Some years ago, there was an odd feature of the scores on the SAT exam that students in the US take in large numbers and that many colleges use as one of their admission criteria. They found that the overall average SAT score was declining but that when the scores were disaggregated by ethnicity, the average scores of every subgroup (white, black, Hispanic, Asian) was rising.

How could that be?

The reason is because of a statistical oddity. The average scores of black and Hispanic students were lower than that of white and Asian students. But the numbers of black and Hispanic students taking the test was rising faster than those of white and Asian students, and hence when the overall average was computed, over time there was a rising number of lower scores entering the mix, that dominated over the fact that the subgroup scores were rising.

I recalled this when thinking about the current presidential election. Donald Trump has been steadily alienating large subgroups of the population with his remarks. He has denigrated women, Hispanics, blacks, Muslims, the disabled, young people, and immigrants, and his support among them is now pretty low. Even his support among Catholics has taken a dive recently, and his criticisms of pope Francis have not helped.

Given that, you would expect his overall poll numbers to be cratering. But they have not. He has stayed within around 5% or so below that of Hillary Clinton. How can that overall gap be relatively stable when his support among so many subgroups is falling? I suspect that it is because of the reverse of the SAT phenomenon. The increasing percentage of his favored demographic group who are saying that they will vote may be compensating for his losses elsewhere.

Trump’s basis of support is highest among older white men and this is still a large group, with them voting at the highest rates. Overall, Americans vote in low numbers, at rates of of less than 65% since the beginning of the 20th century. Older and white people vote at the highest rates of around 70% so there is still a large pool of untapped voters out there who might be receptive to his message. If he can energize that group and bring out more of them as likely voters, that rise can compensate for the losses he is experiencing among all the other subgroups.

This may explain his doubling down on his divisive rhetoric. He needs to really inflame people with the idea that things are really going to hell in the US and persuade those who may have not voted in the past to come out this time to save the nation from the hordes of immigrants who, thanks to president Obama and Hillary Clinton and the media and political establishments, are roaming the streets and murdering good, law-abiding, patriotic Americans.(Or at least save us from the prospect of having taco trucks at every street corner.)

On the other hand, he has to reassure them that they are not racists for supporting him and this explains his occasional forays such as visiting Mexico and black churches and saying that he employs many Hispanics and that he loves everybody and that everybody loves him. The contradictory nature of his campaign may be a feature not a bug.

It is not entirely a crazy strategy and it may be the only one available to him. Joshua Holland looks more closely into the makeup of the white electorate and especially the geographical divide and between those with more and less formal education and how each subgroup might respond to Trump’s message.

But as his efforts to fire up his supporters so that more of them decide to vote this time gain steam, we can expect to hear more frightening talk from him.