Predicting Political Involvement through Demographics, Overall Involvement, and Political Interest

A series of regression analyses were run to examine the relationship between political involvement, political perception and preference, age, gender, fascination with politics, strength of feelings about politics, number of days per week searching out political information (radio, television, internet and newspapers), political interest, partisan strength, non-political club membership, contribution to clubs, time available for clubs or leadership, and religious service attendance frequency. Table 1 shows the regression weights for the various models. The full model had an R²=0.558, F(13,326)=31.595, and p<0.001. The following variables had significant regression weights: age, political fascination, number of days per week getting politically informed, political interest and non-political club membership had significant regression weights. The number of days per week getting politically informed had the largest individual contribution to the full model.

The first research hypothesis was that a model including just demographics variables (political perception, age and gender) would not perform as well as the full model. This reduced model had an R²=0.158, F(4,326)=15.961, p<0.001. Age and gender had significant regression weights, while age had the largest contribution to the model. In support of the research hypothesis, this model did not perform as well as the full model, R²-change=0.4, F-change(9,326) =15.904, p<0.001, suggesting that political and general interests are necessary for predicting political involvement.

The second research hypothesis was that a model including just general involvement variables (being a member of a non-political club, being a contributing member, time available for clubs or leadership and frequency of religious service attendance) would perform just as well as the full model. This reduced model had an R²=0.079, F(4,335)=7.164, p<0.001. Club membership and contribution had significant regression weights; membership had the largest contribution. Contrary to the research hypothesis, this reduced model did not perform as well as the full model, R²change=0.479, F-change(9,326)=39.254, p<0.001.

The third research hypothesis was that a model including just political interest variables (political fascination, strength of feelings about politics, number of days per week searching out political information (radio, television, internet and newspapers), political interest and partisan strength) would perform as well as the full model. This reduced model had an R²=0.511, F(5,334)=69.821, p<0.001. Contrary to the research hypothesis, the reduced model did not perform as well as the full model, R²-change=0.047, F-change(8,326) =4.333, p<0.001.

The fourth research hypothesis was that political interest model would be able to predict better than the other predictors, and that the strongest single predictor of political involvement would be a part of the political interest model. This research hypothesis was supported-the political involvement model had an R²=0.511, and the other reduced models had R²=0.158 (general involvement) and R²=0.079 (political interest). As seen in Table 1, number of days per week getting politically informed had the strongest unique contribution to involvement in the full model, β=0.491, p<0.001, and is in the political involvement model.

Finally, the predictive utility of the three reduced models was compared, using Steiger’s Z-test. The correlation between the demographic model and the political interest model was r=0.353, p<0.001, and political interest accounted for larger amounts of variance among political involvement Z=6.866, p<0.001. The correlation between the demographic model and the general involvement model was r=0.148, p=0.006, and accounted for equal amounts of variance among political involvement, Z=1.786, p=0.074. The correlation between the general involvement model and political interest model was r=0.204, p<0.001 and political interest accounted for larger amounts of variance among political involvement, Z=9.133, p<0.001.