Gut rapid pattern recognition subverts rational assessment

This is a great example of the care you have to have in balancing gut instinct pattern recognition and closer, more precise analytical consideration.

The example starts with this tweet from Pew.

The American middle class is shrinking in most metropolitan areas. 1 of 16 striking findings of 2016. See the rest: https://t.co/IxNsruVGoE pic.twitter.com/qdYECjrvGH

— Pew Research Center (@pewresearch) December 21, 2016

My first instinct was "Wow, that looks like the 2016 election results." And it does.

Click to enlarge

Surfing from that instantaneous spark of recognition, you might then conclude that those areas where there was the greatest shrinkage of the middle class are the areas where there was the greatest support for the Democratic candidate for President. All very reasonable.

But then you have to override your gut pattern recognition and ask yourself some more probing questions and deal with actual facts.

Here are a series of facts that have to be fit to the maps.

Democrats basically won the urban core (see the red-blue map) and urban cores are the areas where there is the greatest inequality. Perhaps the issue is not shrinkage of the middle class but rather the rise of inequality.

The middle class has been shrinking since the late 1970s. But the cause of the shrinkage is interesting. Some people have indeed moved down and out of the middle class. But many more have moved up into the upper-middle and into the wealthy according to Pew (in a different report).

If the rich are more conservative, then a shrinkage of the middle class that is occurring because people are getting richer should mean that those areas with the greatest shrinkage ought to be aligning with Republicans, but that is not the case. That supports the supposition that it is not about middle class shrinkage but about urban inequality.

Finally, while the maps look near identical at first glance, the longer you look at the details the more you see discrepancies. Alaska and Hawaii both have little middle class shrinkage but both are, respectively, solidly red and blue. The fit for Ohio, Iowa, Missouri and others also looks fairly spotty. I would guess that there is a correlation between middle class shrinkage and party dominance but that the r-value would be pretty low.

Instantaneous pattern recognition sees a high degree of correlation but measured and considered observation and awareness of context suggests that there is probably, in fact, quite a low correlation between voting affiliation and shrinkage of the middle class.