Influential News: Impact of Print Media Reports on the Fulfillment of Election Promises

Abstract

Building on the mandate theory of democracy and literature on media coverage of elections, this article theorizes why information regarding party promises that is transmitted through the media could affect the former’s fulfillment. Utilizing a unique data set composed of 2,676 promises issued by 14 legislative parties over a 15-year period in post-communist Bulgaria, the study is among the first to longitudinally analyze the role of media in pledge fulfillment, while controlling for institutional and other explanations. The conclusions demonstrate that media reporting of election promises affects the fulfillment of pledges made by coalition parties, when more than one outlet has printed a promise, and under conditions of strong ideological divisions within the cabinet. Furthermore, the impact of media reporting is greater for pledges that do not otherwise have a high likelihood of being fulfilled.

KEYWORDS:

• election pledge fulfillment
• media
• post-communist parties
• mandate theory of democracy
• Bulgaria

Disclosure statement

No potential conflict of interest was reported by the author.

Supplemental material

Supplemental data for this article can be accessed on the publisher’s website at https://doi.org/10.1080/10584609.2018.1541032.

Notes

1. Most post-communist countries do not fit neatly into major classifications of democratic systems such as Lijphart’s Patterns of Democracy (Fortin, 2008).

2. Pledges made during the first two elections after the end of communism (1990 and 1991) are excluded from the analysis as those elections resulted in the same parties participating in two governments of different types before the next vote took place. This makes it more problematic to assign distinctions between parties in government and those in opposition. For example, after the 1991 elections, SDS first formed a short-lived single party government, and then it took part in a broader coalition with all legislative parties. Including pledges from the 1990 and 1991 elections does not affect the results but focusing on the elections of 1994 and afterward makes for a more straightforward discussion of the analysis. Consistent with existing research, caretaker governments are also excluded from the analysis.

3. Reliability analysis was performed on identifying the pledges from party election platforms and on determining if a pledge was covered in the news. The Cronbach alpha values for agreement between the coders were .92 and .72, respectively. Further details are reported in the Appendix.

4. The full data set is available through the journal’s website, and the replication files are uploaded to Harvard University’s Dataverse site at https://dataverse.harvard.edu/.

5. More precisely, this variable reflects whether or not a pledge has been at least partially fulfilled. Some pledge researchers (e.g., Artés, 2013; Naurin, 2011; Thomson, 2001) analyze separately the shares of pledges that are fully and partly fulfilled. Such distinction would add additional nuance into the ability of parties to act on their election promises. Unfortunately, there are relatively few partially fulfilled pledges in the data set. Furthermore, only a small fraction of partially fulfilled pledges were previously reported by multiple media outlets, five pledges (full data) and three pledges (government parties subset), respectively, making it not feasible to compute interaction coefficients in these regressions. Consequently, the main text of the article presents the analysis with a dummy dependent variable and not an ordinal one.

6. The length of the election campaigns in the country drives the number of weeks prior to each election, for which data are analyzed. Campaign dates are defined by law and reported by international election monitors such as the Organization for Security and Cooperation in Europe (OSCE). Media coverage for such a length of time is consistent with other research (Allern, 2007; Harris, Fury, & Lock, 2006), and it should be noted that in parliamentary democracies parties do not issue election platforms outside of campaigning. Thus the pledge data (and its coverage by the media) include all promises made by parties in the course of an election.

7. The six newspapers selected for this study were chosen to include the most widely read and prominent venues in their respective categories. 24 Chasa and Dneven Trud are the dailies with the widest circulation in the country, and most consistently so for the length of period under study (per European Journalism Center and World Trends Data). These two newspapers are routinely included in comparative media studies such as the Providing an Infrastructure for Research on Electoral Democracy in the European Union (PIREDEU) database on coverage of European Parliament elections. Duma and Demokratsia are the only two party-owned presses that are considered as important among the country’s print media outlets. The same applies for the weekly Kapital and Pari, regarding coverage of economic news.

8. This is a conservative way to estimate how often a pledge is published in the news, as the variable does not account for pledges being mentioned multiple times by the same story (and such data were not collected). It is unlikely that accounting for multiple mentions within the same story would change the main findings, and the results are presented with a more conservative account of a pledge mentioned at least once in a given story.

9. Analyses with a different version of this variable (reported by the media), which reflects if a pledge was mentioned at least once by at least one outlet, are reported in the Appendix.

10. The data are available from https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/YJUIBI .

11. Additional information on coding decisions and examples is provided in the Appendix.

12. Main competitors is excluded from Models 2 and 3 due to collinearity with the variables of interest in these models.

13. The measure for ideology, which equals zero for single-party governments, is perfectly correlated with type of government (single or coalition), and thus this variable can only be included in the analysis for coalition pledges. Furthermore, due to high levels of correlation with the main variables of interest in Model 3, econ growth, term, and econ policy x media are excluded from the estimation.

14. The interpretation of interaction coefficients of odds ratios requires multiplication of the relevant terms (UCLA Statistical Consulting Group, n.d.). In the case of the results of Model 2, when the media variable is at zero, the odds ratio for the impact of coalition on fulfillment is .97. When the media variable equals 1, the effect is derived by multiplying the odds for coalition to the odds for the interaction term: 5.69 (5.87 x .97).

15. Calculations are done using the prvalue command in Stata 14.0 following the respective logit estimations.

Additional information

Notes on contributors

Petia Kostadinova

Petia Kostadinova is Assistant Professor in the Department of Political Science at University of Illinois at Chicago.