Are There Shy Trump Voters?
Something that keeps me up at night (literally I was working on this when I should have been sleeping) is the shy Trump voter theory. That Trump voters are ashamed to say they are afraid of Trump because liberals are so very scary. So, I did some research to see if that holds up. I analyzed the polling in the 16 closest states from 2016 (except Minnesota where there wasn't sufficient data), comparing what the final polls predicted to the ultimate results, how many undecided voters there were, how far off they were, and how much they understated or overstated each candidate's support. The results look like this...
State
|
RCP Average
|
Actual Results
|
Error
|
Total Decided (2 way)
|
Clinton RCP Average
|
Clinton Actual
|
Clinton Error
|
Trump RCP Average
|
Trump Actual
|
Trump Error
|
OH
|
T- 2.2
|
T• 8.1
|
C+ 7.9
|
90.2
|
44.0
|
43.8
|
- .2
|
46.2
|
51.3
|
5.1
|
IA
|
T• 3.0
|
T• 9.5
|
C+ 6.5
|
85.6
|
41.3
|
41.7
|
.4
|
44.3
|
51.2
|
7.9
|
WI
|
C- 7.5
|
T• 0.7
|
C+- 7.2
|
87.1
|
46.8
|
46.5
|
-.3
|
40.3
|
47.2
|
6.9
|
MI
|
C• 3.6
|
T• 0.3
|
C+• 3.9
|
90.4
|
47
|
47
|
0
|
43.4
|
47.3
|
3.9
|
NM
|
C• 5.0
|
C• 8.3
|
T+3.3
|
85.6
|
45.3
|
48.3
|
3
|
40.3
|
40
|
.3
|
NC
|
T• 0.8
|
T• 3.6
|
C+ 2.8
|
92.2
|
45.7
|
46.2
|
.5
|
46.5
|
49.8
|
3.3
|
PA
|
C• 2.1
|
T• 0.7
|
C+2.8
|
91.5
|
46.8
|
47.5
|
.3
|
44.7
|
48.2
|
3.5
|
TX
|
T+11.7
|
T+9
|
T+ 2.7
|
88.3
|
38.3
|
43.2
|
4.9
|
50
|
52.2
|
2.2
|
ME
|
C• 5.5
|
C• 2.9
|
C+ 2.6
|
87.5
|
46.5
|
47.8
|
1.3
|
41.0
|
44.9
|
3.9
|
NV
|
T• .8
|
C• 2.4
|
T+ 2.4
|
91.6
|
45.4
|
47.9
|
2.5
|
46.2
|
45.4
|
- .8
|
CO
|
C• 3.0
|
C• 4.9
|
T+1.9
|
85
|
44
|
48.2
|
4.2
|
41.0
|
43.3
|
2.3
|
FL
|
T• 0.4
|
T• 1.2
|
C+.8
|
93.6
|
46.6
|
47.4
|
.8
|
47
|
48.6
|
1.6
|
GA
|
T• 4.8
|
T• 5.1
|
C+ .3
|
95.4
|
44.8
|
45.4
|
.6
|
49.6
|
50.5
|
.3
|
AZ
|
T• 4.0
|
T• 3.5
|
T+.5
|
90
|
43
|
44.6
|
1.6
|
47
|
48.1
|
1.1
|
VA
|
C• 5.3
|
C• 5.4
|
C+.1
|
91.3
|
48.3
|
49.8
|
1.5
|
43
|
44.4
|
1.4
|
NH
|
C• 0.3
|
C• 0.3
|
0
|
87.1
|
43.7
|
46.8
|
3.1
|
43.4
|
46.5
|
3.1
|
Average Error
|
|
C+1.5
|
2.856
|
89.5
|
|
|
-1.48
|
|
|
-2.875
|
Here are what I think are the important conclusions we can draw.
1. It's hard to argue there were a ton of shy Trump voters influencing the results. Of the polls in the 16 states, 9 of them either had the race accurate to within a percentage or actually overestimated Trump. In 13 of the 16 states, polling was in the margin of error. Only in 3 states was Trump underestimated by a statistically significant amount.
If you want to argue for shy Trump voters, you'd have to explain why they only seemed to be shy in the rust belt. Why were they shy in Ohio, but not VA, Michigan, but not New Hampshire, or Wisconsin, but not Maine? The theory doesn't quite jive.
2. Even if there were shy Trump supporters there were not lying Trump supporters. There may have been some embarrassed folks who said they were undecided when they were actually determined to vote Trump, but there were not voters who said they were voting for Clinton who wound up voting Trump.
In virtually every state the actual percentage of Clinton voters was higher than the polling average predicted, which is expected as third parties decide. On average Clinton did 1.48% better than the final polls, and Trump did 2.875%.
Meaning that of the undecided voters, about 2/3 ultimately decided on Trump. Is this because they were actually Trump supporters all along and were just so darn bashful? Or is it because of things like the Comey letter coming at the 11th hour. Can't really say.
What does this mean for 2020?
If these patterns hold up in 2020, then Biden wins.
Take Pennsylvania for instance. If the RCP average stays stagnant till election day, Biden has 49.5% of the vote, and Trump has 45.9. This means 95.4% of voters are decided on one of the two major parties, and 4.6% are either undecided or voting for a third party.
If all of those undecided voters vote for Trump, he ties. If .1% of them go to Joe Biden, Jo Jorgenson, or any other candidate, then Biden wins.
In Michigan, Biden has 50% of the vote. Trump has 43.5%. Again, Trump needs every undecided voter, and 0% third party support if, as in 2016, Biden is not likely to get less than his current RCP average.
TL:DR Version
If there were any "shy Trump" voters in 2016, they reported themselves as undecided. The number of undecided voters this time around is, in most states, is too small to make a difference. Even if 90% of them are actually shy Trump supporters, that won't be enough to win.
This time around, for Trump to win, there would actually have to be voters who are lyin' about Biden. We didn't see anything like that in 2016, so it's unlikely we'll see it now.