Question

Study C

Many political scientists predict election outcomes for fun.  It turns out that with a handful of variables, it is possible to estimate the two-party vote share of an election years out.  For instance, a model accounting for only GDP growth rate in the first quarter of the election year and whether the sitting President is running again will yield fairly accurate in its forecasts.  The table below presents the results of a linear regression model predicting two-party vote share (100 = incumbent party received all the votes; 0 = incumbent party received no votes; 50 = the two parties split them equally).  

Variable	Model #1	Model #2
Q1 GDP growth (percentage points)	2.0 (0.4)	1.0 (0.25)
Incumbent Running (0 = No; 1 = Yes)		6.0 (0.5)
Constant	48 (3.0)	45 (2.0)
N	18	18
R-sq	0.30	0.63

Numbers in parentheses represent standard errors.

According to Model #1, what is the relationship between first quarter GDP growth and two party Presidential vote share?

According to Model #2, what is the relationship between first quarter GDP growth and two party Presidential vote share?

How much more predictive of Presidential vote share is Model #2 than Model #1?

In 2016, the incumbent President was not running and first quarter GDP growth was 0.7 percentage points.  What was Hillary Clinton’s expected vote share according to Model #2?

What variables would you add to the analysis to predict vote share and why?

Answer 1

1. According to Model#1, for every 1 point rise in GDP vote share of Incumbent party rises by 2 points

2. According to Model #2, for every 1 point rise in GDP vote share of Incumbent party rises by 1 point

3.Model#1 predicts vote share by 30% (r-sq) whereas Model -2 by 63%

4. Model 2 : Incumbent vote share = const + b1* GDP +b2* incumbent running

Hillary Clinton vote share = 45 + 1*0.7 = 45.7

5. Factors like Corruption, employment growth, inflation control, Asset price growth etc , factors influencing voters decisions can be taken into account.