New Model to Predict Unemployment Sees Rate at 8.1%

By Ben Casselman

The big monthly jobs report doesn’t come out until Friday, but no need to wait until then to know September’s unemployment rate: 8.1%.

At least, that’s the best guess of economists Regis Barnichon and Christopher Nekarda, who in a paper last month laid out a new method for forecasting the unemployment rate. Mr. Barnichon, of the Spanish research institute CREI, and Mr. Nakarda, of the Federal Reserve, plan to start posting unemployment projections based on the new technique in the coming months. Until then, they provided Real Time Economics with a look at their forecast for Friday’s jobs report.


The projection itself, made last Friday, isn’t particularly remarkable. An unemployment rate of 8.1% would represent no change from August, and is also the consensus forecast of economists surveyed by Dow Jones Newswires. Messrs. Barnichon and Nekarda also forecast a very gradual improvement in the months ahead: 8% in October and November, 7.9% in December, and an incremental drop to 7.1% over the course of the next year (although the authors caution in their paper that their projections become less accurate as they go further out in time).
More interesting than the numbers themselves is the author’s methodology. Traditional attempts to forecast unemployment are based on economic fundamentals such as gross domestic product — figure out where the economy is going, the theory goes, and you can extrapolate the unemployment rate. The problem is that while such theories work well over the long term, they’re lousy at near-term predictions, especially when the economy is at a turning point — which is, of course, exactly when such predictions are most valuable.

Messrs. Barnichon and Nakarda take a different approach. They look at the jobs data itself, ignoring most outside indicators and instead using past months’ data to project the future unemployment rate.
To do this, the authors exploit the fact that the unemployment rate isn’t really a single number. It’s a function of two separate processes: people entering unemployment (primarily by losing their jobs) and people exiting (by finding jobs or by leaving the labor force).
Separating out the two pieces is key because they often move independently, with inflows to unemployment — that is, layoffs — leading the way, and outflow — hiring — lagging behind. Companies tend to start laying off workers at the first sign of a downturn, and stop laying them off when things stabilize, even if they aren’t ready to start hiring yet. By modeling the inflow and outflow rates separately, the authors come up with a far better forecast than they could from looking at the unemployment rate as a single entity.
It’s also much better than most existing models, easily outpacing the Fed’s Greenbook forecast and the consensus forecast of professional economists, at least for short-term projections. And it is much better at predicting turning points. In 1982 and 1990, for example, the Fed’s forecast expected unemployment to peak much sooner — and at a much lower level — than the new model, which correctly predicted unemployment would continue to rise.
Because it looks at the individual flows into and out of unemployment, the new model also provides insights into other key measures of the labor market that are ignored by traditional forecasts. In recent years, for example, economists have focused less on the unemployment rate, which can be skewed because it only measures people who are actively looking for jobs, and more on the “labor force participation rate,” which measures how many people are working or looking for work. In August, for example, the unemployment rate fell not because more people found jobs but because 368,000 people dropped out of the labor force.
Messrs. Barnichon and Nakarda’s model predicts that trend to reverse in coming months, with people gradually returning to the labor force, pushing the participation rate slowly but steadily upward. If they’re right, that means the unemployment rate will be falling for the “right” reasons: people actually finding jobs.