Section 3: Feature article

View the most recent version.

Cyclical changes in output and employment

by Philip Cross

A recurring question during economic downturns is the relationship between output and employment. Do changes in employment lag output growth? Do employers cut output faster than jobs during recessions? And have these relationships changed over time? This paper tries to answer these questions by comparing data on monthly and quarterly GDP and employment (from the Labour Force Survey). It also compares Canadian results with the US.

Comparing the year-over-year growth of monthly real GDP and employment since 1982 shows the two change direction in tandem most of the time (constant dollar monthly GDP data begins in 1981). More specifically, turning points in the growth of output and employment appear to have been virtually the same over the past three decades.

Turning points

The graph on year-over-year changes in output and jobs provide an overview of their relationship. However, a detailed examination of the timing of turning points requires the scrutiny of the actual monthly and quarterly data (just as the dates of peaks and troughs in the economy cannot be determined precisely by looking at year-over-year changes in GDP or jobs).

Since 1976, there have been two complete cycles when both GDP and jobs contracted in absolute terms. ¹  The quarterly turning points for output and employment were close in both the 1981-82 and 1990-91 recessions (Figures 2a and 2b). The initial drop in GDP in the third quarter of 1981 was accompanied by no change in employment, while the 0.3% fall in output in the second quarter of 1990 slowed employment growth to 0.1%. This suggests a slight lag of employment behind output during the onset of a recession. Thereafter, output and employment both fell in every quarter of these recessions, before turning up in unison (in the first quarter of 1983 and the second quarter of 1991).

While the turning points were almost identical, the initial contractions in real GDP were more pronounced than for employment. In the first two quarters of recession in 1981, real GDP declined 1.2%, double the 0.6% drop in jobs over the same period. And in the first two quarters of the 1990 downturn, real GDP fell 0.8% versus a 0.1% dip in employment. However, employment contracted nearly as fast or even faster than output for the duration of these recessions. A similar pattern appears to be occurring in the current cycle, with a sharper retreat in GDP than jobs late in 2008, but job losses matched the drop in output early in 2009.

Table 1 compares monthly turning points in GDP and employment. On average, turning points in output and jobs occurred within one month of each other during the recessions of 1981-82 and 1990-91.

The behaviour of employment in the early 1990s complicates the identification of its trough. After steady declines in the second half of 1990 and the first quarter of 1991, jobs grew by 0.2% in the second quarter of 1991 and held onto this gain in the third quarter. This accompanied a slow recovery in real GDP. However, by early 1992 GDP growth stalled, followed by anemic gains of 0.2% and 0.4% in the second and third quarters. Such marginal increases in GDP could not support any employment gains given even modest productivity growth, and employment fell slowly but steadily between the fourth quarter of 1991 and the third quarter of 1992 (almost entirely in Ontario). This may appear as lagging behaviour, but the renewed weakness of jobs late in 1991 was a response to faltering GDP growth at that time, not a delayed response to lower GDP early in 1991. This is supported by the increase in jobs for nearly six months after GDP bottomed in March 1991 (Figure 3).

Relative GDP and job growth

Figure 4 compares the year-over-year change in quarterly GDP from the Income and Expenditure Accounts ²  with employment since 1977 (the current LFS began in 1976). The quarterly data are smoother than the monthly data, as would be expected, which allows the underlying relationship between the two to become more apparent.

During both recessions and slowdowns, employment growth closely tracks GDP growth. A separation between output and job growth occurs mostly during periods when GDP grows 3% or more, when productivity gains are the easiest to achieve. The only periods when employment growth exceeded output were during the 1980 slowdown and the initial stage of the downturn in 2008. Both these episodes followed an extended period of labour shortages in some parts of the country, which may have caused employers to retain workers longer until the severity of the slump became clear.

The regional pattern of job growth supports the idea that shortages made employers hesitant to lay off workers in these two cycles. The slower response of employment in Canada to the weakening of the economy in 2008 originated largely in western Canada. Employment in central Canada peaked in the second quarter followed by a slow decline. In western Canada, jobs increased 0.8% and 0.5% in the first two quarters of 2008. However, growth stalled in the second half after commodity prices retreated. By the first quarter of 2009, jobs in the west were disappearing almost as fast as the 1.5% loss in central Canada. The 1980 slowdown also accompanied high commodity prices and shortages in western Canada.

The closeness of output and job growth in cyclical downturns may surprise some analysts. For a generation of economists, it was widely accepted that firms ‘hoarded’ labour during recessions. ³  That is, firms were said to not cut jobs as much as output (irrespective of whether these cuts lagged output or not). This theory was based on the costs of firing labour (such as severance pay) as well at the cost of searching out people to hire and train in the subsequent recovery. A variant of this theory was that there was an optimal time for firms to profitably hoard idle workers, after which they would be let go. ⁴ 

Instead, comparing output and job growth in Figure 4 implies that since 1977 firms in Canada appear not to have hoarded labour during recessions, nor even during mild slowdowns. This is consistent with other measures of firms becoming more efficient in reducing costs, such as the long-run drop in the inventory-to-sales ratio of firms in recent decades.

GDP and jobs in the US

Looking at output and employment growth in the US is revealing (Figure 5). Changes in non-farm payroll growth in the US clearly lag turning points in output. Moreover, the lag between quarterly changes in output and employment in the US has increased over time, from less than 1 quarter before 1973 to 4.1 quarters since.

Indeed, in the previous two cycles in 1991 and 2001, jobs did not begin to recover until a year after GDP started to rebound, although employment began to decline soon after GDP receded (Figures 6a and 6b). Again, the US experience is different than the Canadian, where there was little or no lag. The greater lag in the US data, which is widely scrutinized, may explain why some analysts persist in believing there is a definitive lag in Canada.

As for labour hoarding, since the 1982 slump firms in the US cut jobs more than output during recessions and the early stages of recovery. During 1991, 2001 and 2008, the gap between output and job losses in Figure 5 becomes progressively larger. This appears to reflect the ability of the US to raise productivity during recessions and especially early in recoveries, a result squarely at odds with the idea of labour hoarding during downturns. By contrast, employers in Canada have been able to keep productivity stable during slowdowns, but not to increase it as rapidly in the first year of recovery as in the US. The counterpart is that productivity growth later in expansions usually was greater in Canada than the US (although not in the current decade). This helps explain why productivity growth in Canada and the US is comparable over long periods of time. ⁵ 

Has the relationship between output and jobs changed over longer periods? While the revamped Labour Force Survey in Canada starts only in 1976, US payroll data extend back to World War II. Comparing GDP with employment clearly shows that firms in the US did not hoard labour during downturns in the 1950s and 1960s. During the six recessions between 1949 and 1969, the drop in employment without exception exceeded the decline in real GDP (Table 2). The two fell in tandem in 1974-75, partly because the memory of shortages kept employment growing early in the recession before severe job losses after September 1974. The mild recession of 1980 was the only post-war downturn where the drop in output clearly exceeded the employment decline.

The most recent downturn suggests that employment in the US continues to change in relation to GDP. Employment fell in December 2007 (and every month since), six months before GDP began to contract in the second half of 2008. As noted by Barry Bosworth, a productivity expert at the Brookings Institute, “Nowadays, [companies] seem to anticipate a decline in output and lay off workers ahead of time.” ⁶  Moreover, the drop has been steeper for jobs than output, continuing the trend set in earlier recessions.

Conclusion

Cyclical turns in output and employment appear to occur at about the same time in Canada. There often is a slight lag of employment early in a recession, especially after a period when employers have faced labour shortages. There was a lag between output and jobs in the US, especially in recoveries, but jobs led output into recession in the current cycle. Turning points in both GDP and jobs in Canada are closely synchronised with cycles in US GDP, but not employment in the US. There is little evidence from past cycles that jobs languish in Canada when output recovers significantly. However, the experience of the US suggests that behaviour may be different in the current cycle.

During downturns after 1980, employment essentially fell at the same speed as output in Canada. In the US, firms consistently have cut jobs faster than output as a result of downturns since World War II. In neither country is there strong evidence that employers hoard labour during downturns.

Next | Previous