A political scientist wanted to account for the percent of the total vote collected by various candidates for the U.S. House of Representatives. She gathered data on Congressional contests in 2010. Her data include the following variables:
~VOTE: Percent of the total vote received (0 – 100).
~INCUMB: Coded 0 for non-incumbents, and 1 for incumbents.
~SPENT: Candidate’s campaign expenditures (in thousands of dollars).
~OCCUP: Candidate’s occupation prior to candidacy, coded 1 = elected public official, 2 = appointed public official, 3 = lawyer, and 4 = other. (Incumbents were coded 1).
Table 1: Predicting Vote Share in the 1990 Congressional Races
Variable Coefficient S.E.
Intercept 25.4 4.1
INCUMB 12.0 4.3
SPENT 1.2 0.5
OCCUP -1.5 1.3
***N = 714
***R-Squared = 0.42
She regressed VOTE on the three predictors and got the results shown in table 1. Based on this analysis she reached the following conclusions. Correct any errors of interpretation:
A) The vote share of incumbents is predicted to be larger than the vote share of non- incumbents by 12 percentage points, even if the non-incumbents outspend the incumbents.
B) The t-statistic on OCCUP is non-significant, so prior occupation is probably unimportant as a determinant of VOTE.
C) If a candidate for the U.S. House of Representatives spent no money at all; he or she would be expected to win 25.4% of the vote anyway.
D) Since the coefficient on INCUMB is significant at less than the .01 level, there is less than a .01 chance that the hypothesis of no effect is true.
E) Incumbency is roughly ten times as important as campaign expenditures.
F) The estimated equation predicts victory for incumbents spending over $12,000.
G) The equation accounts for the vote totals of 42 percent of the candidates.