Confusing Employment Data
May Mask a Recession

Providence Business News, Opinion Page
February 2008
 

   by Leonard Lardaro

We recently received the December 2007 payroll employment data for Rhode Island. There wasn’t much to celebrate within these data, which shouldn’t be surprising since the national economy continues to slow and based on my Current Conditions Index, Rhode Island is in the early stages of a recession. But there is a lot more going on with the employment data than has been appreciated by the media up to this point. There are two points that need to be discussed to better understand this.

First, a strange statistic has emerged of late in Rhode Island. In order to computer the annual change for a given year, a number of persons have decided that it is appropriate to simply compare December of that year to the prior December’s value. Using this, we Rhode Island had a 3,300 job increase for 2007. While this is one method for approximating annual change, it is not the correct way. To do this, it is necessary to compare average employment for all of 2007 with the average for 2006. Doing this, we arrive at an annual change of +4,850, which provides a very different picture of how well employment has been holding up here.   

The second point is a great deal more substantial. The data that was just released is produced using a fairly small number of firms called the Current Employment Survey (CES). Eventually (in February each year), the CES data are revised based on a much larger and more accurate survey, the Establishment Survey. The CES data exist and are used as a “stop gap” since the Establishment Survey data lag by several quarters. Over the years, the CES data have produced many false impressions, especially for the final three months of the year. For example, a year ago, the CES data indicated what appeared to me to be an overly pessimistic jobs count for Rhode Island. Using econometric (statistical) methods, I simulated what the likely revised employment count would be using non-survey based data. My prediction was for an upward revision of 2,000. The actual change turned out to be an increase of 2,400. As you might have guessed, I have just completed my simulation for this past year.

Accurately estimating employment is very complicated around turning points, which is where I believe Rhode Island currently finds itself – the early stages of a recession. The US Bureau of Labor Statistics uses a “birth-death” model in estimating employment to account for jobs added by newly created firms (births) and the loss in employment by firms that go out of business (deaths). Unfortunately, this birth-death adjustment tends to work better in the early stages of a recovery than when a recession is just beginning, as births tend to be overstated then while deaths are understated. The net result is downward adjustment to released data.

And, downward revision is what I see for Rhode Island’s labor market data. My simulation this year suggests that the average employment level in 2007 will be about 750 less than what had been stated with the CES data. While that doesn’t seem like a very large change, the major difference occurs with employment at the end of 2007 – my simulation sees actual December 2007 employment as being 2,000 less than the recently released CES value. That is a big difference!

Should my simulation prove to be correct there are several important implications. First, job gains here have been quite a bit less than previously thought, which is consistent with the data from last year on layoffs and long-term unemployment. Second, existing forecasts of employment for this year will be wide of the mark, being overly optimistic. For example, a Rhode Island forecast by a regional group predicts a job increase of 4,000 for 2008, apparently on top of the currently published CES values. Guess again! My forecast calls for an increase in payroll employment of 1,000 from the value I have predicted, which means virtually no change from the existing CES value for all of 2007. Finally, Rhode Island is contemplating changing its welfare system to a “Work First” system. A period of weak or declining employment is not exactly the best time to make such a change. So much for timing!  This is a further indication that balancing our budget will be extremely difficult, even when implementing changes that we believe are appropriate for the long term.