When Business Insider announced its call yesterday on nonfarm payrolls, which projected an increase of 285,000 jobs, commenters were quick to rebut our estimates.
The prediction, which used a blend of data including commercial and industrial loans and five regional Federal Reserve employment sub-indexes, among others, is substantially higher than the consensus estimate at Bloomberg, which calls for 210,000 new nonfarm jobs in February.
With more than 90 economists making predictions for this report and a standard deviation of 25,000, the probability that nonfarm payrolls will increase by 285,000 is just 0.63 per cent.
But Wall Street estimates are often far from perfect, especially with nonfarm payroll data.
Over the last decade, actual nonfarm payroll growth and contraction fell outside of 2.58 standard deviations 41.7 per cent of the time. In statistics, 2.58 standard deviations represents a 99 per cent confidence interval.
Why does that matter? Statistically, if you accept the notion that the Street’s consensus is “correct,” a result falling more than 2.58 standard deviations should happen only once every 100 times.
That’s not the case. In fact, just look at January’s results. Before the announcement, Wall Street had modelled nonfarm payroll gains of 140,000. The actual growth: 243,000, 3.55 standard deviations from the mean. Since job creation accelerated in September 2011, the resultant nonfarm payroll data has been 4.58, 0.89, 1.52, and 1.92 standard deviations from consensus.
Taken together, the Street has continued to underestimate job growth during the recent recovery.
It’s important to note that the estimates do not necessarily fall perfectly under a normal curve. However, the distributions do generally take the bell shape and cluster together centrally around the mean.
Below we present actual NFP results compared to estimates. The meaningful point to look at is the t-test standard deviation, or how far the actual NFP number was from consensus estimates (a line highlighted in blue represents a result outside of the 99 per cent confidence interval).
Photo: Eric Platt/Business Insider, Data: Bloomberg