NN IP: A closer look at the US labour market

Until a year or two ago, analysing the US labour market was pretty easy: The unemployment rate was still clearly above any reasonable NAIRU (non-accelerating inflation rate of unemployment) estimate but was declining rapidly on the back of strong employment growth. Because the latter took place against the background of a growth momentum which hovered around 2% the economy was characterized by low productivity growth of around 0.5% on average since 2011. However, a widely held view was that productivity growth could strengthen if and when an investment recovery pushed the amount of capital per worker higher. This process could well receive an extra boost once the economy nears full employment, because this may increase the incentive for investing in labour-saving technologies. Meanwhile, labour market slack and to some extent also low productivity growth explained why nominal wage growth remained muted. All this makes it immediately clear that the nexus between the employment growth, labour market slack, wage growth and productivity growth is pretty vital for the outlook for the economy as a whole as well as for central bank policy.

Now that the US economy is closing in on full employment, this nexus presents a few interesting puzzles for policymakers and markets. The main thread in these puzzles is that employment growth remains surprisingly strong, but that decline in the unemployment rate has decelerated substantially on the back of a pick-up in labour supply growth. What's more, wage growth has increased moderately but remains well below pre-crisis levels. Also productivity growth shows some signs of accelerating, but remains well below pre-crisis levels as well. This raises the question if and when the growth rate in these variables will shift towards a substantially different path. In particular, at some point employment and labour supply growth should slow down while wage and productivity growth should pick up. Still at which point and to what extent this will happen remains surrounded by a lot of uncertainty.

This week we will mostly focus on the nexus between employment growth, labour market slack and wage growth and leave the outlook for productivity growth in the background. If anything the productivity puzzle is by far the biggest and most important one facing us at and therefore warrants a separate discussion which we will publish in the near future. For now, we note that our base case is that productivity growth will continue to improve moderately in the foreseeable future on the back of more investment as well as an increase in risk taking and economic dynamism.

How much further can labour supply pick up?

One of the most remarkable features of the US labour market is that monthly payroll growth has remained well above any reasonable estimate of break-even payroll growth, i.e. the monthly payroll increase needed to keep the economy at full employment once it gets there. Still, the 6-month and 1-year trend in monthly payroll growth has been decelerating moderately from around 250K in mid-2014 to 160-200K in March 2017. Meanwhile, the unemployment rate has hovered in a 4.5% - 5% range over the past one-and-a-half years and has only displayed a very slight downtrend over this period. This is very different from the around 0.8-1.2 pp per year declines seen before that period. The reason for this is a remarkable stabilisation in the participation rate, following years of rapid decline. Because the participation rate is defined as the labour force as a fraction of adult (i.e. over 16) population, it is heavily influenced by population ageing and a large part of the downtrend since the crisis can be explained by that. It is therefore better to look at Demographically Adjusted Participation Rate (DAPR). The DAPR declined by around 1.5 pp between 2009 and mid-2015 and has recovered by around 0.5 pp since then, which leaves it still 1pp below the 2006-2007 average.

If this gap were to be closed entirely, the economy could easily stomach 200K monthly payroll gains for another year or two without pushing down the unemployment rate. In that case we would probably not see a significant rise in wage inflation from current levels and we could even see a decline in unit labour cost growth, which exerts downward pressure on core inflation to the extent that productivity growth picks up. In such a scenario, the Fed should probably not be tightening at a pace of 25bp per quarter but just let this supply side miracle happen instead. Unfortunately, real life is not that simple and it pays to delve deeper into the cause of the decline of the DAPR. When decomposing the prime age labour force into men and women, it is striking that the DAPR for both sexes is lower than the OECD average. For male workers the gap is around 4pp and the gap is the biggest in the older cohort (45-55). For women the story is a bit different. The participation rate of US prime age women used to be above the OECD average until around 2006 but has fallen below that number now. The reason is that other countries have seen a marked increase in female participation rates.

The next question is then why the US performs subpar in this respect. An interesting explanation is given by Nobel Prize winner Angus Deaton, whose research shows that US prime age workers face more serious health problems than their counterparts in other DM countries. This holds especially for middle-aged men with a low education level. As a result, the decade-long declining trend in mortality rates in the 45-55 age group flatten out from around 2000 onwards in the US while it continues to decline in other DM economies. What's more, mortality rates in this age group have been significantly higher than the DM average over the past 25 years anyway.

Furthermore, Deaton also shows that the middle-aged men who do survive but are not in the labour force suffer severe health problems. Around half of them take pain medication every day, which is once again substantially higher than in the rest of DM space. Apart from a clear difference in "health performance" the US also scores badly on another issue as it features a much larger share of the population which is imprisoned and/or convicted of a serious crime. The latter makes it is relatively difficult to find a job after the prison sentence has been served and if a job is found it is often a low paid one. As a result, former convicts easily become chronically discouraged workers.

The interesting findings go to show that economists should definitely look to other disciplines (sociology, political science, etc.) sometimes to explain certain phenomena. Of course, these are all secular trends while the fall in the DAPR is a cyclical phenomenon. This may be the case, but cyclical forces can become secular (and secular forces can influence the cyclical picture). The key issue to note here is that the social safety net in the US has a lot more big holes in it than in many other DM countries. Hence, once people have been unemployed for a long enough time they can fall in a poverty trap that is less easy to escape than in other countries. This poverty trap is conducive to inducing health problems and may push some people into criminal activities.

Having said that, it is certainly not the whole story. Over the past 8 years, the DAPR has received a boost from the fact that older workers work longer before they retire. In the early days after 2008 this could be explained by the sharp decrease in net wealth, which forced people to work longer due to a fall in pension savings. Since then, net wealth has recovered big time but the effect persists anyway. As for the drags on the DAPR since 2009 it seems that increased school enrolment played an important and increasing role until 2012 but that drag has diminished moderately over the past two years.

The same story can be told for workers claiming disability, except that drag has been smaller and has only started to stabilize since early 2015 after which it improved slightly. Discouraged workers acted as a rapidly increasing drag since the start of the crisis but this drag has waned a lot since late 2013. In fact, the majority of the rise in the DAPR can be ascribed to a return of discouraged workers. Still the drag in this category relative to the pre-crisis base line is still some 25bp and it makes sense to think that a strong labour market will cause the larger part of this drag to disappear. To what extent this will hold for the drags exerted by school enrolment and disability remains to be seen. They have both started to wane a little over the past few years and this trend may or may not continue.

Adding it all up, we see a good chance that the overall participation rate will at least stabilize for the rest of this year as the DAPR will rise somewhat further. In this case a slowdown in payroll growth to around 150K on the back of an increase in underlying productivity growth could deliver an unemployment rate which more or less moves sideways around the March level of 4.5%. Of course there are many scenarios which can be envisaged that deliver the same outcome. The point is simply that the interaction between changes in the participation rate and employment growth cast a lot of uncertainty on the course of the unemployment rate for the rest of this year. In our view, the risk is skewed towards the participation rate exerting a continued dampening influence on the decline in the unemployment rate as the DAPR will rise further. In addition to this, we expect employment growth to slow down somewhat on the back of a rise in productivity growth. In short, we are not that worried about labour market overheating in the near future.

The degree of slack is surrounded by a lot of uncertainty

Of course, to determine labour market overheating one needs to compare the current state of the labour market to the full employment benchmark. This is easier said than done. The most conventional measure is the difference between the unemployment rate and the NAIRU but this is a very crude measure indeed. We have often pointed to the existence of shadow slack in the form of involuntary part-timers as well as factors which may push the NAIRU lower such as a rise in productivity growth. The former still has some room to run before the % of involuntary part-timers is back at the 2006-2007 average. This is the reason why the U6 unemployment rate is still well above the average seen in that period. Meanwhile, a pick-up in productivity growth would not be immediately reflected to the same extent in wage growth, which is why the unemployment rate can fall further before triggering wage acceleration. Finally, downward wage rigidity has prevented real wages in some sectors from adjusting downwards to the full extent when the unemployment rate was sky high. As a result, real wage levels have remained "too high" over the past few years and nominal wage growth in these sectors will remain relatively irresponsive to a reduction in labour market slack at least until the "equilibrium" real wage rate is attained. In this respect, we note that underlying inflation has done little to speed up this process. Indeed, it is a telling sign that the share of workers receiving zero wage gains is still well above the 2006-07 average, even though the unemployment rate is now at roughly the same level seen back then.

A final point which is not often mentioned is that shifting demographics imply that comparing unemployment rates over time to assess the amount of slack can be tricky business. In other words, demographics can change the NAIRU over time. For instance, a decline in the share of the labour force for a subgroup which tends to have a relatively high average unemployment rate over time will cause a decrease in the overall unemployment rate without changing overall pressure on wages. In this respect, one important trend over the past decade has been a decline in the share of young workers in the labour force which is a direct result of population ageing. Young workers tend to be inexperienced and thus have a relatively low productivity level while they also switch jobs more often, all of which is conducive to this age group displaying a higher average unemployment rate than older age groups.

However, this is not the whole story as the afore-mentioned increase in school enrolment also plays a role. This has caused young people to enter the labour force at a later age than previously, but when they do they have a better education and thus more chances of finding a job. This "delayed labour force entry" thus exerts a direct downward effect on the unemployment rate by lowering the fraction of young workers in the labour force. However, there is also an indirect effect because the increase in human capital has caused the unemployment rate for young people to decline, all else equal. Research from the San Francisco Fed suggests that this combined effect may have lowered the unemployment rate by around 2pp since the 1970's. This could be an important explanation why estimates of the NAIRU are much lower today than they were back then.

In the end the proof of our labour market story will ultimately reside in the behaviour of wage growth. Also here the story is not straightforward as there are different measures which all have their strengths and weaknesses. Looking through all these measures, wage growth seems to be on a moderate upward trend. Some pundits argue that this may change rapidly once the unemployment rate falls well below the NAIRU (which once again itself may well be a moving target). We cannot exclude such a non-linearity in the Phillips curve of course. Indeed, it makes sense to think that wages will accelerate more strongly once the economy hits hard labour supply side constraints. Having said that, slack is not the only factor driving wage growth. Inflation expectations also play an important role. Historically, rapidly accelerating wage growth has often gone hand in hand with a marked increase in inflation expectations which triggered a wage-price spiral. Such a development can certainly not be excluded in the more distant future but is unlikely to happen in the near future, given the fact that core inflation has been below target for years and measures of inflation expectations remains close to historical lows.

Emerging markets: good Chinese data

Risks of a deterioration in US-Chinese trade relations declined further after the US Treasury confirmed that China was not considered a "currency manipulator". This news came a few days after the Trump-Xi summit where both leaders appeared committed to at least keeping the status quo in their relationship.

Meanwhile, Chinese trade data for March came in strongly, far above expectations. Export growth reached 16%, partly thanks to a low base effect, but also because global demand has been strengthening since November of last year. Import growth reached 20%, mainly due to strong commodity imports. This reflects the still solid construction activity growth linked to the real estate boom.

Important for future domestic demand growth in China is the current pace of credit growth. In March it fell by almost a full percentage point to 15.2%. Broad credit growth has been on a declining trend since January of last year and is now at an 8-year low. Nevertheless, the absolute level of credit growth remains above nominal GDP growth.

Q1 GDP growth came in a bit stronger than expected, with headline growth at 6.9%. The quarter-on-quarter growth rate was weaker than expected, at 1.3%. Meanwhile, industrial production growth for March accelerated to 8%, from 6% in January/February. Power consumption growth, one of the more credible data points, also accelerated to 8%. Fixed-asset investment growth strengthened to 9%.

All in all, Chinese growth data have continued to be strong. We believe that growth is close to its peak or that it already has peaked by now. The real estate sector has been a key driver of growth and here we see clear signs that the tightening measures are having effect. New home sales volumes, for instance, fell from 24% in January/February to 11% in March. The slowdown in broad credit growth also suggests that economic growth should start slowing anytime soon.