Secular Stagnation in the US

Robert E. Hall 22 April 2015

a

A


This column is a lead commentary in the VoxEU Debate "Secular Stagnation"


The disappointing post-crisis performance of the US economy and even more disappointing performance of continental Europe and Japan have revived interest in the possibility of secular stagnation. Under stagnation, real incomes fail to grow or even shrink, and the economy’s output falls farther and farther below its earlier upward trend. Rising unemployment may also occur. Summers (2014) ignited interest in the possibility of secular stagnation.

One important factor in stagnations is the inability or reluctance of the central bank to lower interest rates as low as would seem to be appropriate, given the ability of low rates to stimulate output and employment. The Federal Reserve and the Bank of Japan have kept rates slightly positive since the crisis, while the ECB did the same until recently, when it pushed the rate just slightly negative. All three economies had combinations of high unemployment and substandard inflation that unambiguously called for lower rates, according to standard principles of modern monetary economics.

Extreme slack persists in continental Europe and Japan, but in the US, several labour-market indicators, such as low short-term unemployment and high levels of unfilled job openings, indicate the end of the period of slack that followed the crisis, while others, such as long-term unemployment and involuntary part-time work, still show slack but are declining and will probably reach normal levels in the coming year.1 Forecasters believe that the Fed will unpin the short-term interest rate in the middle of 2015 or a bit later in the year. Markets for forward rates agree.

European versus US secular stagnation: Demand versus supply factors

Thus a consensus is forming that inadequate demand will no longer be a factor in whatever US stagnation occurs in coming years. In Japan and Europe, on the other hand, the case for boosting demand is strong and inadequate demand is almost surely a main cause of the stagnation.

Despite the resumption of normal conditions in the US labour market and the consensus that slack is gone, the US economy is stagnated in the sense that the standard of living stopped growing around 2000. Family purchasing power today is just the same as in that year. Figure 1 shows that it grew briskly during the 1990s, slowed markedly prior to the crisis, dropped below its 2000 level as a result of the crisis, and grew slowly in recent years.

Figure 1.

Two episodes of low purchasing-power growth despite a growing economy appear in the figure. From 2002 through 2007 (recovery from the 2001 recession), and 2010 to 2013 (the recovery from the 2008-09 Great Recession). The unemployment rate reached 4.8% in 2007 – well below the long-run average rate of 5.8% and is right at that long-run rate today. The evidence is strong that inadequate demand is not behind the general stagnation of purchasing power, though it was a factor in the period immediately following the Crisis. As of 2014, the US has had a decade and a half of a new kind of secular stagnation, one associated with declining supply.

Causes of US secular supply stagnation

Four factors account for the stagnation of purchasing power in the US economy: 1) declining labour share; 2) depleted capital; 3) reduced productivity growth; and 4) declining labour-force participation. I will discuss indexes that capture each of these factors in turn using indices that all start at unity in 1989. An index of total purchasing power from earnings is the result of multiplying the four indexes together.

  • Labour’s declining share of income.

Figure 2 shows an index of labour’s share (including fringe benefits) of total US income. It tends to be level in recessions, fall during the first half of ensuing expansions, then rise back to a high level at the next recession. But superimposed on that pattern is a general decline that cumulates to about 10% over the period. Like the general declining trend in earnings, the decline in the share seems to have started around 2000. Economists have pursued multiple explanations of the decline, but no consensus has formed.

Figure 2.

  • Slow overall productivity growth.

Figure 3 shows that productivity grew rapidly from 1989 to 2007. The Great Recession caused a dip in productivity, as did past recessions (due mainly to idle facilities). Though productivity grew at normal rates during the recovery, it did not make up for the shock of the crisis, so the average growth since 2006 has been below par.2 Household earnings suffered proportionately. See Fernald (2014) for further discussion of productivity.

Figure 3.

  • Depleted capital per household.

Figure 4.

Figure 4 shows the third factor – the amount of capital available to equip the average worker. With more plant, equipment, and software, workers earn more. Capital per household rose rapidly during the 1990s, but more slowly after 2000. Capital per household actually fell during the Great Recession, and its more recent growth has not come close to placing capital per household where it would have been if the trend of the 1990s had continued.

  • Low labour-force participation.

Figure 5 displays the average household’s involvement in the workplace as measured by an index of annual hours of work of household members. Hours per household grew rapidly until 2000, fell as usual during the recession of 2001, flattened but did not grow during the boom of 2002 through 2007, unlike previous booms, collapsed in the Great Recession, and have risen during the recovery that is still underway. The decline in hours since 2000 is the single biggest factor in the decline in household earnings.3 Recent growth in hours per household offers some hope for the return of earnings growth in coming years.

Figure 5.

Why hours worked declined

Because declining hours account for the biggest part of the stagnation of earnings, I will dig deeper, by breaking them down into three components: Labour-market participants per household; fraction of participants working; and hours per worker.

  • Figure 6 shows an index of participants per household.

Figure 6.

As the chart illustrates, participation rose during the 1990s, especially in the second half of the decade, but has fallen since. The Great Recession depressed participation only slightly and does not appear to have been an important determinant of the overall decline in involvement in the labour market. Of course, the recession was a time when fewer participants were actually working and more were looking for work.

Economists have been working hard on trying to understand the surprising decline in participation, which exceeds forecasts that were made in earlier years. Most research agrees that the slack labour market had a relatively small discouraging effect. Another suspect that has been found to have at most a small role is changes in the composition of the working-age population – the negative effect of aging of the population on participation just offsets the positive effect of higher educational attainment. A large increase in the fraction of households subject to taxes imposed on families benefiting from food stamps, disability, and other safety-net programs may be a factor.

  • Figure 7 shows an index of the fraction of participants who were actually working – the remainder were unemployed and actively looking for work.

This factor was flat on average, falling in recessions and rising in the ensuing recoveries. It has risen recently, as unemployment has fallen to the upper-five-percent range. It is not an important element of the stagnation of earnings as of today.

Figure 7.

  • Figure 8 tracks hours of work per week for the average household.

It was quite constant over most of the period, but fell sharply during the Great Recession and recovered only about half of the decline since. It is too early to judge whether hours per worker will return soon to its earlier level or remain as an element of the stagnation of earnings.

Figure 8.

Work versus other time uses

Some indication about the changing balance between work and other uses of time comes from the American Time Use Survey, which began in 2003. Table 1 shows the change in weekly hours between 2003 and 2013 in a variety of activities.

Table 1.

  • For men, the biggest change by far is the decline of 2.5 hours per week at work, a big drop relative to a normal 40- hour work week.

A small part of the decline is attributable to higher unemployment—the unemployment rate was 6.0% in 2003 and 7.4% in 2013.

  • The decline for women is much smaller, at 0.8 hours per week.

For both sexes, the big increases were in personal care (including sleep) and leisure (mainly video-related activities). Essentially no change occurred in time spent in education. Women cut time spent on housework.

Is there hope for a return to normal growth of household purchasing power?

Capital seems likely to continue to return to its historical growth path, as Figure 4 suggests. For the three other major categories, forecasting is a challenge. There has been no sign of a reversal of the decline in labour’s share of total income and no body of research that supports the idea that it will. Productivity growth is definitely under way, at rates similar to those in the 1970s and 1980s, but well below the rates of the 1950s, 1960s, and 1990s. In particular, there is no sign that a burst of productivity growth will make up for the complete stall in productivity growth around the crisis, as Figure 3 shows.

Most importantly, there is no sign suggesting a departure from the decline in labour-force participation shown in Figure 6. Some commentators have declared a turnaround in participation based on recent monthly data, but Figure 9 suggests this is wishful thinking. Participation has declined along a straight line during the period of improving conditions in the labour market, suggesting a complete disconnect between participation and the state of the labour market.

Figure 9

One possibility for growth in purchasing power is that unemployment may dip below 5.5% -- the level that some believe defines full employment. The unemployment rate reached 3.8% in 2000 and 4.4% in 2007, in both cases at the ends of long expansions, without triggering inflation much above the Fed’s target of around 2%.

I reiterate that these conclusions apply to the United States. In continental Europe, the case is strong that demand has far from recovered. In Japan, unemployment is at low levels but the performance of the economy is substandard.

References

Fernald, John (2014). “Productivity and Potential Output Before, During, and After the Great Recession”, NBER Macro Annual, 2014, forthcoming.

Lawrence H. Summers, “U.S. Economic Prospects: Secular Stagnation, Hysteresis, and the Zero Lower Bound” Business Economics Vol. 49, No. 2  National Association for Business Economics

Footnotes

1 Complete backup for all of the calculations is available from my website, stanford.edu/∼rehall

2 See Fernald (2014) for a discussion of the evidence.

3 See Foote and Ryan (2015).

a

A

Topics:  Macroeconomic policy

Tags:  secular stagnation, US labour market performance

Robert and Carole McNeil Hoover Senior Fellow and Professor of Economics, Stanford University and President-elect of the American Economic Association

Vox Talks