Uneasy Money

Nick Rowe and Scott Sumner have recently had an interesting little debate about whether the slowdown in real GDP growth and labor productivity since the 2007-09 downturn is the result of cyclical or secular factors. Nick argues that successful inflation targeting in the two decades before the 2007 downturn had given rise to entrepreneurial expectations of stable aggregate demand, thereby providing a supportive macroeconomic environment for long-term investment that generates rising labor productivity over time. By undermining confidence in macroeconomic stability, the 2007-09 downturn diminished the willingness of businesses to continue make long-term investments and thus compromised one of the institutional pillars supporting long-term investment and productivity growth. Despite a recovery, expectations of future aggregate demand are now held with less confidence – higher perceived variance – than previously, thereby reducing entrepreneurial willingness to commit to the long-term capital expansion that increases productivity.

Scott is skeptical of the argument, because productivity growth had already started to decline after the 2001 downturn. Of course, one could argue that geopolitical uncertainty after the 9/11 attack and the invasions of Afghanistan and Iraq could have had a similar depressing effect on investment well before the 2007 downturn. So the decline in productivity growth that was underway at the time of the 2007 downturn is not necessarily inconsistent with Nick’s basic story. But Scott at least partially defends himself against that response by showing that real long-term investment as a share of GDP rose sharply after the 2001 downturn and was well above the levels of 1950s and 1960s.

Seeing no reason why the pace of productivity growth couldn’t have been affected by both cyclical and secular forces, I am happy to agree with both Nick and Scott. But I also have my own theory about the slowdown in productivity growth, which I have discussed previously, so this seems like a good time to weigh in again on the topic. As I pointed out in a 2015 post, one characteristic that distinguishes the 2007-09 downturn from earlier downturns is that it was associated with relatively large sectoral shifts in demand. Thus, the 2007-09 downturn was characterized by a higher percentage of jobs lost in the downturn that were not subsequently restored than was the case in earlier downturns. In earlier downturns, the decline in aggregate demand caused workers to be laid off temporarily from their jobs when demand and output fell, but a large percentage of laid-off workers were later rehired by their former employers when demand and output recovered. And even many of those laid-off workers that weren’t rehired by their previous employers still eventually found jobs doing work very similar to what they had been doing before losing their old jobs.

The depth and the severity of recessions can be measured not just by the unemployment rate, but also by the long-term unemployment rate. What set the 2007-09 downturn and the recovery apart from earlier downturns — even the 1981-82 downturn, in which the unemployment rate rose to almost 11 percent, higher than the 10 percent rate at depth of the 2007-09 downturn – was a long-term unemployment rate substantially higher, followed by a slower rate of decline, than in any post-World-War II downturn. I quote from a recent article on long-term unemployment

In January 2017, there were 1.85 million long-term unemployed. The number first dropped below two million in May 2015. That means 24.2 percent of the unemployed have been looking for work for six months. That’s better than the record high of 46 percent in the second quarter of 2010.

Sadly, it’s barely better than the darkest days of the 1981 recession. At that point, 26 percent of the unemployed were out of work for more than six months. On the other hand, total unemployment was worse than it is today. There was a 10.8 percent overall unemployment rate. In other words, the Great Recession created a higher percent of long-term unemployment.(Source: “Potential Causes and Implications of the Rise in Long-Term Unemployment,” The Federal Reserve Bank of Richmond, September 2011.)

Here’s how I put it in 2015.

[T]he 2008-09 downturn was associated with major sectoral shifts that caused an unusually large reallocation of labor from industries like construction and finance to other industries so that an unusually large number of workers have had to find new jobs doing work different from what they were doing previously. In many recessions, laid-off workers are either re-employed at their old jobs or find new jobs doing basically the same work that they had been doing at their old jobs. When workers transfer from one job to another similar job, there is little reason to expect a decline in their productivity after they are re-employed, but when workers are re-employed doing something very different from what they did before, a significant drop in their productivity in their new jobs is likely.

In addition, the number of long-term unemployed (27 weeks or more) since the 2000-09 downturn has been unusually high. Workers who remain unemployed for an extended period of time tend to suffer an erosion of skills, causing their productivity to drop when they are re-employed even if they are able to find a new job in their old occupation. It seems likely that the percentage of long-term unemployed workers that switch occupations is larger than the percentage of short-term unemployed workers that switch occupations, so the unusually high rate of long-term unemployment has probably had a doubly negative effect on labor productivity.

Long-term unemployment has adverse effects on health and many other metrics of well-being, effects that aren’t confined to the unemployed, but extend to their families, friends and communities. An increase in long-term unemployment, even if originally caused by an aggregate demand shock, is associated with a long-term negative supply shock. So it’s not surprising that the unusually and persistently high rate of long-term unemployment after the 2007-09 downturn, causing a massive loss of human capital, has depressed the subsequent growth in labor productivity. In my 2015 post, I tried to provide an optimistic interpretation of this phenomenon, but my optimism was misplaced, because the damage inflicted by long-term unemployment is very often irreversible, and rates of long-term unemployment have remained stubbornly high notwithstanding the steady decline in the overall unemployment rate.

Accounting for a disproportionate share of the long-term unemployed, discouraged older workers, chronically unable to find new jobs, have prematurely departed from the labor force. These older workers have presumably been replaced by younger entrants into the labor force, and one would suppose that the productivity of the younger workers is, on average, substantially lower than the productivity of the older and more experienced workers whom they have replaced, though presumably as they gain experience and acquire skills, the productivity of new workers will rise over time. Thus the demographic shift in the labor force is another reason for the low productivity growth since the 2007-09 downturn. But that effect, though largely demographic, has also had a cyclical component, making it difficult to disentangle the cyclical from the secular causes of sluggish productivity growth.

That difficulty is further compounded by another contributory cause of slow productivity growth. In my 2016 post, I discussed the late Walter Oi’s idea that labor is not really a variable factor of production as it is typically treated in simplified models, but a quasi-fixed factor. Here’s how Oi explained the idea:

For analytic purposes fixed employment costs can be separated into two categories called, for convenience, hiring and training costs. Hiring costs are defined as those costs that have no effect on a worker’s productivity and include outlays for recruiting, for processing payroll records, and for supplements such as unemployment compensation. These costs are closely related to the number of new workers and only indirectly related to the flow of labor’s services Training expenses, on the other hand, are investments in the human agent, specifically designed to improve a worker’s productivity.

The training activity typically entails direct money outlays as well as numerous implicit costs such as the allocation of old workers to teaching skills and rejection of unqualified workers during the training period.

So, if the 2007-09 downturn and the recovery was associated with an unusually high flow of workers from old jobs into new jobs, there has been an unusually high level of training expenses incurred by firms as they have brought workers into new jobs. The large investments by firms in training new workers have inevitably caused measured labor productivity to lag below previous trends when the fraction of workers entering the labor force or requiring new training to learn new skills was likely less than it has been since 2009. This idea, at any rate, does provide some reason to hope for at least a modest improvement in productivity and economic growth over time, even if the human cost of almost a decade of extremely high long-term unemployment is now largely irremediable and irretrievable.