Volume 4, No.3
Fall 1998

Go Home

Archive Articles:
On-Going Research in Estimating Local Areas: CES and National Opinion Research Center Collaborating

Where Have All The Miners Gone?

Mass Layoffs

Archives

On-Going Research in Estimating Local Areas: CES and National Opinion Research Center Collaborating
By: Michael F. Macaluso

Current Employment Statistics (CES), one of the programs within the Economic Information and Analysis (EI&A) Division of the Illinois Department of Employment Security (IDES) generates, in partnership with the U.S. Bureau of Labor Statistics (BLS), current monthly estimates of industry employment, hours, and earnings estimates for the State of Illinois and its Metropolitan Statistical Areas (MSAs). CES also provides sample data to EI&A’s Labor Market Economists to produce estimates at the local level for smaller geographical units. These units can be individual counties or clusters of counties that form an economic entity called a Labor Market Area (LMA).

Over the past few years the CES program has been criticized because its sample data is not based on a scientific probability design. To remedy this, in 1995, BLS began the process of converting the sample design. Along with the other large industrial states, Illinois became part of their advisory group. New methodologies were researched to find the best estimators utilizing a scientific probability design. However, the research focused on national, state and MSA estimators. Initially, the LMAs were excluded from this initiative until Illinois’ LMI Director, Henry Jackson, expressed his concern about the need for local data by our Labor Market Information (LMI) customers. Based on this, the decision was made to extend the research to include the LMAs. As part of its work in the group, Illinois contracted with The University of Chicago’s National Opinion Research Center (NORC). In short, estimates for all of the geographic entities, including the smaller ones, i.e. LMAs, are being put on a statistically sound footing.

The purpose of this report is to sketch Illinois’ research relationship with the National Opinion Research Center (NORC). The current status of this relationship, placed in the context of broader national interests, embodies the state-specific need to bring credible, reliable labor market information to local users in their communities. Methodological issues have emerged stemming from NORC’s exceptional thoroughness. The issues, as stated in this report, are intended to bring clarity and focus to our discussions as we move toward applying mathematically and practically sound estimating methods to smaller areas. Through its research relationship with NORC, Illinois has taken the lead nationally to construct estimating methodologies that will produce payroll job estimates for smaller areas consistent with the Bureau of Labor Statistics probability design.

In June 1995, the U.S. Bureau of Labor Statistics began a major design change to the sample used in the national Current Employment Statistics Program. A limitation of the sample is that it is not based on a probability sample design. A probability-based design allows the time-series estimates to be expressed in terms of sampling errors and confidence levels. Another limitation is the difficult-to-measure births and deaths of companies. Technically, the current sample is referred to as a quota sample, i.e. the sample solicitation effort is directed at getting companies to report across the spectrum of industries and size classes within the state, its MSAs, and LMAs. As mentioned earlier, confidence and error measures are not applied. Thus, the need to change.

The U.S. Bureau of Labor Statistics planned the research to span a two-year period followed by production testing and finally implementation. The main research goal focused on national, state, and metropolitan area estimates. To achieve this limited goal, BLS formed a workgroup of members of its national and regional staff plus representatives of 11 large states.* In order to ensure the success of this initiative, BLS contracted with experts in sample design. The subsequent research outcome features the use of the ES-202 Unemployment Insurance-based reports as a variable. Using such a variable considerably increases the power of the estimators utilized. It is believed, as the testing phase will soon confirm, that the estimator will produce statistically valid employment estimates for the larger geographical areas.

However, the states in the workgroup, led by Illinois, voiced concern about the lack of attention given to smaller areas, such as labor market areas. Illinois has committed to providing local labor market information and has demonstrated to BLS the benefit of extending probability research to include local areas.

[Editor’s note: The Workforce Investment Act of 1998 requires implementation of a system of national, state, and local employment statistics.]

Recognizing this need, Illinois contracted with NORC to study these smaller geographical units in order to find an unbiased estimator consistent with those used for the larger areas. Illinois convened a sub-group comprised of California, Connecticut, Iowa, Michigan, New York and Pennsylvania to share the results of NORC's findings and to advise on additional research.

NORC, in concert with Illinois, tested a number of estimators, running simulations to narrow the search for the best unbiased estimator of employment. At this point in time, the estimator that shows the most promise is known as the synthetic estimator. This estimator is built from the monthly sample plus a modeled estimate of establishment levels not in the sample. The estimator uses the ES-202 as the basis for modeling these non-sampled establishments. As the research progresses, questions may arise that will first appear as issues but, in essence, are a by-product of the progress we have made thus far.

Some of the issues are as follows:
The synthetic estimator relies on the ES-202 as an auxiliary source to enhance the sample found in the small areas. Its use is to add in the remaining firms not in the sample. The question, empirical in nature, is at what point should the model be applied? Should it be applied at the statewide level, assuming the relationship is the same at this level as in the smaller areas? Or, should it be applied at a lower level in the state geographical structure as in a Balance of State (the area outside the MSAs) with the counter assumption that this type of area better reflects the dynamics of the firms found in local areas.

A second question orbits around the geographical summability principle. BLS seeks integrity up and down the roster of estimate points to which it grants approval. For the LMAs to be in this BLS family, such a principle would have to be satisfied. NORC and Illinois are currently researching this question. At issue is the quality of the estimates as it relates to the application of techniques designed to achieve this. Summability may be obtained but at what cost?

Also at issue is whether analyst intervention would have an adverse effect on the final product, particularly if the goal is to achieve a uniform product free of human bias. When dealing with local area phenomena, such as firm births and deaths and special employment-influencing events (e.g. labor-management disputes and natural disasters), intervention from a local analyst can add enormous value to the estimate that no model can fully grasp. How do we get the best of both worlds—the scientific and the human factor?

Use of information from other BLS Program areas (unemployment insurance claims, mass layoffs and labor-management dispute documentation) has always been a practice in Illinois. How will such useful information be incorporated under a probability design?

Replication of this research design in other states is clearly desirable. The findings in Illinois are pointing in the direction of the synthetic estimator. However, the following points emerge:

• Is this estimator valid in other states?
• What role will states play in determining the composition and size of sample?

The research nexus of BLS, NORC, and IDES is working to provide local answers to labor market questions. We will keep you posted as this research develops.