Tyler Cowen, my colleague at George Mason is a wonderful thinker and a superb communicator of ideas. But I have not been convinced by his argument in The Great Stagnation that living standards for the average American have either stagnated or grown very slowly since the early 1970’s relative to the earlier part of the 20th century because we have picked all the low-hanging fruit–the best ideas have been exploited and the new ones are less transforming.
The term “stagnation” actually means static, stuck in a rut, not progressing and so on. Tyler often concedes that there has been growth in living standards since say, 1973, but that the rate of growth has slowed. This is an easier thesis to defend. But recently Tyler invoked the lack of change in median income since 1973 has evidence for his thesis. I challenged him in this post to explain what he means by that claim:
Does it mean that the person who was the median or family in 1973 continued onward at a constant standard of living without any gains despite enormous gains in per capita income? This is the way the story is usually told–the rich (as if they were a fixed group of individuals, an exclusive club) somehow managed to gain all of the gains of the intervening 38 years for themselves. This is clearly not true. If you look at any data that follows the same people over time, you will see that their lives improve as they get older and that they are typically better off than their parents. Better off in absolute terms, not relative ones. Some people move up relative to others. Some move down. But the entire distribution moves up.
Or does it mean that the typical family or individual in America today has the same standard of living as people back in 1973? Is the median a surrogate for the middle class? This is a different claim from the first one. The problem with this claim is the types of people in the middle in 1973 are different from the types of people now. There was a major demographic change in the 1970′s. The divorce rate exploded. Suddenly (and it was pretty suddenly) new households were created as couples divorced. The rate of household creation grew faster than population.
In the rest of that post, I argue that the rise in the divorce rate changes the number of families, particularly below the median, lowering the measured median and distorting the measure of progress. I also mentioned that inflation is overstated (thereby understating the rise in living standards) and I closed with a discussion of the returns to education.
Tyler has responded at some length, which I much appreciate. I will quote the middle of his response verbatim and comment along the way:
After introducing my discussion Tyler makes the following arguments:
1. I discuss household size in the footnotes to TGS. Adjusting for it doesn’t make a huge difference and furthermore the rapid-median-income-growth 1960s were a time when household size was falling quite rapidly. I blogged some of the details here.
The size of the households isn’t the issue. It’s the growth in the number of households and the kind of households they are. Yes, households are smaller than they were in 1973, so there is larger per-capita income within the household. That is the effect that is small. But the real issue is the growth in the number of households relative to population that pushes the median down (Taken from the decennial Census here):
The divorce rate exploded in the 1970’s. That’s one reason that 1973 (or sometimes 1978) looks so good in the data. After that time, there was a huge increase in the number of households headed by single adults. Many of the female heads of household were not expecting to work. They now did. They didn’t earn as much as the median and lowered the measured median income. (In this post I show that the same thing is going on in Canada.)
2. Immigration doesn’t seem to shift the median enough to create an illusion of stagnation, I blogged the numbers and details here.
I think Tyler is probably right. It’s a factor but not anything close to decisive.
3. CPI bias has likely fallen over time, which will make the true median income growth differential over time even greater than the numbers indicate. Furthermore CPI measures are getting better over time and doing more to adjust for quality biases; that’s further bad news. Most of all, a lot of CPI bias is offset by ‘wasteful spending on health care, education, defense, and government yet all counted in gdp” bias.
I have no idea if the bias has fallen over time. The estimates of Mark Bils are quite substantial. Here is the conclusion from his paper on quality changes and price measurement:
My results suggest that quality growth for durables has been understated by almost 2 percent per year since 1988. With this addition, quality for durables, even excluding computers, increased by 2.5 percent per year from model changes, with higher rates of 3.3 percent for vehicles and 4.4 percent for consumer electronics. To judge overall quality growth for durables it is important to also include the impact of consumers moving to better products, e.g., from a midsize to luxury sedan or from conventional to plasma television. The analysis of price changes across BLS sample rotations suggests this contributes another 2.3 percent annually to quality. Added to the estimated 2.5 percent from model changes, this yields overall quality growth for durables, excluding computers, of nearly 5 percent per year.
These are huge differences. Yes, the BLS has tried to correct its numbers for quality changes. But Bils’s analysis (and he is not alone) using data from 1988–2006 suggests they have not been very successful. How strange is it that the pace of improvement in the products we enjoy makes it harder to measure the price level, which in turn makes a time of growth and innovation look like a time of stagnation? (Check out the latest EconTalk with Bruce Meyer for more on how the overstatement of inflation has led to an understatement in the growth of the middle class’s standard of living.)
4. Russ doesn’t mention the internet but it’s getting more monetized — and thus more counted in gdp — all the time. The consumer surplus of the unpriced parts, once you eliminate double-counting, probably isn’t much more than two percent of income. Not “two percent growth a year” but two percent period. I could see it being three or four percent, for sure, but that still won’t overturn the basic slowdown.
Could be. I agree that the internet is being monetized. But the consumer surplus remains large.
5. Rising household debt and abysmal job creation since 2000 suggest to me that the quantity data are in line with the incomes data. Around 1999-2000, stagnation suddenly becomes much worse. The only good years since then are the bubble years, whereas across 1973-1998 there are some truly good economic years (partially offset by some very bad ones).
6. 1995-1998 are a poster child for what a non-stagnating period should look like in terms of wages and median income. Lots and lots of years since 1973 don’t look anything like that period. When the growth is real, it shows up in all of the standard numbers and no mystery variables or invocations of biased measurements are needed. I find this comparison illuminating.
I don’t get this. Tyler seems to be saying that when times are good (such as the 1990’s) we shouldn’t have to worry about mismeasuring inflation or distortions from family size. The gains should be obvious. So they’re not. When inflation is badly mismeasured or demographics are changing rapidly, it’s hard to assess progress by the median.
7. I discuss benefits in the book, for the time being I’ll note a) cradle-to-grave private sector jobs, with union-based pension benefits are rarer than they used to be, b) fewer people get health care through their jobs than used to be the case, c) most of the benefits are health insurance but don’t fixate on the size of the expenditure, rather consider that health progress has been slowing down, and d) last year health insurance costs rose by nine percent and no way should that be interpreted as equivalent to an increase in real income, rather it is a sign of system failure. That all said, the text of TGS still leaves room open for a world where virtually all of the benefits of economic growth accrue to the elderly. Such a world still will have a lot of TGS properties.
I don’t have much to say on this issue other than to note that longevity is not the only measure of health and health insurance is not the only measure of health care. I certainly agree with Tyler that by subsidizing health care in all kinds of ways, we have wasted resources.
8. Consumption data often selectively focus on the commodities which have become much cheaper (e.g., flat-screen TVs) and ignore the growth in debt, which now must be paid back.
True. We’ll have to see what happens when people save more to get back to a sustainable path of consumption.
9. The 2000-2011 case for stagnation is stronger and clearer than the 1973-2011 and there also has been more growth along the latter and longer period of time, plus numbers are easier to interpret across shorter time stretches. I will ask Russ if he at least can buy into TGS for the last eleven to twelve years.
Yes, numbers are easier to interpret over shorter periods. When you pick the end of a boom as your start date (2000) and the near-trough of the worst recession since the Great Depression, incomes will fall for many people. That is not The Great Stagnation.
I don’t see panel data as offering a significantly different story from the above but if Russ tosses me a specific citation I will consider it.
Panel data that tell a different story? Go here for one example: when you follow the same people over time, you see that since the late 1970’s, children grow up to have a much higher standard of living than their parents, (even with lousy measures of inflation) and the biggest gains are for the poorest people. This is inconsistent with the Great Stagnation.
The other challenge to the Great Stagnation is that per-capita GDP is way up since the 1970’s. The left argues that the rich got all the gains. The mechanisms they propose to explain this (lower rates of unionization, slow growth in the nominal minimum wage) are not convincing. Unionization rates have been falling steadily since the 1950’s and the minimum wage never covered enough people to make it important. How does the left (or Tyler) then explain this disconnect between national growth and the effect on the middle? They don’t have a convincing story.