A Post Debate/Video "Trickle Down" Effect for Democrats in PA?

Correlation in the Gain in Facebook Followers and Poll Numbers for GOP Presidential Candidates in December

Last week I summarized the change in Facebook followers pre and post Presidential Debate for a sample of the candidates in Pennsylvania.  In the primaries I showed that there was a strong correlation between the change Facebook followings of the Presidential candidates and their poll numbers accounting for 71% of the variability in the poll numbers (see above graph).  

Office	Dem Gain	GOP Gain	Diff	Female candidate
Pres	2.68	0.41	2.27	Y
US Sen	0.81	0.26	0.55	Y
House12	0.32	0.06	0.25	Y
House9	0.18	0.32	-0.14	N
st sen 35	1.44	0.27	1.17	N
st sen 41	2.06	0	2.06	N
strep73	0.32	0	0.32	N
strep72	0.08	0.67	-0.59	Y
strep71	0	2.03	-2.03	N
Median	0.32	0.27	0.32
Mean	0.88	0.45	0.43
SD	0.97	0.63	1.33

This week I'm taking a look at what effect the Presidential debate and the vile Trump audio tape may have had on the down ballot races in terms of the gain in followers (above table) and in the change in the level of engagement (below table) for the period from Oct 7 to Oct 10.  The above table shows the % gain in followers for the period covered.  The above page shows a 2.68% increase for Clinton's PA Facebook page which was higher than Donald Trump's 0.41%.  The mean (or statistical average) shows a higher percent increase for the Democrats than the Republicans but there was higher variability in the Democratic numbers. The median gain for all the Democratic candidates was 0.32% compared to 0.27% for the Republicans.  This difference was not statistically significant.

Office	Dem Engaged	GOP Engaged	Diff	Female Candidate
Pres	26.81	1.37	25.44	Y
US Sen	5.29	1.86	3.44	Y
US House12	3.51	0.45	3.06	Y
US House 9	2.81	3.80	-0.98	N
State sen 35	21.88	8.40	13.47	N
State sen 41	5.31	0.42	4.89	N
State rep73	3.03	0.34	2.69	N
State rep72	3.04	0.67	2.37	Y
State rep71	1.49	0.66	0.82	N
Median	3.51	0.67	3.06
Mean	8.13	2.00	6.13
Std Dev	8.817001	2.491033	7.816291

 
For the percent engaged (a measure of enthusiasm by the followers on Facebook), there was a higher mean (8.13%) and median (3.51%) engaged for the Democrats than for the Republicans (2.00% and 0.67% respectively) with higher variability again for the Democrats. The Wilcoxon Signed Rank test for the difference in medians was significant (p=0.014).  This shows more engagement for the Democrats.  The only race where the Republican had more engagement was the US House 9th District race.

Box Plot Showing Median Engagement by Party

Finally I looked at the % engagement and % gain for races with a female candidate in the race.  There was no effect for % engaged or % gain if one looks at the median for both groups.  There was a higher mean for the races with a female candidate because of high engagement for Hillary Clinton.  I could not test for an interaction between the gain and engagement and the presence of a female candidate because of unbalanced data and unequal standard deviations.  There were three races with a female Democrat and one with a female Republican (the state rep 72nd). 

		female candidate	male candidate
% Gain	Mean	0.85	0.28
	Std Dev	1.20	1.54
	Median	0.40	0.32
% Engaged	Mean	8.58	4.18
	Std Dev	11.25	5.64
	Median	3.25	2.70

 
I have found some evidence that the enthusiasm after the 2nd debate but not necessarily for a gain in followers (but no loss either).  For the Presidential race in PA, the correlation with the Real Clear Politics Poll Average was 0.67 for the post debate engagement.  This means that 45% of the variability in the poll average is accounted for by the post debate engagement.  For election day I will look at how the Facebook numbers will change by election day and whether it will correspond with how they vote.  
 

**Related Posts**

Pat Toomey: In More Than a 30 Second Ad

 

Change in Facebook Followings for PA Candidates Before and After the Debate

 

The Facebook and Twitter General Election 

 

Facebook Primary 2016, August Update, Does it Predict Support?