An international comparison of employment in recovery.

Holland, Dawn ; Kirby, Simon ; Whitworth, Rachel 等

While the global economic recovery remains fragile, output in most of the major advanced economies has been rising since mid-2009. Employment, however, tends to lag production, and unemployment continued to rise well into 2010 in many countries. The ILO estimates that the level of unemployment remains elevated by 30 million worldwide relative to 2007 (ILO-IMF, 2010). After taking into account global population developments, this points to a rise in the global unemployment rate of about 3/4 percentage point. As the advanced economies have been faced with the brunt of the global downturn, unemployment rates have risen far more significantly within the OECD area. The unemployment rate in this region remains 2.9 percentage points higher than at the beginning of 2008.

This note extends two earlier studies of labour market responses to the global economic crisis that have appeared in this Review (Holland, Kirby and Whitworth, 2009 and 2010). While the first two notes discussed employment responses to the downturn, this note focuses on employment responses to the recovery. For comparative purposes we restrict our analysis to the countries studied in the previous two notes, covering thirteen advanced economies.

In Holland, Kirby and Whitworth (2009) we demonstrated a simple rule of thumb between output growth and the unemployment rate in the OECD as a whole, based on Okun's approach. Using a dataset that spans from 1988-2008, this suggests that on average if output rises by 1 percentage point more than trend growth of 0.6 per cent per quarter, this is associated with a decline of 0.6 percentage points in the unemployment rate across the OECD economies. In our analysis of July 2009, at the depth of the global recession, unemployment at the OECD level had risen by somewhat less than anticipated by this simple rule of thumb.

The most recent figures for OECD output and unemployment in the second quarter of 2010 indicate that the level of GDP remains 1.9 per cent below the level in the first quarter of 2008, while the unemployment rate has risen by 2.9 percentage points. If we narrow our analysis to the period of recovery, output rose by 3.3 per cent in the OECD between the first quarter of 2009 and the second quarter of 2010, while the unemployment rate rose by a further 1 percentage point. Figures 1 and 2 put these numbers into context with developments in individual economies and with our rule of thumb.

Figure 1 illustrates the change in output from its pre-recession peak to the second quarter of 2010 against the change in the unemployment rate over the same period in a selection of countries. We include the regression line from the rule of thumb equation reported in Holland, Kirby and Whitworth (2009) to give an indication of which countries have over- and under-performed relative to expectations based on this rule. When assessing labour market performance by Okun's rule of thumb, it is important to bear in mind that there is likely to be a permanent loss of output associated with the financial crisis. Financial crises do not necessarily lead to a permanent loss of output. Barrell et al. (2010) emphasise that only one in four crises in the OECD have left permanent scars on the level of output. However, we assume that the current crisis will induce a permanent adjustment to risk perceptions and appetite, and expect a long-run loss in the level of output of about 3 per cent on average in the OECD (Barrell, 2009). We allow for this adjustment to the rule of thumb regression line illustrated in figure 1. (1)

[FIGURE 1 OMITTED]

[FIGURE 2 OMITTED]

Countries that lie below the rule of thumb line, such as Finland, Italy, Japan and Germany, have exhibited a relatively small rise in the unemployment rate given their output loss, whereas those that lie above the line have exhibited a greater rise in unemployment than expected. Notably, Australia, the US and Spain have all experienced what might be considered excessive rises in unemployment given their output declines. Ireland, Canada, France, Sweden and the OECD as a whole are all clustered closely around the rule of thumb line. This suggests that the unemployment rates in these countries are broadly in line with our estimate of their output gaps.

Figure 2 focuses on the period of recovery, showing the changes in output and unemployment from the trough of the recession in each country to the second quarter of 2010. Most countries in our sample lie above the rule of thumb line, indicating that the labour market recovery has not kept pace with the output recovery. The exceptions are Spain, Ireland and the UK. However, at least in the cases of Ireland and Spain, this is largely a reflection of the weakness of the output recovery rather than a sign of a strong labour market. Output in both countries continued to decline until the end of 2009, and edged up only slightly in the first half of 2010. Employment continued to decline in both countries into the second quarter of 2010.

While the unemployment rate is a useful conventional measure of labour market slack in the economy, a rise in the unemployment rate can indicate either job losses, or it can reflect new entrants into the labour force who have yet to find employment. The two have very different implications for the productive capacity of the economy. Any adjustment in working time will also not be reflected in the unemployment rate, but does affect both income levels and potential output. In order to assess these important labour market developments, it is useful to monitor total labour input as well as the rate of unemployment. We define labour input as the total number of hours worked, or employment multiplied by the average hours worked per employed person in the economy.

Figures 3 and 4 plot the percentage change in output and labour input since the onset of the recession and since the onset of the recovery, respectively. The shading of the labour input columns allows us to distinguish between shifts in the level of employment and shifts in average working time. Output and labour input both remain below their pre-recession peaks in almost all the countries in our sample. Australia and Canada are the two exceptions. Australia suffered very little in the way of output loss during the global downturn, while Canada has already regained its pre-crisis level of output and surpassed its pre-crisis level of employment. Ireland has experienced the sharpest fall in both output and employment in our sample. Both remain depressed by more than 14 per cent since the onset of the recession.

As we discussed in the previous studies, some countries have shown a stronger bias towards cutting employment, while others have favoured adjustment in average hours of work. Germany exhibited very little employment adjustment throughout the course of the downturn despite a very sharp contraction in output, and the level of employment now stands slightly higher than at the onset of the recession. Average hours have declined by more than employment in Australia, the Netherlands, Germany and Japan, whereas the decline in labour input in Spain is entirely down to employment losses. Average hours of work have actually increased slightly in Spain since the start of the recession, while employment has declined by more than 10 per cent.

[FIGURE 3 OMITTED]

The bias towards reducing labour input through average working time partially reflects subsidised short-time worker programmes in Japan, Finland, Italy and Germany. However, Boysen-Hogrefe and Groll point out in their article in this issue of the Review that the short-time worker scheme in Germany can only explain a small share in the decline in average hours of work since the onset of the global crisis. The majority of the change is explained by greater flexibility in working time in all firms, which has increased considerably over the past ten years, rather than just the short-time work institutions, which have been available in Germany for 100 years.

Figure 4 shows that only in Australia, Canada and the UK has the level of employment risen since the trough of the recession in output. Jobs continued to be cut for several quarters after output began rising in many countries. While the level of employment edged up slightly in most countries in the first half of 2010, Spain and Ireland continued to suffer further job losses. The recovery in working time has progressed more rapidly, as might be expected. Firms initially raise labour input to meet production demands by allowing retained employees to recover lost working time, and only subsequently expand their workforce. This appears to be happening in Japan, Germany, Finland, the US, the UK and France, while there is less evidence of a recovery in average hours of work in Sweden, the Netherlands and Ireland. (2) Average working time in Australia has declined in recent quarters. However, this may reflect a shift in preferences towards leisure rather than underemployment, as Australia suffered very little in the way of a recession. Sweden has seen the strongest recovery in output since the low point in the recession, with a rise of 4.8 per cent in the level of output. However, the labour market has not shown an inclination to keep pace with the recovery in output, and both employment and average working time continued to decline in the first quarter of 2010.

[FIGURE 4 OMITTED]

While figure 3 illustrates the rise in labour input needed to restore pre-crisis levels of labour input, labour market conditions also reflect demographic developments that neither unemployment rates nor labour input levels can indicate. The OECD defines a jobs gap as the rise in employment that is needed to restore pre-crisis levels in the ratio of total employment to the working age population. In countries with strong population growth, such as Australia and Canada, faster job creation is needed in order to keep pace with the new entrants to the labour force. They use population levels rather than labour force levels to abstract from any shifts in labour force participation rates. During a recession, labour force participation is likely to fall, as people may withdraw from the labour force if they are unable to find work for an extended period. OECD estimates point to a jobs gap of nearly 18 million in the OECD area, or 3.3 per cent of employment (OECD, 2010).

We construct a similar measure, based on total labour input. Our labour input gap reflects the percentage rise in labour input required to restore the pre-crisis ratio of total hours worked to the working age population. These estimates are illustrated in figure 5. All countries in our sample require some rise in labour input in order to restore this ratio. In Germany, Australia, Japan, Canada and France the magnitude of the required rise is small, at less than 2 per cent. A rise of more than 5 per cent is needed in Finland, the US, Spain and Ireland, while more moderate rises are required in the UK, Italy, the Netherlands and Sweden.

[FIGURE 5 OMITTED]

As we have emphasised in the previous studies, output is not the sole determinant of labour demand. Employment levels may also be maintained if employees are willing to accept wage cuts as an alternative to layoffs. In a downturn, firms bring in lower levels of revenue, and if they cannot find a way to reduce costs many will go bankrupt. There are four routes through which firms can reduce their labour costs: reducing employment levels; reducing average hours worked per employee; reducing average wages; and raising the average productivity level per employee by investing in more productive technology.

In Holland, Kirby and Whitworth (2010), we calibrated an estimate of the expected change in employment, given actual country-specific developments in output, average hours worked and real wages, by running a series of simulations using NIESR's global econometric model, NiGEM. (3) In this note we take a slightly different approach to calibrating the expected change in employment. We use a long-run labour demand relationship, drawing on the framework developed in Barrell and Pain (1997), which hinges around the marginal product condition of a (CES) production function of the form:

Q = [[gamma] [[delta][K.sup.-[rho]] +(1-[delta])[([Le.sup.tech]).sup.-[rho]].sup.1/[rho]] (1)

where Q is output, K is the capital stock, L is labour input, tech is the rate of labour-augmenting technical progress, [gamma] is a scaling factor, [delta] is a distribution parameter and the elasticity of substitution, [sigma], is given by 1/(1 + p). Profit maximising firms will aim to set the marginal product of labour equal to its cost, which is the real wage paid by producers. This first-order condition reduces to a simple log-linear labour demand relationship of the form:

In(L) = [theta] +ln(Q)- [sigma] ln(W)+([sigma] -1)tech (2)

where W is the real wage and [theta] is a constant term based on a function of the parameters in equation (1). Taking first differences, splitting labour input in employment (E) and hours (H) and assuming an elasticity of substitution of 0.5 leads to the key equation that forms the basis of our analysis.

[DELTA]ln(E)= [DELTA] ln(Q) - 0.5 [DELTA]ln(W) - 0.5 [DELTA]tech - [DELTA]ln(H) (3)

Given equation (3), we can calculate an expected change in employment for a given change in output, real wages, hours and labour augmenting technical progress. For the first three, we use the observed actual changes. The change in labour augmenting technical progress is more difficult to estimate. As a first estimate, we can assume that the average rate of technical progress is unchanged from the pre-recession period. We use the average growth rate of hourly labour productivity over the period 1997-2007. This is reported as [DELTA][ln(E).sup.e1] in the table above. However, if the economies suffer a permanent loss of output as a result of the financial crisis, this would also be reflected in a decline in trend productivity levels. While the bulk of this can be expected to be effected through capital shallowing, there may be a slowdown in the rate of technical progress as well. Our second estimate assumes no change in the level of technical progress over our sample period. This is reported as A [DELTA][ln(E).sup.e2] in the table.

Table 1 reports the actual change in employment between the onset of the recession and the second quarter of 2010, followed by our two estimates of the expected change in employment derived from equation (3), using the changes in hours, output and real wages reported alongside. To estimate [DELTA][ln(E).sup.e1] we convert the average annual productivity growth to an average quarterly rate, and adjust by the number of quarters between the pre-crisis peak and the second quarter of 2010 in each country.

Figure 6 illustrates our estimates of labour hoarding and labour shedding over this period. This is calculated as the difference between the actual change in employment and our two estimates of the expected change in employment from table 1. The bar labelled 'without tech adjustment' is derived from [DELTA][ln(E).sup.e2] while 'with tech adjustment' is based on [DELTA][ln(E).sup.e1].

According to our indicator, most of the countries in this sample have been hoarding labour since the onset of the recession. Labour hoarding in Finland has been more significant than elsewhere. The exceptions are the US, Australia and Spain, where labour shedding predominates. This is consistent with our findings illustrated in figure 1 above. The evidence is less clear for Sweden and Ireland. After allowing for technological advancement at the same average rate as in the preceding decade, it appears that firms in these countries may have been hoarding labour on average rather than shedding it. Allowing for technology advancement has the effect of increasing the estimate of labour hoarding, or reducing the estimate of labour shedding, as a more productive technology requires less labour input to produce the same amount of output.

The results reported in this paper differ somewhat from those in figure 9 of Holland, Kirby and Whitworth (2010), which indicated a greater tendency towards labour shedding in our sample of countries. There are two key differences to bear in mind when comparing the two approaches. While both studies are based around the long-run labour demand relationship described in equation (2) above, the earlier study allowed for estimated short-term dynamics around this equation within a full macroeconometric model. If dynamics of adjustment tend to be slow, firms will hoard labour over the short term, but begin to shed labour as they adjust to their long-run equilibrium production path. The approach adopted in this note shows us where the economy is relative to its long-term equilibrium, rather than where the economy is relative to where we would expect it to be given the normal dynamic patterns of adjustment within the economy. The other key difference to bear in mind is the time frame. With three quarters of additional information, the adjustment towards equilibrium has progressed from where it appeared to be in January 2010.

[FIGURE 6 OMITTED]

DOI: 10.1177/0027950110389770

REFERENCES

Barrell, R. (2009), 'Long term scarring from the financial crisis', National Institute Economic Review, 210, October, pp. 36-38.

Barrell, R., Davis, E., Liadze, I. and Karim, D. (2010), 'The effects of banking crises on potential output in OECD countries', NIESR Discussion paper no. 358.

Barrell, R., and N. Pain (1997), 'Foreign direct investment, technological change, and economic growth within Europe', Economic Journal, 107, pp. 1770-6.

Holland, D., Kirby, S. and Whitworth, R. (2009), 'Labour markets in recession: an international comparison', National Institute Economic Review, 209, July, pp. 35-41.

Holland, D., Kirby, S. and Whitworth, R. (2010), 'A comparison of labour market responses to the global downturn', National Institute Economic Review, 211, January, pp. F38-F42.

ILO-IMF, (2010), The Challenges of Growth, Employment and Social Cohesion, Discussion Document from Joint ILO-IMF conference.

OECD (2010), OECD Employment Outlook 2010--Moving beyond the Jobs Crisis, Organization for Economic Co-operation and Development, Paris.

NOTES

(1) We make no such adjustment to figure 2, as the bulk of the output adjustment is assumed to have been completed by the time the economy starts growing again.

(2) Although, as discussed in Barrell, Kirby and Whitworth in this Review, there is evidence that firms are not increasing working hours fast enough to meet demand from employees, as indicated by the increase in people working part-time because they have been unable to find full-time work.

(3) For further detail on the structure of the NiGEM model, see the discussion in Appendix A of the World Economy chapter and further details on http://nimodel.niesr.ac.uk/.

Table 1. Actual and expected employment change

 [DELTA]ln(E) [[DELTA]ln(E).sup.e1]

Ireland -15.2 -19.5
Spain -11.0 -8.3
US -4.8 -3.5
Finland -3.7 -11.7
Sweden -3.3 -3.7
Italy -2.2 -4.9
Japan -1.8 -4.9
UK -1.6 -5.5
France -1.4 -4.8
Netherlands -1.3 -4.5
Germany 0.4 -3.5
Canada 0.8 -1.1
Australia 2.7 4.4

 [[DELTA]ln(E).sup.e2] [DELTA]ln(H) [DELTA]ln(Q)

Ireland -14.8 -3.9 -14.4
Spain -8.1 0.7 -4.7
US -1.1 -1.2 -1.3
Finland -9.1 -2.3 -7.0
Sweden -0.7 -1.4 -3.1
Italy -4.5 -1.1 -5.8
Japan -2.9 -3.0 -4.4
UK -3.0 -1.7 -4.8
France -2.6 -0.4 -2.1
Netherlands -2.8 -2.2 -2.7
Germany -1.8 -1.7 -2.7
Canada 0.5 -0.1 -0.1
Australia 5.9 -2.5 3.7

 [DELTA]ln(W) Ave. annual
 productivity
 growth
 1997-2007

Ireland 8.8 3.7
Spain 5.4 0.1
US 2.0 2.0
Finland 8.7 2.6
Sweden -2.1 2.4
Italy -0.4 0.3
Japan 2.9 1.8
UK -0.1 2.2
France 1.8 1.9
Netherlands 4.7 1.5
Germany 1.4 1.6
Canada -1.0 1.3
Australia 0.5 1.7

Note: log changes are multiplied by 100.
[[DELTA]ln(E).sup.e1] is adjusted for expected labour augmenting
technology growth, based on average historical productivity growth.

[[DELTA]ln(E).sup.e2] assumes no change in the state of technology
over this period.