Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
The Fed is always and everywhere obsessed with the US labour market. If Phillips curve assumptions around the relationship between unemployment and price inflation are not enough to focus minds on the data, maximum sustainable employment is the second explicit part of the Fed’s dual mandate.
And so Non-Farm Payrolls day tends to be one of the biggest, perhaps the biggest day in bond traders’ calendars. For non-bond aficionados, here’s most of what you need to know about the first Friday of each calendar month:
— Big jobs number => Tighter labour markets => Fed more likely to put up rates or not cut rates => bond goes down (yield goes up) => Bondholders sad 😢😢😢
— Small jobs number => Looser labour markets => Fed more likely to cut rates or not hike => bond prices rise (yield goes down) => Bondholders happy 😊😊😊
Of course, it’s not just bondholders who care. Stock prices react too, though we’re reluctant to go down the ‘when does bad data being good for stocks turn into bad data being bad for stocks’ rabbit hole (today, at least).
As well as non-farm payrolls, a slew of other official employment data gets released at the same time. These come from two sources: the Current Employment Statistics release (aka the payroll survey) and the Current Population Survey (aka the household survey).
The payroll survey is used to estimate hourly earnings, hours, income and jobs by sampling a bunch of firms and government agencies. The household survey estimates the unemployment rate, among other things by sampling a bunch of households. Both come from the US Bureau of Labor Statistics. Both sample in the week leading to the 12th of the month. Together, they tend to create a mosaic of the labour market that analysts and policymakers can read.
But for the last couple of years, they’ve each given a very different message about job growth. Sure, they measure slightly different things, but the ‘payroll-adjusted’ household measure tries to correct for this and is showing an even bigger divergence than the straight household measure of employment.
According to the household survey, total employment bounced back strongly from Covid through autumn last year, and has since been a bit rubbish. But according to the payroll survey, job growth has continued to go, uh, gangbusters:
This is not new territory for Alphaville. We started writing on this around the time it started to be a thing last autumn, and we did a further dive into the differences in June. But to put some numbers on the discrepancy, the household survey reports around 700,000 fewer workers from the peak last November, while the payroll survey reports around 1.6 million more workers. It’s more than a bit weird that a working population around half the size of Ohio’s appears to have fallen down the back of the sofa. This is no rounding error.
Why should you care? In short, because the Fed’s next move is going to be based — in part — on which version officials choose to believe.
In a recent note for clients, Barclays economists — Jonathan Millar, Marc Giannoni, Pooja Sriram, and Colin Johanson — come down hard on the side of the payroll survey.
Analysts, clients, and the media are quick to complain that it’s the payroll survey that’s misleading, the Barclays gang say (FTAV covered Standard Chartered’s gripes with the payroll survey in depth a while back). And *checks notes* Barclays’ gripes here.
It’s a reasonable position to hold: the survey’s methodology of estimating births and deaths of firms is opaque, and it is subject to periodic revisions when compared to slower, better, measures of employment. Moreover, these revisions have invariably been downward. The implication is that the go-go jobs data that is keeping the Fed from cutting should be downplayed, and more and faster rate cuts should be the order of the day.
But Millar et al reckon that payroll survey haters have it all wrong. The household survey, they say, is the unreliable narrator of labour markets. As they put it (emphasis theirs):
While it is reasonable to speculate that the CES [the payroll survey] overstates job gains somewhat, our analysis shows beyond the shadow of a doubt that the household survey has severely understated job gains since 2022. Though it is conceivable, at least conceptually, that this mismeasurement has not contaminated the measured unemployment rate, it is also quite plausible that it has risen less than advertised.
Over five and a half thousand words of detailed nerdery, they make their case. The tl;dr is that the payroll survey, for all its faults, is kept honest by being trued up to an actual census — the Quarterly Census of Employment and Wages (QCEW) — in January each year. This results in those revisions about which critics get so het-up.
The household survey, by contrast, never gets trued-up. After cross-referencing the household survey (and its adjusted version) to the QCEW (and its adjusted version), as well as to data released by ADP (a private payroll provider), Barclays reckons the household survey has massively undercounted jobs. We’re talking millions upon millions of undercounted workers. Furthermore, they think the household survey’s smaller size, deteriorating response rate, and re-scaling methodology makes it less statistically trustworthy, and that incorrect population estimates mechanically constrains gains in household employment.
They conclude:
the fact that the CPS overstates cooling in the pace of job gains, and in the overall labor market, is hard to escape.
They don’t go quite as far as to definitively say that the rise in the US unemployment rate is a nonsense, but they get pretty close. It’s worth saying that other analysts looking at this issue reckon that because population-count problems affect both numerator and denominator of the household survey, the unemployment rate is going to be a fairish reading.
Who cares? A big deterioration in the labour market according to the household survey is the sort of thing that gets the Fed’s attention. Former NY Fed President Bill Dudley last week argued for immediate rate cuts largely based on this deterioration. If this deterioration never happened, the case for a cut (or the three that are priced in by year-end) falls a lot.
We know exactly what you’re thinking: couldn’t this all be summed up via a Hamiltonesque cabinet rap battle [ed: uh?]? Great minds — with a little help from a combination of ChatGPT and AI song generator Suno, here’s the upsum [ed: 🥴].
