The county is awash in hundreds, if not thousands, of election analyses.
Be careful which you cling to. Many of the tools being employed are laden with problems and pitfalls and much of the data is subject to multiple interpretations.
For example, increasingly I see “analysts” trying to divine how segments of the population voted based on county and sometimes community returns which are regressed on, graphed against, or cross-tabulated by demographic data.
Statisticians have a name for trying to derive conclusions about individual behavior from group level data like county and precinct returns. It’s called the ecological fallacy. As you can tell from that second word, the method is fraught with analytic danger.
Simply put, you cannot assess how groups of individuals changed their voting behavior by looking at data from counties or even precincts unless the group in question makes up almost all of the county or precinct, or you employ some very complex and still controversial math that approximately none of these analysts are actually using.
That is, you can’t tell how Latinos, African Americans, Jews, non-college whites or any other segment changed their vote by looking at county or precinct data unless that group makes up almost the whole of that county or precinct.
And you can’t generalize from those places to the voting behavior of members of that group in the country at large, because people who live only with people like them are different from group members who live in more highly integrated places.
Take 2004 for example. George W. Bush won the country’s 15 poorest states, while Democrat John Kerry, won nine of the 11 wealthiest states.
Did that mean the rich were voting for Kerry and the poor for Bush?
Not then. Sixty-two percent of those who earned over $200,000 a year voted for Bush, but only 36 percent of those who made $15,000 or less cast ballots for him.
Yet, I’ve seen analyses purporting to demonstrate that this or that group moved toward or away from a candidate based on changes in counties and communities where the segment under discussion comprises just 10, 20 or 30 percent of the population.
Nonetheless, now that all the votes have been counted, we can extract some lessons by examining the results, though we need to put those facts in context and carefully consider alternative explanations.
First and foremost, Donald Trump’s claims notwithstanding, he did not win a landslide popular vote victory. His popular vote margin is now less than 1.5 percentage points. That’s the 10th lowest in history. Nineteen presidential elections produced true landslides of 10 points or more. Trump was not among them.
Second, while overall turnout declined by 3,606,356 votes or 2.3 percent, the results were hardly uniform. In the seven swing states, turnout actually rose 2.8 percent, while it declined by 3.5 percent in the rest of the country.
Those results are consistent with several theories. It is possible that being repeatedly told by the media that your vote doesn’t count because you don’t live in a swing state depresses turnout in those locales.
It is also possible that all the campaign ads, candidate visits, direct mail, texts, canvases and other activity that flooded swing states — along with the sense that you could be decisive — actually stimulates turnout.
What about votes for individual candidates?
The total number of votes cast for Harris declined by 9.5 percentage points compared to Biden’s 2020 performance in non-swing states, while Trump’s total increased by 3.5 points.
In the all-important swing states, however, the picture was different. Harris’ vote was just three tenths of a point lower than Biden’s, while Trump increased his support by 6.3 percent.
These facts too are broadly consistent with several competing theories. It may be that Harris largely held the Biden coalition together in in swing states while running a campaign that ostentatiously ignored the rest of the country hurt her there. Or Trump may have found new voters while Harris relied on those with a history of turning out.
Voter file analysis, which won’t be available for some time, will help sort through some of these questions.
That Harris did better in the swing states may testify to the relative effectiveness of her campaign tactics where they were deployed. Though obviously they were not effective enough to win, or even to improve on Biden’s 2020 vote total. (See my post-election column on the fundamentals)
The fact that both candidates did better in swing states than in non-swings may suggest that intense campaign activities do matter.
Or it may be that swing states are different for reasons having little to do with campaigning.
Whatever your favorite theory, be cautious about the data, method and logic you use to support it.
Mark Mellman is a pollster and president of The Mellman Group, a political consultancy. He is also president of Democratic Majority for Israel.
