Author: Andrew Berger-Gross
An earlier post on the LEAD Feed introduced the Local Area Unemployment Statistics (LAUS) program, which is responsible for producing the official unemployment rate for states and local areas. The LAUS program generates timely estimates for thousands of geographic areas on a monthly basis and does so at a fraction of the cost of other federal statistical programs with similar levels of geographic coverage. However, these estimates are only preliminary when they are first released and may not precisely reflect conditions on the ground. This is why LAUS has a monthly and annual revisions process.
Monthly revisions are required because some of the data inputs that support the LAUS estimation model are often incomplete at the time of estimation, and need to be updated in the following month. These revisions are typically very small.*
The annual LAUS revisions are much larger. In addition to incorporating further updates to data inputs, the annual revisions take advantage of a larger amount of input data and a more powerful — and expensive — estimation method. LAUS estimates are revised in each subsequent year, but the first- and second-year revisions are the most consequential.
Annual revisions revealed that the unemployment rate in North Carolina was higher than originally predicted in 2010 and 2011, while the rate turned out to be lower than initially reported in 2012 and 2013.
Although the LAUS unemployment rate is an important snapshot of current labor market conditions — arguably the best estimate we have — the picture that emerges from the preliminary numbers is blurry. The revisions process allows this picture to develop over time into a clearer depiction of economic reality.
General disclaimers:
The LAUS estimates are based on an estimation model and are subject to model based and nonmodel based error. Any mistakes in data management, analysis, or presentation are the author’s.
Footnotes:
* Estimates for the month of December are omitted from these graphs. This is done for two reasons. First, monthly revisions do not occur in December, making a comparison with other months problematic. Second, dropping the December data creates visual breaks between years that make the extent of the annual revisions easier to interpret.