What we really need to know, we don't know: the data we have is not the data we need.
The issues and confusions clustering around the US labour market hold the key not just to the sustainability of the US recovery, but to its medium term trajectory, and the development of monetary policy. If labour participation rates rebound, so too do estimates of potential output and the current output gap, with major implications both for monetary policy, and asset pricing. But if the post-crisis fall in participation rates are not reversed, how soon will effective full-employment be reached, and when should monetary policy tighten to head off potentially inflationary consequences? And if the c12mn gone missing from the labour market are never coming back, can policymakers be satisfied with the current recovery?
Every facet of this debate is muddied by the chaotic divergence between the two main monthly assessments of labour markets: the establishment survey, which is responsible for the non-farm payrolls data, and the survey of households, which is used to tally movements in the labour force, employment and unemployment. Over the last 18 months, the divergence between these two, and the sheer volatility of the household survey, defies easy analytical interpretation – frankly, it looks chaotic. To take the latest data: April's non-farm payrolls survey reported jobs expanding 288k mom, whilst the household survey found employment falling by 73k mom. And that's by no means the most dramatic of the deviations between the two found over the last 18 months, as the chart shows.
Such wild an unexplained volatility directly undermines the credibility of the household survey – ie, the credibility of data on unemployment and labour participation. But it also indirectly compromises the credibility of the establishment survey too. As a result, on the most important question about the US economy, we are effectively flying blind.
In particular, they no longer seem a reliable guide to how tight, or loose, labour markets really are.
One way of assessing the tightness of the market may be to look at the extent to which people are changing jobs, and the difficulty of replacing them when they do change jobs. There are two linked ideas here. The first is simply that people are keenly aware of the state of the job market, and are noticeably more reluctant to leave their current job if they think it will be difficult to find another. If so, the rate at which people quit their jobs might turn out to be a good indicator of labour market conditions.
This thought was elaborated (some time ago) by Fed governor Yellen, who thought the state of the job market – including the extent to which it needs to attract new entrants – may be examined by looking at the length of the 'vacancy chain'. The idea behind the 'vacancy chain' is that when a person quits a job, it creates a vacancy. If that vacancy is filled by someone else who moves jobs . . . . the vacancy chain grows.
The logical extension to this is that the longer the vacancy chain, the more people have to be persuaded to move jobs, and since that often involves bidding up the price, consequently the greater the upward overall pressure on wages.
Every month, the Bureau of Labor Stats releases a further survey, of job vacancies (the JOLTs survey). Because it is released rather later in the month than the non-farm payrolls and household survey, it tends to get filed under 'old news'. But it tracks the number and rate of vacancies, and the number and rate of quitting, with the quitting rate calculated as the number of quitters as a percentage of total employment.
What does it tell us? The quit rate has clearly risen slowly but quite steadily since the nadir of 2010, but has not yet recovered to its pre-crisis levels. In fact, March's 1.8% rate is still 1.7SDs below the 2001-2007 average: the labour market has not returned to 'normal' levels, with employees' confidence that their prospects may be improved by moving job still historically depressed.
We can get an idea of changes in the length of the vacancy chain by comparing changes in the vacancy rate with changes in the quit rate. The higher the vacancy rate relative to the quitting rate, we may assume the longer the vacancy chain, and the greater the likelihood that wages are raised in order to encourage people to switch jobs. In the next chart, I compare the 12m changes in vacancy minus quitting rate (which I'm calling the net openings rate), and the change in average wages.
Several things about this chart are worth noting:
- First, between 2001 and 2011, the changes in the net openings rate (which I am using as a proxy for the length of the vacancy chain) did indeed march amost in lockstep with changes in wages. The relationship held both during recession and recovery.
- Second, the recovery in the net openings rate has been sharp, and has now topped levels seen at the pre-crisis peak.
- Third, nevertheless, from mid-2011 onwards there has been a complete and unexpected divorce between this measure of labour market tightness, and wages. Why this happened is unclear: the Fed has argued that it reflects companies attempts belatedly to compensate for wages 'stickiness' during the recession. Since nominal wages did not actually fall during 2008-2009, so during the recovery they will now rise less than expected.
- Fourth, the openings minus quitters rate now seems to have peaked, and in fact seems to be rolling over.
Of these, the last two are obviously the most important. If the net openings level continues to stay at the currently elevated levels, it seems reasonable to expect that upward pressure on wages will eventually emerge. After all, the Fed's explanation of why wages remain relatively depressed isn't completely convincing: before the crisis compensation as % GDP peaked out above 55%, before dropping to around 53% in 2010. But the fall hasn't stopped there: by 4Q13, it had fallen to 52.5% - it is the lowest in recent history.
But set against that, the fact that the net openings/quitters rate now seems to have peaked may be a warning that the recovery of the labour market may be in question.
These measures, then, don't directly answer the questions that need answering. But they do suggest that labour markets have tightened in a way which is not yet expressed in rising wages – to an extent which is genuinely puzzling. Perhaps the market is anticipating a cyclical recovery in the participation rate? Second, however, the topping-out of the net openings rate also suggest that we should treat the recent rise in non-farm payrolls with some caution – labour markets may not have been as robust over the last six to nine months as that survey suggests.