Pay for Oppression: Do Workers in Fairer or Safer Jobs Make Less Money?

The debate over worker rights on the job has taken an interesting turn with Tyler Cowen’s defense of the compensating wage differential argument.  This, for those who have not run into it before, says that workers, whether through bargaining or the labor market, get a certain amount of utility.  They can take that utility in money or in better working conditions, but the sum has to add up the same.  This means that workers in safe jobs, all other things being equal, will make less than workers in dangerous ones; workers subject to bosses’ sexual advances will make more than those in more respectful organizations; and so on.  There are four conclusions you would draw from this if it were true:

1. Dangerous or demeaning work is not a problem per se.  The workers in those jobs are just as well off as they would be in a better setting.  Don’t worry about them.

2. Employers have a financial incentive to make jobs better—safer, fairer, etc.  That way they have to shell out less cash.  In fact, they have just the right incentive, the amount of money workers are willing to give up in order to get better treatment.

3. Regulation can’t make anything better, but it can make things worse by taking away an option that would otherwise be available to workers.  Some workers would rather have the money and put up with the dangers and indignities of a crummy workplace.

4. You can measure the monetary value of such intangibles as the value of life, the value of not being harassed, of being able to pee when you want, etc.  It’s simply the difference in wages between jobs that offer more versus less, scaled by how much change in risk, harassment or whatever the worker is being compensated for.

Now it happens that this is a topic I know something about.  In fact, I wrote the book on it.  (OK, a book.)  For the full story, read the book.  Here I will make a few brief comments about the evidence.

I agree that fringe benefits that are essentially monetary in character, such as pensions and insurance, are subject to this sort of process.  Workers really do trade money in one form for money in another, and unions bargain explicitly over this.  It is also true that public insurance, like workers comp, is largely financed out of wages too, no matter how the laws are written.

It is not true that nonmonetary costs and benefits of work are compensated—not fully at any rate, and sometimes not at all.  For an empirical demonstration, see this old study I wrote with Paul Hagstrom that has, to my knowledge, never been rebutted.  Tyler links to a slightly less old Viscusi/Aldy lit review that cherry-picks shamelessly, not only in its selection of what counts as a valid study, but (especially) in which results of the authors they choose to report.

For those who don’t want to go to the sources, here are the two fundamental empirical problems with studies that claim to show wage compensation for things like occupational safety: (1) They use industry-based measures of which jobs are risky, but they ignore all the other industry-level determinants of wages, like concentration, capital intensity and percent unionized (usually).  (2) They show signs of potential publication bias, where specifications are selected that yield the “right” result.  What signs?  They don’t provide summary data on the full range of specifications tested (“taking the con out of econometrics”), so that the reader can determine whether the reported specifications are outliers or not, and the results are hardly ever tested on subsamples.  Think for a moment about this last point: if there are compensating wage differentials for women exposed to sexual harassment, this should apply to black women and white women, unionized women and nonunion women, women in blue collar jobs, white collar and pink collar, and other plausible breakouts.  If not, you would need to have a story to explain why some get it and not others.  (Or why some subsamples even have exacerbating differentials, which I found with occupational safety—the “sweatshop effect”.)  There is a small literature on subsamples in wage-risk regressions, and they are all over the map.

Note that the Joni Hersch study Tyler links to is vulnerable on all these counts.  It doesn’t consider interindustry differentials.  There is only one specification reported.  No subsamples.  Not convincing.

The bottom line is that skeptics have every reason to remain skeptical.  Actually, a world of compensating wage differentials would be better than the one we live in.  Some jobs are unavoidably difficult, dangerous or unpleasant, and they should pay more.  But there are also human rights, like freedom from abuse and freedom of expression, that shouldn’t be for sale, even when you’re on the clock.