Author: Andrew Berger-Gross
Welcome, labor market watchers, to March 2016! Temperatures are creeping upward, the days are getting longer, and you’ve finally dragged that Christmas tree out for curbside pickup. Now for the fun part: annual data revisions.
State and local unemployment rates are estimated and published every month by LEAD and our partners in the federal-state statistical system, and these monthly estimates are then revised in the following month. Each year around this time, we conduct additional revisions that impact historical unemployment rates from previous years. While the monthly revisions to the unemployment rate are very small, the annual revisions have been quite large in recent years, often painting a substantially different picture of our economic health than the data initially published (and reported in the news media) on a monthly basis.
For example, according to the data published at the time, North Carolina saw its unemployment rates spike upward during the first half of 2015. The upturn in North Carolina’s rate was reported extensively in local news media, and some commentators went so far as to attribute this apparent worsening in labor market conditions to changes in public policy.
However, all of this hand-wringing was in vain. Newly-released annual data revisions demonstrate that North Carolina experienced hardly any uptick in unemployment at all in 2015. Rather, the unemployment rate remained remarkably stable, varying only one-tenth of a percentage point during all of 2015.
In 2015 we also saw a sharp increase in North Carolina’s labor force participation rate after fifteen years of declines. Some observers even attributed the apparent increase in unemployment rates to a fast-growing labor force. Again, these conclusions turned out to be largely premature; our state saw a nearly flat labor force participation trend throughout 2014 and 2015 after revisions. Although the initial estimates showed a 1.5 percentage point increase between December 2014 and December 2015, this was revised down to a 0.2 percentage point increase.
Another prominent feature of the 2015 labor market was a sharp increase in the number of employed North Carolinians. Many economists have been watching the data for signs of “escape velocity” – a self-sustaining cycle that would bring us the rapid employment gains seen during previous recoveries. However, it turns out that our 2015 employment surge was mostly a mirage; revisions show employment increased only 1.0% between December 2014 and May 2015, rather than the 2.8% originally estimated.
So why do economic trends change – and even reverse direction – upon later revision? This occurs because our knowledge about the labor market at any particular point in time is incomplete, and any attempt to estimate current conditions in the economy is bound to be clouded by uncertainty. LEAD has written previously about the numerous sources of error in the unemployment rate, all of which create some degree of uncertainty about actual conditions on the ground. Some of these sources of error are temporary and are resolved over time through a process of revisions.
State and local unemployment rates are produced by the Local Area Unemployment Statistics (LAUS) program. The LAUS annual revision process involves updating data inputs that feed into the LAUS estimation model and running a more powerful version of the model to incorporate a more complete set of input data. As in previous years, the process of running the estimation model on a larger set of data was likely responsible for the bulk of revisions to the LAUS unemployment rates.
So how should a labor market watcher such as yourself interpret the economic data that are reported in the news media every month when these data are likely to be revised at a later date? Here are some guidelines that can prevent you from prematurely jumping to conclusions:
- Consult data from different sources. For example, if the unemployment rate is increasing, are we also seeing declines in job creation or a slowdown in other economic indicators? If the answer is “no”, then it is possible that the unemployment rate data are erroneous and will be revised downward at a later time.
- Pay careful attention to published measures of uncertainty (such as the margin of error). These measures usually depict one particular source of error – e.g. sampling error – and do not account for every conceivable problem that might occur in the process of data estimation. However, they can give you a general sense of how confident we are in the accuracy of the data.
- Ignore the month-to-month movements in economic data and focus instead on long-term trends. Monthly economic data are often noisy, subject to revision, and (most importantly) provide little information about the overall direction of the economy. Long-term trends are much more stable, less affected by data revisions, and provide a wealth of information about what is happening in our economy and what we can expect in the future.