Jobless in Seattle
Sorry to use that headline, which is no longer original (see here, here and here), but it’s such fun I couldn’t resist. Seattle is a Mecca for US progressivism, having raised its minimum wage aggressively over the last three years on the way to taking it all the way to $US15, the current lodestar of most minimum-wage enthusiasts. Unfortunately, as a comprehensive new study suggests, the effect of the sharply rising minimum wage has been sharply rising joblessness among Seattle’s low-wage workers, as many economists had feared.
The closest we come to a law in economics is the “law of demand,” which says that if the price of something rises then, all else equal, people will buy less of it. The average 10-year-old can figure out that relationship if you ask the question right so most people are underwhelmed when they learn it’s such a big deal in economics.
Of course, our biggest deal is the “law of unintended consequences”: You do what you think are good things and they turn out not quite the way you were hoping. Combine the two laws—law of demand and law of unintended consequences—and the minimum wage becomes a more dubious proposition than it might at first sound. Raise wages by fiat and employers seem bound to cut back on employment, at least some. Workers who keep their jobs and hours will benefit. Workers who lose jobs and hours are out of luck. Whether that’s a good thing or not is now much harder to say.
The important question is the exact terms of the law of demand: Just how much of an employment effect is there? That’s a much-studied question. One famous paper, published in the 1990s and focusing on fast-food restaurants in New Jersey, suggested the employment effect wasn’t very big. If not, then raising the minimum wage mainly means higher incomes for low-wage earners and may help reduce poverty, as its backers hope.
The new Seattle study comes to quite a different conclusion. And it does so using more complete data than most earlier studies. As part of its unemployment insurance program Washington state requires employers to provide data both on what workers earn and on how many hours they work. Only three other US states do that. The Seattle study is therefore able to analyze hours and wages for all insured workers in the city, which is most workers.
One advantage of the data set is that you can tell exactly how much a person’s wage went up in the period following the minimum wage increase. Without individual wage data studies typically assume wages went up by the full percentage increase in the legislated minimum, which probably isn’t what happened. Not everyone whose wage rises because the minimum rose was at the previous minimum. For instance, if the minimum wage goes from $9 to $12, it’s not the case that everyone who is earning $12 now was earning $9 before. Some will have been earning $10 or $11. But if you do assume everyone had a $3 increase you’re effectively over-estimating the wage increase. If you detect only a small employment effect, you might think: Hmmm, big wage increase, moderate employment effect, therefore the demand for labour isn’t very sensitive to wages. But if in fact many people’s wages didn’t go up by the full $3 what you conclude is: moderate wage increase, moderate employment effect, therefore the demand for labour is more sensitive than we would have thought without this information.
A second neat thing the Seattle researchers did was to purposely dumb down their data. Many previous studies had looked at the effect of minimum wage increases on employment in the restaurant sector and, as mentioned, some hadn’t found much of an effect. The trouble with those studies was that they had to look at employment effects on all workers in the restaurant sector, not just minimum- and low-wage workers. The Seattle researchers began by looking at the effects on poorer workers and found pretty big employment effects, both across industries and within the restaurant sector. But then they asked: what effects would we get by doing what these other studies did and looking at ALL workers in the restaurant sector? Their result was: not much of an employment effect. Got that? When their data allowed them to focus on the workers most likely to be affected, they found big effects. But when they replicated the data gaps earlier researchers had been forced to work with they found not much of an effect, which suggests earlier studies may have found the results they did mainly because of their incomplete data.
At some stage in most policy arguments people say “That’s an empirical question,” which usually implies it has an empirical answer. But empirical inference isn’t easy. Without exactly the right data you may get a misleading or even no empirical answer.