I can identify (at least in principle) with the economist Steven D. Levitt, whom his Freakonomics co-author describes as “an intuitionist.” Dubner continues, “He sifts through a pile of data to find a story that no one else has found. He figures out a way to measure an effect that veteran economists had declared unmeasurable” (Dubner & Levitt, An Explanatory Note). Like Einstein, Levitt sees an objective reality under the subjective experience each of us brings to a discussion, “the belief that the modern world, despite a surfeit of obfuscation, complication, and downright deceit, is not impenetrable, is not unknowable, and—if the right questions are asked—is even more intriguing than we think. All it takes is a new way of looking” (Ibid).
Still, for me at least (much more so than for Dubner & Levitt), economics (and numbers in general) remains a wholly moral endeavor. After all, as even Dubner and Levitt points out in Freakonomics, Adam Smith was first and foremost a moral philosopher not an economist (per se). Though Smith is much better-known as the author of The Wealth of Nations, he actually preferred (and spent much of his life revising) an earlier work, The Theory of Moral Sentiments, which he believed was far superior (and which I look forward to reading).
But rather than merely talk about these things in principle, for this first stats post I’d like to leave you with two examples (both from FiveThirtyEight’s political coverage) of how numbers can help reveal trends, while also pointing us to the right kinds of trends to look for (which may not always be in the usual numbers we seek).
Our first example comes from last October, when Congressman John Boehner announced that he would be stepping down as Speaker of the House. In a way, this made intuitive sense since congressional gridlock has prevented most of the legislative initiatives of both parties. But as Nate Silver points out, such gridlock was not merely a Republican-Democrat dynamic. Instead, Boehner was faced with the daunting challenge of preserving a semblance of unity between Chamber-of-Commerce and Tea-Party factions within the Republican Party, while simultaneously sending bills to a skeptical Senate and (when they cooperated) on to a Democratic President.
Even more ironic, is that Boehner in large part succeeded in holding these disparate factions together throughout his tenure as Speaker. As Silver demonstrates, under Boehner’s leadership, the GOP was more united in votes on the floor of the House since the 1910s. But this is not entirely good news: Republicans were almost unanimous on partisan issues (94.6%), but that consensus weakened on bipartisan votes (to 90.4%; see below left).
His title, however, is misleading at best. Though congressional Republicans were indeed “more united than ever” under Speaker Boehner, they are also more divided than at any point since 1973. As Silver himself also seems to understand, then, the story of Boehner’s resignation is much more about division than unity, and his title should make that clear. In this case, numbers help us take a longer view toward his resignation by putting things into historical perspective, but also confirms what we generally recognize: the Republican Party is in a state of flux.
The second major example, is what FiveThirtyEight refers to as “the Endorsement Primary.” It’s hard to imagine a presidential primary season without opinion polling, electioneering, or, well... primaries. But primaries (and their more party-centric cousins, caucuses) have only come to the fore of the American political scene through a long and dynamic history. Elections have been part of American society since its earliest colonial days, and so has electoral change, especially in terms of universal suffrage and the gradual standardization of election dates. Many of these reforms are a matter of constitutional, federal, or state law, while others have accrued gradually through internal reforms within the major political parties.
In US presidential elections, the most significant reforms since 1787 (perhaps even more so than the 12th Amendment) have occurred in this manner, driven from within the parties. Early on, names for the presidential ballot were essentially decided by officeholders at the state and federal levels. By the 1830s, national conventions were adopted, to which state parties would nominate delegates who would then represent their state’s views on the national platform and nominees.
But it wasn’t until the 1960s that primaries and caucuses became the official way for states to determine whom their delegates would support as the party’s nominee. Further reforms have maintained this primary/caucus form, but have also opened up the process considerably, more or less tying a state’s delegates to the results of the popular vote in statewide elections (and paralleling changes in the ways states allocate their votes in the Electoral College). And this has, in turn, driven much of the changes in election polling and presidential campaigning over the last few decades.
However, as Bycoffe points out, the results of these polls are notoriously difficult to draw inferences from (even with good samples of the American electorate) much less predict the actual outcomes of popular votes. In fact, they have become so volatile that Gallup has decided to focus on issues rather than candidates. For the data, Bycoffe refers readers to the work of Cohen, Karol, Noel, and Zaller in their book, The Party Decides, in which they “evaluated data on endorsements made in presidential nomination contests between 1980 and 2004 and found that ‘early endorsements in the invisible primary are the most important cause of candidate success in the state primaries and caucuses’” (Bycoffe). He also provides an interactive on the data from the last twelve primaries for readers to visualize the phenomenon.
This is important, of course, because both parties are in the midst of primaries for the 2016 Presidential Election. FiveThirtyEight has therefore built a qualitative measure of success in this “invisible primary” by assigning points to each candidate based on the endorsements they have received to-date, weighted by the perceived influence of each office (10 points for sitting governors, 5 for Senators, and 1 for Representatives). The take-away: Secretary Clinton currently dominates her party with 457 points; Governor Bush (46 points) and Senator Rubio (43 points) have slight leads in the GOP, while most officeholders are yet to endorse anyone; and despite his strong showing in polls, Donald Trump has received no major endorsements to-date.
There doesn’t appear to be any question whether “the establishment” will accept their party’s final nominee (historically speaking, they almost certainly will). The real question is whether officeholders will await the winnowing effects of the early voting states (as in the 2008 primaries, the Democrats in 2004 and 1988, and the Republicans in 1980), or whether they will seek to shift public opinion themselves. Personally, I favor the former view, though that latter is still possible.
Numbers are powerful things. Most of us approach any particular issue with an already-developed set of personal values and presuppositions that can (if we don’t recognize them) affect the way we make decisions. Numbers, however (whether in experimental or observational studies), can assist us by highlighting trends that may at first seem counterintuitive and forcing us to explain why or why not this new data makes sense, and from what perspective (maybe even changing our own minds in the process). But as Einstein also pointed out, “Not everything that can be counted counts, and not everything that counts can be counted.”