Crises and joint employment-productivity dynamics: a comparative perspective for European countries.

Marelli, Enrico ; Signorelli, Marcello ; Tyrowicz, Joanna 等

INTRODUCTION

The joint dynamics of employment and productivity are of crucial importance especially in the short term. Reductions in labour demand can be severe during an economic crisis and can increase unemployment. They lead to political pressure that frequently involves policy changes designed to address voter demands. The timing between cyclical fluctuations and crisis management has changed over the past decades, with an attempt to devise and implement so-called pro-active policies. The political pressure for such policies is stronger if other countries have already adopted similar policies in the past (Mukand and Rodrik, 2005). However, in certain cases, such efforts may not only be premature but also useless, as the employment response may differ, depending on national institutions such as labour market flexibility, active and passive labour market policies and so on as well as on the depth and the duration of the down cycle.

In this paper, we shed some light on short-term joint employment-productivity dynamics in the EU economies along the lines set out by Mulligan (2009). We have broken down changes in output into changes in employment and in productivity over a horizon of up to 2 years from the peak of the business cycle. Such analysis over a relatively long period, in this paper the last two decades, of economic fluctuations, provides some typical patterns for many of the EU countries. The actual combination of both types of adjustment, employment and productivity, do indeed differ according to the causes of the output decrease as well as according to the policies targeting labour market adjustments. On the basis of the institutional framework and the nature of economic slowdowns, we were able to provide economic justification for the emerging groupings of EU economies. We have also documented the peculiarities associated with the recent global economic recession.

The labour market in the EU is frequently described as being less flexible than other markets such as the United States. However, this is not true for all EU countries. Moreover, the recent economic slowdown has shown a large number of peculiarities that are different from previous adjustment patterns. We have identified four main groupings of EU countries, based on similar adjustments within groupings and different adjustments between groupings, depending on the employment-productivity path they followed in the recent recession. Inquiring into the reasons of such divergence is beyond the scope of this paper, but we nonetheless suggest some potential explanations.

The policy implications of this empirical exercise are clear. First, before any actual policy is adopted, governments should identify the most probable type of adjustment their economy is expected to follow and see whether the policies already implemented are sufficient. Some EU governments in the recent crisis seem to have 'jumped' prematurely to adopt new instruments to save jobs although these policies have also proven to be detrimental to prospects for recovery. Second, some EU countries seem to have developed long-term stable solutions that have helped cushion employment adjustment after the recession. Popular 'crisis packages' seem to be particularly redundant in such cases as well as for countries where unemployment increases are typically delayed and where the risks of persistence are higher. Whether they are due to additional automatic stabilizers or pro-active safety nets, adverse labour market adjustments in such environments are weaker on the employment side and less detrimental to productivity dynamics.

The paper is structured as follows. Section 'Review of the literature' briefly reviews the relevant literature. In section 'An empirical investigation for a sample of EU countries', we discuss in detail the method and the data used, before analysing and discussing the main empirical results. The concluding remarks are presented in section 'Conclusions'.

REVIEW OF THE LITERATURE

This review is divided into three sub-sections. First, we analyse the links between employment and productivity, especially from a business cycle standpoint (Employment and productivity: Key empirical studies). Then we consider the main features of European business cycles (Business cycles in the EU). Last, we illustrate the key characteristics and labour market effects of the latest global crisis (Impact of the recent financial crisis).

Employment and productivity: Key empirical studies

Many of the empirical studies of the links between economic dynamics and labour market performance refer to Okun's law (eg, see Padalino and Vivarelli, 1997; Lee, 2000). The intensity of the relation, ie, 'Okun's coefficient' varies between countries and is unstable over time, depending on structural conditions and the legal framework, especially for labour market institutions, reflecting in particular labour hoarding practices. (1) Moreover, there are many difficulties in detecting a stable threshold of GDP growth rate that allows a decline in unemployment and an increase in employment as opposed to so-called jobless growth. Furthermore, the findings of Boulhol and Turner (2009) seem to support this view.

The timing of the labour market response may also be relevant. For example, the IMF (2010, Chapter 3) showed that the responsiveness of the unemployment rate to changes in output has increased over time in many countries as a consequence of looser employment protection and greater use of temporary employment contracts. However, it has been estimated that in normal recessions it takes three quarters after output has begun to recover for employment to start increasing, and an additional two quarters for the unemployment rate to peak. Thus, the evolution of employment and unemployment patterns are not necessarily alike. Unemployment can continue rising even after employment growth has become positive.

Another important factor is the symmetry of responses. If adjustments in employment or unemployment are symmetric in the response to output growth changes, higher unemployment responsiveness to changes in output should help in the recovery. However, Arpaia and Curci (2010) detect an asymmetry over the cycle, with recessions being characterized by more job destruction than job creation in subsequent recoveries.

Analysing the co-movement between output and employment yields, we can begin to understand, at least indirectly, productivity dynamics. Whenever 'labour hoarding' practices prevail, productivity may actually decline or its growth rate decelerates at least in the short term. (2) In the recovery phase of the cycle, labour hoarding may cause a lower productivity gain and much slower recovery (see Caballero et al., 2008). On the other hand, if labour hoarding is a forward looking strategy for both employers and employees, a kind of productivity dividend may occur, since '[w]hen demand begins to revive, output can often be boosted without causing commensurate increases in the payroll. Firms respond by using some idle plants and equipment and by redirecting existing labour to production-related tasks' (Fulco, 1984, p. 3). In addition, hoarded employees may be those with greater experience, more training and firm-specific skills who could further contribute to increased productivity. It is also quite self-explanatory that this phenomenon is short run (eg, Beaudry and Collard, 2002; Becket and Gordon, 2008). In particular, Pichelmann and Roeger (2008) consider this potential trade-off within the Lisbon Strategy. Belorgey et al. (2006) showed that employment rate changes negatively affect the productivity growth rate, supporting the hypothesis of diminishing returns for the employment rate.

In the medium run, a trade-off between employment and productivity is just one possible outcome, but there are other possibilities. For example, Marelli and Signorelli (2010) identified four different models of growth for EU countries over the last two decades: intensive, extensive, virtuous and stagnant. (3) In Eastern Europe, all centrally planned economies followed an extensive model with a high employment rate and low productivity levels and dynamics before the beginning of transition in the early 1990s; rationalization and restructuring of their economies then significantly reduced employment rates, often accompanied by an increase in the shadow economy and irregular employment but with more or less important productivity gains. (4) On the other hand, many old EU countries have moved in the opposite direction since the mid-1990s. Following two decades of jobless growth, they have shifted to an extensive growth model. For instance, Italy and Spain, following certain reforms that introduced greater labour flexibility, benefited from a considerable increase in employment rates, but accompanied by worsening productivity growth as many new jobs have been for unskilled and low-wage workers, in many cases taken up by immigrant workers. (5)

Finally, it is widely recognized that, in the long run, sustained productivity growth is beneficial also in preserving and expanding employment. Growth accounting approaches have tried to calculate the contribution of capital, labour, human capital and so on to growth and show that employment increases are compatible with rises in total factor productivity. (6)

Business cycles in the EU

A preliminary consideration is that, in the world, business cycles have been highly synchronised across countries over the past two decades. This fact, together with the opening of countries and increasing trade, is one of the most evident effects of globalization. (7) Kose et al. (2003) demonstrate that increasing trade intensity contributes to increase the transmission of both peaks and troughs of business cycles across the globe.

Studies of business cycles in the EU conducted over the last decade have focused on the degree of synchronization, above all to find whether the fundamental conditions for an effective monetary union have been satisfied by current Euro-zone members or will be satisfied by candidate countries. These empirical investigations refer to correlations of output, GDP, industrial production or employment and, in some cases correlations of exports, consumption and services are also investigated. An intensifying correlation of real variables implies that the response to shocks has become more symmetric across European countries, a characteristic of optimal currency areas and a pre-requisite for a successful currency zone.

The results of empirical studies on cycle correlation show that Euro-area countries correlate among themselves more than they do with the rest of the world. Empirical studies have shown that synchronicity has increased not only within the Euro-zone or the old EU countries, including some peripheral countries, therefore making the concept of a core of European countries less meaningful, but also between old and new Europe. (8) As regards output correlations, Hungary, Poland and Slovenia seemed to be the most correlated with the Euro-area, whereas the lowest correlations, close to zero, were found for the Baltic states (Babetskii, 2005; Fidrmuc and Korhonen, 2006). More recently, Darvas and Szapary (2008) argued that, in terms of GDP, as well as imports and exports, the so-called new and old member states have become more and more synchronized. However, consumption and investment cycles remain unsynchronized.

Impact of the recent financial crisis

The collapse of the Lehman Brothers bank in the fall of 2008 is widely recognized as the highlight of the so-called sub-prime crises in the United States. Actually, many of the weaknesses of the financial system had arisen since the second half of 2007, but the process was initiated by increases in the federal fund rates in the United States in late 2005. While the chronology of the events is well known and has been widely discussed, the real effects are still largely controversial.

First, the business cycle peak in the United States, as well as in other leading global economies, had come to a turn already in the second half of 2007. As a result, some real adjustments should have been expected even without the financial storm caused by the collapse of Lehman Brothers and subsequent bailout of other financial institutions. Second, the magnitude of the real adjustments was surprisingly large, with unprecedented declines in production, income and international trade. In addition, employment fell and unemployment rates gradually rose. The United States experienced the highest unemployment rate since the Great Depression, while, for example, the Baltic States experienced an annual contraction of between 15 % and 18 % of GDP. (9) In fact, the real effects varied greatly across countries. Naturally, the size of the shock as well as the magnitude of adjustments depend on various factors such as country reliance on international trade, dependence on natural resources, financial liberalization of the banking system, fiscal resources at the government's disposal and so on (see, for eg, European Commission, 2009a, b). In terms of the labour market impact, the IMF (2009) investigated various employment adjustments and labour hoarding phenomena. A first general result was that during the latest crisis, there was a much bigger (negative) impact on productivity per worker in most countries (both advanced and emerging), suggesting that labour hoarding was much higher during this recession. (10)

In most developed economies, deterioration of the employment rate was slight, notwithstanding sometimes large output reductions. On the other hand, recent events show that labour hoarding incentives by governments are more common in countries specialized in industrial activity, where income-support policies are also more frequent. These countries often developed crisis-specific legislation such as subsidies for part-time work in Germany or extending income support for workers formally maintaining job contracts at reduced working time or at 0 hours in Italy. Thus, the fall in labour demand was accompanied by either labour-hoarding practices, with consequent reductions in productivity, or by a fall in labour supply thus dampening the impact on unemployment rates.

The opposite cases include countries where employment fell heavily, also yielding exceptionally high unemployment rates, but with smaller or even no effects on productivity. The IMF (2009) partially explains this heterogeneity in performances by considering the multifaceted dimensions of labour market flexibility, such as employment protection legislation (EPL), the types of wage-bargaining arrangements, the level and duration of unemployment benefits and the diffusion of temporary contracts. The stronger employment response in low EPL economies compared with medium/high EPL economies is consistent with the literature, suggesting that employment protection reduces both inflows to and outflows from employment. For medium to high EPL countries, the reduction in employment during this crisis has been similar to that during previous cycles despite substantially bigger GDP declines, confirming this higher degree of labour hoarding.

The employment-productivity adjustment path was also investigated by Beegle et al. (1999), who analysed the feasibility of intensive margin adjustment instead of reductions in jobs and an increase in the unemployment rate. Moreover, sectoral and regional reallocation of labour, for example workers moving back to agriculture or from urban to rural areas, are usually important, especially in developing countries, as is movement into the informal sector and towards subsistence activities. Analysing the real effects of the financial crises in East Asia and Mexico during the 1990s, Fallon and Lucas (2002) found that aggregate employment fell by much less than production and even increased in certain cases despite different sector and regional responses.

Eichhorst et al. (2010) offered a new explanation of the different patterns of adaptation in European labour markets during the latest crisis. They considered the complex role of labour market institutions during a negative shock by better incorporating in the analysis different types of labour market flexibility and also by focusing on interactions between institutions, shocks and policies. In particular, they considered: (i) external numerical flexibility (dependent on EPL, the benefit system affecting labour supply, labour taxes); (ii) internal numerical flexibility (working time adjustments); (iii) external functional flexibility (occupational mobility, influenced by active labour market policies); (iv) internal functional flexibility (changing organization of production); (v) wage flexibility. In some countries, enterprise-level strategies regarding working time adjustment, often favoured by ad-hoc policy interventions, reduced the negative impact of the crisis on employment. In addition, in certain countries this effect seems to he stronger now than in the past.

AN EMPIRICAL INVESTIGATION FOR A SAMPLE OF EU COUNTRIES

Before presenting data, methodology and empirical results, we briefly illustrate certain possible key employment-productivity dynamics in the simple but useful framework proposed by Mulligan (2009). The nature of the boom-slump short-term GDP evolution calls for changes in the dynamics of labour demand. After passing the peak of the GDP growth rate, which can only be identified ex post, labour demand can only slow or decline. However, if employment is not restrained, the rate of growth of labour productivity will start to drop. Alternatively, labour shedding may actually maintain the labour productivity growth rate. Finally, if there is both employment and productivity adjustment, there is not only a shift of the labour demand curve, but also a simultaneous movement along it. These choices are made at the firm level, and, they are conditional on existing labour market institutions and crisis-management policies and they may vary across sectors and industries. However, depending on which effect dominates, either job-preserving adjustment or changes to maintain labour productivity can be found.

Recent policy instruments implemented in certain EU countries have targeted either both or at least one of these adjustments. For example, instruments favouring labour hoarding change the nature of the shift, maintaining employment at the expense of productivity.

Data and sample

We used country-level quarterly data on GDP and employment based on SNA measurement standards. (11) Employment was measured in full time equivalents. Data were collected from Eurostat and cover the period I990ql-2009q4 or the most recent available data for EU countries. The selection of countries was based on data availability. The variables of interest crucial for the analysis are output and employment. Output was measured as gross domestic product, adjusted for seasonality and working days. Employment was based on national accounts, also adjusted for seasonality and working days. However, for certain countries and years, both variables were not available. Table 1 reports the total number of observations available for each country.

Labour productivity was calculated for each country as GDP divided by employment. This measure of labour productivity is far from ideal in many respects. Conceptually, it should be defined as output per hour of work, but the number of hours worked is unavailable for most European countries. Resorting to total employment as a proxy is naturally a second best option for two main reasons: (i) it does not permit a comparison of intensive versus extensive margin adjustments, and (ii) in the case of considerable changes in the number of hours worked due to large-scale shorter working time regulations implementation, the measure naturally underestimates the scale of any adjustment. On the other hand, in the approach followed in our study, the crucial element is to analyse changes in productivity and demand for labour using the same categories for all countries.

Moreover, the very means of defining labour productivity as a ratio between output and amount of labour used is also controversial. There has been a long debate in empirical studies on alternative measures (Bernard and Jones, 1996). Some studies suggest that either gross value added or estimated marginal product of labour should be used instead. However, the very estimation of MPL remains controversial, with different functional forms yielding different estimates of MPL and MPL changes (Growiec, 2009).

Analytical strategy

To observe the nature of adjustments in selected EU countries, we followed the method developed by Mulligan (2009), consisting of three steps:

1. identify the peak quarters to determine the location of the anchor;

2. compute the anchored changes in labour productivity and employment for each of the peaks and for each of the countries in the sample;

3. graph changes in employment against changes in labour productivity

In order to identify the peaks, we used the Hodrick and Prescott (1997) filter (henceforth HP filter) applying it to logs of seasonally adjusted GDP time series. There is a long and heated debate on the use of the HP filter as opposed to Baxter and King (1999) or Christiano and Fitzgerald (2003) filters: see, for example, Kaiser and Maravall (2001). (12) The HP filter is quite simple to implement. However, it has also potentially large drawbacks, namely it significantly distorts the higher moments of the data, a point discussed by King and Rebelo (1993) and Cogley and Nason (1995). However, the actual reliability of filtering to establish the size of cyclical and trend components is of secondary importance in our study as long as the timing of the events is not distorted. In a recent Monte Carlo simulation, Corbae and Ouliaris (2006) demonstrated that while alternative filters differ in the value of trend departure, they do essentially suggest the same moments. (13) In total, we identified 51 peaks in the sample, but the employment data coverage allowed only 42 of them to be analysed. Table 2 reports the number of analysed peaks for each of the countries in the sample.

For each of the identified peaks and for both time series, employment and labour productivity, we computed cumulative changes anchored to the quarter of the peak over the window of eight consecutive quarters. (14) Thus, the dynamics are not annual or quarterly dynamics, but the cumulative changes from the best economic performance of up to 2 years later. The window, eight quarters, was chosen for two main reasons. First of all, in the sample this was roughly an average duration from peak to bottom. Second, the economic situation later than eight quarters from the peak is only rarely driven by the same initial impulse resulting in real economy adjustments.

Intuition behind the empirical approach

In our sample, virtually all peaks overlapped within 1 year (four quarters). We therefore did not identify separate peaks and crises, focusing rather on the waves across Europe. These include the wave of the late 1990s with peaks within 1997q4 and 1998q4 in 7 countries, early 2000s with peaks within 2000ql and 2001ql in 10 countries, and the recent crisis with peaks within 2007ql and 2008ql in 14 countries. In addition, there were isolated cases of additional shorter and smaller cyclical departures in the case of Denmark, Finland and France. We therefore mainly focused on the following three cases: late 1990s, early 2000s and late 2000s.

We analysed the cumulated changes in employment and labour productivity, where the base for comparison was always the peak quarter. In order to facilitate comparisons across time and countries, the series were additionally logarithmized. In the remainder of the paper, we always show the most recent crisis with a solid dark line, dashed black line for the previous crises of the 2000s and finally dashed and solid grey lines denote earlier events. This graphical representation allows us to compare both the type and magnitude of the changes.

The severity of the economic slowdown differs significantly across EU countries. While most EU member states experienced negative GDP growth rates, Poland being the only exception, the Baltic countries experienced two-digit reductions in annual GDP growth, while the EU-27 saw a decline of -4.2% in 2009 and a 2.0% growth in 2010. Since the severity of the crisis differs across countries, partly due to different labour market institutions and policies, the scale and shape of employment-productivity adjustments were significantly different across countries or clusters of countries. Moreover, transition countries were typically characterized by larger responses to the initial shock. Therefore, the adjustment patterns are likely to be different.

Empirical results and discussion

Our analysis reveals that previous crises were associated with a fall in employment accompanied by an increase in productivity. This pattern seems to have been dominant in the 1990s (solid grey lines) as well as, for most countries, in the late 1990s or early 2000s (dashed black lines). However, more recent adjustments frequently depart from this pattern (solid black lines). We can find four main patterns of adjustments following the latest crisis:

(a) a reduction in employment together with an increase in productivity;

(b) reduced (or absent) change in employment because of the key role of labour hoarding together with reductions in productivity;

(c) a simultaneous reduction in both productivity and employment, consistent with profound economic changes and recession;

(d) a growth or preservation of both employment and productivity, suggesting only minor real adjustments in the economies.

We discuss these four patterns below. Four groups of countries have been identified with reference to the adjustment pattern followed after the latest crisis.

As already mentioned, reduced employment associated with increased labour productivity occurred in the US economy and was to a large extent independent of the financial turmoil. As discussed by Mulligan (2009), the adjustments had already started in the second half of 2007. For European countries, however, this pattern was only been followed by two economies: Hungary and, to a greater extent, Spain (Figures 1 and 2). While Eastern European countries had typically much more volatile labour market indicators, Spain had much greater adjustment. Employment decreased by 5% in Hungary and by 10% in Spain, but the accompanying growth in productivity was comparable in both countries.

Interestingly, while the origin of the Hungarian economic slowdown involved an external adverse shock coupled with fiscal imbalance, the main source of Spanish difficulties emerged from the real estate and construction sector bubble of the past decade. The huge employment loss in Spain was the result of a high share, more than 30%, of temporary contract employment, which declined by more than 20%.

It is also interesting to note that the late 1990s recession wave was associated with increased productivity and a simultaneous growth in employment in both countries. As regards the surprising dynamics of this variable, it should be noted that the Spanish adjustments of the early 1990s, the peak was identified at 1992ql, started a long period of prevailing net job creation, improving on a very low initial employment rate, while the Hungarian experience at the end of the 1990s, the peak was identified at 1998ql, in the aftermath of the so-called Ruble crisis, was conditioned by a dominant trend of net job creation after the huge negative employment decline of the early transition years.

[FIGURE 1 OMITTED]

[FIGURE 2 OMITTED]

[FIGURE 3 OMITTED]

This type of adjustment is typical of countries with a high degree of labour market flexibility. In anticipation of an economic downturn, firms can promptly reduce labour demand, thus keeping up the growth rate of labour productivity. In fact, neo-classical theory of real business cycles (RBC) predicts that the adjustment should follow a backbone pattern. (15)

On the other hand, when certain labour market rigidities are at work, they can lead to a minor initial adjustment in employment and obvious productivity declines, only subsequently followed by a contraction in labour demand. This is the pattern followed by the sub-group of countries shown in Figures 3-7.

In addition to Hungary and Spain, after an initial period of labour hoarding, a similar pattern of decreasing employment and increasing productivity also occurred in Denmark, Finland, the Netherlands and the United Kingdom (Figures 3-6). This group of countries seem to have followed the Hungarian and Spanish adjustment patterns, but with the lag of a few quarters. While an absolute recovery of productivity occurred only in Denmark, the other three countries seem to have entered the path of employment reduction and productivity growth.

[FIGURE 4 OMITTED]

[FIGURE 5 OMITTED]

Huge differences compared with previous crises for all of these countries emerge from the graphs. During past crises, the well-known ability of these economies to create jobs was also maintained in bad times, probably due to the positive role of their labour market institutions. The worst impact of the latest crisis can be partly explained, at least in some cases, by the different size of the downturn interacting with low internal flexibility. (16)

[FIGURE 6 OMITTED]

[FIGURE 7 OMITTED]

As we know, Denmark, Finland and the Netherlands have highly innovative labour market institutions, including the advanced flexicurity instruments as well as quasi-automatic stabilizers such as the Finish Employment Fund In the past, these countries were able to maintain employment constant or even with a moderate growth after the business cycle peak. The exceptional scale of employment adjustment in Finland, and also in Denmark, seems to suggest that the current shock was indeed stronger and raises the question of the extent to which policy instruments in addition to regular stabilizers and pro-active instruments were used in these countries.

On the other hand, the United Kingdom was able to preserve employment levels in the past, thanks to a flexible labour market with relatively low constraints on firing and hiring. However, over the past two decades, it has also developed a wide system of active labour market instruments and institutions, such as the Netherlands and Denmark. (17) The case of the United Kingdom, where the cumulated reduction in employment and productivity reached roughly 2%, is indeed surprising. This country should in principle closely follow the US pattern. A possible explanation is that, in the recent past, both GDP and employment grew almost constantly in the United Kingdom. The employment level changed from roughly 21.5 million to 25.5 million, or over 25% in only 15 years. At the same time, the United Kingdom experienced relatively intense immigration and an internationally open labour market. The decrease in both productivity and employment may thus be explained by a combination of the migration outflow and type (b) or very weak type (a) adjustment patterns.

A fifth country in this group, the Czech Republic, is a special case. It followed the typical RBC pattern during the previous crisis. Currently, however, despite a considerable drop in output, productivity continued to grow for the first four quarters after the peak. This may be explained by relatively early peak in Czech Republic, in 2007q3, and initially very mild GDP adjustment.

Since the adjustments in Germany and the Netherlands are relatively small, these countries seem to be a borderline case between type (a) and type (b) adjustments. We have chosen to include the former country in the next group (Figure 10) and the latter in the previous group (Figure 5). However, the typical examples of type (b) adjustment are France and Italy (Figures 8 and 9). The current processes in these countries involve predominantly the maintenance of employment even at the expense of productivity. It should also be noted that this form of adjustment is new for Germany, which previously followed a type (a) pattern.

[FIGURE 8 OMITTED]

[FIGURE 9 OMITTED]

France and Italy are considered to be examples of the so-called Southern European labour market model. Namely, they both have relatively low participation and employment rates and high entry barriers, especially for young people, although both features are much clearer in Italy than in France. At the same time, they are relatively highly unionized. Observing the adjustment patterns in these two countries, it is worth noting that the relation between productivity and employment is actually quite similar in these countries. Indeed, the business cycles of both these countries are fairly synchronized. Moreover, Germany, France and Italy used and increased their internal flexibility with significant reductions in hours worked induced by policy interventions in the so-called crisis packages of these countries, which sought to minimize employment reduction. These policy instruments proved to be strongly detrimental to productivity.

[FIGURE 10 OMITTED]

The third group consists of countries where adverse huge employment adjustment was accompanied by decreased productivity. This type (c) is depicted in Figures 11-13. This group comprises the three Baltic countries, the most heavily hit EU economies. Estonia, Latvia and Lithuania experienced a double-digit reduction in GDP, which is by far the largest since the onset of the economic transition. As a consequence of fairly flexible labour markets, the impact on employment was also considerable.

As regards past crises, the Baltic States seem to have followed a RBC adjustment. Estonia and Lithuania would be classified as type (a) patterns, while the type (b) pattern in the case of Latvia may be explained by relatively mild economic crises in previous years. However, this puzzle may be justified as follows: the reduction in employment in the Baltic States, at strikingly high rates of 10%-20% in total, is a consequence not so much of a recession, but rather of a collapse of the economy. For a relatively small economy, the populations of these countries range between 2 and 4 million, a drop in the size of the economy by one-fifth is a shock that cannot be absorbed without negative adjustments in both productivity and employment.

[FIGURE 11 OMITTED]

[FIGURE 12 OMITTED]

[FIGURE 13 OMITTED]

The final type of adjustment involves simultaneous increases in both productivity and employment: type (d) is represented in Figures 14 and 15. This group comprises Belgium and Poland. In general, such behaviour is actually contrary to the predictions of any crisis theory, let alone the RBC approach, and it seems to suggest that there is no economic crisis. Indeed, of these countries, Poland maintained positive GDP growth rates, while the recession in Belgium was relatively moderate. Moreover, these countries seemed to have conformed with type (a) pattern in the periods of more pronounced economic slowdown. The earlier crises demonstrated increased productivity and reduced employment, although the extent of adjustment was much lower for Belgium.

Summarizing, although the adjustment patterns across countries seem to have been highly heterogeneous and even divergent, there appear to be four main adjustment patterns. In addition, it is very interesting to highlight that the employment-productivity dynamics occurred in past crises often differed from the latest crisis dynamics.

[FIGURE 14 OMITTED]

[FIGURE 15 OMITTED]

CONCLUSIONS

The objective of this paper was to verify whether the current crisis has had any peculiarities in how productivity-employment adjustments have taken place. One of the potential responses is that both employment and productivity fall after a big economic crisis, such as the one that began in 2007-2008. Productivity dropped much more, with initially small changes in unemployment, in countries with rigid labour markets or where internal flexibility substitutes for external one. On the other hand, the RBC approach suggests that the post-peak adjustment should involve reductions in employment accompanied by increases in productivity. Such a pattern was followed in the United States over the recent crisis, which prompted some researchers, among them Mulligan (2009), to claim that the current economic recession is rooted in the real, rather than in the financial sphere.

We have produced empirical evidence on different adjustment patterns among EU member states in terms of productivity-employment adjustment. Indeed, the paper identifies highly divergent adjustment paths. We analysed the first 2 years after the peak of the economic performance and considered the past two decades of data, identifying for each country between two and three such events. In all countries, the peak quarter occurred before the collapse of Lehman Brothers and the adjustment was not accompanied initially by any crisis-oriented policy packages.

This comparison allowed us to divide adjustment paths into groups coherent within and different from each other. A comparison of adjustment patterns over time and across the institutional frameworks highlights two types of peculiarity. The first is associated with the atypical adjustments in many European countries over the recent crisis. The second encompasses the positive or detrimental role the institutions and policies may play in absorbing the consequences of a global economic crisis.

Our findings are consistent with Eichhorst et al. (2010) who not only used a different methodology, but also gave policy implications based on a cross-country comparison of crisis responses and policy responses of EU members. Like the US adjustment pattern, a reduction in employment together with an increase in productivity, consistent with the RBC theory, occurred only in Hungary and Spain. France, Italy and Germany formed a second group, where the processes involve predominantly maintaining employment at the expense of productivity. Countries known for labour market flexibility, Denmark, Finland, the Netherlands and the United Kingdom, uniformly experienced a short joint decline in both employment and productivity. An adverse huge employment adjustment was accompanied in the Baltic States by a very strong reduction in productivity. These economies experienced a double-digit reduction in GDP during the latest crisis and the recent major shock is the second within less than a generation.

The employment-productivity adjustments in countries adopting the flexicurity model based on external flexibility and effective active and passive labour policies were quite similar to the past. Employment dynamics were more favourable in the countries taking advantage of higher internal flexibility and related policies, which permitted a flexible reduction in hours worked with minor job destructions. However, the price paid in terms of productivity reduction was very high, reaching 4-6 percentage points in just six to eight quarters after the business cycle peak. (18) More recent developments have affected growth performance and employment dynamics, but they are beyond the scope of this analysis. (19)

Summing up, it seems that accompanying external flexibility measures with internal flexibility practices offers a greater chance for swift recovery of labour market performance without penalizing productivity dynamics. It is also crucial, however, to design instruments to respond to the processes actually occurring and suited to the country in question.

APPENDIX

Identifying the 'peak' quarters for the analysed countries

[FIGURE A1 OMITTED]

[FIGURE A2 OMITTED]

[FIGURE A3 OMITTED]

[FIGURE A4 OMITTED]

[FIGURE A5 OMITTED]

[FIGURE A6 OMITTED]

[FIGURE A7 OMITTED]

[FIGURE A8 OMITTED]

[FIGURE A9 OMITTED]

[FIGURE A10 OMITTED]

[FIGURE A11 OMITTED]

[FIGURE A12 OMITTED]

[FIGURE A13 OMITTED]

[FIGURE A14 OMITTED]

[FIGURE A15 OMITTED]

REFERENCES

Arpaia, A and Curci, N. 2010: EU labour market behaviour during the great recession European Economy Economic papers 405, February.

Babetskii, I. 2005: Trade integration and synchronization of shocks: Implications for EU enlargement. Economics of Transition 13(1): 105-138.

Baxter, M and King, RG. 1999: Measuring business cycles: Approximate bandpass filters for economic time series. Review off Economics and Statistics 81(4): 575-593.

Beaudry, P and Collard, F. 2002: Why has the employment-productivity trade-off among industrialized countries been so strong? NBER Working paper no. 8754.

Becket, ID and Gordon, RJ. 2008: The role of labor market changes in the slowdown of European productivity growth. CEPR Discussion paper no. 6722.

Beegle, K, Frankenberg, E and Thomas, D. 1999: Measuring change in Indonesia. Labor and population program. Working paper series, 99-107, RAND: Santa Monica, CA.

Belorgey, N, Lecat, R and Maury, TP. 2006: Determinants of productivity per employees: An empirical estimation using panel data. Economics Letters 91(2): 153-157.

Bernard, AB and Jones, CI. 1996: Comparing apples to oranges: Productivity convergence and measurement across industries and countries. American Economic Review 86(5): 1216-1238.

Boeri, T and Garibaldi, P. 2009: Beyond eurosclerosis. Economic Policy 24(59): 409-445.

Boulhol, H and Turner, L. 2009: Employment--productivity trade-off and labour composition OECD Economics Department Working paper no. 698.

Caballero, RJ, Hoshi, T and Kashyap, AK. 2008: Zombie lending and depressed restructuring in Japan. American Economic Review 98(5): 1943-1977.

Chari, V, Kehoe, P and Mc Grattan, E. 2006: Business cycle accounting. Research Department Staff Report 328, Federal Reserve Bank of Minneapolis.

Christiano, LJ and Fitzgerald, TJ. 2003: The band pass filter. International Economic Review 44(2): 435-465.

Cogley, T and Nason, JM. 1995: Effects of the Hodrick-Prescott filter on trend and difference stationary time series Implications for business cycle research. Journal of Economic Dynamics and Control Elsevier 19(1-2): 253-278.

Corbae, D and Ouliaris, S. 2006: Extracting cycles from nonstationary data. In: Phillips, PC, Corbae, D, Durlauf, SN and Hansen, BE. (eds). Econometric Theory and Practice: Frontiers of Analysis and Applied Research. Cambridge University Press: Cambridge, UK, New York.

Darvas, Z and Szapary, G. 2008: Business cycle synchronization in the enlarged EU. Open Economies Review 19(1): 1-19.

Eichhorst, W, Feil, M and Marx, P. 2010: Crisis, What Crisis? Patterns of Adaptation in European Labor Markets IZA Discussion paper no. 5045.

European Commission. 2009a: Economic Forecast. European Commission: Brussels November.

European Commission. 2009b: Employment in Europe 2009. European Commission: Brussels November.

Fallon, PR and Lucas, REB. 2002: The impact of financial crises on labor markets, household incomes, and poverty: A review of evidence. World Bank Research Observer 17(1): 21-45.

Fidrmuc, J and Korhonen, I. 2006: Meta-analysis of the business cycle: Correlation between the Euro area and the CEECs. CESIFO Working paper 1693.

Frankel, J and Rose, AK. 1998: The endogeneity of optimum currency area criteria. Economic Journal 108(449): 1009-1025.

Fulco, L. 1984: Strong post-recession gain in productivity contributes to slow growth in labor costs. Monthly Labor Review 107(3): 3-10.

Gomez, V and Maravall, A. 1997: Programs TRAMO (Time Series Regression with Arima Noise, Missing Observations and Outliers) and SEATS (Signal Extractions in ARIMA Time Series)--Instructions for the User Banco de Espana, Working paper SGAPE-9700.

Growiec, J. 2009: On the measurement of technological progress across countries. MPRA paper 19321, University Library of Munich: Germany.

Hodrick, RJ and Prescott, EC. 1997: Postwar U.S. business cycles: An empirical investigation. Journal of Money, Credit and Banking 29(1): 1-16.

ILO. 2010: Global Employment Trends. International Labour Organization: Geneva.

IMF. 2009: World Economic Outlook. Sustaining the Recovery. IMF: Washington.

IMF. 2010: World Economic Outlook. Rebalancing Growth. IMF: Washington.

Kaiser, R and Maravall, AM. 2001: Measuring Business Cycles in Economic Time Series. Springer-Verlag: New York.

King, RG and Rebelo, ST. 1993: Low frequency filtering and real business cycles. Journal of Economic Dynamics and Control 17(1-2): 207-231.

Kose, AM, Prasad, ES and Terrones, ME. 2003: How does globalization affect the synchronization of business cycles? The American Economic Review 93(2); Papers and Proceedings of the One Hundred Fifteenth Annual Meeting of the American Economic Association, May 57-62.

Lee, J. 2000: The Robustness of Okun's law: Evidence from OECD countries. Journal of Macroeconomics 22(2): 331-356.

Marelli, E and Signorelli, M. 2010: Employment, productivity and models of growth in the EU. International Journal of Manpower 31 (7): 701-712.

Mukand, S and Rodrik, D. 2005: In search of Holy Grail: Policy convergence, experimentation and economic performance. American Economic Review 95(1): 374-383.

Mulligan, C. 2009: What caused the recession of 2008? Hints from labor productivity. NBER Working papers series no. 14729, February.

Padalino, S and Vivarelli, M. 1997: The employment intensity of economic growth in the G-7 countries. International Labour Review 136(2): 191-213.

Pichelmann, K and Roeger, W. 2008: Employment and labour productivity in the EU: Reconsidering a potential trade-off in the Lisbon strategy. In: Eichengreen, B, Landesmann, M and Stiefel, D. (eds). The European Economy in an American Mirror. Routledge: New York, pp. 128-142.

Rose, A. 2000: One money, one market: the effect of common currencies on trade. Economic Policy 15(30): 7-45.

(1) Note that, during a recession a decline in (the growth of) GDP causes a decline in labour demand and employment, hence an increase in unemployment, that is, Keynesian unemployment due to a lack of aggregate demand. The rise in unemployment can however be partly reinforced by 'wage rigidities' ie, classical unemployment.

(2) An indirect consequence is that nominal unit labour costs are bound to increase despite a deceleration in wage dynamics because compensation per employee will adjust, at least in the variable component, as shown by Arpaia and Curci (2010). The resulting loss of competitiveness will dampen foreign demand and exports, especially in countries, such as the European ones, that cannot benefit from large increases in domestic demand because of fiscal and macroeconomic adjustments. This in turn will worsen recovery prospects.

(3) The paper also provides econometric corroboration of the diminishing returns for the employment rate hypothesis.

(4) The employment rate reversed its tendency to decline only a few years after the huge transition recession of the early 1990s. It should be noted that, as for transition countries, the last 2007-2008 crisis caused a second negative job shock in less than a generation.

(5) The end of eurosclerosis (see Boeri and Garibaldi, 2009) has, however, been unbalanced because of the diffusion of temporary contracts, the spreading of dual labour markets and the consequent effects on differentiated productivity and wage dynamics.

(6) This approach is sometimes also followed in business cycle studies. For example, Chaff et al. (2006) found that, both in the Great Depression and in the 1982 recession, the efficiency and labour wedges together accounted for essentially all fluctuations.

(7) The clearest example is the Great Recession, which in 2008-2009, after the financial crisis, hit all countries in the world although in China, India and other emerging economies there was just a slight deceleration in their previous growth rates.

(8) Concerning tire mentioned core of countries, a different conclusion can be reached if we look at the different financial conditions of peripheral countries compared with core countries or at some other real aspects of convergence such as competitiveness, trade balance, inflation differentials and so on. Previous studies on cycle correlation referred to the integration effect of currency unions (see Frankel and Rose, 1998, Rose, 2000).

(9) In the United States, the unemployment rate in March 2010 stood at 9.7% compared with 6.2 % in September 2008 and declined to 9.1% in August 2011, It increased in the EU-27 to 9.6% in March 2010 against 7.1% in September 2008 and was still at 9.5% in August 2011.

(10) The cyclical behaviour of unemployment during recessions and recoveries is also discussed in IMF (2010, Chapter 3). After the crisis, the highest increases of unemployment rates in the world were recorded in developed economies, the EU and the remaining countries of Europe, while working poverty dramatically increased in many regions, especially in South East Asia and the Pacific, South Asia, North Africa, and Sub Saharan Africa. See ILO (2010)

(11) The annual data could be acquired for a larger number of countries from the GDRC database, but quarterly data are more appropriate for this type of analysis. The insufficient data availability from Eurostat was complemented, when feasible, by data acquired from the central statistical offices of the analysed countries. However, the changes in SNA definitions and data collection standards make it impossible to extend the data set beyond the analysed sample of countries and crisis episodes.

(12) The Hodrick-Prescott filter estimates trend by smoothing, in effect, by taking a weighted moving average of the original series, where the moving average is symmetric and centred. In practice, one must choose how smooth the resulting trend should be. Hodrick and Prescott (1997) suggested a smoothness parameter of 1,600 for quarterly data, and that parameter value is most commonly used in practical applications.

(13) Indeed, using other filters did not alter the identification of peak quarters. Filtering was performed with standard [lambda] = 1600 and without any additional modifications.

(14) In some cases, eight quarters of data were not available. In such instances, all the latest available data were used.

(15) In fact, this is the pattern followed by the US economy, which suggests the observed slowdown, irrespective of the additional adverse contributions from the financial sector, was originally neo-classical in nature. As to RBC theory see King et al. (1993).

(16) See Eichhorst et al. (2010) and section 'Impact of the recent financial crisis'.

(17) Indeed, the United Kingdom followed a path similar to Spain and Hungary in the previous crisis (the peak was identified at 1999q3).

(18) An important exception is provided by Germany, the only EU country where unemployment did not increase (despite the huge recession in 2009) and where nonetheless productivity gains were realized thanks to the recovery commenced in the second part of 2009.

(19) We just mentioned that in certain EU countries, unemployment started to increase again at the end of 2011, because of the effects of the 'sovereign debt crisis'.

ENRICO MARELLI [1], MARCELLO SIGNORELLI [2] & JOANNA TYROWICZ [3,4,5]

[1] Department of Economics, University of Brescia, via San Faustino 74/B, 25122 Brescia, Italy. E-mail: emarelli@eco.unibs.it

[2] Department of Economics, Finance and Statistics, University of Perugia, via A. Pascoli, 20, 06123 Perugia, Italy. E-mail: signorel@unipg.it

[3] Faculty of Economics, University of Warsaw, ul. Dluga 44/50, 00-241 Warszawa, Poland

[4] National Bank of Poland, ul. Swietokrzyska 11/21, 00-919 Warszawa, Poland

[5] Rimini Centre for Economic Analysis, Rimini, Italy. E-mail: jtyrowicz@wne.uw.edu.pl

Table 1: Size of sample by country

                                                        Total number
Country          Beginning of sample   End of sample   of observations

Belgium                1990q2             2010q4             84
Czech Republic         1995q2             2010q4             60
Denmark                1990q2             2010q4             84
Estonia                1995q2             2010q4             60
Finland                1990q2             2010q4             84
France                 1990q2             2010q4             84
Germany                1991q2             2010q4             79
Hungary                1995q2             2010q4             60
Italy                  1990q2             2010q4             84
Latvia                 1995q2             2010q4             60
Lithuania              1995q2             2010q4             60
Netherlands            1990q2             2010q4             84
Poland                 1995q2             2010q4             60
Spain                  1990q2             2010q4             84
United Kingdom         1992q2             2010q4             75

Note: For the Czech Republic, only data from the labour force survey
were available. In the case of France, employment based on the
System of National Accounts was no longer reported as of 2003, and
we therefore used the labour force survey data. Employment data are
not available for 199982 and 199983 for Poland. If seasonality and
working days adjusted data were unavailable, in three additional
countries, the procedure of adjusting for seasonality was
implemented using the Tramo/Seats procedure (Gomez and Maravall,
1997), with the assistance of Demetra.

Source: Eurostat

Table 2: Number of peaks by country (a)

                     Number of
Country           peaks in sample   Peaks identified

Belgium                  2          2000q4, 2008q2
Czech Republic           2          1998q2, 2007q3
Denmark                  2          2000q4, 2008q2
Estonia                  2          1997q4, 2007q4
Finland                  4          1990q4, 1998q2, 2001q1, 2007q4
France                   3          1992q1, 2001q1, 2008q1
Germany                  3          1995q3, 2002q1, 2008q2
Hungary                  2          1998q1, 2008q1
Italy                    3          1992q1, 2001q1, 2008q2
Latvia                   2          1997q4, 2008q1
Lithuania                2          1997q4, 2008q2
Netherlands              3          1992q1, 2001q1, 2008q1
Poland                   2          2000q2, 2008q1
Spain                    2          1992q1, 2007q4
United Kingdom           2          1999q3, 2008q1

(a) Please consult the Appendix for the graphical representations of
the cyclical properties of GDP for each country in the sample.

Source: Own computation based on Eurostat