It’s common to encounter assertions and suggestions that facts are more objective – that facts are more revealing, more important, more real, more trustworthy – than theory. The intended conclusion is that if theory predicts X but the facts show not-X, then the theory is wrong.
It is indisputable that a theory that is inconsistent with empirical data is a poor theory. No theory should be accepted merely because of the beauty of its logic or because it leads to conclusions that are ideologically welcome or politically convenient. Yet it is naive in the extreme to suppose that facts – especially the facts of the social sciences  – speak for themselves. Not only is it true that sound analysis is unavoidably a judgment-laden mix of rigorous reasoning (“theory”) with careful observation of the facts; it is also true that the facts themselves are in large part the product of theorizing.
For example, what’s the current rate of unemployment in the U.S.? The “fact” – seemingly objective and independent of theory – is today (March 2015) 5.5% . But who counts as unemployed? Twenty-two-year-old college students who choose not to seek work are not counted as unemployed; 22-year-old people who do seek, but cannot find, work do count as unemployed. As sensible and obvious as it is, this classification is the result of theorizing. We humans have decided that it is unreasonable to classify a 22-year-old without a job, but who voluntarily is in that condition, as being “unemployed.” And of course several other theory-driven decisions are involved in deciding who counts as unemployed (as well as in deciding how to gather evidence of which flesh-and-blood people are, and which aren’t, in that category at any given time). A great deal of theory – some of it so obvious as to be uncontested, some of it a matter of legitimate controversy – is indispensable in creating the seemingly theory-free ‘fact’ that we call “the U.S. unemployment rate.”
I rehearse the above because, after talking with Russ earlier this morning about this very matter, I opened an e-mail from a correspondent (an assistant professor of economics) who coincidentally expressed to me his wish that I would be more “humble in face of the facts about the minimum wage.” This correspondent upbraided me for what he takes to be my stubborn, ideologically driven refusal to admit that “the facts prove” (his words) that minimum wages do not have the negative employment effects that standard economic theory predicts. Not surprisingly, this correspondent recommended that I read more carefully the works of Arindrajit Dube – a scholar whose research does indeed offer up evidence that, contrary to standard theory’s prediction, minimum-wage legislation does not harm low-skilled workers.
I have read much of Dube’s work. And I’ve also read much of the work of David Neumark, William Wascher, and others who offer up evidence contrary to that which is offered by Dube and his co-authors. No one who reads this work – and there’s plenty of it – can possibly decide, based on the data alone, whether minimum-wage legislation does or does not reduce the employment options of low-skilled workers. It’s not just that the conclusions drawn from the empirical evidence differ wildly. It’s more that coming to a conclusion requires reasoning about – theorizing about – the data. These data (like all data) do not speak for themselves. It’s just not what data do.
I encourage anyone, such as my correspondent, who pontificates on the need to cast theory aside and ‘bow before the data’ to read the works of Dube and of Neumark, Wascher, et al., carefully. Notice that this research today largely is consumed with theorizing about how best to measure empirically the employment effects of minimum-wage legislation. That is, while these papers appear on first glance to be simply ones that report empirical findings, they are at their core less about reporting empirical findings than about theorizing about how to conduct empirical research into the empirical effects of minimum-wage legislation. These papers are largely exercises in pure reasoning and theorizing.
Here, for example, are some abstracts. The first is from an October 2013 paper by Dube :
Consistent with recent work by Meer and West, I find a negative association between minimum wages and state-level aggregate employment growth in both the Business Dynamics Statistics and the Quarterly Census of Employment and Wages data, and it is sizable for some time periods. However, I show that this negative association is present in exactly the wrong sectors. It is particularly strong in manufacturing which hires very few minimum wage workers. At the same time, there is no such association in retail, or in accommodation and food services—which together hire nearly 2/3 of all minimum wage workers. These results indicate that the negative association between minimum wages and aggregate employment growth does not represent a causal relationship. Rather the association stems from an inability to account for differences between high and low minimum wage states and the timing of minimum wage increases. Consistent with that interpretation, when I use bordering counties to construct more credible control groups, I find no such negative correlation between minimum wages and overall employment growth.
This next abstract is from a May 2014 paper by David Neumark, J.M. Ian Salas, and William Wascher :
The authors revisit the long-running minimum wage–employment debate to assess new studies claiming that estimates produced by the panel data approach commonly used in recent minimum wage research are flawed by that approach’s failure to account for spatial heterogeneity. The new studies use research designs intended to control for this heterogeneity and conclude that minimum wages in the United States have not reduced employment. The authors explore the ability of the new research designs to isolate reliable identifying information, and they test the designs’ untested assumptions about the construction of better control groups. Their analysis reveals problems with the new research designs. Moreover, using methods that let the data identify the appropriate control groups, their results reaffirm the evidence of disemployment effects, with teen employment elasticities near −0.15. This evidence, they conclude, still shows that minimum wages pose a tradeoff of higher wages for some against job losses for others.
These are just two samples among many possible ones that I could give here. Google the authors’ names and find their papers related to the minimum wage. Read those papers. You’ll see that they are largely devoted to theorizing about how best to empirically measure the consequences of minimum-wage legislation.
I am not criticizing the authors of these papers. Quite the contrary. I applaud them. Thinking straight about how to do good empirical work is both important and difficult. My point, instead, is that much of today’s empirical work on the effects of minimum-wage legislation is in fact highly theoretical. For this reason alone (although other reasons can also be cited) it is simply naive to insist that empirical research alone can settle the matter of the effects of minimum-wage legislation. It is uninformed and, frankly, deeply unscientific to accuse those who use theory to question whatever position you take about the effects of minimum-wage legislation as being unwilling to look at the data or as elevating theory above ‘the facts.’