Angus Deaton's Misunderstandings of Price Indices

Price changes, not levels, drive substitution.

David R. Henderson. 

"One item that I didn’t get room to discuss in my recent review of Angus Deaton’s Economics in America is his discussion of the Consumer Price Index. Given the early work that earned him his reputation—see my biography of Deaton in David R. Henderson, ed., The Concise Encyclopedia of Economics—his confusion about price indices is, to put it mildly, surprising.

On page 74, Deaton writes:

When not all prices rise at the same rate, people will tend to move their spending away from the more expensive items and toward the less expensive.

That’s not true. Here’s what he should have said:

When not all prices rise at the same rate, people will tend to move their spending away from the items whose prices have risen by a higher percentage toward items whose prices have risen by a lower percentage.

In other words, what matters is changes in prices, not levels.

Take, for example, steak and chicken in a meat eater’s diet. Steak is typically more expensive than chicken. Let’s say that in 2023, steak was priced at $12 a pound and chicken was priced at $7 a pound.

Then a year later (now), steak is priced at $13 a pound and chicken is priced at $9 a pound. The price of steak has risen by 8.3%. The price of chicken has risen by 28.6%. So you would expect that people would shift somewhat out of chicken and toward steak. But notice that this is a shift from the less expensive to the more expensive.

It’s changes in price, not price levels, that matter.

This post is about Deaton’s misunderstandings. That means more than one. Another one is on quality adjustment. Deaton notes that the Boskin Commission, appointed by the Senate Finance Committee, issued a report in 1996 arguing that the Consumer Price Index systematically overstated the inflation rate. One key factor in this overstatement, argued the Commission, is that the CPI doesn’t adequately adjust for the many quality improvements we see in goods and services.

Deaton doesn’t argue against this view. He writes, on page 78:

I am not arguing here that the Boskin Commission’s estimates of quality improvements are wrong, and I understand that the absence of evidence is not the same as the evidence absence. And as one commission member argued, perhaps a squishy number is better than a firm number that is wrong. Like the commission, I believe that there have been improvements in many goods and services: I am old enough to remember when there were no ATMs and we had to line up inside banks to wait for a teller, and I should (almost) certainly prefer to be treated in a hospital of 2022 than a hospital of 1970.

You can just feel the “but” coming, can’t you? Here it is:

Though if hip replacements are twice as good as twenty years ago, I would still resist being too that I needed only one hip replaced when both were damaged, or being told by my insurance company that it would pay only half the bill.

That’s it? That’s his objection? Obviously, if both hips were damaged, one would want both repaired. That’s completely irrelevant to the quality improvement issue. The question, if the quality of hip replacements has increased, as many friends who’ve had them tell me it has, is whether the price of the replacement should be adjusted for quality. By moving to a nonsensical discussion of “one hip or two?”, Deaton avoids discussing the relevant issue. I could almost picture him pouting. (Maybe I shouldn’t call this a misunderstanding. It’s more like a diversion from the issue.)

By the way, here’s an article I had Mike Boskin write on the issue for The Concise Encyclopedia. Notice his statement near the end that the bias is about 0.8 to 0.9 percentage points per year. Although Mike doesn’t mention it, the Bureau of Labor Statistics did make some changes in response to his commission’s report. Those changes brought the bias down to 0.8 to 0.9; it had been well above 1.0 percentage point per year.

One way that the Bureau of Labor Statistics could have gone was to implement everything the commission recommended. That wasn’t likely to happen. Instead, though, Congress could have mandated that Social Security benefits and other government payments tied to the CPI be “de-indexed” by 0.8 percentage points per year. If Congress had done that in, say, 1998, then Social Security benefits today would be over 18% lower than they are now. That sounds bad (and is bad) for current Social Security recipients like me. But they (we) would have been able to gradually adjust over 26 years and wouldn’t have gotten used to Social Security benefits increasing in real terms. Moreover, it would have been great for people in their 50s and younger because the system would be almost certain to be around for them, without a wrenching adjustment when the cupboard runs bare in around 2033."