Data analysis: More Cautions than Conclusions

Here’s a simple chart. I had fun making it, and am proud of my shrewd code and above-average-but-not-stellar visual design. There’s even a quick message: annual bond returns fair slightly better under Republican presidents than Democratic presidents, but vice-versa for S&P 500 returns.

Done deal, right? Not so fast.

It’s foolish to think markets are exclusively driven by real-time political policy. I don’t know the exact decomposition, but it’s safe to say that many dynamics beyond the political party of the president affect market returns: the political party of the previous administration, the administration before that, the one before that… Not to mention factors like foreign politics, international supply and demand, and influenza pandemics, to name a few.

I invite you to ponder other analyses that would paint an insightful picture. Personally, I’d like to know the temporal correlation: how does the previous administration’s party affect the current return? The party before that? I also would like to know how returns change with the tenure of the president: does their last year bring better markets than the first? Oh, another! Instead of categorizing by political party, what about regressing on age? Education level?

The list goes on, but splicing on so few data points isn’t the only problem here. The problem is ignoring the plethora of factors that affect the incredibly complex financial markets beyond presidential political party. As such, I advise against casting votes based on my thin-splicing. But please, appreciate my code.

You can find more data analysis on my website, michaelboerman.com.