VoxEU Column Global crisis Labour Markets

The wage–employment nexus: A tale of persistence

The weak relationship between wage dynamics and unemployment in the euro area since the Global Crisis is widely perceived as one of the main causes of the ‘twin puzzle’ of missing disinflation between 2009 and 2011, and missing inflation thereafter. This column attributes the weak response of nominal wage growth to employment dynamics since 2008 to the countercyclical behaviour of labour productivity, which is driven, in turn, by the exceptionally high persistence of the downturn and the subsequent recovery.

Despite large swings in employment dynamics, throughout the double-dip recession and subsequent recovery, the wage response in the euro area was very limited. The unconditional dynamic correlation between nominal wage growth and changes in employment weakened after the Great Recession (Figure 1). 

Figure 1 Unconditional dynamic correlation between wages and employment before and after the Great Recession

Notes: The figure reports the unconditional correlation between y-o-y log-differenced nominal wages (compensation per employee) and employment at lags 0, 1, …, 8. Data are quarterly and taken from the ECB Statistical Data Warehouse. Blue (red) bars denote the statistics for the period 1995: Q1 – 2008: Q2 (2008: Q3 – 2018: Q2).

A scatterplot of nominal wage inflation and employment growth (Figure 2) strengthens this evidence. It is notable that in the years after 2008, the relation between the two variables is nearly flat.1

Figure 2 The flattening of the euro area wage Phillips curve after the Global Crisis

Notes: The figure reports a scatterplot between y-o-y log-differenced nominal wages (compensation per employee) and the fifth lag of y-o-y log-differenced employment, i.e. the one with the highest unconditional correlation in the period before the Global Financial Crisis. Data are quarterly and taken from the ECB Statistical Data Warehouse. Blue (red) dots denote the period 1995: Q1 – 2008: Q2 (2008: Q3 – 2018: Q2). The black line represents the respective linear fit.

Nominal wage growth in the euro area decreased only modestly over the crisis years and struggled to gain consistent momentum during the subsequent recovery.2 Given their importance for the inflation outlook, nominal wage dynamics have taken centre-stage in the monetary policy debate in recent years (Draghi 2016). Figure 3 compares employment (Panel a) and compensation per employee (Panel b) across the euro area recoveries. The recovery that began in early 2013 certainly cannot be described as ‘jobless’. During this period of recovery, employment has been expanding at its fastest rate compared to all the other expansion periods. However, against the background of a tightening labour market, the trajectory of nominal wage growth has been very muted – so much so that the current upswing has been labelled a ‘wage-less recovery’. The behaviour of nominal wages appears less puzzling in the light of the feeble productivity dynamics (Panel c).

Figure 3 Euro area recoveries 

Notes: The figure reports a comparison among employment (panel a), compensation per employee (panel b) and output per worker (panel c) in all the expansion periods dated by the CEPR business cycle dating committee. Each variable is computed as index with reference period the quarter of the different troughs. The x-axis measures the number of quarters after the troughs. Data are taken from the AWM database.

One possible explanation relies on the phenomenon of ‘pent-up wage deflation’ (Daly and Hobijn 2014), which in turn depends on the presence of downward nominal wage rigidities. During recessions, firms might face limits to cutting nominal wages. As a result, the labour market disproportionately adjusts through the unemployment margin, rather than through nominal wages. If firms face constraints in lowering compensation when the labour market is exceptionally weak, in the earlier part of the recovery they may not then feel the need to raise wages to attract workers. As a result, nominal wages might rise relatively slowly as the labour market strengthens. The theoretical model by Daly and Hobijn (2014) formalises this idea and suggests that this mechanism can be especially pronounced in a low-inflationary environment.

Labour productivity has turned countercyclical

Our recent work (Conti et al. 2019) puts forward a complementary explanation. Since the Great Recession, the euro area has demonstrated a feeble relationship between nominal wage growth and unemployment. This trend has reflected a movement in the response of labour productivity (measured as output per worker) towards an increase in employment, from nil up to 2008 (acyclical), to negative (countercyclical) since then.

Figure 4 shows the impulse responses of nominal wages and labour productivity to a 1% increase in employment, before and after the Great Recession.3 As expected (and consistent with the flattening of the wage Phillips curve), the reaction of nominal wages looks strikingly different across the two samples. Before the Global Crisis, the curves display a hump shaped response, with the peak reached after three years. Since the crisis, nominal wages barely move. This novel result lies in the fact that the vanishing reaction of wages to a change in employment comes together with a change in the behaviour of labour productivity. In the years before the Great Recession, a 1% increase in employment led to a positive (albeit statistically non-significant) increase in output per worker. By contrast, after the financial crisis, the response of labour productivity is large and negative.4 This countercyclical behaviour is crucial for explaining the weak response of nominal wages to changes in employment levels. 

Figure 4 IRFs to 1% increase in employment


Notes: The figure reports the results of a 3-variables BVAR including log(labour productivity), log(employment) and log(nominal compensation per employee) identified recursively. The red line is the median IRF derived from the posterior distribution of the Bayesian VAR (1995:Q1-2008:Q2). The black line is the median derived from the posterior distribution of the Bayesian VAR (2008:Q3-2018:Q2). The dashed red lines and the grey shaded area represent the 16th and 84th percentiles derived from the posterior distribution of the Bayesian VAR (1995:Q1-2008:Q2 and 2008:Q3--2018:Q2, respectively). 

Interpreting the change: The role of business cycle persistence

Why has labour productivity turned countercyclical since the Global Crisis? We argue that the persistence of the business cycle phase plays a crucial role in shaping the response of labour productivity. Similarly, it is through this business cycle channel that we can also explain  the employment-wages ‘pass-through’. 

In the short term, if the labour force cannot be adjusted in a costless manner due to hiring and firing costs, it may pay firms to ‘hoard’ labour over the cycle. As a result, firms may react to changes in demand conditions by varying the intensive margin of labour utilisation (hours per worker or workers’ effort), as opposed to making employment cuts. This is one of the traditional explanations for the procyclicality of labour productivity. The variable utilisation of labour (which is more intense in booms than in downturns) generates the perception of short-run increasing returns to labour. Using a theoretical model, we show that the optimal degree of labour hoarding (and hence the procyclicality of labour productivity) decreases with the expected duration of the cyclical phase. Indeed, since the cost of adjusting the labour force increases faster than the amount of labour to be adjusted, firms want to make small, gradual changes to the number of employees they administer. Thus, firms are more willing to vary employment the longer the cyclical phase is expected to last. The more transitory the cyclical phase is perceived to be, the more firms rely on labour hoarding, adjusting the intensive margin of labour utilisation as a response.5

All else being equal, transitory demand shocks are associated with a procyclical response of labour productivity and a large wage-employment multiplier. Persistent demand shocks generate a countercyclical response and a small wage-employment multiplier (Figure 5). The length of the cyclical phases has been a salient feature in recent years. The Great Recession and the sovereign debt crisis can be viewed as a single (and exceptionally long) contractionary phase, which inflicted persistent scars on the euro area economy, followed by a long and weak recovery. 

Figure 5 Business cycle persistence, the cyclicality of labour productivity and the wage-employment multiplier



Notes: Panel a reports the correlation between labour productivity and employment (y-axis) for different degrees of the persistence of the demand shock. Panel b report the compensation per employee – employment multiplier (measured by the ratio of their cumulated impulse response functions to a demand shock), for different degrees of the persistence of the demand shock. The different lines in each panel correspond to different calibrations of deep parameters other than the persistence of the shock.

The theoretical findings are supported by empirical evidence. In our recent work, we use a Bayesian VAR to identify aggregate demand shocks by means of sign restrictions, taking into account the theoretical predictions regarding the behaviour of labour productivity. We show that persistent demand shocks produce a lower response within nominal wages compared to employment and have become a relevant source of fluctuations in employment and nominal wage dynamics following the outbreak of the Global Crisis.


Conti, A M, E Guglielminetti and M Riggi (2019). “Labour productivity and the wageless recovery”, Temi di Discussione, Bank of Italy Working paper series, 1257, December.

Daly, M C and B Hobijn (2014), “Downward Nominal Wage Rigidities Bend the Phillips Curve”, Journal of Money, Credit and Banking 46(S2): 51- 93.

Draghi, M (2016), “The state and prospects of the euro area recovery”, Speech at the European Banking Congress, Frankfurt, 18 November.

Gali, J and L Gambetti (2019), “Has the U.S. Wage Phillips Curve Flattened? A Semi-Structural Exploration”, NBER Working Paper No. 25476, January.

Ohanian, L E (2002), “Why did productivity fall so much during the Great Depression?", Federal Reserve Bank of Minneapolis Quarterly Review 26(2). 


1 A similar flattening has emerged in the US. See Gali and Gambetti (2019) for an analysis of the flattening of the US wage Phillips curve. 

2 Nominal wage growth has recovered only since the end of 2017. Nominal compensation per employee increased by 2.6% y-o-y in 2018:Q2, the last available observation in our analysis, as in the pre-crisis period (1995:Q1 - 2008:Q2). Nominal wages instead averaged a 1.8% y-o-y growth over the period 2008:Q3- 2018:Q2.

3 The impulse responses are obtained by estimating a small VAR model which includes labour productivity, employment (heads) and nominal wages and is identified assuming the above recursive order: this captures a stylised wage Phillips curve in which compensation per employee depends on employment and productivity.

4 The countercyclical reaction of the labour productivity since the crisis holds across several robustness checks, such as a larger set of variables and different orderings.

5 The role of the expected persistence of the cyclical phase in driving the ‘labour hoarding’ attitude of firms has been somewhat neglected in the literature. A notable exception is Ohanian (2002), who questioned the validity of the labour hoarding explanation of procyclical labour productivity over the Great Depression, based on its duration: “The duration of the Depression, however, raises questions about the plausibility of the labour-hoarding explanation. It is difficult to reconcile the labour-hoarding thesis, which is based on the temporary nature of recessions, with a major depression that lasted well over a decade. It seems unlikely that firms hoarded workers because they mistakenly expected the Depression to end quickly; consumption data suggest that the Depression was expected to last a long time.

1,155 Reads