September 29, 2009

Economic forecasting with household items

Posted in Macroeconomics at 5:44 pm by Eamon Aghdasi

For a while I’ve wanted to just do a quick forecast of employment with “household” items (basically anything you can find on the web). So I took a stab at it. The basic question I sought to answer was what the quarterly employment change (change in seasonally-adjusted, non-farm payroll employment) will be in six months.

Why is this interesting? Well, if you look at the last few months’ worth of employment change data, it’s tough to figure out just what the next few months have in store. Job losses seamed to peak (or trough, whatever) back in January, when the US economy lost about 741,000 jobs on a seasonally adjusted basis. In the next four months we saw this number steadily decline to just 303,000 in May, and everyone got excited. Then the next month it jumped back up to 463,000 in job losses, still a devastating number and pretty disappointing for those that hoped we’d reemerge into positive territory soon.

Now I feel that analysts don’t really know what to expect from the next few months. Will we continue to shed employment at a lower level for much of 2010? Will the numbers reach into slightly positive territory or hover around zero for a while, at a pace of growth slower than that of the labor force? Or will employment rebound strongly in the next few months, enough to bring down the unemployment rate? (Unfortunately, not many people seem to believe in this last scenario.)

emp_change_forecast_workbook

What does a quick forecast say? Well, using just the “household” items, you can get a good historical approximation of quarterly employment changes, two quarters down the line. These “household” items are:

  • Second-order change in employment (basically change in change, from Bureau of Labor Statistics)
  • Inflation (BLS)
  • GDP growth (rates from the most recent three quarters, from the Bureau of Economic Analysis)
  • Growth in durable goods consumption (BEA)
  • Export growth (BEA)
  • Business output per hour (indexed, from BLS)
  • Business output per person (indexed, from BLS)
  • Business average weekly hours (indexed, from BLS)
  • Change in business unit profits (BLS)
  • Interest rate spread (10-year T-bill minus Federal Funds rate, from the Fed)
  • Percent change in M1 (Fed)
  • Percent change in S&P 500 prices (lots of places on the web)
  • Consumer sentiment index (from the University of Michigan… basically because I’m too cheap to pay for consumer confidence data)

The results of this exercise were surprisingly good. Regressing all of these things on employment change two quarters into the future yielded an R-squared value of about .90. And an eyeball test proved pretty satisfying. Take a look:

Forecasted versus real employment change

Using data up to Q2 2009, the model predicts a drop in employment of roughly -673,000 in Q3. Like a lot of other forecasts (by people who are actually paid to do this), this turned out to be much smaller than the actual loss (-1,042,000) reported by the BLS for that quarter.

Unfortunately, the model doesn’t exactly anticipate a robust turnaround in Q4 (shocking), with a predicted employment loss of -613,000 over those three months. So, basically don’t quit your job at the dry cleaners just yet.

Making matters worse is the situation with what is probably my favorite leading indicator of employment: temporary help services. As companies start dipping their toes into the water in terms of once again increasing headcounts, they often start with low-risk temporary employees that are easy to shed. Unfortunately, after completely falling off a cliff the past year-and-a-half, recent data on temporary employment gives us very little to be excited about.

emp3

What’s your prediction for the next few months? How positive or negative are you? What’s the big missing factor from the model? Let me know your thoughts.

Advertisements

1 Comment »

  1. Eamon Aghdasi said,

    September data out today… job loss was 263,000, bigger than last month. And temporary help services lost 1,700; too bad it it wasn’t positive, but at least it’s looking like a local minimum, and maybe that will turnaround in the next couple months.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: