Opinion palls

 

Last night I went to bed expecting a Clinton victory, because although opinion polls can be wrong, the margin of Democrat advantage exceeded the apparent margin of polling error. To further confirm my ineptitude at forecasting, I had also thought Remain would win over Leave in the Brexit referendum, for the same reason.

Today I reflect on some basic psychology, which is that attitudes do not translate directly into actions and, in a further complication, expressed attitudes differ from actual attitudes. This is particularly the case where the attitude is considered socially reprehensible, such as coveting one’s neighbours ox, or wife; being pleased when a rival fails to get a promotion; resenting a newcomer; or disliking someone because a superficial characteristic grates on your nerves.

Of course, I have always known, but often forgotten, that there is a gap between what is thought and said, and a bigger gap between what is said and done.

I also know, but often forget, that public opinion polls habitually use smaller samples than required for proper representativeness,  and that the pollsters then “correct” their raw results using a set of assumptions gained from trying to correct previous errors (turnout rate, for example) whilst also looking over their shoulders at what most other polls are saying, so as to not be the odd one out. Come to think of it, the pollsters’ fascination with finding out what other people are thinking infests the poll companies themselves, and they are guided by what other poll companies are thinking. Polls of polls simply aggregate and confirm the group errors.

Will attention continue to be paid to opinion polls, given their inability to spot what the public really intend to do? The growing awareness that polls are used to influence opinion as much as to measure it has become painfully obvious. They are not trustworthy, and flourish most when people cannot work out their own opinions for themselves. No, I do not want to ban them. I just want to get out of my habit of paying them much attention.

Now, here is a quick check of the predictions I made yesterday against Edison Research exit questionnaires:

Most of all I wonder how years of education, which supposedly indicates an ability to evaluate arguments, will correlate with actual votes, and thereby to test the popular supposition that the better educated voters will shun the Republican candidate. I assume that sex and age will have a minor influence, but the latter might show bigger differences, with older voters more cynical and more likely to vote Republican. Race should not matter at all, because if America is a melting pot then policies not polities should prevail. If races vote en bloc (say more than 65% in one direction) then the woe betide the republic, which will become disunited genetic states.

Years of education

 

Post-graduates voted Democrat, as expected. However, I did not consider income:

There was some effect of low income boosting Democrat votes, but those with income above $50,000 were not swayed by necessity, and voted according to political preferences. Perhaps the well-educated post graduates took the sorts of degrees which did not boost their income. Psychology, anyone?

Here are the sex differences:

In fact, much bigger than I imagined would be the case.

The age differences were also bigger, though I got it half right, in that older voters were more likely to vote Republican than those younger than 45, and particular than those younger than 30.

As regards race, you will see that I put forward a null hypothesis I did not believe in, but wanted to test. I used to follow the majority in describing the US as a melting pot, but have come to think of it as only a dispersal ground.

Black voters are overwhelmingly Democrat, and can be considered a bloc vote. Hispanic/Latinos and Asian also meet precisely my predicted bench mark of 65% to be considered bloc voters. The “other” voters look as if they have been somewhat considering the actual policies, as do the White voters, but both are not that far off the 65% boundary I plucked out of the air as an indicator of race-based voting. Blacks to an extraordinary degree, and Hispanics and Asians to a large degree jumped left, Whites jumped right.

So, is the summary of this election “it’s race, stupid”? If so, I also predicted a Disunited Genetic States. I doubt it. Governments habitually do less than they promise, and have less influence than they imagine. The great flywheel of habit usually reigns supreme.  Usually.  I will consider this matter further after a good night’s sleep.