One of economics’ most famous papers – the 1994 minimum wage study by David Card and Alan Krueger – has just won David Card (pictured) half of a Nobel Prize in Economics. The overall reasons for Card’s award are well explored here and here, and by Card himself here.
The Card & Krueger paper is widely admired. Even more remarkably, it is one of that rare breed of paper which has changed minds in economics. Before its publication, most economists tended to believe minimum wages cost jobs; after they had digested Card & Krueger, some began at least to doubt that was true.
Rather than just building a model, Card & Krueger started with a good natural experiment: the minimum wage went up in New Jersey but stayed the same in Pennsylvania, which acted as a control group. The difference in the difference between the two states before and after the law change should be the effect of the minimum wage. Hence the name of this technique – “difference in differences“.
Essentially, this year’s Economics Nobel celebrated natural experiments, and natural experiments are, in general, a lot better than the armchair theorising which has featured in some previous economics Nobels. They aren’t exactly a new idea; their use in social science goes back at least to the 185os. But Card & Krueger’s natural experiment was a notable and clever one, and it seems to have rekindled enthusiasm for them within economics. As a demonstration of what good social science looks like, I agree with Club Troppo‘s Paul Frijters: this is one of the best picks of recent years.
But here’s the thing I keep coming back to: Card & Krueger’s findings are very often misrepresented as proving something they don’t prove. They’re terrific economics, but they are also part of what we might call the Feynman Social Science Problem.
The problems with Card & Krueger
Card & Krueger’s paper was interestingly counterintuitive in the sense that high prices normally make people buy less of something. Just on that logic, you might wonder why a higher minimum wage would have no effect at all on low-paid jobs. Then again, hey, the world is full of research results that look odd at the start but become perfectly normal once we understand them better.
But Card & Krueger really didn’t give us a solid conclusion about the general employment effects of the minimum wage. I see headlines like “The Nobel Prize winner in economics revolutionized thinking about the minimum wage” (Quartz), and I don’t think it really did. The Nobel Prize citation declares that Card proved that “increasing the minimum wage does not necessarily lead to fewer jobs”, and I don’t think that’s true either. Card & Krueger cleverly gave us evidence that raising the minimum wage might not always have the result many economists expected. That’s useful and sensible, not transformative.
And there are doubts about even that evidence.
Problem 1: taken at face value, Card & Krueger’s paper says that higher wages push employment up. If people everyone took the work at face value, we’d be solving unemployment by lifting the minimum wage higher and higher. But no-one thinks that would work.
Problem 2, which is actually a bunch of problems: the Card & Krueger experiment has a number of possible holes. To take a sample:
- Earlier changes in the minimum wage law are alleged to have caused a bigger drop in teenage employment in New Jersey than in Pennsylvania.
- New Jersey employers reportedly had ample warning of the minimum wage hike, and might have chosen to lower their worker numbers before the wage rise took effect.
- We saw almost inevitable criticisms of the quality of Card & Krueger’ data, some of it reasonably credible.
Problem 3, which is related to the previous two: top professional economists don’t agree on what Card & Krueger shows either. One piece of evidence: in early 2021, the Chicago Booth School’s IGM panels of notable economists in the US and Europe looked at the minimum wage question in the US and could only get 15 per cent of their number to disagree with the idea that raising the US minimum wage to US$15 would lower employment among low-wage workers.
IGM US economists 2021
IGM European economists 2021
(Be aware that the wage hike proposed here was not the same one that Card & Krueger looked at, which may be important.)
You of course might argue that the IGM panels are just a bunch of conservatives. But that seems unlikely to be the case. You can see who was on the panel by going to the US and European poll pages. Also, the same panel mechanism could not get one economist to disagree with the idea that carbon taxes are a better way to implement climate policy than cap-and-trade. And if you think many of the IGN’s economics panellist are behind the times, remember that Card & Krueger published a quarter-century ago.
These sorts of problems are not unusual in economics or in the social sciences more generally. Indeed, we see them everywhere. They don’t make Card & Krueger’s paper bad; rather, they underline that social science experiments are almost always open to question, precisely because they are embedded in complex and messy social systems.
The Feynman Social Science Problem
I can see why Card got the Nobel, and I think this Nobel will give a fillip to the very important principal of seeking out natural experiments wherever we can.
But at the same time, the Card Nobel underlines a huge weakness in today’s social sciences. Looking at the debate over Card & Krueger’s paper gave me a renewed appreciation for the difficulty of drawing conclusions from natural experiments in the social sciences. It’s this: in real-life situations, you often just have more going on than you know how to deal with.
Indeed, the Card & Krueger paper demonstrates what you might call the Feynman Social Science Problem: in most social science situations, it’s nearly impossible to isolate the relevant variables in an economic study with enough certainty to reach a conclusion. In physics, controlling for variables is often possible. In social science, it almost never is.
Most people don’t like to say this very loudly.
The famed physicist Richard Feynman was an exception: he called social science a pseudo-science:
“I might be quite wrong, maybe they do know all these things. But I have had the advantage of having found out how hard it is to get to really know something, how careful you have to be about checking the experiments, how easy it is to make mistakes … I see how they get their information and I can’t believe that they know it – they haven’t done the work necessary, the checks necessary and [taken] the care necessary.”
Feynman’s criticism has proved hard for me to shake. And the Reproducibility Crisis that has gripped the social sciences in recent years hasn’t reinforced anyone’s faith in social science methodologies, either. I have grown increasingly reluctant to trust social science findings in general.
Again, the point is not that Card & Krueger did anything wrong by the standards of the social sciences. Indeed, having found their experiment, they also worked harder than most to check their findings. And they seem to have been pretty modest about the epistemological value of their work. (Plus Card is widely described as a nice guy – and I always like to see nice guys finish first.)
The problem is that people take social science research findings and make more of them than they ought to do.
The continuing doubts on the minimum wage
Since Card & Krueger published their paper in 1994, other results and interpretations have been dribbling in, some of them suggesting that higher minimum wages do cut jobs after all. One big study looked at the minimum wage in the US city of Seattle, which was being pushed up over several years. It estimated that an hourly minimum wage hike from US$11 to US$13 caused a huge cut in low-skilled workers’ hours – 9.4%. In case you’re wondering, Australia’s most recent minimum adult wage equated to around US$14.10 an hour.
We also have some Australian research, though not a lot. The best comes from widely admired economist Andrew Leigh, and it was done before he left the ANU for the national parliament. He is now a Labor federal MP, and doesn’t seem to be talking as much as he used to about the problems with the minimum wage. But some of Leigh’s work, based on Western Australian data, has suggested that a 1.0% increase in the minimum wage could be expected to cut employment by between 0.15% and 0.39%. A 3% minimum wage rise, then, would cut 60,000 jobs or more. This is not a jobs catastrophe, but it’s far from trivial – especially if your job is one of those that disappears.
My own guess is that minimum wage rises probably do cut some jobs, especially in Australia, where the minimum wage is high by world standards. But how sure am I? Not very. How many jobs might they cut? I’ve not much of an idea, and you shouldn’t care about my view much anyway.
My guess is that having read all this, you will probably still have a very strong opinion as to whether a minimum wage is good or bad. My argument here is that you should try to distrust your certainty a little. As Richard Feynman argued: even at their best, the social sciences are nothing like physics.
Note 1: Yes, it’s actually the Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel.
Note 2: Krueger, sadly, died in 2019; Josh Angrist and Guido Imbens, oddly, got a quarter of the Nobel prize-money each.
Note 2: As it happens, both David Card and Andrew Leigh have an extra point to make about the minimum wage: whatever it does to employment, it isn’t the best mechanism for reducing poverty. Many minimum wage earners are students living at home in middle-class suburbs with mum and dad. As Card himself puts it, many minimum wage earners are not in poverty, and many of those in poverty are not connected to the labor market.
Note 3: This post repeats elements of a post I wrote for Troppo in 2017. It started as an edit, but after a certain point it made more sense to make it a new post.