Furloughed employments are having complex effects on the labour market
As central bankers in the US and the UK worry about inflation, eyes are also turning to the somewhat peculiar developments in labour markets, including increasing wages. In a previous blog, I coined the phrase “pay puzzle”, to join a set of conundrums in the UK economy, which already included the well-established “productivity puzzle” and the post-financial-crisis “employment puzzle”. The intriguing question this time round? With so many people unemployed and on furlough, how come earnings are going up?
My previous blog tackles some of the key explanations, such as the changing educational, occupational and sectoral mix of employment, which has had a significant impact on the reported earnings figures. There is also increasing evidence of labour market mismatches: yes, there are a lot of people available for work, but those people may not have the skills required for the vacancies that are being advertised.
I also mention a constant interpretation challenge: when commentators highlight changes in any variable — be it GDP, employment, vacancies, or earnings — after a dramatic change in underlying conditions, even small increases can appear huge in percentage terms. This is certainly the case for job adverts and vacancies, for example. Yes, based on ONS data, the number of job adverts increased by more than 200% from May 2020 to May 2021. However, the absolute number of vacancies in the period from February to April 2021 was still 20% below that of February to April 2019.
These puzzles add to what was already a well-known question among economists ahead of the pandemic: has the Phillips curve — a rule of thumb used by many to describe the (assumed) relationship between inflation and unemployment — become extinct? I won’t go into more detail here, other than to note that, as far as I can tell, such a curve hasn’t really existed for a while in the simplistic format that most people talk about it. Indeed, in textbooks, one can find at least 3 different formulations of the curve.
One of the key reasons that the Phillips curve often doesn’t fit empirical data is that inflation is obviously driven by many other factors beyond labour markets (even though, perhaps legitimately, monetary policy makers are most concerned about potential wage-price spirals, so monitor this closely). But another one is the same reason why the aggregregate unemployment and furlough figures seem at odds with reports of labour shortages: workers are not perfect substitutes for each other. Even within a sector, a bartender cannot overnight become an accomplised chef, or a bus driver turn into a lorry driver.
It is therefore useful to look at workers in groups — by qualification, sector, occupation, region, or other characteristics — within which on can assume more substitutability. Unfortunately, I don’t have access to micro-data that would allow to do this in a sophisticated way (or that would be very up to date). We can, however, look at unemployment, vacancies, furloughs and wage growth on a sectoral basis.
Turns out that unemployment on its own — whether nationally or within each sector — has indeed been a pretty bad predictor of earnings growth since the start of the pandemic (e.g., R-squared of 0.04). Mapping vacancies against pay increases for each sector (on a monthly basis) provides a much more robust relationship (e.g., R-squared of 33% and P-value of less than 0.0001). The best explanatory power, however, is found when we use the combined number of furloughed and unemployed workers as a measure of labour market tightness or slack (see chart).
This kind of high level analysis (as I’ve mentioned before) is only indicative. Nevertheless, I’m fairly confident that we can learn something from this, too. [Note: I have also played with the data extensively in the background, but don’t have space to share all the findings here — but do get in touch if you’d like to see it!] So, what can we draw from the graph that I have plotted?
First, both the data for 2020 and 2021, and monthly data I have looked at from 2000 onwards, suggests that it is easier to interpret if we consider each of the sectors more or less its own labour market. In other words, there seem to be significant barriers to employers hiring unemployed workers from sectors other than their own, at least in the short term. For example, even in “steady state” economic conditions, there are some sectors with consistently high levels of unemployment — such as hospitality and admin and support services— and others with low levels of unemployment — such as the public sector (including education and health) and financial services.
Second, since the COVID-19 pandemic started, it has been misleading to look at unemployment statistics on their own, due to the impact of furloughed workers on labour market dynamics: the ONS estimates that in early May, 2.7 million people were still furloughed, corresponding to 10% of the business workforce in the UK. Once we add these statistics into the picture (x-axis of the graph), fairly strong patterns emerge.
For example, just within the hospitality sector (orange marks), wages dropped a lot in April to June 2020, at the height of the first lockdown, while they increased as the economy was opened up in late summer and the autumn of 2020. A similar pattern is true essentially in most sectors (but the link is fairly weak in some, such as the public sector, financial services, and the information and communication sectors).
Overall, at least measured in this way, the Phillips curve seems to be alive and well. The lower the labour market slack, the higher the wage increases. Of course, 2020 and 2021 have created pretty extreme “natural experiments” in the economy. But it is precisely for this reason that data from this period is incredibly interesting and can shed light on economic phenomena. I consider this is a small silver lining, in the midst of the otherwise significantly negative consequences for people’s lives, livelihoods and life satisfaction.
