I’m surprised that the famous 1994 Card-Krueger study of the minimum wage caused Scott Sumner to shift his thinking about the consequences of that policy. (Scott says in the comments that his thinking about the minimum wage has not shifted so much that he favors it.) I reveal below just why I’m surprised, but first I must saddle you with a longish preface.
The principal practical economic argument against the minimum wage is that it puts some low-skilled workers out of jobs. The core of economic theory informs us of this result, in the same way that the core of physics tells us that a dime dropped into an Olympic-sized swimming pool causes the water level of that pool to be higher than it would be had the dime not been dropped into it – and no amount of failure to detect empirically the resulting rise in the water level will cause physicists to doubt that a dime dropped into a body of water causes the water level to rise. (Howls would greet any physicist who said “Well, physiometricians keep studying the dropping of dimes into big pools – real-world pools that have swimmers and divers constantly going into and out of them – and these physiometricians very often, although admittedly not always, find empirically that none of these dropped dimes has any effect on the water level of the pools. So, being a data-driven physicist, I conclude that a dime dropped into a real-world big swimming pool does not displace water in those pools. I am, I repeat, driven by the facts and not dogmatically by theory!”)
Unlike with physics (because the economy is a vastly more complex phenomenon than is the physical universe), what the core of economic theory does not reveal is the magnitude of the job losses. Will the implementation today in Someplace, USA, of a minimum wage of $X.YZ destroy 1% of the current number of jobs in Someplace? Ten percent of these jobs? Eighty percent? One-twentieth of one percent? That’s an empirical question that no amount of a priori theorizing can answer. Empirical investigation is necessary – although keep in mind three facts:
(1) Even the best such investigation can never reveal the correct answer with complete certainty, because no empirical investigation of reality can adequately control for the many factors in addition to the implementation of the minimum wage that affect the labor market in Someplace; not only does the demand for, and supply of, labor change for reasons having nothing to do with Someplace’s minimum wage, but, also, employers and workers have many practically unobservable margins in addition to the job-vs.-no-job margin on which to adjust to a minimum wage (such as reducing the amount of leisure time de facto allowed to workers while workers are formally on the job);
(2) the correct answer is time-dependent; the number of job losses caused today in Someplace by the implementation today in Someplace of a minimum wage will almost certainly be fewer than are the number of job losses caused by this minimum wage in Someplace over the course of the two months following its implementation – a number which itself will be different (likely smaller) than is the number of job losses caused in Someplace by this minimum wage over the course of the year following its implementation; the full effects of a minimum wage take time to play out; and
(3) whatever is the correct answer for the job-losses caused in Someplace, USA, by the implementation of a minimum wage (over whatever particular span of time is considered), that answer is unique to Someplace, USA, and to that particular time in Someplace, USA; the correct answer for the job-losses effect of the minimum wage imposed today in Someplace, USA, – even if that answer were given to us by all the angels in heaven and confirmed by god himself – tells us very little about the magnitude of the job losses caused by minimum-wage implementations in Elsewhere, USA, Somewhere, Canada, and Ourtown, UK; indeed, this correct answer doesn’t even tell us very much about what the job losses would be in Someplace, USA, had the minimum wage been implemented there at a different time (say, last year, or three years from today).
So, now to Scott’s reaction to Card-Krueger – namely, to Scott’s being persuaded by Card and Krueger that the magnitude of job losses caused by minimum wages is less than he’d previously believed: My main problem with Card-Krueger is, as it has always been, that that paper is very short-run. As explained above in point number (2), the impact of the minimum wage takes time, and nothing in economic theory says that that time will be ‘short’ when measured on a calendar. This fact, combined with the reality that minimum wages have been around in the U.S. since before WWII (meaning that by the early 1990s employment and production practices were already adjusted to their existence, and employers by then plausibly expected continuing, occasional hikes in minimum wage), greatly diminishes the information content of a measured change in employment following a minimum-wage hike, even when that change occurs in one particular jurisdiction (New Jersey) that adjoins another (Pennsylvania).
I’ve always believed that what Gertrude Stein said of Oakland, CA, applies to the Card-Krueger study: “There is no there there.”
* By the way, this truth applies even if empirical researchers discover that implementing a minimum wage caused the employment of low-skilled workers toincrease because of monopsony power in Someplace. Such a finding, even if it is flawless, does not tell us that Elsewhere, USA, also has such monopsony power that a minimum wage in Elsewhere will result in no job losses there. Nor does this finding tell us how long the minimum wage in Someplace can remain in place before it starts to jobs to be fewer than otherwise. Most importantly, this finding does not tell us even that the minimum wage is a sound policy in Someplace: it might well have been the case that, had no minimum wage been imposed in Someplace, that market forces would have responded to the monopsony power and created better jobs than were the ones that were created in response to the minimum wage.