These pages look best viewed in Netscape!
Copyright 1994 by the Wyoming Department of Employment, Research & Planning

New Labor Force Estimates: Measures to Change, But Are They Better?

by: Tom Gallagher

Beginning with the January 1994 Labor Force estimates, several Bureau of the Census and Bureau of Labor Statistics initiatives will change the way employment, unemployment, and, consequently, the unemployment rate are measured and estimated. The net impact of these changes, either individually or collectively, is not clear.

One of the changes to the way the labor force is measured may substantially increase Wyoming’s rate of unemployment.

The most notable change to expect in January is that the estimate of both the number of persons employed and the number of persons unemployed will increase as a result of adjusting the labor force estimates to an estimate of the 1990 population undercount. The second change expected to emerge is that the labor force estimates will be less prone to month to month unexplained volatility. Several other changes will be introduced at the same time and, while their impact is unclear, it appears that the net effect on the rate of unemployment will be self canceling. This article identifies the major changes the federal government is introducing and their impacts on levels of employment and unemployment, the rate of unemployment and volatility in the time series itself. These changes are summarized in the table: Changes to Expect in the January 1994 Labor Force . . ..

Labor force estimates in the most populous states are taken directly from the monthly household Current Population Survey (CPS). The sample of 59,000 households, however, is not large enough to create reliable estimates of the proportion of the population in the labor force and the proportion unemployed for the 39 least populous states.

In the early 1980's the Bureau of Labor Statistics (BLS) attempted to make the labor force estimates for small states more reliable. The Bureau designed estimation techniques which included not only information from the household survey, but also information from the establishment survey of employment (also known as CES Nonagricultural Wage and Salary Employment) and Unemployment Insurance claims levels (found in the Wyoming Economic Indicators Table). After several years of experience with the original estimation techniques, revised estimation techniques are scheduled for implementation with the January 1994 estimates.

1st Change: Increase the Level of Change Between Preliminary and Revised Labor Force Estimates.

Since the new labor force estimation techniques for Wyoming also rely on establishment survey data, it is important to understand that a change is under way here as well. Research & Planning currently carries approximately 3,700 establishments in this survey. We have normally had 80 percent of the monthly employment responses available for the preliminary estimates. Preliminary estimates were usually completed during the middle of the third week of the month. This has meant that differences between preliminary and revised estimates have been minimal. The BLS has moved the due date for preliminary estimates forward by six working days. This change is expected to drop the proportion of participating employers available for the preliminary estimates to 60 percent. Since the establishment survey is used in the labor force estimates, the overall impact of changing the delivery date will be to introduce an added but unknown level of change between preliminary and revised labor force estimates.

2nd Change: Reduction in the Level of Month to Month Volatility.

Monthly estimates of employment, unemployment, and the rate of unemployment are frequently used to monitor national, state, and local economic progress. One of the unique characteristics of these time series is that given a long enough period of observation, one can create statistical profiles of seasonal change, trend, and normal random statistical movement. Such random statistical movement, or error, is a function of sample size relative to the size of the population for which labor force estimates are created. The new estimation techniques are aimed directly at measuring and then removing random statistical movement, or error, from the estimates. The BLS uses sophisticated modeling techniques frequently associated with various types of forecasting to accomplish this task. One of the objectives of removing random error is to smooth the level of change in the labor force estimates from one month to the next.

3rd Change: Lower the Rate of Unemployment.

During the past nine months, Research & Planning has been running the new estimation techniques in parallel with the “official” estimating techniques. During this period, the new estimation techniques reduced variability in the employment estimates by 3 percent. At the same time, the parallel technique increased variability in the estimates of the number of persons unemployed by 8.1 percent compared to the “official” published estimates for Wyoming.

During the nine months ending with September, the new estimation techniques also dropped the State’s unemployment rate from 5.6 to 5.0 percent. The evidence to date suggests that the new estimation techniques tend to reduce the estimate of the rate of unemployment.

4th Change: Raise the Rate of Unemployment

But there is more. The CPS has undergone substantial redesign. That redesign, including a new questionnaire and data collection assisted by laptop computer, appears to have just the opposite effect from the new estimation techniques for Wyoming. A parallel test among a U.S. sample of 12,000 households using the new CPS system, which began in July of 1992, yields a national unemployment rate of 7.6 percent compared to an unemployment rate for the existing CPS of 7.0 percent. Figure 1 shows the U.S. Unemployment Rate for both the current CPS and the Parallel Survey from September 1992 to August 1993. The Bureau of the Census and BLS are unsure of the reasons for the increase.

The impacts of the new CPS on individual state estimates in January 1994 are not clear. However, if the results of the 1990 Decennial Estimates are any guide, the net effect on Wyoming may be minimal. While the April 1990 Decennial Census unemployment rate estimate for the state was 5.9 percent, the estimate from the BLS approved computational technique was 5.8 percent. The differences for most neighboring states were minimal, however, Montana’s decennial rate was 7 percent while the BLS computational technique provided an estimated unemployment rate of 5.3 percent. The widest gap was found in Louisiana with a BLS rate of 5.5 percent and a decennial rate of 9.6 percent. In sum, the future impact of the new CPS may be similar to the impact of the 1990 Decennial Census in that differences between states when moving from the old to the new CPS may vary substantially.

5th Change: No Discernible Evidence That the Measurement of Labor Market Hardship is More Valid With the New CPS.

Differences between decennial estimates of employment and estimates using the old CPS may reflect different techniques in asking survey questions or variances in other procedures, or they may reflect differences in the true level of labor market hardship. The same thing may be true when comparing the new with the old CPS. However, while there are relatively inexpensive ways to validate the labor force estimates, there is no program in place in the federal statistical system to conduct such validations.

Is the new CPS more accurate than the old CPS? The new CPS, like the old, represents merely the respondent’s verbal report of his or her behavior in the labor market. There is no program component as part of the CPS which checks administrative records, such as quarterly Unemployment Insurance records on employment and earnings, against behavior reported in the CPS. Citizen-reported behavior in either CPS has never been organized to be checked against direct measures, such as unemployment insurance records. Without such validation it cannot be determined that the new CPS is a good investment.

Is the new CPS any better than the old CPS? One of the reasons cited by the BLS for the implementation of a new CPS is that “...there have been many societal changes, such as the ... more prominent role of women ...” in the labor market since the last redesign was implemented in 1967. To the extent that the new CPS contains no mechanism to adapt to future social changes - such as the increase of minorities and foreign born individual in the labor force, the change in the cognitive meaning of questions, or the social acceptability of unemployment or dislocation and the willingness to report it - the new CPS is obsolete on the day it is introduced. While the new CPS may represent an improvement in some regards, it fails to address key problems used as a justification for the investment.

6th Change: Restoration of Wyoming’s Sample Size Reduces Month to Month Volatility.

The CPS is a survey of households and is designed to estimate the proportion of civilian adults in the labor force and, of that proportion, the proportion of persons unemployed. However, the number of households in each state is constantly changing, and the federal statistical system is constantly playing a game of catch up. Some adjustment in how the household sample is drawn each month can be made between each decade’s Decennial Census with the use of such information as the issuance of building permits. But the major adjustment in the size of each state’s sample is made following each Decennial Census.

Each state’s sample size is determined by the proportion of each states households needed to meet the federal criteria of an 8 percent coefficient of variation (that is, variation in the estimates of less than 8 percent are not considered to be statistically significant) in the labor force estimates. Sampling theory dictates that the smaller the population, the larger the relative share of the population required to attain a constant minimal level of error. Thus, in April 1994, the CPS will begin using new samples based on the 1990 Decennial Census. Because Wyoming lost in the number of households between 1980 and 1990, the sample will be increased from 690 to 956 households in an attempt to restore the estimates to an acceptable level variation. The CPS is a sample of households and remains unadjusted for the Decennial undercount of persons. Since the population undercount tends to be associated with minority populations, the CPS continues to be structurally biased against minorities.

Change 7: Increasing the Size of the Labor Force.

The CPS is an estimate of proportions (%). Each month the CPS estimates are, in turn, adjusted to a population level. Monthly population levels are estimated using the Decennial Census as a baseline, a subsequent count of each state’s births and deaths, and an estimate of migration. In January 1994, the BLS will adjust the baseline, and subsequent monthly estimates, upward to reflect the 1990 Decennial undercount. Following the Decennial Census, the Census Bureau uses two techniques to estimate the number of persons missed in the decennial canvass. In 1992, the Census Bureau decided not to adjust the decennial counts to the estimated undercount but suggested that certain of their clients, the BLS among them, could choose to do so. On November 11, 1993, the BLS announced that in January 1994 it would introduce the population undercount estimates and carry that adjustment back to 1990. Adjusting monthly estimates of the size of the population upward will increase the estimated size of the civilian population and, with it, the level of employment and unemployment. Despite the fact that the levels of employment and unemployment will increase beginning in 1994, and retrospectively to 1990, current rates and prior rates of unemployment will remain unaffected on an annual average basis.

Depending upon which of two techniques the BLS uses (or an average of the two available techniques) it is most probable that the level of the labor force will be adjusted upward in Wyoming in the vicinity of 2.2 percent. Applying this increase to the April 1990 BLS sanctioned state estimates serves to increase the number of persons employed by approximately 5,000 and the number of persons unemployed by 300 with no impact on April’s unemployment rate. Re-computing the labor force estimates for the period 1990 forward using the new estimation techniques, and adjusting those proportions to the undercounted population levels, will cause differences to occur between the new series and what has already been published.

Of all of the changes introduced by the Bureau of Labor Statistics and the Bureau of the Census, the upward adjustment of the population levels to reflect the undercount will be the most noticeable. However, since the new CPS has the potential to significantly raise the unemployment rate, it is evident that Wyoming could experience a radical departure from our relatively low unemployment rate to a new and higher rate. This increase should be interpreted as a function of one or all of the changes in the way the labor force is measured, not as a turn in Wyoming’s economy.

Tom Gallagher is the Manager of Research & Planning.

Button Bar 4
Go Back To:
Table of Contents
Labor Market Information
Employment Resources
Send Us Mail

These pages designed by Gayle C. Edlin.
Last modified on January 16, 2001 by Valerie A. Davis.