Media logic in the coverage of election promises: comparative evidence from the Netherlands and the US

This article analyzes the occurrence of media logic in the coverage of election promises in the Netherlands and the US. Whereas studies of media logic commonly focus separate attention on one of its various manifestations, we believe a comprehensive understanding requires a more inclusive approach. In response, we include five aspects of media logic in our study of news coverage: the occurrence of (a) the strategy frame, (b) the game frame, (c) the conflict frame, (d) personalization, and (e) negativity. Our study contributes innovatively to the existing literature by taking an approach that, rather than starting from campaign manifestos, analyzes election promises as they are reported on in newspapers. We take this approach because the media are the primary source of information about election promises for citizens. The results of our study indicate that media logic is ubiquitous in the coverage of election promises, but that media logic does not always behave across different media and political systems in the way the literature predicts. Notably, the results show that, in contrast to our expectations, coverage of election promises is more negative in the Netherlands than in the US.


Introduction
The practice of presenting policy views to voters via party manifestos before elections is well established in Western democracies. The rationale behind this is that voters can decide to which political actors they want to give their mandate to represent them. This is also known as the party mandate model (Louwerse 2014), on which there is an elaborate strand of research that analyses to what extent political actors fulfill their pledges. Studies focusing on pledge fulfillment show that politicians are in general inclined to fulfill their pledges. These include the seminal metaanalysis by Thomson et al. (2017), who utilize the data sets of 11 scholars and synchronize research from recent decades that covers 57 election campaigns in twelve countries. Pétry and Collette's research (2009, p. 77) is also relevant since their meta-review led them to conclude that, "contrary to popular belief, political parties are reliable promise keepers".
Despite these findings, opinion polls and surveys show that most voters are skeptical regarding politicians keeping their election promises (ISSP Research Group 2008). Around 50% of the Dutch population and 83% of the US population do not believe that politicians keep these promises (Louwerse 2014;Rasmussen Reports, 2014). In light of the party mandate model, this can be problematic since the legitimacy of the system draws on the ability of voters to choose their policy preferences.
This difference between how voters and scholars assess the credibility of what politicians promise may stem from the fact that the two groups form their opinion on the matter in different ways. While scholars often inform themselves by taking pledges made in election manifestos as their starting point, voters tend to use the media as their main source of information (Strömberg 2001). The public therefore receives information about election promises in a different way from scholars. In addition, this information has been selected and edited by the media (Cushion et al. 2016) since, according to the principles of media logic, the media are inclined to question politicians' intentions and highlight their failures (de Vreese and Elenbaas 2008; Lengauer et al. 2012).
The literature about media logic shows that the resulting style of reporting may have various behavioral effects, including cynicism and consequences for voter turnout and participation (de Vreese and Boomgaarden 2006;Iyengar et al. 2004). Therefore, if media logic plays a part in the coverage of election promises, it could result in negative reporting about pledge fulfillment, which could in turn fuel citizens' skepticism regarding pledge fulfillment and establish legitimacy issues within the democratic system. As a key step in this line of research, it is important to assess the extent to which media logic is present in coverage about election promises (van Aelst and de Swert 2009;Vliegenthart et al. 2011). However, such studies remain rare.
In addition, this study contributes to the field of media logic studies in that, as suggested by Takens et al. (2013) and Magin (2015), it looks for a more comprehensive understanding of the broader phenomenon by including multiple manifestations of media logic at the same time and by taking into account multiple context factors simultaneously. While we acknowledge that 'media logic' is a particularly complex concept and that no broadly accepted definition exists, as discussed below, we do observe a coherence in the literature that allows us to analyze various manifestations at the same time, which, in turn, allows us to gain a more comprehensive understanding of how media logic behaves across different contexts.
The current article explores the role of media logic in the coverage of the promises made by national politicians during election campaigns in the Netherlands and the US. In the style of Strömbäck and Dimitrova (2006), we conducted a comparative study of media coverage in two countries with different political and media systems. Additionally, in contrast to many other studies, our research usefully differentiates between campaign periods and governing periods, which is also seen as an important context factor since existing research shows that the occurrence of media logic may differ between these two stages of the election cycle (van Aelst and de Swert 2009). The main research question is: 'To what extent do the content characteristics of media logic occur in the coverage of election promises in the Netherlands and the US in different stages of the election cycle?'

Media logic as part of mediatization
In the analysis of how the relationship between media and politics has become transformed in recent decades, 'mediatization' is an often-used concept that describes the increase in the influence that the media have on democratic processes as well as on other societal domains (Driessens et al. 2018;Hjarvard 2013). It is broadly accepted that mediatization involves the process in which the media as an independent institution exert influence on other social institutions and actors (Peleg and Bogoch 2014). In the context of politics, mediatization refers to the "changes in the decision criteria and action rationales of political institutions" (Esser 2013, p. 161).
As mediatization is too complex a concept to be studied in its entirety (Magin 2015), many empirical studies on this issue have focused on media content (see Magin 2015, Peleg andBogoch 2014). This study follows this tradition.
In their seminal work on the influences on and characteristics of political communication, Blumler and Kavanagh (1999) describe the development that Western media outlets made towards becoming independent institutions by loosening their ties with political organizations. This development also changed political coverage, because the media could no longer rely on large audiences, and had to compete with each other for attention. Consequently, commercial principles became increasingly important for media outlets (Brants and van Praag 2006). The way in which these came to direct the editing and selection process of coverage is commonly referred to as 'media logic' (Mazzoleni 2008).
The next section gives our conceptualization of media logic, and, accordingly describes five different manifestations of media logic in coverage, relating them to election promises and explaining their possible impact.

Conceptualization of media logic
Originally, Altheide and Snow (1979) described media logic as the way in which news is organized, presented, and emphasized. When analyzing more recent definitions, we observe that Hjarvard (2008, p. 113) emphasizes the format requirements of the media. Strömbäck (2008, p. 233), in turn, broadens the definition by adding the aspect that news values and storytelling techniques should not only fit the format of the medium but should also be directed to "capture people's attention". In other words, the media focus on "what they assume the public […] is interested in and/or enjoys" (Brants and van Praag 2017, p. 5).
We derive three core elements from the literature on media logic. Media logic is coverage that is 'shaped' or 'framed' in a way that (a) fits the format, (b) is consistent with news values, and (c) uses media storytelling techniques (Strömbäck 2008). Illustrations of format requirements include the use of soundbites (Brants and van Praag 2017). 'News values' are values through which journalists think they will make stories more attractive to the public, such as covering news from a negative angle. According to Strömbäck (2008, p. 233), storytelling techniques include "simplification, polarization, intensification, (…), visualization and stereotyping". He also mentions personalization and the framing of politics as a strategic game or "horse race". Hence, it is possible to distinguish between format elements of media logic and content elements of media logic. This study addresses only the content elements of media logic, which are elements that focus on the shaping of content. The reason why format characteristics are excluded from this study is that they overemphasize the technological role of the media and underexpose the communicative practices (Brants and van Praag 2017).
In the literature, there is criticism of the use of the media logic concept. The main criticism is that technological and social developments imply that no single media logic exists and that we should take a more differentiated approach (Brants and van Praag 2017;Lundby 2009;Thimm et al. 2018). Although these are fair comments, the study of the content of news coverage is still useful because, as Esser (2013) argues, traditional media still shape their content according to the above-mentioned principles and remain an important source of information for citizens about political news in general and election promises in particular.
Even though media logic has been regarded as having multiple manifestations, research has focused on only one or a small number of aspects of it (see Landerer 2014;Magin 2015;Opperhuizen et al. 2019;Takens et al. 2013). And, despite theoretical expectations on media logic, empirical work does not indicate a clear expected pattern in which different aspects of media logic manifest themselves (Takens et al. 2013;Vliegenthart et al. 2011). Therefore, Takens et al. (2013) and Magin (2015) advocate studying multiple aspects of media logic in different contextual situations at the same time because "it is still unclear to what degree the assumptions of increasing mediatization apply to Western democracies under different structural conditions" (Magin 2015, p. 416). We, therefore, adopt an inclusive approach that looks at various content elements of media logic simultaneously, which allows us to gain a comprehensive understanding of how media logic behaves in different contexts.
We take as a starting point the study by Takens et al. (2013), who argue that and explore whether contest coverage, personalization and, negativity are content elements that form part of a single media logic. In their study, however, 'contest coverage' is an amalgamation of multiple elements that we believe to be analytically distinct (see also Aalberg et al. 2011;Schuck et al. 2013). In response, we subdivide this element into three separate factors and treat them as individual aspects, that is the strategy frame, game frame and conflict frame. Hence, the resulting five content elements of media logic that we include in our study are the use of (a) the strategy frame, (b) the game frame, (c) the conflict frame, (d) personalization, and (e) negativity. We will elaborate below on our conceptualization of these content elements and on what they could mean in the context of coverage about election promises, the operationalization of which we describe in the methodology section.

The strategy frame
The first aspect of media logic is the 'strategy frame'. This category includes frames that characterize politics as a strategic game. Within such frames, the focus is on politicians' political strategies in a campaign as well as the tactics that political actors use to win elections or gain political advantage. In reports about election promises, the coverage implies that, rather than pursuing solutions to solve important social problems, political parties and politicians make pledges in order to gain votes. Personality, style, motivation, and instrumental action are all at the core of a strategy frame (de Vreese and Elenbaas 2008).

The game frame
The 'game frame' places political issues within a competitive frame. The game frame is often referred to as a 'contest frame' or 'horse-race frame' (Schuck et al. 2013). Since the focus is on winners and losers, this type of coverage frequently uses sports or war metaphors, whereby journalists make use of election results and/or opinion polls (Aalberg et al. 2011). In coverage about election promises, this would mean that a particular election promise is connected with the performance of political actors in elections or polls.

The conflict frame
When a 'conflict frame' is applied, a news report stresses the existence of two or more opposing sides. The attraction of the conflict frame is that discussing both sides of a conflict allows journalists to uphold an appearance of practicing balanced journalism (Schuck et al. 2013). In coverage about election promises, this could manifest itself by criticism of the promise when it is actually made. But it could also show itself by disagreement about the fulfillment of the promise.

Personalization
The fourth aspect of media logic, that is personalization, refers to a shift in media attention towards individual politicians at the expense of political parties and/ or other collectives. Rahat and Sheafer (2007, p. 67) define personalization as "a heightened focus on individual politicians and a diminished focus on parties, organizations and institutions". In the coverage of election promises, personalization is, therefore, present when election promises are attributed to an individual politician rather than to a collective.

Negativity in tone
The fifth aspect of media logic is an emphasis on the negative. Drawing on the work of Lengauer et al. (2012), in this article, negativity is measured by analyzing the tone of coverage. Leading questions in deciding on the tone of election promise coverage are: has an election promise been fulfilled and can we speak of success, or has the promise been broken, and is the message one of failure? And, does the coverage suggest that the election promise is to be realized in the future or is it skeptical about its fulfillment?
Studies have shown that media logic can have an impact on voter behavior. The content elements which are subject to this study are associated with depressing knowledge on substance. Thus, where they are found in coverage about election promises, this could mean that they cause a misrepresentation of substantial information about election promises among citizens. Some of the elements, such as the strategy frame, are associated with inducing cynicism (de Vreese 2005; Iyengar et al. 2004). Other studies have shown that the focus on opinion polls (game frame) may make politics more attractive, and therefore increase participation and turnout (Iyengar et al. 2004).

Two cases: the Netherlands and the US
Since there is no pre-determined absolute threshold at which politics can be considered as being mediatized, it is preferable to compare relative levels of mediatization (Landerer 2014). Some studies of media logic have achieved this by focusing on similar countries over time (Magin 2015;Zeh and Hopmann 2013), while others have based their investigations on comparing multiple countries (Arbaoui et al. 2016;Esser and Umbricht 2014). The contribution of our study lies in examining multiple aspects of media logic under different structural conditions and throughout the election cycle. Therefore, in the style of Strömbäck and Dimitrova (2006), who analyzed mediatization in the US and Sweden, for the current study we selected the US and the Netherlands as two countries with different political and media systems (Vliegenthart et al. 2011).
According to Esser and Umbricht (2014), the differences in the historical-institutional 'context' of the press systems is an explanatory condition that could explain differences in political news coverage. This contextual condition leads us to expect significant differences between the US presidential system and the Dutch parliamentary system since certain news practices developed earlier in the US than in European countries such as the Netherlands, which reshaped and adjusted some of these practices according to their customs and traditions (Esser and Umbricht 2014).
The aim of our strategy is to find commonalities in the occurrence of media logic regardless of the systematic differences, which could be indicative of broader trends above and beyond these two countries (Vliegenthart et al. 2011); it also aims to find differences that can be useful for generating ideas and new hypotheses regarding contextual influences such as political and media systems and election cycles on the occurrence of media logic.
In addition, since it is argued in the literature that differences can be expected in the occurrence of media logic in coverage within an election cycle (van Aelst and de Swert 2009), we also differentiate between campaign periods and governing periods. This increases our contribution to the literature since we add a contextual element to examine how multiple aspects of media logic behave in coverage. We will elaborate below on both aspects of our study, and introduce our hypotheses.

Impact of the systems
We selected the Netherlands and the US for this study because these two countries have distinct political and media systems (Vliegenthart et al. 2011). We expect that many aspects of media logic will occur more often in the US than in the Netherlands because the US has a stronger tradition of commercialization than the Netherlands, which leads to this expectation since commercialization provides the incentive for increasing market value and attracting more news consumers (Kriesi 2011).
The political-institutional differences between the Netherlands and the US strengthen our expectation that media logic will occur more often in coverage in the US than in the Netherlands. In the Netherlands, where governments are typically formed by a coalition of multiple parties, it is seen to be more difficult for political actors to steer coverage towards partisan purposes by using government means and channels than it is in the US, where considerable executive power rests with a single actor, namely the President of the United States (Brown 2011). Consequently, we expect the strategy frame to occur more often in the US than in the Netherlands (Aalberg et al. 2011). In addition, the fact that the presidential system and the electoral system in the US create clear 'winners' and 'losers' makes it easier for journalists to apply a game frame than in the Netherlands, where it is harder to point to winners and losers after an election (Lijphart 1999). Furthermore, the US presidential system appreciates personal leadership more (Hendriks 2010), which leads to the expectation that personalization occurs more often in the US than in the Netherlands. Finally, political actors in the US more readily express a negative attitude towards opponents, which facilitates the use of conflict frames by the media, but also leads to the expectation that negativity is more common in coverage in the US (Walter 2012). These expectations lead to Hypothesis 1.

H1
The characteristics of media logic are more common in election promise coverage in the US than in the Netherlands.

Impact of election cycles
One of the leading theories in the literature about the plausible effect of election cycles on the occurrence of media logic suggests the phenomenon is more prevalent during a campaign period than during a governing period (Schuck et al. 2013). This expectation originates from the assumptions that political actors are more active during a campaign period, that the media take more care to be neutral and objective, and that the public are more critical of possible media bias (van Aelst and de Swert 2009). According to this theory, journalists use media frames when political parties make an effort to attract media coverage (Aalberg et al. 2011). This effect is strengthened by the fact that the public pay more attention to the coverage of political news during a campaign period than during a governing period.
The alternative theory is that, today, there is no difference between campaign periods and governing periods as regards the occurrence of media logic in coverage. This postulate originates from the observation that political parties constantly strive to win voters in a 'permanent campaign', which makes it likely that the media continually report on election promises (van Aelst and de Swert 2009).
In this article, we consider the latter theory less likely to be corroborated, for three reasons. First, particularly in the Dutch consensus democracy, the principles of coalition building imply that during governing periods the political parties that are in power have to defend the interests of the coalition as a whole rather than partisan interests (Brown 2011). Second, political actors face financial limitations. This is true in the Netherlands in particular since a vital difference between the US and the Netherlands is that political actors in the US have significantly more resources at their disposal, in part since US campaign laws are less strict regarding the permissibility of private funding of political campaigns (BPC 2018; Walter 2012). Such differences make it unlikely that Dutch political parties can afford a permanent campaign, which is likely to reduce media attention in some periods. And, even in the US, there is evidence against the permanent campaign theory (e.g., Alaimo 2017). Third, it is more rational for political actors to focus their campaign efforts on a campaign period since voters tend to be more receptive of political coverage during campaign periods (van Aelst and de Swert 2009), and many people postpone their voting decision until the very last moment (McAllister 2002). From these expectations, we derive Hypothesis 2.

H2
The characteristics of media logic are more common during a campaign period than in a governing period.

Methodology
For our study, we conducted a quantitative content analysis of newspaper articles. Using the LexisNexis database, we made an inventory of articles that mention the term 'election promise' or its direct equivalents in both Dutch and English. 1 The use of these explicit references increases the validity of our measurements since it ensures that readers are likely to associate coverage with election promises and pledge fulfillment. For each country, we analyzed coverage of the most recent and completed campaign period and the governing period that followed. Mirroring the approach of Benoit et al. (2013), the American campaign period we analyzed ran from 20 July until 4 November 2008; the governing period ended on 6 November 2012. And, in line with the approach used by Strömbäck and Dimitrova (2006), The New York Times and The Washington Post were selected as leading broadsheet newspapers, and USA Today as a leading popular newspaper.
Following Brants and Bos (2014), we analyzed the campaign period in the Netherlands that ran from 22 August to 12 September 2012; the governing period ran until 15 March 2017. To increase the availability of data to an acceptable level, we included almost all national Dutch newspapers, that is 12 publications. 2 Our initial search produced 1695 hits. After removing identical newspaper articles as well as articles that did not fit our research purpose, e.g., articles that discussed local elections, and election promises from other countries, 1245 items remained. As a unit of analysis, we took any text segment of any length that discussed a separate election promise. A single article could thus contain one or more items.
To test whether additions to our search string would generate relevant results, we ran an additional query with less explicit reference to election promises. 3 This search generated over 6000 hits. From a manual evaluation of the relevance of the first 1000 'most relevant' hits as defined by LexisNexis for both countries, we concluded that for the US only 90 articles included relevant items, which implies a precision of 0.09% (see Stryker et al. 2006). For the Netherlands, this number was 52, or 0.052%. From this result, we concluded that we were not missing out on large quantities of relevant text segments and therefore decided not to broaden our search along these lines since this would require great effort with very little result, whereas the original dataset already showed considerable validity.
To increase validity and reliability, all variables were coded in a binary system according to the presence or absence of media logic by means of a detailed codebook, which was based on existing measurement instruments from the literature (Potter 2009). A summarized codebook is provided in Table 1.
In the media logic literature, two main patterns can be detected as regards to the operationalization of frame research (Aalberg et al. 2011). First, there are studies that investigate a dominant-frame basis (see Patterson 1993), where scholars look at whether or not a particular frame is the dominant frame in media coverage. Second, there are studies that investigate media frames on a present/absent basis (see Karidi 2018). Since we analyzed the occurrence of multiple frames in coverage at the same time, we adopted a present/absent approach. A dominant-frame approach would make the use of some frames 'invisible' because it would code a frame absent if it is not the dominant frame, whereas we are not interested in giving frames a relative weight.  (2008) Game frame -References to opinion polls and/or election results as a competition -Reporting in terms of winners and losers -Presenting politics as a game, sport or war Personalization The promise is attributed to an individual politician, rather than to political parties and/or other collectives Present: the election promise is attributed to an individual official/politician (e.g., name, he, she) Absent: the election promise is attributed to a collective (institution/coalition/government/political party) Rahat and Sheafer (2007) Table 1 (continued) Key characteristics

Coding instruction
Origins Negativity -Framing the fulfillment of a promise as a failure etc.
-Framing a pessimistic prospect as regards the future fulfillment of a promise Present: Indications of negativity are when a promise is framed as being a failure, political failure, broken, unfulfilled, fiasco, disaster, crisis, frustration, cynicism, disappointment, scepsis Present: Impressions of negative developments, the promise is connected with a negative prospect of pessimism, negativity, hopelessness, critical developments and potential threats Absent: Indications of a positive tone are political success, solutions, improvements, progress, gain, welfare, achievement, enthusiasm, hope, sustainability and completion Absent: An optimistic prospect of an election promise is when positive realistic developments, prospects and scenarios are given, or when there are potential gains or promising expectations Lengauer et al. (2012) To measure intercoder reliability, the literature recommends that 10% of all items should be recoded (de Swert 2012). In our study, 16% of all items were recoded by a second and senior researcher. Here, we use Krippendorff's Alpha (Kalpha) to interpret the results in the statistical analysis platform SPSS. The Kalpha values, as shown in Table 2, were measured per variable. The mean of all variables is 0.79 Kalpha, indicating a high coding consistency. Only the conflict frame scores somewhat lower, but at 0.73 is still acceptable and above the 0.67 minimum (de Swert 2012).
We use a Chi square test to assess the statistical significance of our findings and Goodman and Kruskal's Tau as an association measure because of the asymmetric relationship between our dependent and independent variables.

Results
The results clearly show that media logic is present in the coverage of election promises. Below, we first discuss the results relating to H1 by presenting the occurrence of the three media frames, personalization, and negativity in both systems, in the US and the Netherlands. Subsequently, we use the same approach when discussing the results relating to H2, which addresses both stages of the election cycle.

H1: Comparing systems
In H1, we stated that we expected media logic to occur more often in the US than in the Netherlands. We argued that the differences in political and media systems would lead to this finding. Table 3 shows that for the game frame and conflict frame our findings are in accordance with H1, but the differences are not statistically significant. The strategy frame contradicts H1 more clearly, since it is the only frame that occurs more often in the Netherlands than in the US (GKTau = .000, p = .000). However, there seems to be no association, meaning that the media and political systems could not explain the occurrence of the strategy frame in coverage. In addition, we find that during a governing period the strategy frame is significantly more present in the US than in the Netherlands (GKTau = .042, p = .000). The association here is very weak. Comparing campaign periods, we see no significant difference. The fact that the strategy frame occurs more often in the Netherlands in general and during a governing period in particular contradicts H1.

H1: The use of media frames across systems
When analyzing the appearance of the game frame, we do not find any significant difference between the US and the Netherlands, (p = .050) or between the governing periods of both countries (p = .243). This is not in line with H1. However, we do find the presence of the game frame significantly more often during a campaign period in the US than in the Netherlands (GKTau = .220, p = .006); the association is weak. This is in line with H1.
The findings regarding the occurrence of the conflict frame are also in line with H1. We see that the conflict frame occurs significantly more often in coverage in the US than in the Netherlands (GKTau = .015, p = .000). The association is very weak. The conflict frame is also present significantly more during a campaign period in the US compared to the Netherlands (GKTau = .143, p = .001) and during a governing period in the US compared to the Netherlands (GKTau = .009, p = .001). The association is weak for both periods. Table 4 shows that personalization is significantly more present in coverage from the US than from the Netherlands. The association is moderately strong (GKTau = .350, p = .000). The results also indicate that personalization is significantly more present in coverage during a campaign period in the US than in the Netherlands (GKTau = .441, p = .000). This observation holds for a governing period too (GKTau = .207, p = .000). The association is moderately strong during a campaign period and weak during a governing period. These findings are in agreement with H1. Table 5 shows that negativity is significantly more present in coverage from the Netherlands than from the US (GKTau = .067, p = .000). The association is very weak. When looking at negativity during a campaign period, we do not find any significant differences between the US and the Netherlands (p = .373). However, during a governing period negativity is significantly more present in coverage from the Netherlands than from the US (GKTau = .084, p = .000). The association is very weak. This finding is not in line with H1, since it was expected that coverage in the US would be more negative than coverage in the Netherlands.

H1: Negativity across systems
To sum up, the hypothesis that media logic occurs more often in the US than in the Netherlands does not hold since the results are mixed. Some of the media frames and negativity even contradict the existing literature, which calls for a reconsideration of some of the theories on media logic. But, with the exception of the occurrence of personalization throughout the election cycle and the presence of the game Table 4 Comparing systems: personalization *Significance < 0.05 , *significance < 0.01 , **significance < 0.000 NL total (N = 477) 46.9% (360) 45.7% (21) 55.6% (20) 76.1%*** (328) 46.4% (340) frame during a campaign period, the associations are weak. This means that we should refrain from making too strong claims about the explanatory power of media and political systems as regards to the occurrence of media logic. Even so, our findings do indicate that the occurrence of media logic differs between the Netherlands and the US, particularly during campaign periods: governing periods appear to have more similarities.

H2: Comparing stages of the election cycle
By contrasting the two different periods, we investigated whether the expectation expressed in H2 is corroborated by our research, i.e., that the characteristics of media logic occur more often during a campaign period than during a governing period.

H2: Media frames across periods
As Table 6 shows, when analyzing the occurrence of media frames, significance is absent for the strategy frame in the Netherlands. This result is not in line with H2.
Although the game frame occurs significantly more often during a governing period in the Netherlands than in a campaign period, there is no association (GKTau = .000, p = .016). The conflict frame occurs significantly more often in a governing period in the Netherlands. Again, the association is absent (GKTau = .000, p = .000). These results contradict H2 and signify that coverage in the Netherlands during a campaign period is more 'one-sided' than during a governing period.
For the US we observe that the strategy frame occurs more often during a campaign period than during a governing period. However, the significance is weak and association is absent (GKTau = .000, p = .017). This result is in line with H2. For the game frame and the conflict frame, the differences between the periods in the US are not significant, which contradicts H2. Table 7 shows that personalization is more common in coverage during a governing period than in a campaign period (GKTau = .041, p = .000). This finding is significant. However, the association is very weak. In the Netherlands, personalization was found to occur significantly more often during a governing period than during a campaign period. The association here is absent (GKTau = .000, p = .000). This finding deviates from H2, but is in line with van Aelst and de Swert (2009), who state that during election periods politicians position themselves as spokespersons for their parties rather than drawing attention to themselves as individuals. Whereas we did not find a significant difference in the presence of personalization in coverage from the US, personalization was found to occur more than 90 percent during a campaign period and governing period, which agrees with the literature that suggests that the presidential system in the US favors a high degree of personalization (Hendriks 2010).  (924) 50.0%*** (23) 74.5% (321) 86.1% (31) 82.4% (603)

H2: Negativity across periods
Finally, Table 8 shows that we find no significant difference in negativity between a campaign period and a governing period (p = .189). However, the result is different for the Netherlands individually.
In the Netherlands, during a governing period, 76.1% of all election promise coverage can be characterized as being negative in tone. During a campaign period, this figure drops to 45.7. This finding is significant, but the strength of the association is very weak (GKTau = .041, p = .000).
For the US, we observe that during a governing period, 46.5% of all election promise coverage can be characterized as being negative in tone. During a campaign period, the percentage is 55.6. However, this difference is not significant (p = .285). This could be an indication of a permanent campaign, but we should take into account that the absence of significance could also be caused by the small N (36).
Overall, we have to discard H2, which holds that the characteristics of media logic are present more often during a campaign period than during a governing period. After all, most of the elements of media logic do not occur significantly more during a campaign period than during a governing period. Especially with regards to the Netherlands, the findings contrast with H2, which means that they occur more often during a governing period than during a campaign period. The fact that associations were very weak most of the time, though, prevents us from making strong claims about the explanatory value of the difference between periods. Nonetheless, our finding that the occurrence of media logic during campaign periods differs from the occurrence during governing periods is an indication that the theory of a permanent campaign does not hold in the context of the coverage of election promises.

Conclusion and discussion
This article set out to answer the question: 'To what extent do the content characteristics of media logic occur in the coverage of election promises in the Netherlands and the US in different stages of the election cycle?' We find that media logic does not always behave in the way the literature predicts. Our results show that contextual factors affect the occurrence of media logic. One of these factors is the election cycle since our results reveal a difference in both countries in how often media logic occurs during a campaign period and how often it occurs during a governing period. We should mention here that our associations are not very strong, which suggests that the explanatory power of the contextual elements regarding the occurrence of media logic is limited. Nevertheless, our findings justify a call for more fine-grained theories of media logic that take into account various contextual factors such as differences in media and political systems, and the different stages of the election cycle.
Taking into consideration the field relevant literature, which suggests that media logic occurs more often in a campaign period than in a governing period (Schuck et al. 2013), it is somewhat of a surprise to find that in the Netherlands multiple aspects of media logic, in particular negativity, are more dominant in the coverage about election promises during a governing period than during a campaign period.

Table 8
Comparing periods: negativity *Significance < 0.05, **significance < 0.01, ***significance < 0.000 Campaign total (N = 82) Governing total (N = 1163) Campaign NL (N = 46) Governing NL (N = 431) Campaign US (N = 36) Governing US (N = 732) Negativity 50.0% (41) 57.4% (668) 45.7%*** (21) 76.1% (328) 55.6% (20) 46.4% (340) A plausible explanation for this finding could be found in the agenda-building literature (e.g. Brandenburg 2002). Here, it is suggested that effects of politics on media are stronger during election times, since journalists are willing to give politicians the space to express their views in those times. In routine times, this consideration might be less relevant. If then, during routine times, the media-politics interaction is more determined by media and thus follows a media logic, it makes sense that news is more negative. Another explanation could be that campaign periods in the Netherlands are more future oriented and relatively optimistic, possibly because strategic policy-making usually happens during governing period, when it is more likely that promises get broken. When campaign periods were more retrospective, they could be expected to be more negative. However, these explanations do not necessarily hold for the US since we did find that negativity, albeit without significance, occurred more often during campaign periods than during governing periods. Future research could provide more clarity by focusing on such questions. In addition, whereas the literature suggests that negativity is more common in countries such as the US (Walter 2012), our research shows that negativity is more dominant in coverage from the Netherlands. This is a relevant finding since Thomson et al. (2017) find that the US performs only slightly better than the Netherlands with respect to promise fulfillment. There is, thus, no great difference in pledge fulfillment, which could have provided a plausible explanation for the relatively high degree of negativity in the Netherlands compared with the US. A possible explanation for the difference in negativity could, therefore, be that the consensus building that is typical of the Dutch political system, especially when compared with the US, is more easily framed by the media as politicians 'breaking an oath'. During the coding, we observed that coverage from the US often had to be coded absent for negativity because a particular promise was declared a later priority. In contrast, in the Netherlands, politicians commit themselves to a coalition agreement early on, and are therefore less inclined to state that a promise has become a later priority, as coalition agreements make it clear which promises will be partly or completely abandoned so as to facilitate the necessary consensus building between coalition partners.
This negative media picture of pledge fulfillment could strengthen the opinion of voters that they do not have much to choose from, and this, in turn, could lead to political cynicism (Iyengar et al. 2004). Politicians may then be motivated to be tougher during coalition negotiations, which could, in turn, harm the political governability of the Netherlands (Hendriks 2010). Our study does not allow us to state that the presence of media logic in the coverage of election promises does indeed lead to such effects. For such a claim, further research is necessary on the direct link between coverage and behavioral effects (Kunkel 2009). At the same time, we do find a strong presence of multiple aspects of media logic in coverage of election promises. Hence, there is reason for concern (Iyengar et al. 2004).
Despite the fact that different countries and periods were studied, and various measures were taken to increase the reliability and validity of our research, it would be preferable to conduct comparative research across a broader range of contexts and across a wider selection of countries. This would also remove possible doubts that were expressed in this study about the lack of significance where this might have been caused by a small N. In addition, the search terms used for this research may have produced a bias towards media logic. The term 'election promise' and the linguistic variants that were used may have led to a specific type of coverage of election promises. Moreover, the use of such terms by journalists may in itself be an expression of media logic since it suggests that governing revolves around fulfilling one's earlier promises and also invites the use of the language of completion and failure. We nonetheless decided to take the current approach as a way to increase validity, which we believe outweighs the disadvantages. In addition, to increase comparability and to balance broadsheet and popular newspapers, in our selection of US newspapers we mirrored Strömbäck and Dimitrova's (2006) approach. A different selection may have produced different results.
Lastly, our research does give reason to continue studying election promises in an innovative way based not on campaign manifestos but on media coverage since the clear presence of media logic in coverage might have profound effects on the views of voters about promise fulfillment. It is important, therefore, to pay more attention to how people learn about election promises and pledge fulfillment via the media. This could also create possibilities for improving our crucial understanding of political cynicism.

Compliance with ethical standards
Conflict of interest On behalf of all authors, the corresponding author states that there is no conflict of interest.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creat iveco mmons .org/licen ses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.