How Goldman Calculates Its 100,000 NFP Forecast For Tomorrow

Tyler Durden's picture

There was a time, long ago, when economic data mattered, and when Goldman's NFP forecast was considered one of the best on the street due to the proximity of The Pound and Pence to both 85 Broad and 33 Liberty. Then Goldman went to 200 West, central planning took over, and Bizarro world was the result, where a huge NFP beat would mean a collapse in the stock market once the prospect of QEternity actually ending returns. In other words, Goldman lost its touch. Yet their insight can still be valuable. Which is why below we present the argumentation that Goldman's Sven Jari Stehn uses to expect a BLS payroll number of 100,000 tomorrow, translating into an 8.1% unemployment rate.

Payroll Preview: Another Sluggish Report

  • The labor-market news received since the disappointing August employment report has been mixed. While the ADP measure of employment, the manufacturing ISM survey and households' perception of job availability surprised on the upside, initial jobless claims and job advertising have been volatile, and the employment sub-index in the non-manufacturing ISM weakened notably.
  • We therefore expect another sluggish jobs report in September, with a 100,000 gain in nonfarm payrolls and a flat 8.1% unemployment rate. The possibility that the uncertainty associated with the fiscal cliff has started to weigh on firms' hiring decisions poses a downside risk to our forecast.

The August employment report was a disappointment as payrolls grew by only 96,000 and the unemployment rate ticked down to 8.1% due to a drop in labor force participation. The labor-market news since then has been mixed:

  1. Jobless claims. Although initial jobless claims were higher in the September survey week (at 385,000) than in the August survey week (369,000), claims are now back to 367,000.
  2. ADP. The ADP measure of private employment increased by 162,000 in September, ahead of expectations for a 140,000 gain but down from (a revised) 189,000 gain in August.
  3. ISM surveys. While the employment component of the manufacturing ISM survey improved in September (from 51.6 to 54.7), the corresponding sub-index of the non-manufacturing ISM weakened (from 52.0 to 49.5). We have generally found the latter to be more closely correlated with overall payroll growth, not surprisingly given that it covers a much larger part of the economy.
  4. Perceptions of job availability. The differential between respondents who view jobs as "plentiful" versus "hard to get" in the Conference Board survey improved 1.8 points in September.
  5. Online help-wanted advertising. The Conference Board measure of help-wanted advertising declined in both July and August, but rose in September. The index was up in seasonally adjusted terms in August (September data are not yet available). We have generally found that online help-wanted ads have the most predictive power for payrolls at a 1-2 month lag, so we would view this as a split verdict.

Our payroll models interpret these conflicting signals as another sluggish employment gain in September. We therefore expect a 100,000 gain in nonfarm payrolls and a flat 8.1% unemployment rate. A sluggish employment report would also be consistent with the state of the economy more generally. Economic momentum since early August has softened, as our GDP tracking estimate for Q3 GDP and the August CAI have fallen from 2.3% to 2.0% and from 1.1% to 0.5%.

A downside risk to our forecast for the September jobs report is the possibility that the uncertainty associated with the fiscal cliff has started to weigh on firms' hiring decisions. In a simple test conducted in late August (i.e. before publication of the August employment report) we found no evidence that the fiscal cliff--or, to be more precise the looming automatic spending cuts (or so-called "sequester")--had already started to weigh on employment. We reached this conclusion by exploring whether industries with meaningful exposure to government spending had experienced weaker employment growth between May and July than industries with little exposure to government spending. We found no such relationship (with an r-squared of 0.003). Exhibit 1 shows an update of this analysis, showing the relationship between industry exposure to government spending against employment growth in August. Although the relationship is far from watertight--with an r-squared of only 0.026--a hint of a negative relationship has emerged. Such an effect would also be consistent with the very weak durable goods orders numbers in July and August. We therefore view any uncertainty effects from the fiscal cliff as a downside risk to our payroll forecast.