When David and Goliath campaign online: The effects of digital media use during electoral campaigns on vote for small parties

ABSTRACT

Challenger parties might have an advantage online compared to mainstream parties, since digital technologies increase their visibility at a low cost and connect them with niche audiences. Previous research has examined this phenomenon focusing on parties’ use of the Internet, yet we need to focus on voters’ behavior to quantify the effect. To this end, we use data covering 21 national and regional elections from four countries. Our results confirm that digital media use during the campaign boosts uncertainty about one’s own vote choice and, ultimately, increases the chances to change one’s voting intention from mainstream to small parties.

KEYWORDS:

• Digital media
• online campaigning
• challenger parties
• elections

Disclosure statement

The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the article.

Notes

1. Each regional survey has a sample size of around 1,000 respondents, making a total of 2,000–3,000 observations per region, with the exception of Bavaria, where we have about 6,000, and the Canadian provinces of Quebec, Ontario, and British Columbia, where we have only one survey and about 1,000 respondents. Note as well that the study included data for Zurich and Lucerne, but we have excluded these cases for one main reason: Switzerland is the epitome of a consociational democracy, whose main characteristics—according to Lijphart—are grand coalitions and mutual veto between elites; all of which hinders referring to “big” or “small” parties, as a small party can be in fact a ruling one. Finally, note that the national election included for the French case was the one meant to select the 14th National Assembly, held a little over a month after the French presidential election in May.

2. Our criterion is also partially consistent with Norris’s conception of major and minor parties (Norris, 2003). According to Norris, major parties have at least 20% of the seats in the lower house of parliament, while minor parties have less than 20%. Our classification fulfills Norris’s categorization in all cases except in two—both criteria conflict. These conflicting cases are all federal elections in idiosyncratic regions such as Quebec and Catalonia, where major electoral parties at the regional level become minor parliamentary parties at the federal level. In these cases, our classification diverges from Norris’s categorization because we use vote share and not seat share to classify parties as either big or small. The fact that the level of party nationalization of the Canadian electoral system is very low in comparison with the rest and that seat share cannot be used to determine which parties are big and small in European elections are two additional reasons to choose vote share and not seat share as our main classification criteria.

3. Note that the use of composite measures (scales) for both off-line media attention and digital media exposure during the campaign reduces the amount of measurement error and simplifies our models, reducing the risk of multicollinearity. We still acknowledge that using two dichotomous variables (albeit in a composite index) to tap digital media use is not ideal. Frequency of exposure would have been a better measure of digital media use and more indicators of digital media use—including, for instance, social media use—before and during the electoral campaign would have helped to tap more accurately this phenomenon. Please find the exact question wording behind these variables and their descriptive statistics in appendix I and II, respectively.

4. More precisely, we will test a generalized linear model via maximum likelihood by means of the gsem STATA 13 command.

5. See Appendix III.

6. We are not computing the total or indirect effects for the “stick to big parties” category, as this would be beyond the scope of the paper.

7. The AIC (Akaike Information Criterion) and BIC (Bayesian Information Criterion), both maximum likelihood estimates that assess model fit, should be interpreted against a reference model. Should we drop the mediators, both model fit measures would improve, although our goal is not to find the most efficient predictive model of voting behavior, but to test the causal path through perceived viability of small parties and indecision.

8. Each coefficient is represented under its respective arrow.

Additional information

Notes on contributors

Carol Galais

Carol Galais is a postdoctoral fellow in the department of Political Science and Law at the Universitat Oberta de Catalunya. Her research interests include the psychological factors that lead to participation, the civic duty to vote, the role of digital media in public opinion, and the political socialization process.

Ana Sofía Cardenal

Ana S. Cardenal is a professor of Political Science at the Universitat Oberta de Catalunya. Her research fields include comparative politics and political behavior. Her recent research interests include electoral politics, public agenda, political mobilization, political parties, and uses of the Internet.