Beyond the Phillips curve: Understanding low wage growth in the euro area

Wage growth in the euro area over 2013 to 2017 was subdued despite notable improvements in the labour market, leading some to claim a breakdown of the output–inflation relationship. This column presents comparative analyses of wage developments in the euro area, showing that the Phillips curve is alive and well and can be used to explain much of the weakness in wage growth during 2013-2017. Other factors also found to have played a role include compositional effects, the possible non-linear reaction of wage growth to cyclical improvements, and structural and institutional factors.

The euro area labour market improved strongly from 2013 onwards, while wage growth remained subdued until recently. From the second quarter of 2013 to the second quarter of 2019, the unemployment rate in the euro area decreased from 12.1% to 7.6%. Employment increased by 11.2 million in this period and many more people are now in employment than before the crisis. Wages, on the other hand, barely grew from 2013 to 2017, and wage growth has picked up only recently (see Figure 1). This observation holds true regardless of how nominal wage growth is measured or defined.

Figure 1 Measures of wage growth over the cycle

At the same time, wage growth was persistently and substantially over-predicted by international institutions and professional forecasters (see Figure 2), while employment grew more strongly than expected. This raised concerns over whether the relationship between slack and wage growth had changed. Over-predicting wage growth was a widespread phenomenon in this period, also affecting other jurisdictions beyond the euro area and the EU (IMF 2017, 2019, OECD 2019, European Commission 2019).

Figure 2 Wage forecast error comparison across different forecasters for one calendar year ahead (autumn forecasts – percentage points, annual data)

What can explain low wage growth and the persistent over-prediction? Our work presented in Nickel et al. (2019) sheds new light on these questions, taking strong country heterogeneity of labour markets across euro area countries and the far-reaching differences in economic and institutional forces behind the wage formation process into account.

Figure 3 Decomposition of latest wage growth into its main drivers in the euro area (deviations from mean in year-on-year growth terms, percentage point contributions)

Slack in the labour market, relatively weak productivity growth, and the impact of the prolonged period of low inflation have been holding back wage growth in the euro area. Decomposing wage growth into the contributions of its main determinants based on a benchmark Phillips curve specification shows that there are three key factors explaining wage growth developments in the past few years, with changing importance over time (see Figure 3). First, labour market slack exerted a substantial negative drag on wage growth until the end of 2016, while more recently increasing labour market tightness has pushed wages up. Second, in 2016 17 low past inflation became the dominant driver of low wage growth, but that drag has been dissipating as well. Third, between 2014 and 2016, as well as in 2018, low productivity growth also contributed to low wage growth – albeit to a smaller degree than the other two factors, especially during the subdued wage growth period. In 2016-17 persistent negative residuals remain in the decomposition of wage growth – also highlighting the importance of other factors beyond cyclical drivers as discussed above. In the course of 2018, wage growth was above the mean and the negative residuals disappeared.

Despite some cross-country variation, wage growth appears to be responsive to the cyclical position of the labour market, as well as to productivity and past inflation in most euro area countries. Based on a simple benchmark Phillips curve specification using the headline unemployment rate, backward-looking inflation expectations and productivity growth, a cross-country exploration of the wage drivers shows that the average country-by-country estimated coefficients are broadly consistent with euro area coefficients (see Figure 4). This specification is similar in spirit to the price Phillips curve proposed in Ciccarelli and Osbat (2017) and Bobeica and Sokol (2019).

Figure 4 Estimated coefficients of the considered wage drivers (coefficient)

The relative importance of the various wage growth drivers differs markedly across countries. The euro area Phillips curve decomposition might conceal country-specific developments, which could offset each other at the aggregate level. To this end, Figure 5 focuses on the relative average contributions of each driver for each country over the euro area low wage growth period (2013-17). For most countries a combination of considerable labour market slack, low productivity growth, and below-average inflation (expectations) kept wage growth below its sample mean.

Figure 5 Decomposition of wage growth into its main drivers across EU countries over the period 2013 17 (deviations from country model-implied mean in year-on-year growth terms – percentage point contributions, averaged over 2013- 2017)

Standard Phillips curve-type mechanisms do not paint the entire picture, given sizeable residuals in the euro area as a whole and certain euro area countries in particular, especially in the period from 2016 to 2017. Factors beyond the standard Phillips including compositional effects (such as in Christodoulopoulou and Kouvavas 2018 and Kouvavas et al. 2019), non-linearities in the reaction of wage growth to slack (such as in Bonam, et al. 2018 and in Byrne and Zekaite 2018) but potentially also structural drivers like digitalisation, globalisation, or ageing (for the latter see Bodnar 2018 and Dossche and Koester 2018) are found to also have played a role for low wage growth in the euro area.

Figure 6 Comparison of one-year ahead forecasts from benchmark Phillips curves and Eurosystem/ECB staff at different vintages

Wage Phillips curve also give reliable signals in real time, making them a valuable tool for cross-checking wage forecasts. Figure 6 shows forecasts for the next year at different forecast vintages. This means that values shown for the first quarter of 2012 relate for example to the March 2012 exercise forecast for 2013 – the first year of the “low wage period”, which was characterised by sluggish wage growth despite an improving labour market. For the medium-term (forecast of wage growth between the current and the next year) a benchmark Phillips-curve specification would have under-predicted wage growth in the first phase of the low-wage period in 2013/14 (see Figure 6) and slightly overestimated wage growth in the later part (2015 17), while the Eurosystem/ECB staff projections over-predicted wage growth basically for the whole low-wage period. This supports the view that a relatively straightforward benchmark wage Phillips curve (and some other well-performing specifications) could have acted as an important cross-check and counterweight to the Eurosystem/ECB staff wage projections over the low-wage period – and especially in its first two years. More recently both medium term Eurosystem/ECB staff and Phillips curve forecasts have been largely in line with wage growth outcomes in the euro area.

Looking ahead, our results can be seen as a motivation for investigating several issues further. These include attempts to better integrate analyses of different driving forces into a more holistic framework, which might help to quantify the relative role of different factors more accurately. Also, the importance of non-linearities in the wage Phillips curve deserves further in-depth analyses. It also seems worthwhile to investigate further the pass-through of wages to prices and its determinants in more detail.

Authors’ note: The views expressed are the authors and do not necessarily reflect those of the ECB.