VoxEU Column Productivity and Innovation Labour Markets

Labour adjustment and productivity dynamics: Cross-country evidence and policy implications

Labour productivity is more procyclical in OECD countries with lower employment volatility. To capture this new stylised fact, this column proposes a business cycle model with employment adjustment costs, variable hours, and labour effort. In the model with variable effort, it shows that greater labour market frictions are associated with procyclical labour productivity as well as stable employment. In contrast, the constant-effort model fails to replicate the observed cross-country pattern in the data. Labour market deregulation has a greater effect on labour productivity cyclicality and employment volatility when effort can vary. 

Using data from 1984 to 2019, Figure 1 shows that countries with lower employment volatility are characterised by more procyclical labour productivity, that is, by a higher correlation between output and measured labour productivity (i.e. output per hour worked).

This pattern holds also for the Great Recession period from 2008 to 2013, which is widely believed to have been driven by deficient demand. Replacing employment volatility with unemployment volatility does not alter our key empirical message, suggesting that the participation margin does not play a large role for the stylised fact shown here. 

Figure 1 Relative employment volatility and cyclicality of labour productivity


Notes: Sample period 1984q1-2019q4. Labour productivity measured as quarterly real output per hour. Employment volatility measured relative to output. Cyclical component of log productivity, log employment and log output extracted with HP filter (Hodrick and Prescott 1997). Volatility measured as standard deviation, cyclicality measured as correlation between cyclical components of output and labour productivity. 
Data sources: OECD, Eurostat, Ohanian and Raffo (2012), ILO, National Offices of Statistics. 

What does economic theory say?

Economic models struggle to explain the procyclicality of labour productivity when demand shocks are important drivers of business cycles. This is because those shocks lead to countercyclical movements in labour productivity, while technology (or supply-side) shocks give rise to a co-movement of output and labour productivity.

First, a natural candidate explanation for the pattern in Figure 1 is that technology shocks are the dominant source of business cycle fluctuations in countries with highly procyclical productivity, while demand shocks are more important in countries where the cyclicality of labour productivity is low. This is consistent with employment being rather stable in the former group of countries, and more variable in the latter. However, the Great Recession was a large shock that hit several countries simultaneously and, arguably, in similar ways. This suggests that the large cross-country variation in business cycle moments shown in Figure 1 can be traced to structural differences across economies, which in turn led to differences in shock transmission, rather than the shock mix itself being idiosyncratic. 

Second, the Great Recession is widely believed to have been driven by deficient demand (e.g. Christiano et al. 2015). In a demand-driven recession, a drop in labour productivity is difficult to explain with standard business cycle models in the absence of variable factor utilisation. With unchanged technology, we expect firms to cut back their labour input as demand for their products declines. Labour productivity goes down only if labour falls by less than output. With capital fixed in the short run, this means that labour is utilised less intensively – i.e. effort falls – during the downturn. Variable capital utilisation as in Christiano et al. (2005) could, as an alternative model feature, generate procyclical labour productivity without the necessity to endogenise labour effort. However, Lewis et al. (2019) show that effort clearly outperforms capital utilisation in terms of explaining the euro area business cycle.

Employment adjustment frictions… 

In a new paper (Dossche et al. 2021), our aim therefore is to develop a model that can replicate the evidence in Figure 1 without relying on cross-country differences in the relative importance of technology versus demand shocks. Rather, we focus on differences in labour market adjustment, coupled with variable labour utilisation, as the key candidate explanation. In particular, we attribute the procyclicality of labour productivity to variations in effort, which in turn result from a reluctance of firms to adjust the workforce. This idea of labour hoarding dates back to Okun (1963) and Oi (1962). To model labour adjustment adequately, two ingredients are important: labour market rigidities and variable hours per worker.

Employment protection remains restrictive in many countries, especially in the euro area. The OECD's employment protection legislation (EPL) index is defined over a range from 0 (very little protection) to 5 (very stringent protection). Its value for 2019 ranges from 0.09 in the US to 3.6 in the Netherlands. Spain is a special case. While workers on permanent contracts enjoy a high degree of employment protection, temporary contracts with low firing costs are widespread. This dual labour market gives rise to US-style labour market flexibility, reflected in high employment volatility (see Figure 1). There is also large variation in redundancy pay across countries; data from the World Bank's Doing Business Report 2017 range from zero in the US to 27 weeks of salary in South Korea, and even higher numbers for non-OECD countries. 

High costs of laying off employees in times of low demand discourage labour adjustment along the extensive margin. Already Nickell (1979) found that fluctuations in hours were higher and employment fluctuations lower after the 1966 Redundancy Payments Act increased the cost of dismissal in the UK. In a sample of 20 OECD countries over the period 1975-1997, Nunziata (2003) shows that stricter employment protection and looser working time regulations were associated with a lower variability of employment over the cycle. This finding is confirmed in more recent data by Gnocchi et al. (2015).

…and flexibility at the intensive labour margin

Today, around one half of the adjustment in total hours worked in the euro area is through changes in hours per employee rather than changes in employment (Dossche et al. 2019). Short-time work (STW) schemes and working time accounts, used extensively in Germany for example, make hours worked more flexible. Lydon et al. (2019) show that the short-time work take-up rate among firms is positively related to greater firing costs and more stringent employment protection. 

Our final crucial model ingredient is variable labour utilisation, or effort. Labour effort cannot be observed directly. However, several indirect measures suggest that it is positively correlated with the business cycle – workplace accidents, sick leave, and indicators of bad health outcomes are all procyclical. Firms report that they pay more for labour in recessions than is strictly necessary. Evidence from workers’ time use surveys and self-reported effort point in the same direction. According to the prominent ‘shirking’ theory, effort results from the fear of being laid off when caught slacking off at work. Workers exert more effort during downturns when the probability of finding another job is relatively low. Indeed, a couple of papers find evidence of countercyclical effort in a single firm, suggesting that shirking might play a role at the micro level. At the macro level, though, the evidence overwhelmingly supports procyclical effort. 

A model with three labour margins

We develop a business cycle model with capital and three labour margins: employment, hours per worker, and effort per hour. Importantly, firms face employment adjustment costs, which use up part of their output. Workers are expected to provide a certain amount of effective labour; they choose the combination of hours and effort per hour that minimises their disutility from working. 

Figure 2 Baseline model responses to a 1% technology shock


Notes: Blue dashed lines show constant-effort model, red solid lines show model with variable labour effort. Impulse responses measured as percentage deviation from steady state. Horizontal axis shows time horizon in quarters.

Consider Figure 2. In the standard model without effort, depicted by the blue dashed lines, employment and hours rise in response to a positive technology shock, while measured labour productivity increases. In the presence of variable labour effort (red solid lines), effort also expands. Hours increase by more, and employment by less, than in the constant-effort model. This allows firms to economise on employment adjustment costs. Labour productivity responds in a procyclical fashion, rising by more in the model with effort.

Figure 3 Baseline model responses to a 1% demand shock


Notes: Blue dashed lines show constant-effort model, red solid lines show model with variable labour effort. Impulse responses measured as percentage deviation from steady state. Horizontal axis shows time horizon in quarters.

An expansionary demand shock is depicted in Figure 3. Also here, the presence of variable effort shifts part of the adjustment away from the extensive margin and towards the intensive labour margin. Employment moves by less, while hours and effort adjust by more in response to the shock. Labour productivity is countercyclical conditional on the demand shock. With variable effort, measured productivity drops by less than in the constant-effort model. As a result, the drop in the wage is visibly reduced.

We show that, in a model with labour effort, greater employment adjustment frictions imply more procyclical labour productivity along with more stable employment, consistent with the observed cross-country heterogeneity. The constant-effort model fails to replicate the pattern in the data. As a consequence, labour market deregulation – a reduction in firms' employment adjustment costs – reduces the cyclicality of labour productivity by more when effort can vary than in the case where effort is constant. Variable effort is thus relevant for evaluating the effect of such a reform. An important lesson from this exercise is that the cyclicality of labour productivity does not in itself reveal the relative importance of technology versus demand shocks as the dominant source of fluctuations.

The labour adjustment process with its associated costs is complex, multi-faceted, and heterogeneous across countries. Further research – ideally focusing on individual countries – is needed to analyse how different types of labour market institutions affect both employment volatility and the cyclicality of labour productivity.


Christiano, L, M Eichenbaum and C Evans (2005), “Nominal rigidities and the dynamic effects of a shock to monetary policy”, Journal of Political Economy 113(1):1-45.

Christiano, L J, M S Eichenbaum and M Trabandt (2015), “Understanding the Great Recession”, American Economic Journal: Macroeconomics 7(1):110-67.

Dossche, M, A Gazzani and V Lewis (2021), “Labor Adjustment and Productivity in the OECD”, CEPR Discussion Paper 16202, ECB Working Paper 2571, Bundesbank Discussion Paper 22/2021. 

Dossche, M, V Lewis and C Poilly (2019), “Employment, hours and the welfare effects of intra-firm bargaining”, Journal of Monetary Economics 104: 67-84.

Gnocchi, S, A Lagerborg and E Pappa (2015), “Do labor market institutions matter for business cycles?”, Journal of Economic Dynamics and Control 51: 299-317. 

Hodrick, R J and E C Prescott (1997), “Postwar U.S. business cycles: an empirical investigation”, Journal of Money, Credit and Banking 29(1): 1-16.

Lewis, V, S Villa and M H Wolters (2019), “Labor productivity, effort and the euro area business cycle”, Discussion Papers 44/2019, Deutsche Bundesbank.

Lydon, R, T Y Mathä and S Millard (2019), “Short-time work in the Great Recession: Firm-level evidence from 20 EU countries”, IZA Journal of Labor Policy 8(1):2.

Nickell, S (1979), “Unemployment and the structure of labor costs”, Carnegie-Rochester Con-ference Series on Public Policy 11: 187-222.

Nunziata, L (2003), “Labour market institutions and the cyclical dynamics of employment”, Labour Economics 10(1): 31-53.

Ohanian, L E and A Raffo (2012), “Aggregate hours worked in OECD countries: New measurement and implications for business cycles”, Journal of Monetary Economics 59(1): 40-56.

Oi, W Y (1962), “Labor as a quasi-fixed factor”, Journal of Political Economy 70: 538-538.

Okun, A M (1963), “Potential GNP: Its measurement and significance”, Cowles foundation paper, Cowles Foundation for Research in Economics at Yale University.

945 Reads