What is the unemployment rate calculation, how is it calculated, and is it accurate? Since the measurement is an estimate of the number of unemployed with respect to the labor force, it is more proper to call it a ratio but we will continue to use rate for consistency. The official unemployment rate reported by the Bureau of Labor Statistics (BLS) might not be an accurate representation of the unemployment situation.
We will examine the source of the data used to calculate the unemployment rate. Then we examine whether including more factors might provide a more accurate representation of the jobs situation.
Source of the data: the BLSThe data comes from a report called the Current Population Survey (CPS), also known as the Household Survey. The Bureau of Labor Statistics (BLS) is the agency that collects and publishes the data, including the CPS itself, and the economic measures derived from it. The BLS also calculates several alternative measures of labor underutilization.
The CPS is a monthly survey conducted by the BLS (BLS, 2008f). It is a survey of 60,000 households where the data is collected (BLS, 2003). “The sample is selected to reflect the entire civilian non-institutional population. Based on responses to a series of questions on work and job search activities, each person 16 years and over in a sample household is classified as employed, unemployed, or not in the labor force” (BLS, 2008b).
First, to dispel a myth, unemployment insurance is just that, insurance. It is not part of the official unemployment rate calculation. Not all unemployed people are eligible or apply for benefits and many remain unemployed after their benefits run out. There are no questions on unemployment insurance in the survey.
The answers respondents provide place them in a specific category. Based on the BLS definitions of employment, unemployment, discouraged, labor force, and marginally attached (see appendix for definitions), one can easily see what segment of the population one would be in at various times in one’s life. The definitions are not necessarily a problem but interpretation of them can be.
An example of how interpretation of the data and definitions can be an issue is how financial markets eagerly await the release of the data every month and react to it. Based on whether the reported unemployment rate rises or falls, the stock market often moves in the opposite direction. Many people make decisions based on the reported unemployment rate, perhaps giving it a greater importance than it should have.
Considering the sample size, 0.01% is within the margin of error, yet that small a change can create a disproportionate reaction. The BLS recognizes this and often refers to such a change as statistically insignificant. In addition, the underlying data is based on what survey respondents say in answer to the specific questions. Depending on many factors, that alone can skew the results. Misunderstanding and differences in precision may change the results.
The Labor Force Participation Rate & the Marginally AttachedTo calculate the unemployment rate, the BLS starts with total civilian non-institutional population 16 years old and over and takes away those “not in the work force”; which results in the official “labor force”. Then, taking the number of “unemployed” divided by the total labor force, we get U-3, the official unemployment rate (BLS, 2008a). However, issues arise with the definitions of “labor force participation rate” and “marginally attached”. These terms are closely related since the marginally attached are part of those not participating in the labor force.
To illustrate potential issues, we look at a couple of those definitions. In 1994, there was a change in the definition of both “discouraged worker” and “employed part-time for economic reasons”. This change required that, in order to be classified as either, participants must be available for full-time work in the week they were polled (Employment and Earnings, 1996).
According to Polivka and Miller, the changes in the definition of “discouragement”
and “involuntary part-timers” created a significant drop in those populations. “In fact, fully half of those previously classified as discouraged would no longer qualify for that category” (as cited in Epstein, 2006).
Another issue may arise with the preciseness of a definition. For example, individuals who did not look for work in the previous four weeks, for any reason, are not counted as unemployed; they are “not participating in the labor force.”
About 1.6 million … were marginally attached to the labor force in September, 336,000 more than 12 months earlier. These individuals wanted and were available for work and had looked for a job sometime in the prior 12 months. They are not counted as unemployed because they had not searched for work in the 4 weeks preceding the survey [italics added]. Among the marginally attached, there were 467,000 discouraged workers in September; the number of discouraged workers has increased by 191,000 from a year earlier.… The other 1.1 million persons marginally attached to the labor force in September had not searched for work in the 4 weeks preceding the survey for reasons such as school attendance or family responsibilities (BLS, 2008c).
Discouraged workers are one aspect affecting the participation rate but the age distribution of the labor force is another, more significant factor. If, as Robert F. Szafran’s (2002) research indicates, the participation rate should be considerably higher than reported (2002) the “missing” workers must be “not in the labor force.”
The overall labor force participation rate is significantly impacted when large groups move into or out of a population segment with either above or below average rates for the group. There will be more distortion on the reported participation rates by age groups with a higher percentage of the population. Groups with higher participation rates will tend to raise the reported rate while groups with lower participation will lower it. One way around this issue is to adopt a technique demographers have long used and take a weighted average of the rates for each age group. Using age-adjusted participation rates removes the effects of the relative size of each age group on the overall rate (Szafran, 2002)
Other FactorsLabor Force GrowthThe total eligible population, the labor force, and its components are not static. The population is constantly growing while the employed, the unemployed, and labor force participation fluctuate. If the labor force growth is negative, it is likely that the participation rate is as well. Although we cannot tell whether this is due to fewer entrants, an increase in those not counted in the labor force, or both.
Hidden participantsThe data presented also miss a small portion of the labor force where the wars in Iraq and Afghanistan are involved, hiding more than an additional 100,000 members of the labor force (Defense Manpower Data Center, 2008). These are civilians on active duty not normally excluded from the labor force. When the wars are over, they will be returned to the civilian population and will presumably look for work unless they are lucky enough to return to the job they had when activated.
Churn & SpinAnother fluctuation is the movement of people between the three states of employed, unemployed, and marginally attached. These movements, known as “churn” occurs naturally as people gain or lose jobs and enter or leave the labor force for whatever reason.
Some industries normally experience a great deal of churn all the time. Industries like retail, food services, and construction account for the majority of churn in total non-farm employment while accounting for a much smaller portion of total employment (BLS, 2008d).
Simply, market economies tend to create and destroy a lot of jobs simultaneously, as employers collectively allocate scarce resources for additional labor in some sectors and occupations while cutting back in others. Even in strong job markets, some jobs are being lost, just as some jobs are gained in the midst of a recession…. Each year millions of people nationally lose their jobs. But in most years, an even larger number of people find new jobs…. But over time, it is this very churn that helps struggling economies transfer resources to better uses, while keeping healthy economies robust in an ever changing world (Fitzgerald, T., Holland, M., 2005).
Perhaps the most influential factor of the effect of the unemployment rate on the markets and people’s lives, however, is “spin”. Gene Epstein’s book “Econospinning: How to read between the lines when the media manipulate the numbers” describes this phenomenon in detail and this factor may be the most compelling argument for a change in the reporting of the unemployment data.
The need to milk the numbers for suspense and drama blows their importance way out of proportion. The excessive focus on the minutia of the report ensures that meaningless figures will keep getting mentioned, as when Liesman obsesses about differences of a hundredth of a percentage point in the unemployment rate. Yet the Bureau of Labor Statistics (BLS) keeps warning that a tenth-point difference is within the range of statistical error (Epstein, 2006).
Alternative Or Supplemental Measures & A Possible Alternative ApproachSo far, all the data and calculations we have discussed come from the Household Survey. This section will introduce additional sources of pertinent information that is also collected and disseminated by the BLS along with suggested new ways to present the data.
For a complete picture of the employment situation, we might consider starting with the total population growth aged 16 and over, then show the institutionalized, breaking out the military, reserves (or “civilian military”) on active duty, the civilian non-institutionalized population aged 16 and over, those not in the labor force, and the labor force as trends. Since the data on the institutionalized, military, and reserves was not readily available from the sources used, an alternative is presented in Figure 1:
Figure 1 The linear trend line of month-to-month changes in civilian adult population, labor force, employed, unemployed, and those “not in the labor force” over the past year. (Dube, 2008)(BLS, 2007-2008)Using averages may impart more information than simple numbers. In “Econospinning”, Gene Epstein suggests a 6-month moving average of the unemployment rate (2006). This is analogous to a moving average of a stock, which provides a frame of reference for where the stock or, in this case, unemployment is heading. Is it trending up or down? How does the current average compare to the past year? Just as with stocks, one can overlay moving averages of various time frames for even more relative trend information. Figure 2 shows three moving averages along with the published monthly unemployment rate.
The chart in Figure 2 indicates that the trend at the beginning of 2001 was down and the 6-month average crosses both the 3-month and the 12-month average on the way up in the second half of 2001, signaling a long trend up to a peak in early 2004. In the second half of 2004, the trend turns back down and the 3-month average signals the change by crossing the 12-month, then the 6-month averages.
Figure 2 Unemployment rate and moving averages Note: Monthly rates derived from the Bureau of Labor Statistics CPS data published unemployment rates per month 1998 – 2008 (Dube, P., 2008) (BLS, 1999 – 2008)The 3 and 6-month averages both cross the 12-month average again mid-year in 2007 and the gap keeps widening to the most recent report. One interesting aspect that this graph shows is the near simultaneous crossing of the 3 and 6-month averages over the 12-month average on the way up but a decided gap in the crossings on the way down. This happens in both the visible upswings.
Following that with the composite age-adjusted participation rate with a breakout of the participation rate by age (data not currently available), followed by the Job Openings and Labor Turnover Survey (JOLTS) data, discouraged workers, and the unemployment rate together, as trends would depict the churn and give us a sense of where the job market was heading.
The job openings rate, from the JOLTS program, measures the unmet demand for labor. As a complement to the unemployment rate, it gives us a more complete picture of the U.S. labor market. “It is the flip side of at the unemployment rate, which measures the excess supply of labor” (BLS, 2002).
Participants in the Current Population Survey are interviewed several months in a row. Approximately three-quarters of those interviewed in any given month were also included in the survey the previous month. Because of this, their movements between the segments can be tracked. This effectively tracks the churn that goes on as a normal part of the labor market and is called Labor Force Flows, or “gross flows” (BLS, 2008g). The trend of this data may provide a useful measure of how the job market is changing over time. Including JOLTS data may help clarify the data.
Figure 3 Unemployment rate and JOLTS data Note: Monthly rates derived from the Bureau of Labor Statistics CPS data published unemployment rates and JOLTS published rates per month (Dube, P., 2008) (BLS, 2001 – 2008)If we plot the unemployment rate against the JOLTS rates of Hires, Job Openings, and Separations, the chart in Figure 3 appears to show the Hires and Separations remaining fairly flat. This could be an indication of the natural churn in the labor market. The coincidence of these trends in November 2007 and February 2008 seems to coincide with an upward change in the slope of the unemployment trend and the crossover in May of 2008 is accompanied by a steep jump in unemployment. These rates cross again in September 2008 and, if they flatten out, may signal the establishment of a new, lower churn rate. This chart clearly shows the job openings trend is almost a mirror image of the unemployment trend.
Figure 4 The Alternative Measures of Underutilization rates plotted. Rates derived from the BLS published report. (Dube, P., 2008) (BLS, 2000 - 2008a)The chart in Figure 4 shows the trend of the BLS calculated Alternative Measures of Labor Underutilization. By plotting the trends, we see that all of the trends are strikingly similar. This suggests that the official rate may be adequate, despite the common feeling that it leaves out too many people, as long as it is viewed as a trend. Using just the rate compared to the previous month is akin to using the closing price on any given day compared to the day before to determine a stock’s value. It is just insufficient information to base a judgment on.
Perhaps using Gene Epstein’s U-7 would be better than U-3 for two reasons. First it doesn’t really matter much which rate is used if it is a trend and we have shown the trends follow each other closely. Second, including those who reportedly want a job should satisfy those who claim that U-3 is inaccurate since it includes discouraged workers.
Releasing this data, as suggested here, prior to the release of the CPS data might just work for giving the media, thus the public useful information about the health of the economy and job market This might allow consumer confidence to more accurately reflect the state of the economy, thus giving the American public more confidence in the government’s collection and reporting of the data.
When the Information Technology industry started, the systems produced data. In the 1980’s, the industry started to distill that data into “information”. This evolution provided for better decision-making capability. In the 1990’s, the industry continued to evolve and began moving toward “knowledge” and “knowledge management”.
What this paper suggests is following the example of the Information Technology industry and evolving what, and how we report.
While this paper raises some interesting questions, more study is required to reach a conclusion. However, it appears that using composite graphs and moving averages presents a more complete picture of the jobs situation than the unemployment rate alone. Perhaps this is what we need to evolve from providing data to supplying information and deriving knowledge.
References
Bureau of Labor Statistics. (1998 - 2008). Labor Force Statistics (CPS) [Data file]. Available at: http://www.bls.gov/webapps/legacy/cpsatab1.htm
Bureau of Labor Statistics. (2001 - 2008). Job and Labor Turnover Survey (JOLTS) [Data file]. Available at: http://www.bls.gov/jlt/data.htm
Bureau of Labor Statistics. (2002). What is JOLTS? Retrieved Nov. 13, 2008, from http://www.bls.gov/jlt/jltwhat.htm
Bureau of Labor Statistics. (2003). Handbook of methods. Retrieved Nov. 15, 2008, from http://www.bls.gov/opub/hom/homch1_b.htm
Bureau of Labor Statistics. (2008a). Alternative measures of labor underutilization. Retrieved Oct. 18, 2008, from http://www.bls.gov/news.release/empsit.t12.htm
Bureau of Labor Statistics. (2008b). Employment Situation Technical Note. Retrieved Oct. 18 2008, from http://www.bls.gov/news.release/empsit.tn.htm
Bureau of Labor Statistics. (2008c). Employment situation summary. Retrieved Oct. 25, 2008, from http://www.bls.gov/news.release/empsit.nr0.htm
Bureau of Labor Statistics. (2008d). Job openings and labor turnover summary Retrieved Nov. 13 2008, from http://www.bls.gov/news.release/jolts.nr0.htm
Bureau of Labor Statistics. (2008e). Labor force characteristics. Retrieved Nov. 7, 2008, from http://www.bls.gov/cps/lfcharacteristics.htm
Bureau of Labor Statistics. (2008f). Labor force statistics from the current population survey. Retrieved Nov. 7, 2008, from http://www.bls.gov/cps/home.htm
Bureau of Labor Statistics. (2008g). New research series on labor force status flows from the Current Population Survey. Retrieved Nov.16, 2008, from http://www.bls.gov/cps/cps_flows.htm
Defense Manpower Data Center. (2008, August). Contingency tracking system (CTS). Available at http://www.defenselink.mil/news/Aug2008/d20080806ngr.pdf
Dube, P. (2008) Raw unpublished data
Epstein, G. (2006). Econospinning: How to read between the lines when the media manipulate the numbers [Electronic version]. Hoboken, New Jersey: John Wiley & Sons, Inc. Available from http://skillport.books24x7.com/toc.asp?bookid=16751
Fitzgerald, T., Holland, M. (2005, November). Helicopter churn: The macro view on job loss and growth. Fedgazette, 17(6), 3. Retrieved October 18, 2008, from ABI/INFORM Global database.
Szafran, R. F. (2002, September). Age-adjusted labor force participation rates, 1960-2045. Monthly Labor Review, 125(9), 25-38. Retrieved October 25, 2008, from ABI/INFORM Global database.