Section 3: Feature article

View the most recent version.

The changing cyclical behaviour of labour productivity

by Philip Cross ¹ 

As the economy slumped in 2008 and 2009, labour productivity in Canada fell slightly as the combined reductions in employment and the average workweek did not match the drop in output. This marks a departure from recent recessions in Canada and the US, when labour productivity increased during recessions.

This paper compares the behaviour of labour productivity during the recent downturn to the previous three cycles in Canada and to the US. It finds that since 1981, labour inputs had fallen by as much or more than output during recessions in both Canada and the US. Employers initially reacted to a downturn by shortening the workweek, but then resorted to job cuts if the recession was prolonged. The time between the onset of lower output and job losses has shortened over time, to the point that early in 2008 employment fell before output receded in the US. In Canada, employers did not reduce labour inputs as fast as output in 2008-2009. As a result, labour productivity in Canada declined for over a year during a recession for the first time in over three decades. This extends the weak trend of productivity growth in Canada that has persisted since 2003, while productivity growth in the US has increased steadily, irrespective of the business cycle.

An overview of changes in output and employment

The relationship between output and employment during cyclical downturns has changed over the decades (output and employment are used as a proxy of labour productivity before 1981, when quarterly productivity data begin). From the 1950s through to 1980, employment in Canada did not decline as fast as output, if it fell at all. Employment declined much less than real GDP in the recessions of 1953-1954, 1957, and 1960-1961, and with a notable lag for the latter two. Employment did not drop at all during the mild downturns in 1970 and 1980. The 1974-1975 recession saw two quarterly declines in employment, but these were separated by five quarters of job growth averaging over 1% a quarter, and employment quickly resumed rapid growth in the second quarter of 1975. ² 

The resiliency of employment during recessions in Canada and elsewhere spawned widely-accepted theories that labour was hoarded by firms during downturns. ³  This hoarding occurred because recessions in the post-war period usually were short and related to the reduction of excess inventories. These brief periods of lower output did not justify firms absorbing the cost of shedding workers, which includes the cost of severance pay and the search and training costs of hiring workers a short time after recovery began. As a result, output per employee usually fell during recessions.

Hours worked declined more quickly than employment during most downturns in Canada and the US. The faster response of hours than employment to changes in the business cycle is well-established in economics (and is reflected in the inclusion of the average workweek in manufacturing in our system of leading indicators). Faced with a downturn in demand, employers will shorten the workweek before laying off workers for two main reasons. First, they usually are unsure about the magnitude and the duration of the drop in demand and therefore hesitate to lower jobs. ⁴  Second, a shorter workweek does not entail the fixed costs of letting workers go, such as severance pay. Until they are certain that the severity of a downturn justifies these costs, employers usually resort to shortening the workweek before reducing employment when confronted with a change in demand (total labour input equals the average workweek applied to total employment).

This paper examines what happened to quarterly output and labour inputs during each recession since 1981 using labour productivity data. This differs slightly from total output per employee for several reasons. The productivity data cover the business sector, which accounts for over 80% of GDP in both Canada and the US. The productivity data also make adjustments for hours worked. However, as Figure 1 shows, these differences do not materially alter the relation between year-over-year changes in productivity and output per employee. ⁵  The text flags when there are small differences between the two.

Beginning with the long and severe recession in 1981-82, labour productivity began to behave differently in both Canada and the US during recessions. While demand for labour initially lagged the drop in output in 1981, as the prevailing theory had predicted, by late in 1982 labour inputs had fallen as much (in the US) or more (Canada) than output. This meant that labour productivity was steady in the US and rose in Canada during this recession, something that had rarely occurred during recessions in previous decades.

The trend towards rising labour productivity continued in subsequent downturns in Canada and the US. The only exception was Canada in 2008, when labour inputs fell slightly less than the drop in GDP, resulting in lower labour productivity. As shown in the more detailed analysis that follows, not only has total hours worked followed output down during recent recessions, but the lag between falling output and lower labour inputs has become shorter over time, especially in the US.

The 1981-1982 recession

In 1981-1982, labour productivity in Canada initially stumbled, as employers struggled to adjust their labour inputs in the face of falling output. At first, all of the reduction was made by shortening the workweek. By the third quarter of 1981, most of the drop in labour inputs was coming from fewer jobs. Still, labour inputs did not fall as fast as output, and labour productivity fell through the end of the year. By early 1982, steep cuts to employment and a steady drop in the workweek began to raise labour productivity. By the end of 1982, labour productivity was 2.6% higher than at the start of 1981 despite a 6.0% drop in real GDP in the business sector (Table 3.1). ⁶ 

The comparable statistics for business-sector productivity in the US show that the average workweek also began to decline before employment in 1981. Average hours worked shrank in both the second and third quarters of 1981, while employment edged up and output posted a small net gain. When GDP decisively fell into recession in the fourth quarter, the cuts to labour input quickly shifted to employment. Jobs then bore the brunt of the recession in 1982, as hours levelled off. The net result was to lower labour productivity slightly during the recession in the US, unlike the increase posted in Canada during 1981 and 1982. This mostly reflected much steeper job cuts relative to GDP in Canada than in the US (6.3% versus 2.3%).

The 1990-1992 downturn

In the second quarter of 1990, firms in Canada’s business sector responded to the initial drop in output by reducing employment and average hours equally (for the economy-wide data, hours led the decline in employment). By the third quarter of 1990 and continuing through the worst of the recession early in 1991, the bulk of the reduction in labour inputs was made by lowering employment. By early 1991, productivity was 0.9% below its level a year-earlier.

Although output levelled off over the next four quarters, employment continued to decline until mid-1992 while the workweek was little changed, restoring labour productivity to its pre-recession level by the end of 1991. With total hours worked down 5.8% as employment continued to fall until late in 1992, labour productivity was 2% higher than at the beginning of the recession.

Unlike Canada, employers in the US responded to the onset of the recession in 1990 by trimming hours more than employment. However, by 1991 they had shifted their reduction in labour inputs more to jobs, as average hours levelled off. At the worst of the recession in the first quarter of 1991, productivity had risen 0.7%. By keeping both hours and jobs at a low level as GDP accelerated late in 1991 and into 1992, productivity at the end of 1992 was nearly 8% above its pre-recession level. This compares with a 2% gain in Canada, despite larger cuts to both hours and jobs in Canada. The difference largely reflects the slower growth of business sector GDP in Canada, which by the end of 1992 was still nearly 4% below its pre-recession peak versus a 5% gain in the US.

The 2001 slowdown

In 2001, there was a recession in the US but technically only a slowdown in Canada. This is partly because the downturn originated in the bursting of the ICT bubble in the stock market and in business investment, both of which had a greater weight in the US economy than in Canada. In the US, real GDP fell in the first and third quarters of 2001 (the latter aggravated by the September 2001 terrorist attacks). Employment receded with a lag, starting to decline in the second quarter of 2001 but falling steadily through the end of the year. In Canada, real GDP posted only a marginal decline in the third quarter of 2001 (mostly due a 0.6% drop in monthly GDP in September resulting from the disruption to economic activity from the terrorist attacks).

In the 2001 slowdown, both output and employment in Canada’s business sector continued to rise slowly, except for a mild dip in the third quarter. However, firms lowered the average workweek in the first half of 2001, helping to boost labour productivity by 1.7% over the whole period. Firms resorted to job cuts rather than fewer hours during the third quarter, when there was a great deal of uncertainty about how the US economy would respond to the 9/11 attacks. Real GDP growth quickly resumed in the fourth quarter, and payrolls began to expand. Overall, productivity rose more during the 2001 slowdown than during the first year of the two previous recessions on both sides of the border.

The larger gains in productivity in the US than in Canada first seen in the recession starting in 1990 continued in 2001. Labour productivity in the US rose 3.5% during 2001, versus the 1.7% gain in Canada. This occurred despite a slightly larger gain in business sector GDP in Canada, which usually helps boost productivity. American firms achieved higher productivity by steady reductions in the workweek throughout the year and cutting employment by 0.9% after the first quarter of 2001 (jobs fell another 1.1% after the recovery began late in 2001). It is notable that the drop in employment in the US matched the decline in GDP after only two quarters, faster than the time it took to catch up to GDP in the previous two recessions.

In both Canada and the US, this was the only cycle after 1980 where hours worked fell faster than employment during the downturn. The reliance of employers in both countries on a shorter workweek in 2001 suggests they viewed the slowdown in demand as largely transitory, a view borne out by events. The greater reliance of employers in the US than in Canada on a shorter workweek than on job cuts is noteworthy: the ICT boom in the US that peaked late in 2000 led to an historically low unemployment rate that was unmatched in Canada until the peak of the resource boom in 2007. In both instances, employers in the subsequent downturn resorted more to reducing hours worked than to letting go their hard-won employees.

If firms are hoarding labour less during downturns, this should be reflected in slower growth in productivity during the recovery. This was the case in Canada, where productivity growth in the first year of recovery slowed from 2.5% in 1983 to 1.8% in 1993 to 0.2% in 2002. A similar slowdown over time was evident in recoveries in the US.

The 2008-2009 recession

The most recent recession started in January 2008 in the US when employment began to contract in a wide range of both goods and services industries. Employment in the US declined 0.8% in the first two quarters of 2008, while output was unchanged before it turned down decisively in the third quarter of 2008. This was the first time employment in the US fell in advance of output, and productivity rose. At the end of 2008, the recession in US GDP was nearly equal to the 3.0% drop in jobs. The workweek fell at the same time, raising productivity, but the workweek shrank less rapidly than the drop in employment throughout the recession. By mid-2009 preliminary data showed real GDP had levelled off and then began to increase even as employment fell steadily.

In Canada, the behaviour of output and jobs in 2008 was the opposite of that in the US. Business-sector output dipped 0.6% over the first two quarters of 2008, while employment rose 0.7% (the workweek edged down 0.4% in response to the weakening of output). This was the reverse of the US, where jobs fell while GDP and the workweek only levelled off early in 2008 (total GDP eked out a 0.2% gain in the first half of 2008). And while a drop in jobs preceded the contraction in output in the US, in Canada employment declined two quarters after output began to recede. As a result, productivity in Canada fell 1.0% in the first half of 2008.

Late in 2008 and in the first two quarters of 2009, the drop in output in Canada intensified to 3.0% in the business sector. While employment fell more than the workweek, total labour inputs only matched the recession in GDP, and productivity was unchanged.

Since the fourth quarter of 2007, labour productivity in Canada fell 1.2% through the third quarter of 2009, despite firms having almost two years to adjust to lower demand. Over the same period, labour productivity rose 4.9% in the US. This gap does not represent a new trend that can be attributed to exceptional cost-cutting by US firms during a severe recession. Instead, it represents the extension of a trend that has persisted since the 2001 recession ended. Higher productivity growth in the US first materialized in the 2001 recession. This gap continued at 2% during the recovery in 2002, and has persisted at about 2% a year since. ⁷ 

In the US, productivity edged up 1% in 2008, but then jumped 3.8% in the second and third quarters of 2009. As noted earlier, firms in the US laid the groundwork by cutting employment early in 2008, before GDP and the workweek began to fall. Job cuts in the US intensified throughout 2008 and 2009 (reaching nearly 7%), while the workweek fell 2.3%. In Canada, firms matched this drop in the workweek, but reduced business sector employment only 2.5%.

Why did labour productivity decline in Canada during the most recent downturn, after rising in all the slowdowns after 1981? First, Canada was in a prolonged period of weak growth in output per employee even before the recession started. As discussed in our March 2007 paper, ⁸  the resource boom led to the exploitation of increasingly marginal sources of production in mining (including the oilsands). Another reason may be the legacy of labour shortages from the boom in mining and construction of recent years. Already, these shortages had contributed to unusually weak growth in output per employee in 2006 and 2007 as industries increasingly hired workers with lower skills and productivity, such as youths in Alberta. In light of this experience, firms may have hesitated to let workers go late in 2008 when it was not clear how long nor how severely the recession would affect Canada, especially given the high reserves of cash that companies held and the resiliency of much of our financial system.

Nowhere are these cross-currents more evident than in Alberta, the epicentre of labour shortages during the boom. At the height of the boom in 2007, increasing demand from high-paying industries in Alberta such as natural resources, construction and professional services raised employment in these industries by over 10% from a year earlier. This siphoned off workers from relatively low-paying industries such as accommodation and food, recreation, and public administration ⁹  and other services. When the high-flying resource, construction and finance industries turned down sharply late in 2008, with job losses of over 10% year-over-year, the lower-paying services industries jumped at the chance to replenish their labour force. As a result, even in the midst of a recession, employment in every one of these industries in Alberta rose by over 10% in the year ending in November 2009, according to the labour force survey.

The reverse question applies to the US: why did jobs begin to fall earlier than output in 2008 and faster throughout the recession? As was the case in Canada, this was partly an extension of the recent trend in the US. Output per employee in the US rose at an unusually high rate during the 2001 recession, and this trend continued in the ensuing expansion and into the recession beginning late in 2007. As well, employers in the US relied more on reducing employment than hours worked.

The reasons for the greater reliance of employers in the US on job cuts in 2008-2009 while Canadian employers resorted about equally to changes in the average workweek and employment are not yet clear. Credit flows to firms in the US were more impaired, providing a greater urgency for firms to achieve significant cost savings. As noted earlier, employers initially rely on shorter hours than job cuts in a downturn given the uncertainty about the severity of the recession: the upheaval in the US financial system, which clearly signalled a sharp downturn, meant that there was less uncertainty in that country about the severity of the 2008 recession. Labour laws in the US may make for a lower cost in letting workers go than do the laws in jurisdictions in Canada. And the memory of the difficulty in finding workers in some parts of Canada may have led employers to wait until the severity of the recession was fully revealed (which turned out to be less than widely expected).

Conclusion

Figure 3 provides a summary of labour productivity in the last four cyclical slumps in Canada and the US. In the US, there is a clear trend towards productivity rising during recessions, with not even a momentarily decline in the last downturn. In Canada, productivity by the end of most recessions was higher than at the beginning, with the exception of the 2008-2009 downturn (which may not be over on either side of the border, despite the recent gains in output and employment).

Severe recessions provoked a range of responses in business sector productivity. In some instances, like Canada in 1981-1982 and the US in 2008-2009, firms cut jobs rapidly and raised productivity. In others, such as the US in 1981-1982 and Canada in 2008-2009, the loss of jobs was more muted and productivity declined.

A number of implications follow from the comparison of real GDP and employment and hours worked in cyclical slowdowns after 1980. Employers usually lower output and hours worked before reducing employment (the only exception was in the US in 2008, when employment fell before GDP). This is consistent with economic theory. However, employers implemented reductions to both hours worked and employment more quickly over time, with the exception of the mild 2001 downturn.

As well, this analysis shows that employers in Canada and the US rely relatively more on cuts to employment than on a shorter workweek during all but the mildest recessions. Over the past four downturns, employment in the business sector decreased by an average of 4.1% in Canada while average hours fell by an average of 1.6%, for a total decline in hours worked of 5.7%. In the US, employment on average dropped 2.8% in the past four recessions while the workweek shrank 1.7%, to reduce labour inputs by a total of 4.5%. ¹⁰ 

The relationship between output and employment has changed considerably in the post-war era. Until 1981, employment did not fall as much as output during recessions. Since 1981, output per employee has risen during most recessions in Canada and the US. The only exception to this trend was the 2008-2009 recession in Canada, when preliminary data showed that output per employee declined. This continued a trend of productivity in Canada lagging behind the US that has persisted since 2001.

Next | Previous