New forms of mobilization,new people mobilized? Evidence from the Comparative Study of Electoral Systems

Abstract

Mobilization efforts by parties and candidates during election campaigns tend to reach those who are more likely to vote in the first place. This is thought to be particularly consequential for turnout among the young. Harder and less cost-effective to reach, young adults are less mobilized and vote less often, creating a vicious circle of demobilization. However, new forms of political communication—including online and text messaging—have created expectations this circle might be broken. Is this happening? We examine data from Module 4 of the Comparative Study of Electoral Systems surveys, looking at the prevalence of different types of party contacts in 38 countries, the profile of voters who are reached, and the effects of these efforts on turnout. New forms of party contacting do matter for turnout and partially reduce the age gap in contacting, but still fail to compensate for the much larger differentials that persist in traditional forms of contacting.

Keywords

mobilization party contacts turnout

Introduction

Since Rosenstone and Hansen’s (1993) pathbreaking account, many have studied the effect of mobilization efforts on political engagement. Some, although fewer, have examined conditions under which such contacts affect turnout decisions. Even fewer have looked at different types of contact across countries, even though the technologies available for partisan mobilization have expanded dramatically in recent years. Here, we seek to address this gap using a new and important source of survey data—Module 4 of the Comparative Study of Electoral Systems—that measures different modes of campaign contacting across 38 countries.¹

Explanations of turnout are diverse, including such key explanatory variables as resources (Verba et al., 1995), psychological involvement (Milbrath and Goel, 1977), interpersonal networks (Huckfeldt and Sprague, 1995), social attachments (Putnam, 2001)—especially partisan attachments (Shaffer, 1981)—and institutional factors interacting with all or some of the previous (Anduiza Perea, 2002). Rosenstone and Hansen (1993) focused on mobilization as a determinant of turnout, especially the efforts of political parties and candidates to stimulate voting. We know that such contacts are skewed toward more active and involved voters and are thus likely to increase existing differences in participation (Gershtenson, 2003) notably exacerbating differences in participation between younger and older citizens (Karp et al., 2008; Stevens and Bishin, 2011). The almost universal lower turnout rates among the young (Blais, 2000; Milbrath, 1965) have been explained by fewer resources (Glenn and Grimes, 1968), weaker social attachments (due to geographical mobility, lower rates of marriage and weaker community, and economic ties), and weaker political attachments, especially to party and ideology in particular (Achen and Sinnott, 2007; Strate et al., 1989). But the fact that parties are also less able or willing to mobilize the young seems to add to a vicious cycle that “disconnects the political world from young citizens” (Nickerson, 2006: 48).

In this article, we ask whether the development of new forms of party contacting has reduced this age gap. Comparative Study of Electoral Systems (CSES)’s Module 4 includes measures of self-reported mobilization contacts by parties and candidates not only through such traditional means as face-to-face, mail, and leaflet contacting but also through the use of newer mobilization tools, such as texting, e-mailing, and social network messaging and posting. Online social media have often been described as a potential “leveler in terms of motivating political participation” (Holt et al., 2013: 19; see also Gil de Zúñiga et al., 2014; and Shirky, 2011), including a wide variety of types of political engagement (but see Boulianne, 2015 for a more skeptical view). Party campaigns’ use of these mobilization tools have often been discussed, in particular, in light of their assumed greater potential to reach younger voters (Bosancianu, 2014). In at least some countries, they do seem to mobilize them more effectively (Aldrich et al., 2016). But how widespread are these positive effects? Do these new mobilization tools make young adults easier to reach across a wide variety of societies and political systems? What is their contribution to mitigate or even compensate for well-known age gaps in terms of mobilization, engagement, and turnout?

Voter mobilization: The story so far

Election campaigns have become more professionalized and technologically sophisticated in their mobilization efforts as numerous international studies have attested to (Norris, 2000; Plasser and Plasser, 2002; Farrell and Schmitt-Beck, 2003). These changes have prompted increasing attention to what works and what does not in terms of turning out the vote. While most studies have focused on the case of the United States, the methods used have varied with some scholars using survey data and self-reported contact (Cutright, 1963; Kramer, 1971; Merriam and Gosnell, 1924; Rosenstone and Hansen, 1993; Panagopoulos and Francia, 2009). Others have conducted field experiments, reviving the pioneering approach of Gosnell in the 1920s (1927). Such work has taken off, particularly, since the late 1990s (see Green and Gerber, 2016). These different studies have converged, at least, in one key finding: contacts involving live interaction with voters, especially through face-to-face canvassing, do matter for turnout, particularly when messages appeal to social norms (see Green et al., 2013, Green and York, 2017, and Arceneaux and Nickerson, 2009, for reviews).

Cross-national surveys, such as those analyzed by Karp and Banducci (2007) and Magalhães et al (2016), suggest that being contacted prior to an election increases voting in many political systems. Field experiments by John and Brannan (2008) and Fieldhouse et al. (2013) reinforced the importance of face-to-face contacting by extending the experimental settings to the United Kingdom, and recent work has confirmed these findings elsewhere (Nyman, 2017). However, this line of research has also suggested that party canvassing is not universally effective (Bhatti et al., 2016), while more impersonal methods such as phone and direct mail may be more effective elsewhere than in the United States (Cutts et al., 2009).

The growth in use of modes of digital contacting, especially after the extensive use made of online tools in the Obama presidential campaign in 2008, has reopened these questions. The evidence, so far, has been mixed. Hooghe et al. (2010) reported null findings, Vaccari (2017) got positive findings, while still others (e.g. Aldrich et al., 2016) found heterogeneous results. In terms of specific modes, e-mail messages are seen as one of the least effective prompts (Krueger, 2010; Malhotra et al., 2012; Nickerson, 2007) while text message reminders to vote appear to have only a somewhat stronger effect (Dale and Strauss, 2009; Malhotra et al., 2011). Findings about messages from social networks also range from small (Bond et al., 2012) to null effects (Aldrich et al., 2016; Broockman and Green, 2014).

A few studies examined contacting as a dependent variable. First, not all countries are equal in this respect. Some countries report very high levels of contacting (over 50%), while others drop to nearly zero (Karp and Banducci, 2007). Authors point to a range of explanations. Newer democracies, with less well-organized, experienced, and resource-rich campaigns, report fewer contacts (Birch, 2005). Karp et al. (2008) also show that systems with single member districts (SMDs) lead to higher rates of contact, as candidates are more likely to seek out a direct relationship with a voter than in more anonymous list systems. Furthermore, with lower turnout in plurality systems (Powell, 1986), mobilization has more potential to be effective. Systems where parties are more densely concentrated around the ideological center appear to lead to higher mobilization efforts than polarized systems (Karp, 2012).

Karp et al. (2008) also identify features of voters that make parties more or less likely to mobilize them. One main finding of this comparative work confirms what we already knew about the United States: Citizens who are already active and engaged are most likely to be the targeted. Not surprisingly, campaigns try to maximize the impact of their limited resources by directing their efforts toward those who are most likely to respond positively, that is, those who have previously engaged with politics and who are easier to locate. The significance of other characteristics such as race, socioeconomic status, and organized group membership further underscore the idea that campaigns are prioritizing voters who are already engaged. Recent work by Panagopoulos (2016) has taken this “preaching to the converted” argument a step further by arguing that advances in micro-targeting mean that parties are now increasingly emphasizing base mobilization compared to riskier strategies that focus on persuading harder to persuade independent or undecided voters.

There is also an age gap in contacting. In the United States, younger people are especially mobile and thus more difficult to reach. Parties seem to have internalized that in their judgment about the cost-effectiveness of resource allocation for mobilization (Nickerson, 2006), and the age differential in the probability of being contacted seems to have increased through time (Gershtenson, 2003). Moving to new forms of campaign mobilization, things are less clear. Krueger (2006) finds in that, in the United States, younger people are less likely to be contacted over the Internet. But it is not obvious that these findings will hold either cross-nationally or over time. In the low salience European Parliament elections, for example, Bosancianu (2014) finds that younger people are more likely than older voters to be contacted via the Internet.

Data and results

CSES data allow us, for the first time, to study different modes of contacting across many democracies. In Module 4’s fourth release, data from 38 countries were gathered through national election studies conducted from 2011 through 2016.² That module includes a battery measuring different campaign contacts, several types of which can be broadly divided into “traditional” and “new.” The former specifies if the contact was by mail, phone, or in-person. The latter includes e-mails, text messages, and messages or posts in social networks/micro-blogs such as Facebook or Twitter.

Cross-national variation in levels of contacting

We first present data comparing levels of the new and traditional forms of mobilization. Figure 1 shows the percentages of respondents in each country that report receiving contacts by parties or candidates during the electoral campaign in these different ways.³ Overall, the use of mailing and leaflets remains the most common mode of contacting, followed by face-to-face and phone. Cross-national variations are dramatic, from countries like the United Kingdom or New Zealand (where close to 80% of voters are contacted by direct mail or leaflets) and Ireland or Mexico (where close to half of the electorate reports a face-to-face contact) to countries such as Portugal or Bulgaria (where very few voters report a contact of any kind).

Figure 1.

Prevalence of different types of party contacting (%).

Figure 2 shows a more direct comparison between being (1) contacted in any way, (2) through any “traditional” mode (face-to-face, mail, or phone), or (3) through any “new” mode of contacting (texting, e-mail, or social networks).⁴ First, there is very large variation in the rates of overall contacting. The United Kingdom has the highest level of citizens reporting being contacted in some way (close to 90% in the 2015 election). More than two-thirds of respondents in five other countries—New Zealand, (in both elections), Canada, Switzerland, Mexico (in 2015), and Australia—reported a partisan contact. At the other end are countries such as Romania, Slovenia, Poland, Portugal, or Bulgaria where fewer than 10% of voters reported any type of contact.

Figure 2.

Percentage of respondents contacted by parties or candidates in traditional, new, or any ways.

Second, many more voters report being reached through traditional modes than through the newer modes of contacting (as already suggested in Aldrich et al., 2016, and Bosancianu, 2014). In no country was as much as a third of the electorate reached by texting, e-mail, or social networks. Finally, Figure 2 already suggests that the new forms of party mobilization mostly seem to reach people who are also contacted in traditional ways. This is more clearly visible in Figure 3. The overwhelming majority of contacted respondents report either traditional modes alone or both traditional and new forms of contact. With the partial exception of Taiwan, South Korea, and South Africa, very few individuals are reached exclusively through new forms of mobilization.

Figure 3.

Percentage of respondents reporting different combinations of contacts by parties or candidates.

Correlates of party contacting

What drives party mobilization? One part of the answer pertains to the social and electoral context in which individuals are embedded. Greater levels of party contacting have been found in electoral systems with SMD and in older, more established democracies. Other studies point to the use of winner-take-all electoral rules and party system polarization. We know of no studies that have investigated whether these patterns carry over into the newer forms of mobilization. On the one hand, the relationship between SMDs and mobilization hinges partially on the ability to connect voters to a particular territorial location. While that connection is clear with the traditional modes, it is much more difficult for parties to be confident of territorial location with mobile phone or online contacting. On the other hand, the ability of parties to use these newer modes is likely contingent on the technological readiness of a nation. For example, while in highly economically developed countries such as Finland, the United Kingdom, or Norway, over 90%, of the population has access to the Internet, countries such as Kenya, Thailand, and Mexico have 50% penetration or less. Thus, we should expect, ceteris paribus, that these new forms of party contacting should be more prevalent in the more developed nations.

In Table 1, we take a look at aggregate-level correlations between the prevalence of different types of contact and a series of country-level features, such as the use of winner take all/plurality rules,⁵ the use of SMD,⁶ the age of the democracy,⁷ party system polarization,⁸ and economic development,⁹ as well as the Internet penetration rate and mobile phone subscriptions. The first set of variables has been linked to the prevalence of party contacting in existing broad cross-national studies (Karp and Banducci, 2007; Karp, 2012), while Gross Domestic Product (GDP), Internet penetration, and mobile phone subscription rates aim at capturing the extent to which the technological capabilities necessary for particular types of contacting are widespread.

Table 1.

Aggregate level correlations between prevalence of different types of party contacting and a selection of macro-level variables.

	Face-to-face	Mail	Phone	Text	E-mail	Social nets	Traditional	New	Any
Use of winner take all rules	0.00	0.33	0.13	0.03	0.13	0.04	0.29	0.05	0.28
Use of SMDs	0.16	0.48	0.31	0.18	0.25	0.17	0.52	0.25	0.51
Age of democracy (log)	0.10	0.61	0.20	0.00	0.31	0.37	0.50	0.18	0.48
Party system polarization	−0.07	0.25	−0.11	0.02	0.28	0.24	0.13	0.14	0.10
GDP per capita (log)	0.04	0.49	0.15	−0.01	0.57	0.37	0.38	0.33	0.38
Internet penetration				0.05	0.48	0.29		0.29
Mobile subscriptions				−0.02	0.19	−0.11		0.07

First, Table 1 shows that some of the correlates of contacting proposed in the literature are largely supported in this broad array of democracies. The use of SMDs and age of democracy (the latter employed as a proxy for the level of professionalization of parties in previous studies—Karp and Banducci, 2007) appear to be particularly relevant in this regard. Results also reveal that, within the traditional modes of contact, mail and leaflet contacting is most related to these macro factors. In contrast, the prevalence of new modes of contacting seems to be weakly related to most of those factors. The only partial exception is GDP per capita, particularly, for e-mail contacts.

Turning to the individual level, the primary question here is whether the parties are targeting a different type of potential voter when they use the newer methods. Traditionally, parties tend to target—and to reach—individuals with more resources, and who have stronger social, partisan, and/or ideological ties. A full examination of the role of all relevant individual level characteristics using the CSES data set is not possible because not all election studies included all relevant measures.¹⁰ Given data availability and findings about the importance of the variables, we focus on five core micro-level factors—education, partisanship, marital status, gender, and age.¹¹ As noted above, we pay special attention to age. First, does the probability of being contacted by a party or candidate follow the well-established curvilinear pattern exhibited by the probability of voting itself? Second, to what extent does this hold for the new modes of contacting?

Table 2 shows the results of logistic regression analyses, where the dependent variable is simply whether a voter reports being contacted each particular way.¹² We show the results of analyses for the six different binary-dependent variables.¹³ Based on the exploratory results in Table 1, the macro-level variables employed include age of democracy, SMD, and GDP per capita for each country/year. In the results presented for texting and Internet-related contacts, GDP per capita is replaced, respectively, by the number of mobile cellular subscriptions per 100 inhabitants and by the Internet penetration rate.¹⁴ Coefficients are standardized by dividing them by two times the sample’s standard deviations.

Table 2.

Correlates of party mobilization: Multilevel logistic regression standardized coefficients.

		Traditional			New
	Face-to-face	Mail	Phone	Texting	E−mail	Social network
Age of democracy (log)	0.10 (0.32)	1.30* (0.56)	0.34 (0.47)	0.25 (0.54)	0.02 (0.40)	0.23 (0.34)
Single member districts	0.34 (0.29)	1.14** (0.38)	0.99* (0.43)	0.35 (0.55)	0.44 (0.32)	0.18 (0.27)
GDP per capita, 000 s (log)	−0.25 (0.31)	1.70** (0.51)	0.30 (0.46)	—	—	—
Mobile phone subscriptions (per hundred)	—	—	—	0.12 (0.48)	—	—
Internet penetration rate	—	—	—	—	1.27** (0.41)	0.74* (0.35)
Female	−0.11*** (0.03)	0.06* (0.02)	0.08* (0.03)	−0.14*** (0.04)	−0.25*** (0.04)	−0.21*** (0.05)
Age	0.99*** (0.17)	1.07*** (0.15)	1.60*** (0.20)	2.54*** (0.31)	1.34*** (0.29)	−0.45 (0.31)
Age squared	−0.35*** (0.06)	−0.30*** (0.05)	−0.38*** (0.07)	−1.03*** (0.11)	−0.54*** (0.10)	−0.17 (0.12)
Education	0.26*** (0.04)	0.47*** (0.03)	0.45*** (0.04)	0.71** (0.05)	1.07*** (0.05)	0.87*** (0.06)
Married	0.05 (0.03)	0.12*** (0.03)	0.16*** (0.04)	0.12* (0.05)	−0.001 (0.05)	−0.12* (0.05)
Close to party	0.62*** (0.90)	0.41*** (0.03)	0.44*** (0.03)	0.53*** (0.04)	0.74*** (0.05)	0.69*** (0.05)
Constant	−2.83*** (0.77)	−10.80*** (1.04)	−6.99*** (1.13)	−6.71*** (1.77)	−8.56*** (0.67)	−5.82*** (0.57)
Random intercept parameterCountry/election variance	0.75	1.19	1.56	1.93	0.82	0.58
Countries/elections	38	37	36	34	37	37
Respondents	55,830	54,922	53,085	50,192	54,759	54,153

*p < 0.05; **p < 0.01; ***p > 0.001.

The results confirm and provide further detail on the key findings from prior analysis of older CSES data (Karp et al., 2008; Karp and Banducci, 2007). Specifically, they show that individuals living in countries with electoral systems employing SMDs are significantly more likely to be mobilized by parties through mail and phone. Similarly, the age of democracy and GDP per capita variables are significantly related to mail contact. The relationships between the individual-level variables and traditional means of contacting support expectations from the literature. Respondents who are more educated, married, and who “feel close to a party” are more likely to report being contacted. The age variable behaves very much as expected: The signs for age and age squared are both significant and suggest a curvilinear relationship between one’s age and the probability of being contacted.¹⁵ We will examine this aspect in greater detail later. The signs, coefficient sizes, and significance for all remaining individual-level variables are relatively similar, with the partial exception of face-to-face contacts, where women are less likely to report being contacted in person than are men, while marital status and education seem to matter less than for other types of traditional contacts.

In columns 3 to 5, we report the results for the new forms of contacting. E-mail and social network contacting are more prevalent in nations with greater rates of Internet penetration.¹⁶ However, neither the use of SMDs nor age of democracy makes a difference in these types of contacts. The greater sensitivity of traditional contacting to institutional context and democratic longevity is intriguing and suggests that those modes are more strategically aligned with the incentives provided in the wider electoral environment. The newer forms, by contrast, appear to be more untethered and less structured by these broader systemic forces, and remain elusive, in terms of macro-correlates, in this analysis. We will return to this point in the final section. Partisans are more likely to be reached in these new ways than non-partisans, similarly to what happens with traditional modes. Finally, education plays an even stronger role for contacting using these new ways.

A deeper dive into age and contacting

Figure 4 shows plots of the predicted probability of being contacted by parties or candidates at different ages based on the various models estimated in Table 2, that is, after controls are introduced. For these plots, we reestimated the models using only the 32 countries for which we have information about all of the types of contacts.

Figure 4.

Predicted probability of being contacted in different ways, by age of respondent. Multilevel logistic regression, 32 countries.

In almost all modes of contact, the relationship between age and being contacted is non-linear, as usual. The exception is being contacted through an online social network. Here, we find a nearly linear, negative relationship, with young adults being more likely to report contacts compared with their middle-aged counterparts and even more so than the oldest respondents. However, on average, the probability of being reached in this way across our countries is modest. In all remaining modes, younger voters are less likely to be contacted than most other adults. This is especially so for contacts by mail or phone, and young adults are only marginally more likely to receive a text or e-mail than even the very oldest voters.

Each individual can be exposed to multiple forms of contact. Table 3 reports analyses where we look at, respectively, the correlates of reporting any type of contact, a traditional contact, and one or more kinds of “new” contact. Again, we focus exclusively on the 32 countries on which we have responses for all kinds of contact.

Table 3.

Correlates of party mobilization: Multilevel logistic regression standardized coefficients.

	Multilevel logistic
	Any contact	Traditional	New
Age of democracy (log)	0.43 (0.41)	0.50 (0.41)	0.62 (0.50)
Single member districts	1.08*** (0.32)	1.03*** (0.32)	0.37 (0.38)
GDP per capita, 000 s (log)	0.26 (0.39)	0.25 (0.38)	0.31 (0.46)
Female	−0.01 (0.02)	0.01 (0.02)	−0.18*** (0.03)
Age	1.05*** (0.14)	1.27*** (0.14)	1.31*** (0.22)
Age squared	−0.32*** (0.05)	−0.37*** (0.05)	−0.62*** (0.08)
Education	0.47*** (0.03)	0.40*** (0.03)	0.83*** (0.04)
Married	0.06** (0.03)	0.06** (0.03)	0.04 (0.04)
Close to party	0.59*** (0.02)	0.57*** (0.02)	0.65*** (0.04)
Constant	−4.30*** (0.78)	−4.65*** (0.78)	−6.26*** (0.93)
Random intercept parameterCountry/election variance	0.72	0.71	0.98
Countries/elections	32
Respondents	47,609

*p < 0.05; **p < 0.01; ***p > 0.001.

The correlates of being contacted at all and of being contacted through traditional modes (columns 1 and 2) are almost identical, for the simple reason that almost everybody who is contacted in any way is contacted through a traditional mode. Again, use of SMDs, level of education, and our indicators of social and partisan attachment emerge as relevant. With respect to those who report new forms of contacting, no coefficient for any macro-level correlate is significant. There are three other major differences in comparison with traditional contacting. First, females are less likely to be reached. Second, marriage seems to make no difference. And finally, the relationship between the respondent’s level of education and the probability of being contacted is about twice as strong in comparison with what happens in traditional modes.

Figure 5 shows findings with respect to age. While young adults are apparently less effectively reached by parties by traditional means than the other voters, such disadvantage mostly disappears when new forms of contacting are considered. Here, the oldest voters are least likely to be contacted. However, the lower prevalence of the new types of contacts overall, combined with the fact that young adults are not more likely to be reached in these ways than the middle-aged, produces a perhaps surprising overall result: in the end, when considering all kinds of mobilization (“any contact”), young adults remain less targeted than any other type of voter. In other words, the new forms of party contacting have, so far, failed to compensate for the age gap in mobilization.

Figure 5.

Predicted probability of being contacted in any way, by traditional means or by new means, depending on age of respondent. Multilevel logistic regression, 32 countries.

Of course, this is the most general picture. What happens on country-by-country basis? We estimated the individual level component of the multilevel logistic models used for Table 3 separately for each of the 20 countries in which we had complete data about traditional and new contacts, and where rates of prevalence of either were above a minimum threshold of 5% of the sample (in 12 countries, that prevalence was even lower). Then, we estimated the predicted probabilities of being contacted according to age for each country. Figures A1 to A4 in the Online supplemental appendix show the results for four groups of countries. The first, the largest (12 countries/elections), replicates the general pattern: overall, younger people report fewer contacts than middle-aged adults (and in some cases, than all other adults), because they are less likely to be contacted in traditional ways, and new forms of contact fail to compensate for that differential. Then, in Sweden and Norway, exceptionally, younger voters are more likely to be contacted by parties in general, but this results from being more contacted both in traditional and in new ways. In a third group of countries/elections (Austria, Czech Republic, Mexico 2015, and Turkey), age and party contacting seem weakly related. Finally, in two cases, while younger adults were less likely to be contacted in traditional ways, they were more likely to be contacted in new ways. And as a result, the probability of being targeted by parties in any way ends up unrelated to age in Mexico in 2012 or with a differential in favor of younger voters in Finland. However, these are the only two cases where new forms of contacting make up the age gap due to traditional forms of contacting.¹⁷

In sum, with few exceptions, new forms of contact reach small segments of the electorate. They reach a similar profile of voters as traditional contacts, but one even more restricted to men and to the highly educated. Finally, although there is a tendency for younger adults to be less disadvantaged by these new forms of contact, this is not sufficient to overcome the broad age differential in traditional party contacting.

Party contacting and turnout

What difference does this make? Does being contacted increase the probability of voting? And if so, which mode(s) are effective? We have just seen how party contacts are not randomly assigned. Although this may be partially addressed by controlling for known covariates of turnout, both contacting and turnout can still be systematically related with attributes we were unable to measure (see Gerber et al., 2004). The consequence is that the use of observational data such as those collected in these surveys, compounded by the lack of relevant variables in several countries, increases the risk that any estimates of the relationship between contact and turnout will be biased, much more so than if experimental data were available.

However, well-designed sample-surveys such as those in the CSES permit inferences to population values, something that experiments do not support, and it is nevertheless possible to employ strategies that may reduce—albeit never eliminate—the risk of bias. In Table 4, we show the results of three models of turnout, each employing a different independent variable: if the respondent reported being contacted by a party or candidate in any way; if the respondent reported being contacted just through a traditional mode; or if the respondent reported by contacted both in a traditional and a new way. First, we report the results from a multilevel random intercept logistic regression model, using a single macro-level ordinal variable (Compulsory voting),¹⁸ as well as from fixed-effects probit, where the cluster variables (countries/elections) enter the model as dummy variables, thus capturing all variability associated with the cluster level. However, we also estimate a fixed effects, recursive, bivariate probit model. Given that we know that people were not contacted by campaigns randomly and were instead selected (and self-selected) in ways that make being contacted at least partially endogenous, the recursive, bivariate probit procedure is designed to reflect that possibility, by estimating two equations simultaneously, one for the endogenous contact variable and the other for turnout.¹⁹ Table 4 shows the results.

Table 4.

Estimates of the effect of different types of contact: multilevel logistic, fixed-effects probit, and fixed-effects bivariate probit.

	Any contact				Just traditional contact				Traditional + new contact
	Multilevel logistic	Fixed-effects probit	Bivariate fixed-effects probit		Multilevel logistic	Fixed-effects probit	Bivariate fixed-effects probit		Multilevel logistic	Fixed-effects probit	Bivariate fixed-effects probit
Dependent variable	Turnout	Turnout	Turnout	Contact	Turnout	Turnout	Turnout	Contact	Turnout	Turnout	Turnout	Contact
Compulsory voting	0.98** (0.30)	—	—	—	0.98** (0.31)	—	—	—	0.98** (0.32)	—	—	—
Female	0.005 (0.03)	0.007 (0.02)	—	—	−0.002 (0.03)	0.003 (0.02)	—	—	0.001 (0.03)	0.01 (0.02)	—	−0.07* (0.03)
Age	1.76*** (0.16)	1.03*** (0.18)	1.02*** (0.18)	0.64*** (0.13)	1.74*** (0.16)	1.01*** (0.18)	0.99*** (0.19)	0.71*** (0.11)	1.92*** (0.18)	1.12*** (0.16)	1.11*** (0.15)	1.08*** (0.27)
Age squared	−0.42*** (0.06)	−0.25*** (0.06)	−0.25*** (0.06)	−0.20*** (0.04)	−0.41*** (0.06)	−0.25*** (0.07)	−0.24*** (0.06)	−0.18*** (0.04)	−0.48*** (0.06)	0.29*** (0.06)	−0.28*** (0.05)	−0.45*** (0.09)
Education	0.65*** (0.04)	0.34*** (0.04)	0.34*** (0.05)	0.28*** (0.03)	0.64*** (0.04)	0.34*** (0.04)	0.34*** (0.05)	0.14*** (0.03)	0.66*** (0.04)	0.35*** (0.05)	0.34*** (0.04)	0.46*** (0.05)
Married	0.37*** (0.03)	0.21*** (0.02)	0.21*** (0.02)	—	0.36*** (0.03)	0.21*** (0.02)	0.20*** (0.02)	—	0.36*** (0.03)	0.21*** (0.02)	0.21*** (0.02)	—
Close to party	1.12*** (0.03)	0.60*** (0.05)	0.59*** (0.07)	0.34*** (0.03)	1.11*** (0.03)	0.59*** (−0.05)	0.59*** (0.07)	0.27*** (0.04)	1.08*** (0.04)	0.58*** (0.05)	0.57*** (0.06)	0.45 (0.04)***
Any contact	0.46*** (0.04)	0.25*** (0.03)	0.31 (0.25)	—	—	—	—	—	—	—	—	—
Traditional contact	—	—	—	—	0.39*** (0.04)	0.21*** (0.03)	0.32 (0.28)	—	—	—	—	—
Traditional + new contact	—	—	—	—	—	—	—	—	0.84*** (0.07)	0.45*** (0.06)	0.61** (0.22)	—
Constant	−1.14*** (0.17)	−0.04 (0.12)	−0.05 (0.13)	−0.91*** (0.14)	−1.10*** (0.18)	0.00 (0.13)	−0.03 (0.14)	—	−1.24 (0.18)	−0.89*** (0.10)	−0.89*** (0.10)	−2.35*** (0.22)
Random intercept parameterCountry/election variance	0.48**				0.49**				0.51**
ρ			−0.04				−0.06		—	—	−0.09
Countries/elections	32				32				32
Respondents	46,735				42,475				35,972

Note: Standardized coefficients.

*p < 0.05; **p < 0.01; ***p > 0.001.

Consider first the relationship with turnout of the remaining explanatory variables besides contacting. In the multilevel model, the compulsory voting measure is statistically significant and substantively large. Party closeness and education are powerful predictors of turnout in all models. However, their importance is rivaled by that of being contacted by a party in the third set of models, where “contact” is cumulative, that is, having reported being contacted both in a traditional and in a new way. In sum, institutions and the basic “triad” of individual level variables that explain turnout—resources, attachments, and mobilization efforts—are once again found to be consequential.

Figure 6 shows the marginal effects of the different contact variables on the probability of voting. Interpretation is straightforward. In both our multilevel logit and fixed-effects probit models, the probability of turning out is estimated to be about 4 percentage points higher for those who were contacted in any way in comparison with those who were not contacted at all. But strikingly, those who reported being contacted both in traditional and in newer ways were much more likely (8percentage points) to have voted than those not contacted at all.

Figure 6.

Marginal effects of contact on the probability of turning out to vote (95% CI). CI: confidence interval.

The bivariate probit estimations have much larger confidence intervals, as usual, and the coefficients for contacting are not significant for either the “any contact” or “traditional contact” variables. However, the estimated effects of the cumulative traditional + new contact variable are significant and even larger than in the other models.²⁰ Furthermore, the diagnostic statistics mitigate our concerns about the need to address selection bias.²¹

Parties contacting citizens to stimulate mobilization appears to be positively and significantly related to turnout. The 4 percentage point estimated effect of being contacted by a party in traditional ways is similar to that obtained in other comparative studies (Karp and Banducci, 2007; Magalhães et al., 2016). More importantly, the results indicate that new forms of contact may be consequential. Even though they have not greatly expanded the portion of the electorate that was already reached by traditional means, they seem to contribute to a cumulative effect, nearly doubling the propensity to vote beyond what being contacted just through traditional modes are able to achieve, a result that is robust to the estimation strategy employed.

Discussion

Partisan mobilization efforts are known to focus on particular profiles of voters—those who are already engaged and likely to respond. The arrival of digital communication channels has introduced a new and cost-effective way for parties to reach groups less usually contacted—including younger voters—and thus hold the potential to break a vicious cycle of under-mobilization and disengagement. Our analysis, using self-reported contact data from 38 countries, has both positive and negative news in this regard. Whatever the potential for greater youth mobilization through new methods may be, it has yet to materialize to any substantial degree. While younger people are comparatively more likely to be contacted by parties using these newer methods, the overall frequency of such self-reported contact is very limited compared with other more traditional modes. The profile of those contacted is otherwise not particularly different from those contacted in more traditional ways. Indeed, as in the case for education, the sociopolitical selectivity driving the traditional types of contacting appears to be carried over and reinforced by new forms.

These results are even more intriguing given our subsequent findings that the new forms of contact appear to be successful for mobilization. The interpretation of this, however, is not straightforward. Even though traditional methods of contact retain power for mobilizing voters, it is their combination with new modes that seems particularly consequential, virtually doubling the estimated impact of traditional contacts alone. This finding boosts the case for parties to adopt new methods into their arsenal of campaign mobilization weapons. However, given that, at least at this point in their development, the newer types of contact target mostly those who have already been contacted through more traditional means, the most likely outcome of an increasing use of digital methods for the foreseeable future would seem to lead to a greater mobilization of the already engaged.

Several limitations of our study must be acknowledged. First, we relied on observational survey data. Although studies based on experimental data are not devoid of the risk of bias in the estimation of the effects of contact on turnout, that risk is much higher with observational data, despite the estimation techniques employed in this study. Second, we relied on the self-reports of respondents to measure contact. One might argue that, regardless of the potential slippage between actual exposure to partisan messages and voters’ perception of that exposure, it is the latter that should matter to explain voters’ behavior. However, we also know that, particularly in what concerns some aspects of online behavior, self-reports and “objective” measures tend to be discrepant (Revilla et al., 2017), and we can only speculate about the potential effects of such discrepancies between perception and reality for our results.

Finally, our analysis of the macro-level determinants of contact, and particularly the null or weak findings for the new forms of contact, may reflect a “missing variable” problem at the systemic level. On the one hand, Internet penetration rates do not necessarily reflect the extent to which online tools are used for political purposes beyond party contacting in different countries, including political discussion, online petitions, chat room participation, e-mail correspondence with candidates, and so on. Such engagement should, in turn, render individuals easier to target by parties during political campaigns. Unfortunately, we have no measures of such variables across the broad range of countries under examination in the surveys. On the other hand, our institutional variables fail to capture the new and increasingly broader set of protection and privacy rules that shape parties’ micro-targeting efforts, particularly through digital messages. The emerging work on this field has pointed to a relationship between the wider regulatory environment and the incidence of “data-driven campaigning” in a country (Anstead, 2017; Bennett, 2016). Particular attention has been given to the case of the United States, a case where much of the privacy regulation that restricts parties and candidates elsewhere is side-stepped (Bennett, 2016; Bimber, 2014; Hersh, 2015). Unfortunately, a comprehensive cross-national picture of these regulatory controls and data protection regimes operating at the national level across CSES countries also does not currently exist. However, there would clearly appear to be an increasingly compelling case for the development and application of such an index in comparative studies of voter mobilization.

Future work departing from this study, besides addressing the preceding limitations, might well take two main directions. First, it is important to confirm and understand how and why the combined effect of traditional and new forms of contacting actually works. Do traditional forms of contacting receive a boost when followed by e-mail, text message, or tweets or vice-versa? Or is the effect simply due to the sheer volume and diversity of contacts? Thus, an important next step for analyses that follow will be to measure the sequencing as well as the frequency of each type of contacting. Second, we have examined the impact of new and traditional forms of contacting on a singular form of participation, namely voting. It may be that social media and other new technologies are already effectively stimulating participation in other ways, such as mobilizing people to become campaign activists, to donate money, or to seek to persuade others to vote. Whether there are stronger or even different effects of these new methods of contact across political participation is clearly an important next question to address.

Supplemental Material

Supplemental Material, sj-docx-1-ppq-10.1177_1354068818797367 - New forms of mobilization, new people mobilized? Evidence from the Comparative Study of Electoral Systems

Supplemental Material, sj-docx-1-ppq-10.1177_1354068818797367 for New forms of mobilization, new people mobilized? Evidence from the Comparative Study of Electoral Systems by Pedro C Magalhães, John H Aldrich and Rachel K Gibson in Party Politics

Footnotes

Authors’ note

An earlier version of this article was presented at the World Association for Public Opinion Research (WAPOR) annual conference (2017, Lisbon).

Acknowledgements

The authors are grateful to the participants at that conference, as well as to the three anonymous reviewers, for their very useful comments and suggestions. The authors would also like to thank Marta Cantijoch, Rosalynd Southern, and Tobias Konitzer for their help and inputs at the early stages of this project.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Pedro C Magalhães

Supplemental Material

Supplemental material for this article is available online.

Notes

References

Achen

Sinnott

(2007) Taking sides: Learning and voting. In: Annual Meeting of the Midwest Political Science Association, Chicago, IL. Available at: https://pdfs.semanticscholar.org/68da/8d810510dbf09a59a04d6fca5087f3649454.pdf (accessed December 4 2017).

Aldrich

Gibson

Cantijoch

, et al. (2016) Getting out the vote in the social media era: are digital tools changing the extent, nature and impact of party contacting in elections? Party Politics 22(2): 165–178.

Anduiza Perea

(2002) Individual characteristics, institutional incentives and electoral abstention in Western Europe. European Journal of Political Research 41(5): 643–673.

Anstead

(2017) Data-driven campaigning in the 2015 United Kingdom general election. The International Journal of Press/Politics 22(3): 294–313.

Arceneaux

Nickerson

(2009) Who is mobilized to vote? A re-analysis of 11 field experiments. American Journal of Political Science 53(1): 1–16.

Bennett

(2016) Voter databases, micro-targeting, and data protection law: Can political parties campaign in Europe as they do in North America? International Data Privacy Law 6(4): 261–275.

Bhatti

Dahlgaard

Hansen

, et al. (2016) Is door-to-door canvassing effective in Europe? Evidence from a meta-study across six European countries. British Journal of Political Science 1–12. Available at: https://doi.org/10.1017/S0007123416000521 (accessed January 8 2018).

Birch

(2005) Single-member district electoral systems and democratic transition. Electoral Studies 24(2): 281–301.

Bimber

(2014) Digital media in the Obama campaigns of 2008 and 2012: adaptation to the personalized political communication environment. Journal of Information Technology & Politics 11(2): 130–150.

10.

Blais

(2000) To Vote or Not to Vote?: The Merits and Limits of Rational Choice Theory. Pittsburgh: University of Pittsburgh Press.

11.

Bond

Fariss

Jones

, et al. (2012) A 61-million-person experiment in social influence and political mobilization. Nature 489: 295–298.

12.

Bosancianu

(2014) Online mobilization and offline participation in European elections. GeT MA Working Paper Series, n. 4. Available at: https://edoc.hu-berlin.de/bitstream/handle/18452/3736/4.pdf?sequence=1 (accessed June 5 2017).

13.

Boulianne

(2015) Social media use and participation: a meta-analysis of current research. Information, Communication & Society 18(5): 524–538.

14.

Broockman

Green

(2014) Do online advertisements increase political candidates’ name recognition or favorability? Evidence from randomized field experiments. Political Behavior 36(2): 263–289.

15.

Cutright

(1963) Measuring the impact of local party activity on the general election vote. Public Opinion Quarterly 27: 372–386.

16.

Cutts

Fieldhouse

John

(2009) Is voting habit forming? The longitudinal impact of a GOTV campaign in the UK. Journal of Elections, Public Opinion and Parties 19(3): 251–263.

17.

Dale

Strauss

(2009) Don’t forget to vote: text message reminders as a mobilization tool. American Journal of Political Science 53(4): 787–804.

18.

Farrell

Schmitt-Beck

(2003) Do Political Campaigns Matter?: Campaign Effects in Elections and Referendums. London: Routledge.

19.

Fieldhouse

Cutts

Widdop

, et al. (2013) Do impersonal mobilisation methods work? Evidence from a nationwide get-out-the-vote experiment in England. Electoral Studies 32(1): 113–123.

20.

Gerber

Green

Kaplan

(2004) The illusion of learning from observational research. In: Shapiro

Smith

(eds) Problems and Methods in the Study of Politics. New York: Cambridge University Press, pp. 251–273.

21.

Gershtenson

(2003) Mobilization strategies of the democrats and republicans, 1956-2000. Political Research Quarterly 56(3): 293–308.

22.

Gil de Zúñiga

Molyneux

Zheng

(2014) Social media, political expression, and political participation: panel analysis of lagged and concurrent relationships. Journal of Communication 64(4): 612–634. DOI: 10.1111/jcom.12103.

23.

Glenn

Grimes

(1968) Aging, voting, and political interest. American Sociological Review 33(4): 563–575.

24.

Gosnell

(1927) Getting Out the Vote: An Experiment in the Stimulation of Voting. Chicago: University of Chicago Press.

25.

Green

Gerber

(2016) Voter mobilization, experimentation, and translational social science. Perspectives on Politics 14(3): 738–749.

26.

Green

McGrath

Aronow

(2013) Field experiments and the study of voter turnout. Journal of Elections, Public Opinion & Parties 23(1): 27–48.

27.

Green

York

(2017) Field experiments in political behavior. In: Fisher

Fieldhouse

Franklin

, et al. (eds) The Routledge Handbook of Elections, Voting Behavior and Public Opinion. Abdingdon: Routledge, pp. 509–520.

28.

Hersh

(2015) Hacking The Electorate: How Campaigns Perceive Voters. Cambridge: Cambridge University Press.

29.

Holt

Shehata

Strömbäck

, et al. (2013) Age and the effects of news media attention and social media use on political interest and participation: Do social media function as leveller? European Journal of Communication 28(1): 19–34.

30.

Hooghe

Vissers

Stolle

, et al. (2010) The potential of Internet mobilization: an experimental study on the effect of Internet and face-to-face mobilization efforts. Political Communication 27(4): 406–431.

31.

Huckfeldt

Sprague

(1995) Citizens, Politics and Social Communication: Information and Influence in an Election Campaign. New York: Cambridge University Press.

32.

John

Brannan

(2008) How different are telephoning and canvassing? Results from a ‘get out the vote’field experiment in the British 2005 general election. British Journal of Political Science 38(03): 565–574.

33.

Karp

Banducci

(2007) Party mobilization and political participation in new and old democracies. Party Politics 13(2): 217–234.

34.

Karp

Banducci

Bowler

(2008) Getting out the vote: party mobilization in a comparative perspective. British Journal of Political Science 38(1): 91–112.

35.

Karp

(2012) Electoral systems, party mobilisation and political engagement. Australian Journal of Political Science 47(1): 71–89.

36.

Kramer

(1971) The effects of precinct level canvassing on voting behavior. Public Opinion Quarterly 34: 560–572.

37.

Krueger

(2006) A comparison of conventional and Internet political mobilization. American Politics Research 34(6): 759–776.

38.

Krueger

(2010) Opt in or tune out: Email mobilization and political participation. International Journal of E-Politics (IJEP) 1(4): 55–76.

39.

Maddala

(1983) Limited-Dependent and Qualitative Variables in Econometrics. Cambridge: Cambridge University Press.

40.

Malhotra

Michelson

Rogers

, et al. (2011) Text messages as mobilization tools: the conditional effect of habitual voting and election salience. American Politics Research 39: 664–681.

41.

Malhotra

Michelson

Valenzuela

(2012) Emails from official sources can increase turnout. Quarterly Journal of Political Science 7: 321–332.

42.

Milbrath

(1965) Political Participation. Chicago: Rand McNally.

43.

Magalhães

Segatti

Shi

(2016) Mobilization, informal networks and the social contexts of turnout. In: Paul

Gunther

Magalhães

, et al. (eds) Voting in Old and New Democracies. Milton Park: Routledge, pp. 64–98.

44.

Milbrath

Goel

(1977) Political Participation: How and Why do People Get Involved in Politics? Chicago: Rand McNally.

45.

Merriam

Gosnell

(1924) Non-voting: Causes and Methods of Control. Chicago: University of Chicago Press.

46.

Nickerson

(2006) Hunting the elusive young voter. Journal of Political Marketing 5(3): 47–69.

47.

Nickerson

(2007) Does email boost turnout? Quarterly Journal of Political Science 2(4): 369–380.

48.

Norris

(2000) A Virtuous Circle: Political Communications in Postindustrial Societies. Cambridge: Cambridge University Press.

49.

Nyman

(2017) Door-to-door canvassing in the European elections: evidence from a Swedish field experiment. Electoral Studies 45: 110–118.

50.

Panagopoulos

(2016) All about that base: changing campaign strategies in US presidential elections. Party Politics 22(2): 179–190.

51.

Panagopoulos

Francia

(2009) Grassroots mobilization in the 2008 presidential election. Journal of Political Marketing 8: 315–333.

52.

Plasser

(2002) Global Political Campaigning: A Worldwide Analysis of Campaign Professionals and their Practices. Westport, CT: Praeger.

53.

Powell

Jr (1986) American voter turnout in comparative perspective. The American Political Science Review 80: 17–43.

54.

Putnam

(2001) Bowling Alone: The Collapse and Revival of American Community. New York: Simon and Schuster.

55.

Revilla

Ochoa

Loewe

(2017). Using passive data from a meter to complement survey data in order to study online behavior. Social Science Computer Review 35(4): 521–536.

56.

Rosenstone

Hansen

(1993) Mobilization, Participation, and Democracy in America. New York: MacMillan.

57.

Shaffer

(1981) A multivariate explanation of decreasing turnout in presidential elections, 1960–1976. American Journal of Political Science 25(1): 68–95.

58.

Shirky

(2011) The political power of social media: technology, the public sphere, and political change. Foreign affairs 90: 28–41.

59.

Stevens

Bishin

(2011) Getting out the vote: minority mobilization in a presidential election. Political Behavior 33(1): 113–138.

60.

Strate

Parrish

Elder

, et al. (1989) Life span civic development and voting participation. American Political Science Review 83(2): 443–464.

61.

Vaccari

(2017) Online mobilization in comparative perspective: digital appeals and political engagement in Germany, Italy, and the United Kingdom. Political Communication 34(1): 69–88.

62.

Verba

Schlozman

Brady

(1995) Voice and equality: Civic voluntarism in American politics. Cambridge: Harvard University Press.

63.

Wilde

(2000) Identification of multiple equation probit models with endogenous dummy regressors. Economics letters 69(3): 309–312.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.16 MB