TEXT FOR DELIVERY: 9:30 A.M., E.S.T. FRIDAY, FEBRUARY 4, 1994 ___________________________________________________________ Advance copies of this statement are made available to the press under lock-up conditions with the explicit understanding that the data are embargoed until 8:30 a.m. Eastern time. ____________________________________________ Statement of Katharine G. Abraham Commissioner Bureau of Labor Statistics before the Joint Economic Committee UNITED STATES CONGRESS February 4, 1994 Mr. Chairman and Members of the Committee: It is again a pleasure for my colleagues and I to be here to offer a few comments on the Nation's employment situation. Nonfarm payroll employment showed some improvement in January, as 62,000 jobs were added. It is quite likely, however, that employment gains were held down by unusually frigid weather over large portions of the country during the survey reference period. The unemployment rate was 6.7 percent in January, the first month that the estimates were based on a totally redesigned household survey questionnaire and collection methodology. The January employment estimate -2- from the household survey was 1.3 million higher than the figure reported for December. This difference largely reflects the introduction of new population estimates, although the survey redesign also may have affected the employment estimate. This leads me into what is perhaps the topic of greatest interest today. This morning, for the first time, we released data from our revised household survey. Over the past few years, BLS and our partners at the Bureau of the Census have put forth an extraordinary effort to bring about the most comprehensive set of improvements in the Current Population Survey since its inception more than 50 years ago. We are very pleased to report to you that the first full-scale "live" data collection using the revised survey questionnaire and collection methodology went very smoothly. This is a particularly impressive achievement, since the data collection coincided with the earthquake in California and record cold weather across much of the eastern half of the country. I will return to the new Current Population Survey in a moment, but, as I have often said these last few months, changes to this survey require that we rely on our payroll survey even more than usual for our view of current labor market developments. As I mentioned, weather appeared to play a major role in the relatively small rise in payroll employment between -3- December and January. Though the reference week for the survey was the week before the California earthquake and the East Coast snow and ice storms, the extreme cold over much of the Nation during the reference week held down employment. Indeed, the weakness in the payroll survey was quite concentrated in weather-sensitive industries in January, while other industries continued the pattern of moderate growth that has prevailed for some time. Construction employment is most obviously affected by unusual weather. Our seasonal adjustment factors, of course, expect only an "average" month. Thus, particularly bad weather will result in seasonally adjusted construction employment estimates that may mask underlying strength. In January, construction employment was flat, despite general indications of improving health in the industry. Our survey responses suggest larger-than-normal construction cutbacks in the poor-weather areas, while employment continued to grow elsewhere in the country. Similarly, employment in entertainment and recreation services, which also can rise or fall on the basis of unusual weather, fell by 17,000 in January. In addition, we saw employment weakness within retail trade, particularly in eating and drinking places and general merchandise and apparel stores. I should point out that, despite the small employment gain overall, we did continue to see strength in many of the -4- industries that have done well recently. A number of the durable goods manufacturing industries, particularly the construction-related and auto-related industries, had job gains in January, and factory hours remained high. Job growth also continued in finance, transportation, wholesale trade, automobile dealers, and health services. Returning to the household survey, as we have discussed on previous occasions, there are three factors at work to make the reported January estimates, particularly the unemployment rate, different from what would have been obtained had the December survey and estimation procedures been followed. First, the survey redesign can be expected to have raised the measured unemployment rate. Over the whole of 1993, our test of the new survey produced unemployment rate estimates nearly half a percentage point higher than the official estimates for that period, though that does not imply that the new and old survey estimates necessarily would differ by exactly half a percentage point in any particular month. Second, the introduction of new population controls raises the jobless rate by a tenth of a percentage point. Finally, the survey data for January -- and perhaps also for February and March -- may differ from what we would have obtained had the new survey been in place for a longer period of time. As noted earlier, we have estimated an unemployment rate of 6.7 percent for January. The unemployment rate -5- reported for December was 6.4 percent. I would stress again that those two numbers should not be directly compared. We have made efforts in several areas to bridge the gap between the old and new survey data. The first attachment to my statement shows our model-based estimates of what unemployment would have been from September 1992 through December 1993 had we used the new survey methodology and updated population controls. In effect, our model smooths out some of the variance found in the parallel survey conducted during that period and produces estimates that more closely follow the month-to-month pattern of the official survey results. Our model-based unemployment rate estimate for December 1993 was 7.0 percent on a seasonally adjusted basis. While the series shown in this attachment provides a helpful frame of reference, I would nonetheless advise against taking the three-tenths of a percentage point difference between the December model-based figure and the January survey estimate as a measure of the "real" change in the unemployment rate. Aside from the uncertainty associated with any model-based estimate, there may be factors related to the new survey's phase-in -- and thus totally unrelated to economic factors -- that might have affected the January data and may cause the rate to "drift" in upcoming months. In addition, although at least initially we will be using seasonal adjustment factors based -6- on prior years' experience, those factors may not precisely capture the pattern of seasonality in the new series. Another tool that we are providing to analysts is our model of what the January 1994 unemployment rate would have been under the old methods, based on the historical relationship between the unemployment rate and other economic indicators. That rate, as shown in the second attachment to my statement, is 6.3 percent. We will continue to produce a similarly modeled measure for the next few months. As you know, we also are continuing collection of data using the old survey methods with a sample comparable in size to the one we used to test the new questionnaire (12,000 households). Around the middle of the year, we will begin to incorporate the results from this survey into our model-based estimates of what the unemployment rate would have been had the old survey been continued. A last technical item I should note before discussing some of the details of the household survey data is that we have introduced 1990 census-based population controls adjusted for the undercount. We normally introduce new population controls following each decennial census. The introduction of undercount-adjusted controls raises the population level by 1.3 million and the level of employment by about 950,000. And since the new population estimates have a higher proportion of persons with above-average -7- unemployment rates (particularly Hispanics) than the old 1980 census-based population estimates, the jobless rate is a tenth of a percentage point higher than it otherwise would have been. Keeping all of this in mind, we are reporting that the jobless rate for adult women was 6.0 percent in January; the rate for men was 5.9 percent; and that for teenagers, 18.4 percent. The rate for whites was 5.8 percent; that for blacks, 13.1 percent; and that for Hispanics, 10.6 percent. Each of these rates implies a December-January change in unemployment well within the range that we had anticipated based on the results from the parallel survey. Other estimates of particular note are the counts of discouraged workers and persons working part time for economic reasons. The discouraged worker total for January (not seasonally adjusted) was 600,000; we had been reporting about 1.1 million for the prior 2 years. The lower figure reflects a new definitional requirement that, in order to be classified as a discouraged worker, one must have looked for work during the prior year. This change, which implements a recommendation of the National Commission on Employment and Unemployment Statistics, brings the measure more in line with what people have generally perceived it to be -- a count of persons who had given up looking for work due to discouragement over job prospects. Similarly, the number of persons employed part time for economic reasons was about a -8- million lower in January than in December, because respondents so classified now must explicitly indicate that they both wanted and were available to take a full-time job. In summary, today we begin a new era with the release of the first data from the redesigned household survey. Any statistical organization must continue to improve its data to keep up with the times. Nevertheless, we well recognize and appreciate the added uncertainty that this transition imposes on labor market analysts and hope that our continued efforts to bridge the old and revised surveys will ease the changeover. With respect to the state of the labor market, the healthy payroll employment growth that we have seen in recent months appears to have been restrained by unusually poor weather in January, but many of the industries that are less affected by weather conditions continued to expand. The unemployment rate based on the redesigned household survey was 6.7 percent. Despite some uncertainty surrounding this rate, the data released today are consistent with there having been continued underlying improvements in the labor market. My colleagues and I now would be glad to answer any questions the Committee may have.