Following today's sizable miss and significant revision to the ADP data it is perhaps worth taking a step back and looking at some independent research on the adjustments and seasonality issues in forecasting jobs around this time of year and furthermore, why one of the pillars of this extended rally and US decoupling story (a substantially improving jobs market) could be made of salt. Bloomberg's consensus for Friday's NFP at +145k (from +200k prior) and a 30k standard deviation, there is plenty of uncertainty among the economic elite (with 125k to 150k the sweet spot for their guesses) and our favorite outlier Joe LaVorgna near the top at +210k. So while the trend is supposedly improving (though expectations are slightly off December's exuberance), Stone & McCarthy (SMRA) point out a disturbing trend of sizable forecasting errors for the January payroll print with 7 straight years of estimates overshooting by an average of 64k - strangely consistent post the BLS switch to a probability-based sample. But its not just forecasting error, TrimTabs takes a deep dive into the actual daily income tax deposits from all salaried employees (which are historically more accurate than BLS initial estimates) sees the US economy added only 45,000 jobs in January, nearly unchanged from the 38,000 in December. Noting similar forecasting errors as SMRA, TrimTabs points out that the decline in seasonal adjustment factors and the reality of the underlying tax data suggest "It appears that the economy has hit stall speed due to lackluster demand and a deleveraging consumer who would rather save than spend." as wage and salary growth (net of inflation) weakened further to -2.1% YoY in January from -0.5% YoY in December. "The weak job market has us concerned" seems like a truer reality than the establishment trying to keep the dream alive.
"The weak job growth in January has us concerned," says Madeline Schnapp, Director of Macroeconomic Research at TrimTabs [9]. "It appears that the economy has hit stall speed due to lackluster demand and a deleveraging consumer who would rather save than spend."
"We see nothing on the horizon to knock the economy out of its slow growth mode," notes Schnapp. "The economy faces substantial headwinds from negative real wage and salary growth, high unemployment, waning government support, expiring tax incentives, contracting state and local governments, elevated fuel prices, and a sluggish housing market."
SMRA [10]- Why Can't Forecasters Get It Right?
There are clear indications that labor market conditions are improving, including the general behavior of initial unemployment claims.
BUT we have notice a disturbing pattern of January payroll forecasting errors that we might want to reflect upon when thinking about next week's payroll release.
[11]
January payrolls have been weaker than the median estimate in each of the last 7 years. Ever since the BLS switched to a probability-based sample, forecasters have consistently overshoot on January payrolls. Is this because forecasters have not fully appreciated the impact of the large net subtraction stemming from the January birth/death adjustment? Or is this because the BLS has published the annual benchmark revisions since 2004 with the release of the January data, and these revisions add a degree of uncertainty to the change between December and January?
Can't tell. But 7 straight over-shooting may be more than a coincidence. There may be something inherent in the January data that forecasters simply don't understand.
During these 7 years the average overshoot was 64,000, the smallest was 35,000 in 2011, and the largest was 92,000 in 2008. These misses are not especially large, but the uniformity of the overshoot is odd.