Late Night Thoughts on Roe v Wade, Black Swans, Elections, and Uncertainty

I normally let these things sit for a while and run them past a stat professor or two to check my logic and language, but this is a fast-moving story and I wanted to make some points before the experts weighed in.

Dictated to my phone, so beware the occasional homonym. 

1. All predictive models are based to some degree on the assumption that the underlying relationships that held before will continue to hold in the future.

1b. All causal relationships we observe or, more often, fail to observe depend on the ranges of the variables in question. You will find no relationship between age and blood pressure if your study is based on a group of college students.

When we have reason to believe that underlying relationships in changed, particularly when factors have moved out of previously observed ranges, you should be prepared to discard deeply held assumptions about what is likely and what is unlikely.

2. Also remember that most of those nice, straightforward relationships we throw around when explaining and predicting are at best oversimplifications and frequently useful fictions. We constantly gloss over complexity and complication in order to get a clear, concise explanation. Think using a simple linear model when an enormous Bayesian network would be more appropriate. Under normal circumstances, we can usually get away with this -- It may even be the best approach -- but when profound changes are happening these oversimplifications can lead to disastrously wrong conclusions and recommendations.

3. We also have a tendency to put far too much faith in what we have come to think of as fixed points. At least in the social sciences, there are very few values, thresholds, or relationships that we can treat as immutable laws of nature.

The main lesson here is that when we get into uncharted territory, no one can be certain of anything including the pundits and even researchers. 

My gut feeling is that the upcoming SCOTUS ruling is horribly dangerous for the Republican Party. The GOP has staked out a position as the anti-Roe, pro-Putin, anti-vax, pro-insurrection party. On top of that, the governors of their two most populous states have in effect declared war on their own states' economies. My instincts tell me that under these circumstances, bringing any of these issues front and center in the discourse is likely to hurt the Republicans.

But that's just a guess. 

I don't know what I'm talking about, not because this is way out of my field (which it is) and not because I am unfamiliar with the data (though I am, in fact, almost completely ignorant of it*). I don't know what I'm talking about when I make predictions about the upcoming election because nobody knows what they're talking about when they make predictions about the upcoming election. 

We are so far beyond the range of data that no statement can be made with any confidence, not just with respect to Roe v Wade but along a number of other dimensions, the extremism of the rhetoric, the level and precision of gerrymandering, the role of conspiracy theories, the infiltration (sorry, there's no better word) of one party by a hostile power, revelations of coups past and future, and this is not a comprehensive list. 

There are limited sorta-kinda parallels from history books and interesting, perhaps even informative findings from polls and previous elections, but none of it is science, at least not yet. A year from now, you will see some first-rate research coming from smart political scientists that will give us real insight into what is and is about to happen, but for now, no one is making any kind of statistically solid predictive analysis about November. 

*Yes, I know, but the plural just sounds stupid.