On November 2nd a jobs report will be released — just a few days before the Presidential election. The monthly jobs report released by the Bureau of Labor Statistics (BLS) has long been the object of intense focus among denizens of the Beltway and Wall Street due to the impact that the numbers contained therein have on politics, business and the economy. The financial crisis of 2008 and the resultant massive job losses have only increased the scrutiny with which the report is examined. That scrutiny is certainly justified; the jobs report is the most critical monthly release of data related to the economy and second only to the quarterly release of the gross domestic product (GDP) numbers in overall economic importance. Recently, however, some of that scrutiny has gone astray as critics of the current administration, in particular Jack Welch, former CEO of GE, have intimated that the jobs report and the numbers it contains were being manipulated by the for political ends. Although his accusation was quickly showered with derision from both sides of the political spectrum, a quick examination of why the data contained in the jobs report is so difficult to manipulate is critical to assure people of the nonpartisan nature of both the report and those involved in the process of generating the report at the BLS.
The jobs report is the result of two surveys conducted by the BLS, which is an independent agency staffed by career government technocrats using long established statistical methods. The BLS goes to extraordinary lengths to make sure that the data and processes employed in these surveys are shielded from political influence. There is only one political appointee at the agency, the commissioner that heads the bureau. Until recently this was Bush ’43 era appointee Keith Hall, who served as the BLS head from 2008 to 2012. Currently the agency is without a commissioner as Congress has yet to approve Obama’s choice for the post. This would seem to make it difficult for the president and his people to manipulate the numbers given that they lack anyone at the bureau to facilitate such an effort. Indeed, even Mr. Hall, a Republican appointee, defended the September jobs report and the associated processes used to generate it, as he was quoted in an interview as saying that there is no way that someone at the agency could change any of the data from the two employment surveys. “There’s nothing wrong with the numbers,” said Mr. Hall. “The only issue is the interpretation of the numbers. The numbers are what they are.”
An article written by the Washington Post earlier this year detailing the great efforts made by the BLS to protect the integrity of the data used in the two survey reports they issue each month seems to bear this conclusion out. The raw data is confidential. There is an eight day lockdown period on the employees generating the reports prior to their release. Confidentiality agreements must be signed every day of the lockdown period by all participants. All computers used in the process are encrypted and any pertinent data must be locked in a safe even when a worker does something as mundane as go to the bathroom. No one without security clearance is allowed on the floor at the BLS where the data is compiled. Once the surveys are complete, only a select few are given a chance to view them a scant twelve hours before their release with these individuals including the president, the vice president, a handful of economic advisors and the secretary of labor. All information critical to our country’s well being should be as well guarded and cared for as the economic data used by the BLS in its surveys.
Mr. Welch’s claims were, for the most part, derided by critics on both the left and the right and few took his assertions seriously beyond the brief and predictable spate of Twitter and blog posts that occurred in the immediate aftermath of his tweet. The data, it was concluded by those on both sides of the political divide, could not be manipulated by anyone for political gain. The controversy did, however, raise an interesting if somewhat tangential point: how accurate are the processes used in the two surveys and how useful is the data they generate? There are some who assert that the BLS methods employed are, at best, flawed, and should be used with caution when analyzing the economy. These claims are perhaps more plausible and should not be dismissed out of hand.
The two studies related to employment released each month by the BLS are the Household Survey and the Establishment Survey. Each is the result of scientifically random samples meant to represent all U.S. residents and businesses. The unemployment rate is derived from the Household Survey, which uses a poll of approximately 60,000 households as its data source. This sample of households is small compared to the total number of households in the country and as a result the index is notoriously volatile. There are some acknowledged flaws with how the survey is constructed. The unemployment rate that the media and the public pay the most attention to is officially known as the U-3. It is, however, only one of six such indexes that the Household Survey generates. The U-3 does not count people who have given up on finding work, known as discouraged workers, as part of the labor force, nor does it count those who have been unemployed for over 52 weeks, regardless of whether they are continuing on the quest to find employment or not. This latter group is referred to in BLS speak as marginally attached workers. Additionally the U-3 does not differentiate between people who are employed part-time or full time; if you work one hour a week you’re considered employed by the U-3 index. This makes the labor force total suspect at best. When the complicated, controversial and opaque process of “seasonally adjusting” the numbers is factored in, the index that is the end result is deeply flawed and should not be used by the media to inform the general public. Unfortunately the media frequently ignores this point for the sake of expediency which leads to the least accurate metric being the one most people have the greatest level of familiarity with. A second index, called the U-6, does account for discouraged workers as part of the labor force as well as differentiating between those employed part-time and those working full time, with the result that it is often referred to as the “underemployment” rate. This rate is used sparingly by the media leaving many unaware of any other number that can be referenced beyond the U-3 which, regrettably, leaves the public uninformed as to the true nature of the jobs market and by proxy the health of the economy as a whole.
The second survey, known as the Establishment Survey, is the index from which the payroll number, representing the number of nonfarm jobs added to the economy, is derived. This study collects data from 141,000 businesses and government agencies which accounts for roughly 1/3rd of all nonfarm payroll employees in the country. This significantly greater sample size, as compared to the Household Survey, results in a much less volatile and thus more accurate report. However, it too has some significant flaws. The same “seasonally adjusted” methods are applied to these numbers as they are to the Household Survey, with the significant caveat that, unlike the Household Survey, the methodology used is publicly released as are the numbers that the methodology generates. Of greater concern is the decision made by the BLS several decades ago to add 50,000 jobs to the total each month in an effort to account for chronic under-reporting of jobs added to the economy. This decision may have made sense in the boom times of years past, but is probably not improving the accuracy of the index in current context.
The two employment figures that the public is most familiar with, the nonfarm payroll number and the U-3 unemployment rate, are generated from two different surveys of very different compositions, yet are presented in tandem by the media, which can lead to legitimate confusion. The processes used by the BLS in constructing the surveys have known flaws, which leads to data that isn’t as useful as it could be in providing a window of analysis into the economy. Neither the confusing nor flawed nature of these reports is anything new, however, as the core of how the BLS conducts its surveys goes back decades and in some cases generations. Additionally, the general consensus is that it’s nearly impossible for anyone to manipulate the data to their own ends. Jack Welch was one of the most successful CEOs in American corporate history and helped build one of the most powerful international conglomerates the world has ever seen. Jack Welch is perfectly aware of how extremely difficult it would be for anyone to skew the jobs data for any reason. His tweet, therefore, can be viewed as nothing more than a deliberate attempt to poison the public’s faith in the nonpartisan nature of the jobs report. It was irresponsible and malicious and he was justifiably vilified for it.
There are legitimate concerns regarding how the BLS collects its data and constructs its surveys, but these are related to flawed methodology, not deliberate manipulation. The methodologies need to be addressed, but only internally by the BLS, and not through pressure placed upon the bureau by any administration. The agency has shown itself slow to evolve, but perhaps the harsh light cast upon the jobs numbers it generates by this manufactured controversy will yield something positive in the form of more accurate methods being developed in an effort to produce data that can be more usefully employed to assess the economy.