Real Compensation: An Index Number Problem Mankiw and Altig Might Hide

Brad DeLong is puzzled by a suggestion coming from David Altig and Greg Mankiw.

David opened with:

If I had my way, appeals to the BLS average hourly earnings series would be banished from commentary about wages and the fortunes of the workers – unless the commentator explains why that measure is a truer measure of labor compensation than those that include in-kind payments to employees (that is, benefits).

Greg followed with:

I am always surprised when I see economists compare wages and productivity using wage measures that exclude fringe benefits. Theory says that productivity should determine total compensation, not cash earnings.

Brad goes “huh” with the detail being:

I thought that everybody knew that wages and salaries were only one component – albeit the largest component – of total compensation. But I also thought that everybody knew that (except at the very top end) differences between growth rates of wages and salaries and growth rates of real compensation were relatively small. We have pretty good data on short-run and medium-run movements in average hourly wages and median usual weekly earnings. We have pretty bad data on benefits. We know that average total compensation grows about 0.4% per year faster than average wages and salaries, and that the gap is smaller – 0.2%? – at the median. Yes, it would be very nice, it would be better to have median weekly real compensation. But we don’t. So why not use what we have got?

Kash (who has returned to AB blogging) might be happy to see that Greg’s update had to deal with this:

Part of the explanation for this increase is the rising cost of health care. The value of employer-provided health insurance is a key part of many workers’ compensation. It is a benefit that gets more valuable as improvements in medical technology drive up the cost of insurance.

But this was not Kash’s argument, which David later captured:

Not surprisingly, I received several comments arguing that increases in labor compensation have, in the recent past, been driven by rising health care costs. In making sense of all this, I think it helps to consider two different questions we might want to ask when considering payments to workers. The first is, “What does it cost a business to purchase one hour of a worker’s time?” To answer this question, a broad (benefit inclusive) measure of labor compensation is clearly the right one. Now consider the second question: “Does an increase in payments made by firms to workers make workers better off?” On this issue, the answer may very well depend on the difference between compensation paid in dollars and compensation paid in benefits. Maybe, but not necessarily. Consider two workers, both of whom receive the same total amount of compensation, and both of whom spend exactly the same amount on health care. Now suppose that those expenditures rise by 10% for both workers. We’ll further suppose that worker 1’s wages rise by 10%, while worker 2’s employer holds his or her salary constant but picks up the tab for the higher medical costs. Would you say that worker 1 is better off than worker 2? Neither would I.

Yes – David does get around to the index number issue. No – I would not banish reporting on both wages and fringes as separate series even if I agree with David that fringe benefit compensation is part of the story.