Labour market cooling justifies Fed’s dovish lean

One of the drivers of the dovish pivot from the Federal Reserve (Fed) in December was the acknowledgement that the risks to the policy outlook had become more two-sided. In other words, while higher rates were still needed to tame inflation, the Fed saw a risk that staying restrictive for too long would risk damaging a labour market that has so far shown remarkable resilience.

The labour market has certainly shown steady signs of cooling in the past few months – the Fed chair Jerome Powell acknowledged this at the June policy meeting, and followed it up with a comment that risks were becoming “much more balanced” at a central banking forum in Portugal last week. The question for the Fed (and markets), is whether this is a normalisation to pre-Covid trends or the start of a more significant weakening.

There are a number of indicators we can look at to understand the dynamics at play in the labour market at the moment.

One area that the Fed has been keenly focused on is job openings. After peaking at around 12m in 2022, the number of openings has steadily declined with the last monthly reading at 8.1m. Comparing this to the number of unemployed people in the labour force, the ratio between the two peaked at just above 2 and has in the last month dropped to the pre-Covid level of 1.25 (Powell has spoken about the labour market now being as tight as 2019). 

Source: Bloomberg, 31 May 2024

The job openings data has been a key focus for the Fed because it is a guide to how much slack has been coming out of the labour market. Of particular note is the relationship between job openings and the unemployment rate, the so called “Beveridge curve”.

Christopher Waller, one of the more influential members of the board, co-wrote a Fed paper in 2022 which argued that the labour market was able to land softly essentially because there was this slack in job openings. He argued (alongside an economist called Andrew Figura) that the Fed would be able to tighten policy and see a reduction in the amount of job openings in the economy without seeing a rise in the unemployment rate given where the economy was on the curve (i.e. the steepest part on what is a convex curve). This came in stark contradiction to what many economists were saying at the time, including Olivier Blanchard and Larry Summers, who argued that the unemployment rate was going to have to rise significantly and that the chances of a soft landing were practically zero.

Job openings since 2022 have declined by 4m with only a marginal increase in the unemployment rate (at 4% now we are barely off the 3.4% lows), as can be seen in the chart below which plots the monthly job openings and unemployment rates for the last 24 years. A decline in job openings of this magnitude without a recession is unprecedented, but the Fed’s argument at the time was that the labour supply/demand imbalance post-Covid was also unprecedented. Given the convexity of the Beveridge curve the Fed’s cautiousness now looks warranted (Mary Daly, a voting dove, said late last month that the labour market is “nearing an inflection point”). 

Source: Bloomberg, 31 May 2024

Turning to the monthly jobs data, while the vacancy rate has softened from its peak the non-farm payrolls (NFP) data has remained stronger than expected. NFPs averaged 251k in 2023 and are averaging 247k year-to-date, so the pace of jobs growth remains robust and the slowdown gradual. We have seen some divergence in the various surveys however. NFP is a survey asking employers about their employees, whereas the household survey (which goes into the unemployment rate calculation) asks a smaller sample of households about the employment status of workers in the household. Given the size of the sample the household survey tends to be much noisier than the NFP data; smoothing monthly returns over 12 months and comparing the long-term relationship shows the two are historically well correlated, and also that the recent divergence is wide compared to history. There are potentially a few reasons for this, most importantly the household survey might not be properly accounting for the higher levels of immigration the US has seen in recent years. In reality, jobs growth is probably not as weak as seen in the household survey, nor might it be as strong as the NFP data suggests. That said, any historical comparison with the NFP needs to take into account the size of the labour force (i.e. 200k now is lower relative to the total labour force than it was in say 2000, when the labour force was 15% smaller).  

Source: Bloomberg, 31 May 2024

Another topic of discussion recently has been the “Sahm rule”, which tracks the three-month moving average of the unemployment rate and compares it with the lowest unemployment rate of the last 12 months. Named after economist Claudia Sahm, a former Fed economist, the indicator has historically been an accurate indicator of a recession when the level increases above 0.5 (so much so that it is recorded as an indicator in the FRED database). Looking at data over the past 50 or so years, we can see how the 0.5 threshold for the rule has been triggered at the start of each of the past six recessions. Overlaying the yield curve onto this (using the US Treasury 2s-10s spread) highlights how it is a better indicator of the timing of a recession than the bond market (which has been an accurate indicator for recessions but an awful one for when that recession would occur). We do however see a number of instances where the Sahm rule flirts with 0.5 but doesn’t break through, including 1996 and 2003. The 1996 case in particular might be a useful comparison to today’s macro environment given a still strong growth backdrop, a slight weakening in the labour market and what was a shallow cutting cycle. 

Source: Bloomberg, 31 May 2024

Looking at the unemployment data over the past 18 months we can see how Sahm’s indicator has progressed alongside the unemployment rate. The current reading is 0.37 but assuming the unemployment rate stays at 4% in data due later today, this will move to 0.40, not far from the threshold. We expect this to garner more attention after today’s numbers, but it is worth saying that if the unemployment rate stays at that level for the next few months (inline with the Fed’s forecast), the Sahm number would fall back to 0.37 given the 30bp jump in the unemployment rate we saw in August 2023.

Source: Bloomberg, 31 May 2024

Finally, there are other indicators that confirm the slowdown in the labour market. One is initial jobless claims, where the three-month moving average is around 222k vs. 211k at the end of last year. Another is the ISM survey data, where the employment component of Thursday’s ISM services miss was 46.1 (under 50 indicates contraction) and it is averaging 48 this year vs. 51 last year).

All told, we think this points to a clear slowdown in the US labour market, validating the Fed’s messaging in recent months (even while inflation in Q1 was beating expectations). For now, the data to us indicates nothing more than a cooling in the economy, with growth rates that remain below the levels we saw last year but well above the level consistent with a recession.

However, the labour market has historically been prone to weakening quickly once unemployment starts rising. This is firmly in the mind of the Fed’s decision makers, and in our view they will be more sensitive to weakening labour data than to inflation prints beating expectations, at least with the Fed’s preferred inflation measure of core Personal Consumption Expenditure (PCE) below 3%. If these steady yet unalarming trends continue, it is likely the Fed will indeed have the data they deem appropriate to start cutting rates later this year, albeit at a slower pace than in previous cycles.