So far, I have focused on unemployment as a problem of adjustment. In this essay, I review the history of the Dotcom recession, and I focus on the index of aggregate hours worked as an indicator of macroeconomic performance. According to this indicator, the Dotcom recession was quite severe. This in turn raises interesting questions about other indicators and about the policy history.
Below is a table showing the index of hours worked in the nonfarm business sector for the three deepest post-war recessions, starting with the pre-recession peak and moving forward eight quarters (two years) at a time.
Hours Index at Peak
2 years Past Peak
4 Years Past Peak
6 Years Past Peak
Using this measure, the Dotcom recession seems to have been much worse than the previous two, particularly in terms of length. It was not until the fourth quarter of 2006, 6-1/2 years after the peak, that the hours index edged above its peak level. It stayed above the earlier peak for four quarters, and has since slipped back below.
The reason that most people do not look at the Dotcom recession as being a severe one is that the flagship indicators, GDP growth and the unemployment rate, were better behaved during the Dotcom recession. GDP growth was reasonably robust, because productivity growth was high. The unemployment rate remained low, because labor force participation fell.
Overall, my sense is that the Dotcom recession saw a change in the composition of the labor force that favored higher productivity. For example, from the second quarter of 2000 to the second quarter of 2006, the number of employed persons aged 25 and older with a high school degree went from 36.7 million to 36.9 million, essentially no change. Over that same period, the number of employed persons aged 25 and older with a college degree went from 36.0 million to 41.3 million, an increase of more than 14 percent.
Interestingly, the labor force participation rate of college graduates declined by more than that of those with just a high school degree. The participation rate of the latter declined from 64.4 percent to 63.4 percent, while the participation rate of the college graduates fell from 79.8 percent to 77.6 percent.
1. The work force became weighted more toward the college educated. Presumably, those who aged out of the labor force and retired were disproportionately less educated than those who joined the labor force.
2. A relatively high proportion of the college-educated dropped out of or failed to enter the labor force, as indicated by the decline in their labor force participation rate. This decline presumably reflected a more difficult job market during the Dotcom recession.
3. Nonetheless, the composition of employed persons shifted markedly toward more employment of those with a college education. This may account for some of the strong productivity gains. However, I should point out that the more educated labor force tends to have occupations where hours and output are both harder to measure than the classic factory job. Both the numerator and the denominator of output per hour are less reliable.
4. Perhaps what makes the Dotcom recession unusual is the behavior near the peak. It could be that there was an unsustainably high level of employment as a result of Internet companies taking on large numbers of workers in spite of having little or no revenues. In a more realistic job market, more people choose to remain out of the labor force.
5. The severity of the Dotcom recession as measured by the hours index makes it difficult to criticize the monetary policy of 2001-2004 from a conventional Keynesian perspective. If you were to use hours worked rather than the unemployment rate as an indicator of labor market conditions, then you certainly would argue for a policy of monetary easing.
However, in terms of these lectures, to discuss monetary policy is to get ahead of our story somewhat. So far, I have hinted that when there are severe structural imbalances that require wage cuts, expansionary monetary policy might serve to produce the required adjustments by raising prices relative to wages. However, it seems far-fetched to suggest that what we needed from 2000-2006 was inflation in order to cut the real salaries of college graduates so that employers would retain them. The period of 2000-2006 would seem to me to be long enough for a lot of labor market adjustments to work out on their own. If college graduates were not participating in the labor force in 2006 to the extent that they did in 2000, I am reluctant to attribute this shift to inadequate macroeconomic stimulus.
In fact, I am extremely cautious about any conventional macro stimulus theory, given the complexity of labor market dynamics these days. But, again, that is getting ahead of the story.