When ‘data journalism’ doesn’t add up

There is a 42 percent chance of precipitation this evening. The Pirates are favored over the Cubs by two points. With a three-point margin of error, Barack Obama is projected to be the next president of the United States. 

Statistics, probabilities and estimations pervade most decisions we encounter from day to day, whether it’s to don an extra layer before heading out the door, or to stock up the safety bunker in preparation for the next president’s socialist tyranny. But to Nate Silver, the high-profile statistics guru who recently launched the data journalism site FiveThirtyEight, they should be telling us even more. 

In a lengthy preface he penned to inaugurate the ESPN-owned site in mid-March, Silver laid the case for a “nerdier” model of journalistic news analysis. Although he briefly gestured toward grand ideas such as the value of good style and anecdote, he spent much of his 3,500 words criticizing non-quantitative analysis of the news, even unleashing “disdain” for opinion columns. But if there was ever an argument for old-fashioned opinions writing, it is the weakness of Silver’s nerdy analysis. 

Silver’s primary, though often unclear, argument diminishes the value of anecdotal speculation on the part of columnists and analysts. He takes to task the list of failed presidential election predictors that relied more on ideological assumptions than statistical inference. Fair enough, but you don’t exactly need Bayesian statistics to explain why “yard signs” and a feeling of energy aren’t reliable predictive measurements.  

He conflates news analysis and opinions writing, however, to resulting only in muddled mass of confused claims. It makes sense to criticize naive attempts at predicting empirically testable claims without statistics, just as much as we mock someone who doesn’t wear a coat because he can just feel that it won’t be cold. But not all claims worth holding opinions about are empirical or predictive. In fact, the most fundamental ones aren’t at all — that’s why people continue to disagree about them. 

In the popular SportsCenter model of political discourse, normative debates on who ought to be the president give way to predictive debates on who probably will be. Professionally raised in ESPN, it makes sense that Silver operates within this model. It’s not inherently bad, either — someone needs to make predictions about elections and policy preferences. And just like on SportsCenter, it helps to see diagrams and numbers to make sense of the information. 

But there’s a big difference between saying “the Seahawks will beat the Broncos on Sunday” and saying “the United States will be a better country if the Seahawks beat the Broncos.” No such distinction exists in sports discussions. It does, however, in politics.

Therein lies the beauty of opinions writing done well. We all have fundamental beliefs about public issues — the social necessity of welfare programs, the rights of women to receive abortions, the justification for military force. It is unclear how a statistic or a prediction could possibly illicit a meaningful response in these issues. But a well-formulated argument bolstered by personal anecdotes can accomplish what opinions writing does best — elevate the level of democratic discourse, not settle it.

Take, for instance, the contentious debate on the increase of the minimum wage. Statistical analysis supplied by economists and economic journalists should certainly inform any thoughtful stance, but it can only go so far. Even if a graph can show that unemployment will rise, two readers can have fundamentally different reactions. While one may value access to jobs over well-compensated jobs, another may value decent pay for labor as a point of principle. 

To narrow this space, the numbers, graphs and statistics can do nothing. It is in this space, rather, that appeals to anecdote and common values, conjured in clear and elegant writing, can thrive. 

Such values do not appear in Silver’s discussion. As opposed to the rigor behind statistical belief updating, he contrasts the open-ended genre of opinions writing, which, he argues, does not seem “to abide by any standard at all.” 

But why is that a bad thing? The rigid methodologies of statistics are necessary to provide reliable information, but Silver and his colleagues must recognize the limited familiarity of the general public with such methods. So rather than provoking a reasoned response, such statistical arguments serve largely to dazzle those who agree and turn away those who disagree. A layman won’t easily argue with a graph, so he’ll instead disregard it and look for one he likes. 

Though his call to increase the numeracy, as opposed to the literacy, of the American public is admirable, Silver can’t responsibly begin the task simply by inserting graphs without explicating the method.

It would be naive to say that an original voice like Silver’s has nothing to say about journalistic practice. Indeed it has provoked what I hope has been the thoughtful response which you are about to conclude. So far from ending the dispute with elegant graphs, Silver has provoked one with inelegant argument. Maybe he’s more of an old-time columnist than he would care to admit.

Write to Simon at spb40@pitt.edu.