Labour markets in recession: an international comparison.

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

The global recession has driven a surge in the number of unemployed people across the world. The ILO estimates that global unemployment in 2009 will be some 29-59 million higher in 2009 than it was in 2007, a rise of 16-33 per cent. After taking into account global population projections, this points to a rise in the global unemployment rate of 0.8-1.7 percentage points. A rule of thumb that is often used to relate unemployment to output suggests that a decline in GDP of 2-3 per cent is associated with a 1 percentage point increase in the unemployment rate (see for example Knotek, 2007). Known as Okun's rule of thumb, these estimates are based on a simple regression of the percentage change in output against the percentage point shift in the unemployment rate. These estimates vary across time and across countries, and obviously omit important factors that determine potential output. Nonetheless, the framework provides a convenient way of comparing current labour market developments with behaviour in the recent past. In figure 1 we illustrate the relationship between GDP growth and the change in the unemployment rate in the OECD as a whole, using quarterly data from 1988Q2 to 2008Q3.

[FIGURE 1 OMITTED]

There is a clear downward association, and a simple regression suggests the following relationship:

[DELTA]GDP = -1.7173 * [DELTA]U + 0.6243
 (7.4) (20.5)
[R.sup.2] = 0.40; SE = 0.28

where [DELTA]GDP is the quarterly percentage change in GDP, [DELTA]U is the absolute quarterly change in the unemployment rate and t-statistics are reported in parentheses. This indicates an average GDP growth of 0.6 per cent per quarter over our sample period, holding the unemployment rate constant, while a 1 point rise in the unemployment rate is associated with a 1.7 percentage point decline in the GDP growth rate. This is slightly, but not significantly, less than the estimate produced for the US by Abel and Bernanke (2005), who found a I point rise in the unemployment rate to be associated with a 2 point decline in the GDP growth rate.

Since the second quarter of 2008, OECD output has declined by 1.6 per cent, while the unemployment rate has risen by 1.9 percentage points. This suggests that unemployment in the OECD has risen somewhat more than might be expected, given the magnitude of the output decline.

A closer look at developments in individual countries shows very different responses of labour markets to the recession. Figure 2 illustrates the percentage point rise in the unemployment rate between the lowest monthly level recorded since January 2007 and May or June 2009. The series are ordered according to the magnitude of output loss since the first quarter of 20081 in order to show that the size of output loss is not necessarily a good predictor of the rise in the unemployment rate. While there is some degree of correlation, the rank correlation coefficient between the rise in unemployment and the decline in output is just 42 per cent.

The immediately obvious outliers in the data are the Baltic States (Estonia, Latvia and Lithuania), Spain, and to a lesser extent Ireland. The Baltic States have experienced particularly large rises in unemployment since the financial crisis hit, and Estonia has seen the sharpest rise in the unemployment rate of 11.9 percentage points. However the Baltic States have also experienced relatively large magnitudes of output loss, so to a large extent these rises may be expected. Spain however is rather more inconsistent, having suffered a rise in unemployment comparable to those of the Baltic states despite being at the lower end of the scale in terms of output loss.

Similarly the US has experienced a surprisingly large rise in unemployment compared to other countries; once again this is at odds with its relatively small output loss. However this result may be partially explained by the fact that the local trough for the US (i.e. its lowest unemployment rate since January 2007) was in Quarter 1 of 2007. This means that unemployment has been increasing over a longer period of time in the US than in many other countries--the UK, for example, experienced its local trough in Quarter 2 of 2008.

Germany, Japan and Slovenia are interesting examples as conversely they have suffered relatively large losses in output since January 2008, but have seen only small rises in unemployment; the increase in the unemployment rate in Germany was just 0.6 percentage points, compared to a GDP decline of 6.9 per cent in the year to the first quarter of 2009.

[FIGURE 3 OMITTED]

Figure 3 plots the unemployment rise against the output loss since the first quarter of 2008, in order to identify the biggest outlying observations. There has been a bias towards labour hoarding during the recession, with 67 per cent of the countries shown having a smaller rise in the unemployment rate than would be predicted by our estimate of Okun's rule of thumb for the OECD as a whole. Despite this, in the OECD as a whole the unemployment rise since 2008Q2 outstripped the decline in output, as mentioned above. While there is a large cluster of countries close to the regression line, this is largely a consequence of the outlying observations discussed above.

To some extent, differences across countries may reflect labour market regulation and the level of employment protection, as these affect job security and the speed with which firms can hire and fire employees. Many countries, notably in Europe and Japan, have regulations which make it difficult to dismiss workers. However, the empirical evidence on the impact of employment protection on short-term unemployment is mixed (Layard et al., 2005). Of particular relevance to the current climate is the interaction of employment protection with economic shocks. Blanchard and Wolfers (2000) even suggest that in some circumstances employment protection exacerbates the unemployment effect of a negative economic shock. (2) We have investigated the relationship between the rise in the unemployment rate and the strength of employment protection legislation, using 2008 data from the employment protection summary indicator developed by the OECD. This indicator measures the procedures and costs associated with hiring and dismissing workers, and the use of temporary contracts. We have used Version 3 of the indicator, which takes the weighted sum of three broad categories of sub-indicators, including regular employment, temporary employment, and collective dismissals.

Figure 4 plots the rise in the unemployment rate against this measure of employment protection. While Spain is a clear outlier, there is a 56 per cent correlation between these two series if Spain is excluded. Less regulation has clearly been associated with a greater rise in the unemployment rate during this recession.

In addition to employment protection, several European economies (Germany, the Netherlands, France, Belgium and Sweden) have introduced fiscal measures as part of their stimulus packages, offering direct or indirect subsidies to firms to retain employees. This can help explain the minimal rise in unemployment in these countries, especially in Germany and the Netherlands. In contrast, Spain has introduced policies which include a sharp reduction of the ceiling of non-seasonal workers recruited from abroad and from November 2008 financial support for migrant unemployed workers (from outside the EU) to return to their country of origin. Financial assistance schemes for return migration are not normally enough to produce large outflows of migrants and, to date, the uptake target has been missed (OECD, 2009b).

[FIGURE 4 OMITTED]

Where culture or government regulation limits employment flexibility, we may find that economies exhibit greater real wage flexibility. Figure 5 shows the percentage change in real compensation per employee since the onset of the unemployment rise in each economy. A sharp drop in real labour costs can explain at least part of the employment resilience in Japan and to a lesser extent Germany. In some countries, such as Canada, Slovakia and Ireland, real compensation rates have risen sharply, as the sudden drop in consumer prices at the turn of the year was not anticipated and had not been priced into wage agreements. We expect to see more wage restraint next year as a result.

A decline in real compensation per employee can reflect either a decline in the wage paid to each employee, with no adjustment to the labour input provided by employees, or a decline in the number of hours worked per employee, with no adjustment to the wage received. In addition to employment and wages, labour market flexibility may also show up in the average hours worked per employee. During downturns, firms may maintain the same level of employment, but may reduce the number of hours that each employee works in a week. Developments in the unemployment rates discussed above mask certain differences across countries in the developments of labour input, which should bear a much closer relationship to output than the unemployment rate. Labour input also depends on average hours worked and participation rates and we discuss developments in these key series below.

Labour input is given by the total number of hours worked in an economy. This can be disaggregated into total employment, which feeds into the unemployment rate discussed above, and average hours worked per employed person. Figure 6 illustrates employment and hours separately, and shows how labour market adjustment in each country has been allocated between the two. Labour markets in Spain, Hungary and the US have responded much more by reductions in the total level of employment rather than in hours per employee. This is consistent with the high increase in unemployment, particularly in Spain as seen above. On the other hand, labour markets in Denmark and Japan have responded by cutting hours worked per employee more than total employment, though both variables have declined substantially in these countries.

The data for Italy, Belgium and Norway suggest that their labour markets have responded rather differently to other countries, as levels of total employment have actually risen since the financial crisis. But this was at least partially offset by a reduction in hours in the latter two countries. Poland and France however have the opposite pattern, having increased hours per employee and reduced total employment.

Figure 7 shows developments in total labour input, and reflects the pattern observed in the previous graph quite closely. It also illustrates that responses of labour input have, on the whole, been more uniform than figure 6 might suggest. For instance, in figure 6, labour markets in Japan and US are shown as having had very different responses; the former having adjusted more in hours and the latter more in total employment. But when aggregated into a common unit as in figure 7, the difference between the two is much smaller. Spain and Denmark are similar examples.

[FIGURE 6 OMITTED]

[FIGURE 7 OMITTED]

Labour input in Italy, Poland, Belgium, Norway, France, Germany, the Czech Republic and the Netherlands has shown little response to the recession. This may mask some country-specific developments in certain countries that have not shown up in the hours data produced by the OECD. For example, Germany, the Netherlands and Belgium have all introduced subsidised short-time worker programmes. Under the German programme, employees facing unemployment reduce their working hours by up to 90 per cent for up to 24 months. The wage loss to the employees is subsidised by the German employment office at a rate below the original income level but above the unemployment benefit. Participants are encouraged to engage in professional training or education during this period, which may have long-term benefits for the productive capacity of the economy. While this implies a reduction in working hours, it does not seem to be reflected in the data series illustrated in figures 6 and 7. Figure 8 illustrates the ratio of short-time workers to total employment in Germany. This ratio has risen sharply since the turn of the year, reaching 3.1 per cent in March 2009, the highest level since Germany's 1992-3 recession. This suggests that actual hours worked in Germany may be fewer than implied by the OECD series, suggesting that employees remain registered as fully employed during the period of this scheme. While such programs mitigate the rise in unemployment in the short run, they rely on the lack of final demand being temporary. If the weakness of these economies turns out to be more persistent, then these short-term working schemes simply delay the inevitable rise in unemployment (OECD, 2009a).

[FIGURE 8 OMITTED]

Output per unit of labour input gives labour productivity. In general, we would expect labour productivity to decline somewhat during a recession, due to labour hoarding. Indeed government policy can actively promote labour hoarding. Job subsidy schemes and short-time working programmes are two such examples of the active promotion of labour hoarding. However, as shown in figure 9, in some countries productivity has actually increased since the onset of the recession: Spain, US, Denmark and Ireland. A rise in productivity during a recession can have important implications for labour market developments during the recovery. If firms have undergone a structural shift that allows them to maintain a given level of output with less labour input, they will be more reluctant to rehire employees once growth restarts. This has been the pattern in the US in the last two recessions of 2001 and 1990-91. Unemployment continued to rise in the US for 16-19 quarters during the recovery, and the term 'jobless recovery' was frequently used to describe this behaviour. Eventually higher levels of productivity will raise potential output and employment levels should rise, but it may take more than a year for this to feed through.

[FIGURE 9 OMITTED]

Alternatively, higher economy-wide productivity levels may reflect the labour intensity of the sectors that have been hardest hit by the recession. It is interesting to note that the four countries with rising productivity, Spain, the US, Denmark and Ireland, experienced some of the biggest housing booms in recent years, and the labour intensive construction sectors have suffered disproportionately in the recession. The UK also experienced a large housing boom, but it did not experience the rapid expansion of housing supply that other countries exhibited. In countries such as Germany and Japan, manufacturing sectors have suffered most. As the contraction in manufacturing is seen as temporary, unlike the structural transition in the construction sector which has longer-term implications, firms may have been willing to retain experienced workers.

The final labour market factor that we consider in this note is the share of the working age population that participates in the labour force. If participation rates fall, the unemployment rate may temporarily decline, but labour input also declines, leading to a reduction in potential output. Figure 10 plots developments in participation rates since the unemployment rate has started to rise in each country. There has been a sharp decline in participation rates in Ireland, and more modest declines in the US, Portugal and Sweden. On the other hand, participation rates have increased sharply in Japan, as well as in Italy, Spain and France.

[FIGURE 10 OMITTED]

A permanent decline in participation rates can be a damaging way to deal with rising unemployment. For example, offering older employees early retirement as a way of reducing the workforce will leave the economy with a lower level of potential output for an extended period. We simulate the impact of a decline in participation rates in the US using our model NiGEM. We assume that early retirement is offered to employees who are within ten years of retirement. We cut the participation rate by 1 percentage point in 2009, and allow it to return gradually to base over a 10-year period, when this cohort would have retired ordinarily.

Figure 11 illustrates the impact of this shock on the unemployment rate and output. While the unemployment rate initially declines by 1 percentage point, it rises above base by 2012. The short-term reduction in the rate of unemployment slowly disappears as the economy adjusts to a smaller labour supply. The level of employment is reduced to eliminate the increase in upward wage pressure (Layard et al., 2005). The output cost is relatively high, as output remains below base until 2020. Ignoring the fact that a policy of early retirement does not promote a long-term reduction in the rate of unemployment, it is perhaps not the most sensible policy to pursue in a recession. Sharp falls in equity markets due to the recession have had a negative impact on private sector pensions. For those on defined contribution schemes this suggests a smaller pension pot upon retirement than would have been the case before the financial crisis struck.

[FIGURE 11 OMITTED]

A decline in the participation rate does not necessarily imply an early retirement scheme. It may reflect a withdrawal of younger people from the labour force who have found it difficult to find a job and have chosen to return to education or retrain in a new field. This could have a very different impact on the medium-term outlook, as education can raise the productive capacity of the workforce, implying a higher level of output once these students have rejoined the labour force. Expansion of further and higher education places in the current recession is one policy option advocated by Bell and Blanchflower (2009) in order to limit the increase in unemployment due to the fall in labour demand currently being experienced, while simultaneously supporting longer-term growth prospects. However, a decline in the participation rate could also be due to general disengagement with the labour market. Withdrawal from the labour force can be particularly damaging for young people, who will find it difficult to re-enter in later years.

Labour markets have exhibited very different patterns in response to the global recession. The US and Spain, in particular, have shed a large number of jobs relative to the size of their output loss. In the US this appears partially to reflect the low level of employment protection, which makes it relatively easy for firms to lay off workers. While employment protection measures have helped to maintain employment levels in countries such as Germany and France, there is generally a tradeoff in terms of real wages, hours worked or rising budget deficits. Developments in Spain, as well as Ireland and Denmark, appear to be related to the housing market downturns, as workers in the labour intensive construction sector have suffered disproportionately. Japan and Germany have seen little rise in unemployment despite sharp contractions in output. In Japan, this has been offset by a sharp decline in working time, while German employment has been supported by fiscal incentives. At the heart of the problem is a lack of final demand. Bell and Blanchflower (2009) quite rightly indicate that the cause of the increase in unemployment is a lack of final demand. Policies that stimulate economic growth should be effective in limiting the extent to which unemployment increases.

REFERENCES

Abel, A.B. and Bernanke, B.S. (2005), Macroeconomics, 5th edn, Pearson Addison Wesley.

Bell, D. and Blanchflower, D. (2009), 'What should be done about rising unemployment in the UK?', Stirling Economics Discussion Paper 2009-06.

Blanchard, O. and Wolfers, J. (2000), 'The role of shocks and institutions in the rise of European unemployment: the aggregate evidence', Economic Journal, 100, pp. C1-33.

Layard, R., Nickell, S. and Jackman, R. (2005), Unemployment: Macroeconomic Performance and the Labour Market, 2nd edition, Oxford, Oxford University Press.

Knotek, E.S. (2007), 'How useful is Okun's Law', Economic Review, Federal Reserve Bank of Kansas City, fourth quarter 2007, pp. 75-103.

OECD (2009a), Economic Outlook, Paris, OECD.

OECD (2009b), International Migration Outlook, Paris, OECD.

NOTES

(1) See figure I in the Commentary on p. 4 of this Review.

(2) Bell and Blanchflower (2009) dispute their results suggesting that the estimated model is over-fitted.

Figure 2. Unemployment rate rise

Percentage point rise

Poland 1.3
Australia 1.8
Greece 1.2
Norway 0.8
Canada 2.5
US 4.7
Austria 0.7
Spain 10.8
Belgium 1.6
France 1.8
Czech Republic 1.8
Bulgaria 1.4
Portugal 1.7
Romania 0.5
UK 2.2
Hungary 3.1
Italy 1.5
Finland 1.9
Slovakia 2.1
Sweden 3.3
Germany 0.6
Japan 1.6
Ireland 7.2
Slovenia 1.7
Lithuania 10.3
Turkey 4.4
Estonia 11.9
Latvia 10.9

Source: Eurostat and BLS.

Note: Percentage point rise in unemployment rate since recent
trough. Series ordered from smallest to largest output loss
since 2008Q1.

Note: Table made from bar graph.

Figure 5. Real compensation per employee

% change

Japan -2.7

Czech Republic -1.8
Germany -0.7
Sweden -0.6
UK -0.2
France 0.7
Netherlands 1.5
US 2.2
Finland 2.4
Ireland 2.9
Slovakia 3.1
Canada 4.8

Source: OECD.

Note: Percentage change since recent trough in unemployment rate.

Note: Table made from bar graph.