Whether or not increasing the minimum wage drives down employment is one of the most hotly debated (and ideologically driven) questions in economics. This isn’t surprising. The question is a very hard problem. But it also (like many other politicized questions) suffers from a tendency to boil it down to a simple yes or no.
Intuitively, economic theory argues that when wages increase, businesses must either (1) raise prices or (2) cut costs. In competitive markets, raising prices means losing customers, so the only real option is to cut costs, and since businesses produce at the optimum balance of inputs, when the cost of labor goes up, they substitute capital for labor to make up for it. Result: employment goes down.
But the real world is remarkably complex, and much of the empirical work in the past thirty years has failed to find negative employment effects (Katz and Krueger (1992) is a good starting point). There are two major approaches to the question. The first is to use aggregate time-series, which generally suffer from an inability to define a good counterfactual and properly identify effects, which are likely happening at a less aggregate level. Studies prior to 1992, which largely found negative employment impacts, primarily used this method.
The second method is some kind of causal inference, usually difference-in-differences using CPS data or other micro-data. Some of these studies have found impacts and others have not. Yellowitz (2005), for example finds substantial negative impacts in Santa Fe. Potter (2006) did not. Recent studies worth reading are Neumark, Salas, and Wascher (2014) and Dube et al. (2010). Dube’s more recent work suggests that if minimum wages are less than 50% of median wages, their employment effects are minimal.
The research exemplifies a key problem with determining the causal effect of complicated interactions: what control variables and which counterfactuals do you use? This is a (the?) major source of disagreement among researchers who are searching for an answer in good faith.
But setting aside the research questions of correct instrumentation and careful selection of counterfactuals, there are more general reasons why broad reasons why the living wage is so hard:
1. The treatment group and/or treatment effect is probably small
The people and businesses affected by increased minimum wages are likely only a very small portion of the overall population. This can be for a variety of reasons. For one, by the time a minimum wage is raised, the ‘market rate’ may not be very far away. So even if the minimum wage is increased from $6 to $9, if the market is already paying $8, the effective increase is only $1, not $3. Most city-based minimum wage laws also phase in increases over a number of years, so though the eventual minimum wage may be quite a bit higher, it doesn’t happen all at once, and immediate impacts are not as visible.
These ‘small treatment’ issues cut both ways, since they both reduce the impact on businesses, but also don’t do much to increase earnings for low-income workers. It fits well with the results from studies finding both no employment impact and little or no earnings impact (which was the result of my work in Potter 2006).
So the actual treatment can be smaller than indicated by the difference between the old and new minimum wage, but the treatment group itself can also be small, either due to limitations on which businesses the minimum wage applies to or because a significant portion of workers are already earning a minimum wage.
2. The theory is wrong
Or to be more precise, it’s only right under perfect competition and when firms are producing with the most cost effective mix of inputs. But reality is rather different. This doesn’t mean that the theory of how firms react to a change in the price of labor isn’t helpful, but it needs to be thought of in context. The vast majority of businesses are not entirely price takers, are not operating at a zero profit level, and have not conducted the analysis necessary to determine the optimum mixture of inputs (including labor) that would most reduce their costs. There’s rather a lot of room for adjustment under those conditions. While firing employees is one option, other options include raising prices and accepting a lower profit.
It’s the firms and their employees who exist on the margin that are probably most at risk. Businesses that are barely turning a profit or employees that have a job but not a very essential one have less room to adjust, and as a result face potentially stronger impacts. That’s too bad, because those are the populations we most want the minimum wage to help.
3. Ideology is the mind killer
There are many economists in academia who at least strive for an objective approach to economic study, but since economists (as all scientists) are human, they have prior beliefs, and those priors affect, sometimes strongly, the conclusions that they come to. This makes such a politically divisive issue as the minimum wage hard to settle. Like most questions worth studying, it doesn’t lend itself to a soundbyte answer that applies in all cases. And that doesn’t even touch on the vast number of economists who engage in what I like to call “advocacy science”. Pick a conclusion and find work to support it, rather than find a puzzle and work to solve it.
So why is Economics so bad?
Honestly, I don’t think it is. Economics tries to answer extremely complex, often politically charged, problems, in which causality is difficult and perhaps multidirectional, subjects are fickle and imperfect, and tools and data are lacking. There’s certainly a lot of progress to be made, and Economics as a field has grown significantly in the past twenty years. Simple and intuitive theories like rational choice and perfect competition are real world rare, and the field has made the mistake of acting as if they are real world common. Theoretical fads and tribes are endemic and have led to some pretty ridiculous theories (I’ll avoid hoping on the macro criticque bandwagon for now).
I’m actually optimistic though. More and more the economists I see are aware of the complexity of the subject. Younger academics seem to be increasingly pushing for careful pursuit of objectivity. New tools and data are constantly becoming available. We are much better as isolating causal interactions than we were thirty years ago. Winter may be coming, but hope is in the air nonetheless.
So no, we don’t know whether the minimum wage causes employment loss. Almost certainly it does for some value relative to market wages. Whether it’s 50% of median wages as Dube suggests or some other number, or more likely a non-linear function of multiple factors, we can’t say yet, if ever.