Data in Place of Expertise

I usually have a book review on Thursdays, but I am still working through my latest read. Hopefully it will be done by next week. In place of a review, some thinking aloud.

This is more of a rumination than anything driving toward a specific point, but the election has pushed one of my hobbyhorses to the fore. In too many fields in our world, data has begun to replace expertise rather than augment it. The arguments about why Harris lost to Trump have sometimes shown pundits and commentators pushing data as a trump card rather than thinking about what the data means.

One of the truisms of the election is that country moved to the right, despite Harris earning about seven million votes less than Biden and Trump only picking up about 3 million from 2020. To argue that the country turned center right is to ignore the four million voters that did not vote, and why they did not vote. I don’t have a reason for the difference, just plausible theories.

Younger voters were upset by Gaza, and so perhaps they stayed home. Younger voters are also more cynical about the use of the government to correct problems and may have thought there was no point in voting. Maybe people felt Harris’ campaign was too corporate friendly. Maybe they thought it was too trans friendly. Maybe voter suppression was that much better this year. Maybe people in non-swing states simply don’t bother to vote anymore, especially since most of the country is gerrymandered to within an inch of its life. Maybe a combination of all these theories and some I am probably not remembering or thinking about. No one, at this stage, can say for certain.

But that obviously has not prevented people from opining on the meaning of that data, the story it tells. This is at least partially because certain pundits want to push a certain narrative to further their own political opinions. I find it silly that anyone who thinks a Harris campaign that cozied up to corporate leadership and made a big deal out of Liz Cheney campaigning with them was somehow leftists in nature, but one data point allows people to argue otherwise. In a sane world, we would wait for more data to explain the missing voters and the reasons people voted for a given candidate. But because we, and by we I mean in large part our media, have been trained to see data as an argument stopper, the clever use of number can often override more nuanced and correct positions.

Let’s take a slightly less contentious data point and see how this works in practice. Despite the overall vote shifting about six points to the right nationwide, the shift was much less pronounced in swing states, about three percent, where the campaigns were most active. This one data point can and has supported many interpretations. The first is that the Harris campaign was at least somewhat effective, as where it spent the most time and money trying to persuade and educate the electorate it did the best to limit the damage. Another argument is that, no, the Trump campaign was most effective in the swing states as it moved the electorate to its direction in states that it largely lost in 2020. A third argument is that nope, neither campaign mattered much in the swing states because people are motivated to vote in those states knowing how important they are to the election, while people outside those states cannot be bothered as much. Three very different theories all supported by one simple fact.

All of the theories are of course plausible, but all of them get complicated by other facts. For example, Trump did best with people who got their news from Fox, twitter, and their own friends and family. This suggests that the Harris campaign did do well in reaching many people, but that a certain subset of voters was unreachable. One additional fact, and two of the plausible narratives are undermined. I am not picking on those narratives, merely hoping to highlight how one data point does not a rationale make.

Or should not make. We should pay more attention to the social and political scientists who, over the next few weeks, will get more and better data that they can then combine with their expertise to provide us with better understanding of what actually happened. And that of course matters because it can inform how to prevent it from happening again. We should do this, but of course we are not.

The urgency of the situation combined with the inter-factional knifing that is a feature of every Democratic loss, no matter how slight, combined with the incentives of the data-knowers to elevate their numbers over more complicated, nuanced science of actual experts, drives people to short, simple, and above all fast reactions. Data, since we have all been taught over the last decade that data is unassailable, is a good way to win arguments. But data cannot replace expertise, merely help inform it.

If you see an expert dismissing data without good cause, then be wary. But if you see someone making an argument from data, especially from one data point, be even more wary. Life is not a data point, and not everything can be cleanly measured. If you look, you will often find that the data is confused, contradictory, and cannot be understood outside the context in which it was gathered. You need expertise for that, or at least a willingness to look at the entire picture, not just the spotlight your specific piece of data illuminates.

Anything less and you might as well be looking at goat entrails. You’d be just as wrong, but at least then you’d get some nice roast goat out of the exercise.