- State level employment estimates are very important for forecasters, policymakers and the media
- Current monthly employment estimates are over one full year off benchmark
- Using QCEW data through Dec ’11 as guide, the state level estimates for all states will be revised upward over 700,000 (0.7%)
- Revisions to the national data in 2011 appear to be minimal
- 2012 revisions are unknown at this time. 2012q1 data released Sep 27, 2012
What follows is brief summary. For a full set of slides, including more information and graphs, please download the PDF: State Employment Revision Slides
The Bureau of Labor Statistics provides monthly employment estimates for each state and over 370 MSAs, called the Current Employment Statistics (CES). The benefits to the CES data is its timeliness; data is released with less than a one month lag. The downside to the CES is that it is sample based and prone to estimate errors like all surveys. Once per year these figures are revised based on a more complete set of data. In particular, one of the major data series used in this revision process (called benchmarking) is the Quarterly Census of Employment and Wages (QCEW). This data contains all employees covered under the unemployment insurance system so it is a near complete count of all employment, which is nice, however it is less timely and released only every three months, which is less nice. To accurately gauge the health of the labor market it is important to track both of these series to make sure that what is reported on a monthly basis throughout the year is in fact what is actually happening in the economy.
Today, state and local employment estimates are over one full year removed from their last benchmarked data point: June 2011. How are the estimates performing? Based on work our office has been doing on Oregon and also providing some assistance to other states – like Wisconsin – the short answer is: not well in 2011 at least. The following graph shows the year-over-year employment growth in the private sector for three series: the national estimates, the sum of the individual states from the CES and the sum of the individual states from the QCEW.
Immediately following the last benchmarked data, the CES and QCEW growth patterns diverge. This indicates that the state level estimates are significantly under the actual data through the end of 2011 and also that the national data is unlikely to be revised significantly through the end of 2011 given the similarities in the data. Unfortunately any revisions to 2012 data are unknown at this point since additional QCEW data is not currently available. However, given that the national and sum of states CES estimates have converged by mid-2012, revisions to more recent months may be minor. That or the magnitude of revisions will be similar between the national and state CES data.
Using a preliminary benchmark process, which is a quicker and dirtier method than the official BLS process, the monthly sum of states data is likely off by over 700,000 jobs, or about 0.7 percent, in December 2011. Both the national data and the QCEW sum of states indicate that over 2 million private sector jobs were created between December 2010 and December 2011. However if one were to look at the state estimates from the BLS website the figures would only show 1.3 million private sector jobs created.
Some of you may be asking, “Why does this matter?” In particular, having the most accurate information is important for not only forecasters and policymakers but also the media and the larger public discourse. Basing analysis and forecasts off of bad data can be problematic and lead to unnecessary errors. Same for media reports that are unintentionally misleading. Myriad articles in the press are written every single month and direct a lot of the economic (and sometimes political) conversations in many states. When the revisions are small, this is not a big concern, however when they are as large as they currently are in our state, it becomes a much larger concern. As it turns out, Oregon is not even one of the largest expected revision states in the nation.
Based on this method, Oregon ranks 16th highest in percentage terms, while Arkansas and Wisconsin are the largest. In terms of the number of jobs, the larger states dominate, as expected. California, New York, Ohio, however Wisconsin is No. 4 at about +50,000.
So what can be done about this? First off, sometimes the sample self-corrects or the estimates are brought more in-line in real time and the data becomes more accurate. This may be the case today, seeing as the sum of states and national CES data are currently indicating about the same year-over-year growth rate in the private sector (see first graph). This may or may not be the case and the current convergence may be random. However, in the larger picture, shortening the time frame between benchmarks can greatly improve these issues. Some states choose to do their own benchmark on a quarterly basis, which as I understand it is allowed under their agreements with BLS, however some states do not. As a numbers person, having the best possible information is important, however having the data tell the true underlying story is probably the most important issue here. This is the case on both the up and down sides. This is not about cheerleading better data in late 2011, as our office’s in-house, unofficial, preliminary benchmarking process actually started following the 2009 downward revisions to the employment data. We wanted to make sure that we were not caught too far off-guard and forecasting employment growth, even as the underlying data continued to crumble.
For a full set of slides, including more information, graphs and larger sizes, please download the PDF: State Employment Revision Slides