Forbes has an interesting article trying to sum up the state of the economists' empirical debate over the minimum wage and its effect on employment.
Some excerpts to get across the flavor of the conversation:
David Neumark and co-authors have utilized a panel approach and found that minimum wages do lead to lower employment. Arin Dube and co-athors use an alternative approach that takes cross-state border counties as control groups and find there is no impact. They also find that including additional controls in the panel approach, like census region or state trends, makes Neumarks’ results insignificant, allegedly due to pre-existing trends. Neumark replies that the cross border county grouping doesn’t really lead to optimal control groups. The Dube approach, he alleges, really throws out too much identifying information, and the resulting large standard errors don’t give you a precisely estimated zero impact, but a very noisy estimate....
So this is where you are left off. Who has constructed the more reliable counterfactual? Should you specify state time trends and if so how? Does Dube throw out too much useful data? And if so, how does Dube et al identify the impact on separations and hires?.....
Then came Meer and West (2013). This paper extended Dube’s focus on labor market flows and looked at net job creation using the cross-state border counties approach. Much has been discussed about their findings, but I think an important missing contribution of their paper is an explanation for why there would be so much disagreement about the impact of minimum wages if there really was a negative impact. What they show is that if the true effect of the minimum wage was to reduce the growth rate of employment, then looking at the impact in levels would give you two different estimates of the effect: no effect if you included a growth trend, and a negative effect if you didn’t include a control trend.
The article does what it sets out to do fairly well, but it is curious that it never discusses economic theory or why this question is so controversial, and why studies that seem to show no employment effect of a minimum wage might seem curious and worthy of being looked at very carefully.
Science tends to progress not merely by observing the world in an empirical sense, but by trying to abstract out regularities or laws that operate in it. This is, as the language from Forbes above makes abundantly clear, because the world is a crazily complicated multicausal place. Understanding the whys of what you are observing is very difficult if all you do is observe, or observe combined with lots of complicated mathematical attempts to figure out how to account for all the other things that might be going on. Because, as the ongoing debate described above shows, you can never do that in a way that will satisfy all other reasoning and observing minds.
Upon observing an airplane fly, one doesn't assume that apparently the law of gravity doesn't apply to metal plane-shaped things. Rather, you intelligently realize instead that other forces must be in operation on the plane. Similarly, we should not assume that just because one doesn't always see an observable, countable diminution in people employed related to the minimum wage in every specialized situation an academic looks for it, that on the margins the law of supply and demand don't work on labor and that raising the minimum wage is costless in terms of jobs.
If you believe that some people's marginal product isn't worth more than that minimum wage, and if you believe employers would tend (other things being equal) to not want to lose money on hiring people, then you should understand that the minimum wage does cost jobs. How many, and whether you think the benefit to the seen still employed should as a matter of policy outweigh the harm to the unseen unemployed, are different questions.
People involved in the actual science mostly get these distinctions. But that these sort of studies regarding the minimum wage sink into that general background scientistic morass of what the "educated, I read the Sunday New York Times with some attention" believe and "know" about the world is unfortunate, because it privileges some very picayune and minor attempts to look at a portion of the world above the more valuable process of actually understanding on a deep level how the world works.
In short, it allows minor-level smarty pants to confidently "know" that we can raise the minimum wage without causing anyone to lose or not get a job, to let this sort of "sophisticated, empirical" stuff obscure a well-understood fact among economists: you can raise a legal wage above the level at which many people will be employed. Anyone contemplating policy needs to know that.