Over a month ago, I published a piece in response to a CNN article speculating that votes for Gary Johnson and Jill Stein cost Hillary Clinton the election. My earlier analysis showed that CNN’s hypothesis was badly flawed; since that time, however, the election cycle got even crazier as Jill Stein – yes, JILL STEIN – raised sufficient funds for recounts in multiple states!
With many more votes now in and the recounts essentially done, I decided to repeat my analysis with the updated numbers. Has anything changed?
My earlier analysis used election results as of November 12th; this analysis uses the results as of December 20th as shown in The New York Times. Let’s take a second look at my hypothetical scenarios.
[1. What if all Jill Stein voters voted for Clinton?]
My earlier analysis had two states (Michigan and Wisconsin) flip for Clinton and Trump maintaining a 280-258 electoral edge. With the new numbers, however, Pennsylvania also flips for Clinton – by just 4,243 votes – giving her the edge, 278-260!
The plot thickens.
[2. What if all Gary Johnson voters voted for Clinton?]
My earlier analysis showed that this scenario agreed with the CNN hypothesis (FL, MI, PA, and WI flipping for Clinton); with the new numbers, Arizona also flips for Clinton, giving her a 318-220 edge.
[3. What if all Jill Stein voters and all Gary Johnson voters voted for Clinton?]
Nothing changed here; under this scenario it’s still 318-220 Clinton.
[4. What if Clinton received all 3rd-party votes?]
My earlier analysis had AZ, FL, MI, PA, UT, and WI flipping, giving Clinton a 324-214 edge; with the updated numbers, she adds NC for a 339-199 edge.
So does this mean 3rd-party votes cost Clinton after all? The answer is still NO!
Recall my earlier analysis, where I developed a method to estimate the proportionate share of 3rd-party votes Trump and Clinton would each receive, using exit poll data. These are the updated proportionate shares:
[Fig. 1] Exit poll data with updated 3rd-party shares; the data of interest are in the last six columns of this table.
[Fig. 2] Results after adding proportionate share of 3rd-party votes to Clinton and Trump’s updated state-by-state totals.
That’s right – TRUMP STILL WINS by the same electoral margin! Mind you, this is in spite of this model extending Clinton’s popular vote lead by 269,585 (the earlier analysis extended her lead by roughly 100,000).
Next, let’s reexamine the modified CNN hypothesis using the updated numbers; we give Clinton all of Jill Stein’s votes and allocate proportionate shares of all other 3rd-party votes:
[Fig. 3] Results after modifying the table in Fig. 2 to give Clinton all of Jill Stein’s votes.
Trump no longer hangs on! My earlier analysis had Trump maintaining a 290-248 electoral edge; here, Clinton flips the script by just a few thousand votes in Pennsylvania and Wisconsin for a 278-260 edge!
The CNN hypothesis has grown some teeth; however, it still has no legs. Recall the difference-of-means hypothesis tests I ran in my earlier analysis (I ran one set for defectors and another set for Independents). For defectors, it appeared more registered Democrats voted for Trump than vice-versa. Although the evidence for defectors was inconclusive, we saw this paradigm play out in the Electoral College (which honestly surprised me). The evidence for Independents leaning Republican, however, was much stronger – meaning that 3rd-party votes likely hurt Trump more than they did Clinton.
In closing, we can all speculate on what could’ve or should’ve happened (and for stats geeks like me, we can have fun with this!), but the fact remains the Democrats ran a weak candidate and paid the price. Y’all still should’ve Felt The Bern.
N.B.: You will notice asterisks next to some states in the tables above:
- One asterisk: State with at least 99% (but less than 100%) of precincts reporting as of December 20th,
- Two asterisks: State with at least 90% (but less than 99%) of precincts reporting as of December 20th, and
- Three asterisks: For Maine, which splits its electoral votes; in this analysis, the hypothetical electoral counts assume that the 3/1 electoral split between Clinton and Trump, respectively, still holds.