.
The results for radio suggest that ideology and vote are also significant predictors of exposure, especially in the cases of COPE and SER (Table 5).Having a right-wing ideological orientation doubled (in November 2011 and December 2015) or even tripled (in April 2019) the likelihood of listening to COPE (vs.RNE) for political information.Furthermore, PP voters were between three and six times more likely to consume COPE (vs.RNE) in 2011 and 2019, respectively, while being a PSOE or UP voter significantly decreased the likelihood of exposing oneself to that station in April 2019.The importance of the vote is very limited in the case of Onda Cero, and ideology does not show a significant effect.In the case of SER, the importance of left-leaning positions, voting UP, and especially voting PSOE is observed more clearly given that all of these increase the likelihood of preferring that station over RNE during different years.
Finally, the goodness-of-fit of the radio regression model is slightly higher than that of television but clearly lower than that of the print press, as it explains between 14.6-26.3% of the variance.
To compare the influence of ideology vs. party identification on citizens' media preferences, we have specified three regression models (one for each type of media) including time as an independent variable (Table 6).Given that the temporal framework covers an 11-year period , it has only been possible to include the PSOE and PP vote in these models, as these are the only political forces present throughout the period.We left the PSOE vote as a category of reference and included the PP vote as an independent variable.Table 6 presents the regression coefficients of each type of medium: When the regression coefficient is greater than zero, there is a positive association, and when it is less than zero, the association is negative.By using the coefficients in place of beta exponentials, their magnitude can be compared independently from the sign.
The results clearly suggest that the effect of vote on the choice of newspaper, television channel, and radio station is always higher than that of ideology, as the size of the coefficients is greater for all the media analyzed.In other words, voting is a stronger predictor than ideology for Spaniards' media preferences for the three types of media.
Furthermore, it is important to note that in the case of television, voting has more important effects on the selection of Antena3 and La Sexta (vs.Telecinco)-and not for TVE1-than the socio-demographic attributes taken individually.In other words, despite the importance of education level, age, gender, and interest in politics as factors that influence television channel choice, voting is undoubtedly the factor that best predicts consumption of Antena3 and La Sexta, the two major stations that are part of the Atresmedia group.This suggests, as we will see below, a growing segmentation of the television market in terms of politics.

Discussion and Conclusions
This research shows that selective exposure to the media based on ideology and party identification is a stable phenomenon in Spain, which confirms H1.Both variables significantly influence audiences' behavior and do so for all media types.The print press is not the only politically segmented market: radio and television are as well, as recent studies suggest (Valera-Ordaz & Humanes, 2022;Valera-Ordaz et al., 2021).In fact, the model explains approximately twice the variance of television preferences beginning in 2015, probably because of La Sexta's position as a generalist channel focused on the left-wing audience.In this sense, the results suggest that the television industry is moving-since the formation of the television duopoly with the acquisition of Cuatro by Mediaset in 2009 and the merger of La Sexta and Antena3 in 2011-from the classic "catch-all" model in search of large audiences to a more ideologically segmented model thanks to the role of Atresmedia group, whose Antena3 and La Sexta channels are focused on differentiated ideological niches.For its part, the Mediaset  ).
An important conclusion of this study is thus that selective exposure is a structural inclination of Spanish audiences, that is, a phenomenon that remains unchanged over time regardless of the evolution of the political system and changes in the media system.In this sense, our study contributes solidity to the previous cross-sector studies that had identified politically oriented consumption of the media in Spain using bivariate techniques (Fraile & Meilán, 2012;Humanes, 2014;Martín-Llaguno & Berganza-Conde, 2001), and it does so in two ways.First, it confirms that the effect of ideology and party identification on media selection is produced by controlling other important variables associated with media consumption (socio-demographic attributes and interest in politics) through a multi-variable approach.Second, it shows that selective exposure remains a structural characteristic of Spanish audiences over an 11-year period despite two circumstances: (a) the change in the Spanish party system, which is now a multi-party system with a larger variety of political forces; and (b) the notable expansion of the media offer derived from digitalization-and including the creation of two new generalist television chains and the transformation of the television market into a duopoly-and the consolidation of much more fragmented media environments, giving the user a great deal of choice (Prior, 2007;Van Aelst et al., 2017).Despite these profound changes in the political and media landscape, Spanish audiences continue to expose themselves to media based on their ideological preferences and party affinities.
Furthermore, the results suggest that the audiences' behavior should be understood based on partisanship (and not ideology) because party identification is the most important heuristic that guides media consumption decisions for all media types.This aligns with studies of the North American context (Camaj, 2019;Iyengar & Hahn, 2009;Messing & Westwood, 2012).In other words, party identification has more important effects than ideology on individuals' media preferences-which confirms H2-such that the best way to predict which newspaper, radio station, or TV station citizens consume is to learn about their electoral behavior.This result is likely the consequence of the fact that party identification-and voting, its main manifestation-has a notable level of personal relevance because it allows individuals to cognitively orient themselves and make both political (Lau & Redlawsk, 2001) and media decisions.In this sense, our research reveals that the importance of party identification is not exclusive to the North American context and is very much a factor in Spain.In fact, party identification seems to arise as a central heuristic in the context of media hybridization and fragmentation that force the user to choose between many more options.
The results of this research also suggest that some media outlets produce more polarization because they simultaneously inspire exposure and selective avoidance by certain ideological groups.This is the case with La Sexta in 2015 and 2016.UP voters expose themselves to it more frequently, while PP voters selectively avoid it.In 2016, La Vanguardia attracted UP and PSOE voters much more and was selectively avoided by PP voters.El País was preferred by PSOE and UP voters in 2008 and 2016 and selectively avoided by PP voters.The same occurred with TVE1 in 2015, as PP voters exposed themselves to it more frequently and UP voters tended to avoid it.While both selective exposure and avoidance occur simultaneously only at specific times, the existence of these media is susceptible to compromising the exposure to diversity of citizenry in the public sphere, which is vital for promoting social empathy and political tolerance (Huckfeldt et al., 2004).Furthermore, its absence carries risks of polarization.In this sense, the results remind us that political polarization has an important connection to the evolution of media ecosystems and that media and political polarization processes should be conceived of as mutually influencing phenomena (Dahlgren et al., 2019).On the one hand, politically biased journalism has been found to lead to attitude polarization (Levendusky, 2013b;Stroud, 2010).On the other, polarized political debate seems to intensify selective exposure to news media.A recent study on the Catalan independence conflict found that drivers of selective exposure are sensitive to political context.Specifically, the author determined that support for independence became increasingly important as a driver of media consumption as political polarization related to the issue of independence grew (Valera-Ordaz, 2022).
However, future research should delve further into the role of the vote as a driver of media consumption over time in order to observe whether selective exposure and avoidance increase over time and whether political and media polarization processes are similar to those described in the US (Rodriguez et al., 2017;Stroud, 2010) or those that we have begun to sense in Spain (Ramírez-Dueñas & Vinuesa-Tejero, 2021;Valera-Ordaz, 2022) emerge.This is vital in a context in which the fragmentation of media offerings and increase in competition introduce incentives for communication groups to ideologically segment their content to capture and generate loyalty within specific niche audiences (Bennett & Iyengar, 2008).
Finally, it is worth underlining the main strengths of this research and some of its most important limitations.First, this is the first study to analyze selective exposure in Spain from a longitudinal perspective, that is, covering an 11-year time frame characterized by major changes in the political system and the information environment.In this sense, this work adds significant solidity to previous studies on selective exposure in Spain that addressed specific moments (Humanes, 2014;Martín-Llaguno & Berganza-Conde, 2001;Valera-Ordaz & Humanes, 2022), showing that political selective exposure is a well-established and persistent inclination of Spanish audiences.Second, our findings clearly show that party affinity is a stronger driver of news exposure than ideology in Spain despite the major shift that occurred in the Spanish political system.Spain's two largest political parties broke apart in 2015, giving rise to new political organizations (UP, Ciudadanos, and VOX in 2019).In spite of this, party identification was the most important factor of selective exposure to the news media during the period under study.Third, we used postelectoral surveys based on large representative samples to generalize the results to the entire Spanish population.We also used data on habitual news exposure, which is easier to generalize than single exposure measures (Stroud, 2008).
It is also important to highlight a important limitations of the study.First, our work is based on exposure to traditional news media (television, radio, and press).We did not address the consumption of digital news products or social media platforms.As such, we were unable to assess selective exposure online, which has been found to be less powerful than expected despite the fears surrounding digital technologies (Cardenal et al., 2019).Incidental exposure and the weak ties that characterize online networks cause these spaces to serve as diversification forces, helping people to be confronted with novel information that they would not have encountered offline (Barnidge & Peacock, 2019) and providing access to more diverse news sources (Fletcher & Nielsen, 2018).This limitation highlights the need to conduct studies that account for the complexity of current information environments in the context of hybridization and fragmentation and the need to explore selective exposure beyond the limitations introduced by working with secondary data.
Second, the Spanish political and media context changed a great deal during the period under study (2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019), which led to some methodological limitations.For example, we were unable to explore the role of partisanship as a driver of news exposure beyond the PP and PSOE because the other political parties existed for only part of this period.We interpret our results as a sign of the strength of the role of partisanship as a driver of media exposure because party affinity was a significantly more important driver of selective exposure than ideology, which was entirely comparable throughout the same period.Still, future methodological designs will have to delve further into the role of political attitudes in media consumption during larger alternative time frames.
1. Newspaper (El País, El Mundo, La Vanguardia, and ABC) 2. Radio station (SER, COPE, Onda Cero, and RNE) 3. Television channel (TVE1, La Sexta, Antena3, and Telecinco) Notes: The reference category for vote recall is Ciudadanos beginning in 2015, but it did not have a presence in previous elections, so it cannot be used in the regression models for 2008 and 2011.The solution is to use Unión Progreso y Democracia for 2008 and 2011

Table 3 .
Multinomial logistic regression models for print press preferences (reference category El Mundo).

Table 4 .
Multinomial logistic regression models for television preferences (reference category Telecinco).

Table 5 .
Multinomial logistic regression models for radio preferences (reference category RNE).