Disclaimer: this post represents my own views and not those of the Central Bank of Ireland
Two recent CSO releases shed light on the evolution of earnings in the first three quarters of 2020. Alongside employment and hours, understanding the impact of COVID-19 on earnings tells us how household incomes are affected by the shock.
In previous work, using data from the Financial Crisis, we found that earnings in Ireland were sensitive to economic conditions, notably changes in the unemployment rate. The workers most exposed to lower pay when labour demand falls are those with weaker bargaining power. For example, in our paper we focused on the lower wages of new hires during the last crisis.
When looking at changes in average earnings, such as from the CSO’s Earnings and Labour Costs release, it is important to take account of changes in the composition of employment. For example, if changes in employment are concentrated amongst lower paid workers, average earnings could rise when there is a negative aggregate demand shock. In a SSISI paper in 2012, Kieran Walsh showed that these compositional effects can be large. The CSO also noted the potential for compositional effects in the context of COVID-19 average earnings changes.
Tracking the earnings of the same workers in the same jobs can remove some of these composition effects, giving a clearer picture of underlying wage developments. The CSO does something close to this in its Labour Market Insight Bulletin 4/2020, showing changes in average gross weekly earnings conditional on workers being in employment in Q1 and Q3 2020. Earnings includes wage subsidies, where applicable, but exclude PUP payments.
The chart below, from the data in the CSO Bulletin, shows that for all sectors earnings fell by almost 4 per cent for workers in employment in Q1 and Q3. In some sectors, like Administrative & Support Services and Financial, insurance & real estate the changes are double-digit. In others, like Construction and Accommodation & food, earnings are up.
For comparison, during the financial crisis, and controlling for composition effects, average weekly nominal pay also fell by around 4 per cent, most of it between 2008 and 2009 (Lydon & Lozej, Table 2). At that time, the declines were largest in sectors connected to property market, like construction and real estate. The emphasis on nominal pay is important. Between 2008 and 2010, prices (CPI) also fell sharply, by over 5 per cent. This helped offset the fall in nominal earnings, cushioning the impact on households’ purchasing power. Price levels have fallen in 2020, by around 1.5 per cent, which suggests a fall in real earnings in the first three quarters of the year of around 2.5 per cent.
… changes in earnings positively correlated with labour demand, but important to control for hours
The changes in earnings between Q1 and Q3 are generally positively correlated with changes in labour demand, such as changes in employment or job postings. There are some notable exceptions like Accommodation & food – where employment fell by over a fifth, but average earnings rose marginally, by 0.4 percent; or Industry, where employment grew by 2.4 per cent, but average earnings fell by 6.4%.
Despite conditioning on workers in employment in Q1 and Q3, there are likely still many factors affecting earnings dynamics that are not picked up in the conditional averages. One example is hours-worked. As weekly earnings are the product of hours worked and hourly pay, higher or lower earnings could be due to higher or lower hours. This could matter in sectors with seasonal hours, like Accommodation & Food services.
To get at the the change in hours worked, and for a sample broadly aligned to the administrative earnings data, the CSO provided me with average actual hours worked by sector for employed persons interviewed in Q1 and Q3, from the LFS. I use this data to back out change in average hourly pay as the change in weekly earnings minus change in weekly hours worked. Readers should note that earnings data is from administrative sources (including wage subsidies), whereas the hours data is from a survey. Furthermore, the two matched LFS samples are six months apart and may not be exactly representative.
The chart below shows the data. The line in the chart is the change in average gross weekly earnings, corresponding to the bars in Chart 1. Whilst there are offsetting increases in hours in several cases, they are usually small. Furthermore, the direction of the change in hourly pay and earnings is roughly the same for most sectors, with the notable exception of Accommodation and food services. In fact, the increase in hours worked (10.5%, an increase from 32.1 to 35.5 hours per week) offsets a large fall in hourly pay (minus 10.1%). This fall in hourly pay is more closely alinged with the fall in demand (employment and job postings) that we have seen during COVID-19. Looking at the historic LFS data, it is clear that this hours increase in Q3 is not unusual. In fact, it is entirely predictable: the historic Q1 to Q3 change in hours is almost exactly the same as the 2020 figure, at 10.3 per cent.
The third chart below shows the correlation between the estimated change in hourly pay (conditional on working in Q1 and Q3) and the change in job postings by sector from Indeed. Job postings are generally a good indicator of labour demand, and, whilst postings are down across the board, we find that sectors where postings have declined the most have generally see larger falls in hourly pay.
It should be said that three quarters of data is a relatively short time period. Added to this is the fact that the COVID-19 shock has generated a very high degree of uncertainty. For firms in some sectors – such as exporters, industry or multinationals – the demand shock may may turn out to be less bad than initially feared. This might help explain negative earnings growth for workers in Industry, but positive employment growth. It is quite possible that in Q4 or Q1 2021 we may see a strong earnings growth for some sectors as employers unwind pay freezes that were put in place early-on the crisis.
… looking ahead
By combining administrative and survey data in novel ways, the CSO provides timely and granular insights on the COVID-19 labour market. This is crucial information for understanding the impact of the shock, and how policy might help mitigate it.
The decline in earnings in 2020 for employees working in both Q1 and Q3 is similar to falls seen during the last recession, albeit with a different sectoral pattern. There are other differences this time around. The most significant difference is the large and decisive policy response to COVID-19 – both fiscal and monetary. In November, over a quarter of workers were supported by Pandemic Unemployment Payments or Wage Subsidies. Furthermore, the government has committed to these supports remaining in place while restrictions remain in in place. Another important difference is healthier state of household balance sheets going into 2020, a factor which dragged on domestic demand during the last recession.
If the spread of the virus can be brought under control in 2021, this points to a potentially shorter duration shock than before. However, the longer restrictions continue, the greater the potential for behaviour to change – like less business travel or less bricks-and-mortar retail, for example – and the harder it becomes for some businesses to reopen. This would lead to permanent job losses, even after restrictions are lifted. Furthermore, if employment and (real) earnings shocks persist, there is greater potential for precautionary savings, with negative feedback loops for domestic demand. The fact most people who have experienced reduced employment or been laid off due to COVID-19 said they expected to return to the same job suggests a widespread perception of this as a short-term or temporary shock – albeit this was in Q3, before the most recent Level 5 restrictions.
Related to this, in services – the sector most affected by the shock – turnover picked up sharply during the summer easing of restrictions. Although some sub-sectors, like travel and accommodation remained far below pre-COVID levels. Job postings in services track tend to track turnover very closely, rising in the summer, before declining again in the move to Level 5. This suggests that permanent relaxation of restrictions, leading to increased demand, could undo some, but not all, of the labour market damage we have seen in 2020.