Uncertain Politics—A Case for Modesty in Predictions?
In the opening months of his presidency, President Joe Biden unveiled the central planks of Build Back Better, a $4.5 trillion proposal designed to remake American infrastructure and transform the social safety net. Fresh off the smooth passage of $1.9 trillion in COVID-19 relief, some predicted Democrats would unify to pass a budget reconciliation bill containing the proposals before fall. Others were skeptical that moderate Democrats would allow anything to pass in slim majorities. Negotiations on infrastructure failed with one group of Republicans, succeeded with another in June, and ended up as a $550 billion increase to the baseline highway bill. In November, the legislation would pass the House and the president signed it into law. Democrats loaded a $3.5 trillion reconciliation bill with the controversial social spending provisions, before Senator Joe Manchin (D-WV) killed the bill in December and it returned from the dead seven months later as the climate-centered “Inflation Reduction Act.”
In the end, neither camp was right. Given such a tumultuous process, how could they have been? Various evolving factors played a role in the eventual fate of Build Back Better, including the legislative capacity of razor-thin Democratic majorities, a rapidly evolving macroeconomic outlook, and an uncertain incentive structure for GOP lawmakers. When a wide range of outcomes appear plausible, would-be prognosticators should do better to acknowledge uncertainty and resist the temptation to essentialize complicated dynamics.
Before the 2016 election, media outlets reported that a Trump victory was essentially impossible before pivoting to publish articles afterward making it seem obvious in retrospect. While always possible, Trump’s victory does not does not discredit the signs pointing to a Clinton victory. Clinton enjoyed a lead in the polls, albeit one small enough that Trump winning the election by 77,744 votes in three states was highly plausible. Statistically robust models like FiveThirtyEight quantified the possibility as a 71% win probability for Clinton.
Political scientists have coined the term “calcification” to describe a political landscape in which most people are unwavering in political allegiance, but small factors can still shift elections due to razor-thin margins. As such, it is plausible to imagine a scenario in which Clinton could have won Pennsylvania if not for the Comey letter or in which Trump won Minnesota in the absence of the Access Hollywood tape. Even if certain factors appear to have been particularly influential after the election, the identity and relative influence of all of these factors are fundamentally unknowable before the fact. With a wide range of possible outcomes, humble predictions that emphasize a range of possibilities, like the FiveThirtyEight model, are the most reliable in the absence of fortune telling.
Excessive certainty extends to last winter’s Speaker of the House election, where pundits changed their priors based on every twist and turn. If McCarthy lost another vote, he was doomed and a bipartisan coalition was on its way. If he announced a new suite of concessions, he had it locked up and always had things under control. This phenomenon is best encapsulated by a tweet by Bracewell strategist Liam Donovan, who complained about those stating that “McCarthy is dead and Scalise or anybody else can’t get 218 and this will go on forever or end in Dem/unity coalition… who until a week ago thought this was in the bag.”
It can be tempting to assume that conflict in the public eye is theater that obscures some grand plan concocted by a few actors who know how all will end. Such a perception can work to the benefit of leadership. During the aforementioned Build Back Better saga, political pundits assumed that Pelosi had the votes for the bipartisan infrastructure bill because she set a deadline for its House passage. However, “fake-it-till-you-make-it” only gets one so far: deadlines came and went twice because there was no plan. Similarly, fifteen rounds of voting for Speaker indicate that there was no prearranged deal. McCarthy stated he would be Speaker, while the holdouts stated that they would never allow him to become speaker. Outward-facing inflexibility made believing one side or the other tempting, but these statements only show that things were bound to get messy. If McCarthy and the holdouts did not know what would happen, how could an observer?
Even political outcomes that appear highly likely still demand some degree of uncertainty in their prediction. With the exception of leaked opinions, the closed nature of judicial deliberation means that Supreme Court decisions are fundamentally unknowable. Oral arguments, previous decisions, and the statistical modeling of the ideological makeup of the court can be useful predictors. Still, observers are not privy to the fundamentally unknowable internal deliberations of the justices. Are justices partisan hacks using raw power to advance an ideological project or impartial legal scholars with differing yet respectable interpretations of the law? Taking either position as a rule breaks down with curveballs like Bostock v. Clayton County, a 6-3 landmark decision protecting employees from being fired due to their sexual orientation or gender identity. The occurrence of surprising progressive decisions such as Bostock, however rare, breaks down an uncompromising theory about judicial ideology.
Such cases are rare–judicial models typically effectively predict the general bent of a decision, such as the undermining of abortion rights in Dobbs v. Jackson Women's Health Organization. Even so, specificities within decisions can be challenging to predict accurately. Whether the Court only allowed Mississippi’s law to stand or struck down Roe v. Wade entirely in Dobbs is a distinction of tremendous consequence. While an outcome may be likely, incorporating some degree of uncertainty in its specifics is often merited.
Powerful statistical tools and abundant evidence allow for decent predictions, provided that the uncertainty of what is unknown is incorporated. Depending on the situation and available evidence, the level of uncertainty may be so high that an election is a coin flip, or low enough that a bill will probably pass intact. Prediction is consistently more useful and less sensationalizing when a dose of humility is included with an acknowledgment of the range of uncertainty inherent in the outcome of interest. There is a middle ground available, with “probably” or “possibly” instead of “definitely” or “obviously.” Upcoming political outcomes–the budget fight, the 2024 elections, upcoming Supreme Court decisions, and even the medium-term viability of our democracy–stand to have massive consequences for American society. Readiness will hinge on how well people can take a step back and acknowledge that sometimes, no one knows.