The peril of too much data - US primaries

The peril of too much data - US primaries

TLDR: Separate the noise from the signal. But in this information overload world noise is far more prevalent than signal. The result in our industry (macro volatility management), is often bad decisions or decision paralysis. The key to avoid this outcome is not to gather more data. It is to use the existing data better. 



What we’re talking about: It is time to reflect on the lessons of the 2020 primary. Anybody following closely would have experienced one of the most dramatic and volatile races in our lifetimes. Endless sources of analysis. But they, and admittedly we, were probably following a little too closely.


Outside of a promising but short-lived Warren surge, there were only three outcomes worth contemplating: one: a Biden win, two: a Bernie win, three: a contested convention. Countless words were spilt on the Butti-boom, the Klobu-charge, and the…Berg-bloom? But the fact is this was never a very volatile race. 


Despite a historically large and diverse field, new caucus rules, and new magnitudes of spending, anyone who followed the blog posts of forecaster Nate Silver and his FiveThirtyEight site (we are big fans) would not have been surprised at the outcome. 


So despite incredible performances from Buttigieg in the early states, FiveThirtyEight never raised his chances over 5% even after his Iowa win, knowing that candidates without higher national polling support just cannot surge enough to win from his position. They similarly downplayed Bloomberg’s chances, and never gave Amy a hope. 



Lessons to learn: The reason forecasters at FiveThirtyEight are so effective is not because they know more than others – it’s because they use evidence better. They think in terms of “Bayesian priors” – what would I believe without any new evidence? Only on top of that framing do they decide if new evidence is strong enough to knock them off course. In the primary, there is a strong “prior” that you cannot win the Democratic primary without African American support. Should winning Iowa knock you away from that strong prior? No it shouldn’t, and FiveThirtyEight was correct in never giving Pete a chance. They have done similarly well with other parts of the race – when it narrowed to Biden and Bernie after Super Tuesday, should it be considered a close race because it’s down to two candidates and they both have strong bases of support? No, given the demographics of the remaining states and the polling. FiveThirtyEight quickly raised Biden’s odds to 99%. 



But you’d never have known that reading the news (or listening to the indefatigable Bernie supporters) – you’d have thought all along that this was a volatile and unpredictable race – and that Bernie could make a comeback at any point. And sure – that’s a more exciting story. And lots of unexpected things happened. But it wasn’t unpredictable. 


The buzz around individual candidates can still tell us a lot of the state of American politics. Socialism and homosexuality are not the show stoppers they once were, while being a woman might still be. But it told us very little of the actual outcome. 


This is not to say that other outcomes were not possible. A bombshell revelation (looking at you, FBI Director), a heart attack (gerontocracy is after all the new normal), or a phishing attack on a campaign manager (wash your hands and use dual factor authentication) can upend the best forecasts. It is just not the likely outcome. 


All this is to say that the news is generally unhelpful in separating signal from noise. The media’s incentives are always to make things seem maximally exciting, in other words favor noise at the expense of the big picture. In many ways, political risk research firms often create the same problem for clients – analyzing the minutiae of Kazakh politics like it’s the most important issue in the world only add to the noise but doesn’t move the needle in real ways. 


Remembering this – that in reality, only a finite number of issues really matter and few outcomes are really probable - will make you more effective in planning amidst macro volatility.



Post Script

Does this mean we refute the existence of “black swans”? Not at all; be sure to check out our article of grey swans here


Strategic planning

Core Principles