Empirical analyses

As has been previously mentioned, the data was hierarchically structured and thus the data analysis was carried out by means of the statistical technique of multi-level analysis (Hox 2002; Snijders and Bosker 1999). The models' dependent variable is

12 The pictures were sorted according to country and the order of the presentation corresponded to the alphabetical order of the countries' international code on vehicle license plates. In countries with a number of sub-national constituencies the pictures were grouped according to constituency and these sets were presented in a random order. Within the country or constituency, the sets of photographs were also presented in a random order.

Table 1 Front-Runners' Attractiveness by Sex, and by Age. (Source: Own data)

*p < 0.05; **p < 0.01; ***p < 0.001

the share of the votes obtained at the European election 2004 by the various parties or combined lists in the constituencies. For the analyses, the data were weighted with the aforementioned level-2 weighting variable. In the first modeling stage all available level-1 predictors were entered with fixed effects into the model and the intercept was modeled in dependence on the number of available voting options in the constituency (Model 1 in Table 2). The results show that, consistent with our expectation, the vote share tends to fall with an increasing number of voting options in the constituency. Furthermore, it can be demonstrated that front-runners' attractiveness has a clear effect on vote share: with every increase in attractiveness by one scale point, the proportion of votes gained by the candidates' party or combined list rises by approximately 1.5 % points. Consequently, front-runners' attractiveness in this model is able to produce a change in vote share by up to 9 % points. Significant effects also result from the election result at the last national election, membership in the national government, membership in the European Parliament and the age of the front-runner. All these effects point in a theoretically plausible direction: the vote share of a voting option at European elections is heavily influenced by its national vote share. Membership of a national government tends to decrease the vote share and membership of the European Parliament increases it. Finally, older front-runners are able to mobilize more electoral support than younger front-runners.

In an additional modeling stage (Model 2 in Table 2), the effect of candidates' attractiveness on vote share are set free and it is tested whether this effect varies significantly over the 64 level-2 units. Here it is shown that such variation does in fact occur and, thus, the impact of attractiveness differs at the constituency-level. A further modeling stage attempts to explain the strength of the attractiveness effect through the available level-2 variables (Model 3 in Table 2). The only significant finding is that candidate attractiveness has less of an effect in subnational constituencies than in national constituencies. Significant cross-level interactions are, however, not demonstrated with the presence of electoral duty, a barring clause or flexible lists. These variables are therefore excluded from the model in the final modeling stage (Model 4 in Table 2).

In order to achieve the most vivid impression of the impact of front-runners' attractiveness on their parties' or combined lists' vote share, a simple simulation was finally carried out. Based on Model 4 from Table 2 it is calculated how the vote

Table 2 A Multi-Level Model of the Determinants of the Electoral Support for a Party or Combined List at the 2004 European Election. (Source: Own data)

Level-1-Variables Model 1 Model 2 Model 3 Model 4 Level-2-Variables

Fixed effects

Intercept Intercept

0.96

0.96

1.55

1.15

Number of parties in the constituency

0.10*

0.10*

0.09

0.10*

Attractiveness of front-runner Intercept

Subnational constituency Electoral duty

1.49**

1.49**

1.78**

0.49* 0.29

1.57**

0.41*

Barring clause

0.33

Preference voting is possible

0.18

Sex of front-runner: female

2.06

2.06

1.97

2.08

Age of front-runner: (rc = younger than 35 years) 35 years up to 44 years

1.26

1.26

1.40

1.34

45 years up to 54 years

55 years up to 64 years 65 years and older

Result at the last national general election Membership in the national government Anti-European program

Representation in the European parliament

2.32

3.32**

4.36**

0.71**

3.98** 1.21

2.08*

2.32

3.32**

4.36**

0.72**

3.98** 1.21

2.08*

2.53

3.44**

4.53**

0.71**

4.07** 1.57

2.29*

2.43

3.41**

4.45**

0.71**

4.06** 1.40

2.17*

Random effects

Variance of intercept

0.01

0.10

0.11

0.10

Variance of 'attractiveness of front-runner'-slope Deviance

3,441.32

0.02*

3,441.34

0.02*

3,437.53

0.02*

3,439.43

Number of parameters

14

16

20

17

Entries are unstandardized coefficients; unweighted Level 1-N= 527; unweighted Level 2-N= 64

rc reference category

*p < 0.05; **p < 0.01; ***p < 0.001

share of each voting option would have changed if its front-runner had been highly attractive (meaning that, in each case, candidate attractiveness is set at the observed maximum of 5.3 (Monika Benova) for the projection calculation). This results in an average vote share increase of 4.9 % points in national constituencies and 3.8 % points in sub-national constituencies. The maximum vote increase that occurs was 8.2 and 5.8 % points, respectively13. It should also be borne in mind that these vote gains are at the expense of other parties or combined lists and therefore the influence of front-runners' physical attractiveness on the election result is, without doubt, considerable.

When interpreting these findings, however, the following restriction must be kept in mind: Our multilevel model determines the effect of physical attractiveness while keeping constant all other independent variables. At the same time, our empirical analysis has shown that the physical attractiveness of the candidates is not independent of their age (Table 1): Attractive candidates tend to be younger than unattractive candidates. According to the results of our multilevel analysis younger candidates get fewer votes than older candidates (Table 2). The effect of physical attractiveness is therefore partially compensated by the effect of age.

 
< Zurück   INHALT   Weiter >