Workers Displaced from 2007-2011 have Struggled to Find Employment

Every two years the Bureau of Labor Statistics (BLS) surveys workers who were displaced from “long-tenured” jobs in the previous three years.  A “long-tenured” job is one which lasted at least three years prior to the job “displacement.”  A job “displacement” is a separation that occurred because:  “a plant or company closed or moved, there was insufficient work” or the “position or shift was abolished.”  Generally these longer term jobs are better jobs and “displacements” are job separations for economic reasons that occur disproportionately in recessions.  The BLS survey is conducted in January or February of even-numbered years.  The most recent data from the January 2012 survey indicates that the number of displaced workers was much higher during the 2007-2009 recession and the rate at which workers found new jobs during the recovery has been much lower than after the recession of 2001.  However the earnings received by displaced workers who were re-employed at the time of the BLS survey are similar for the recessions of 2001 and 2007-2009.

The following bar chart indicates the number of workers displaced from “long-tenured” jobs per year, over the previous three years, in BLS surveys from 1994 to 2012.  The line graph above the bar chart indicates the re-employment rates for these workers as of the survey.  On average the survey occurs about 18 months after the typical worker lost their job.  The 2010 and 2012 surveys include workers displaced in the 2007-2009 recession while the 2002 and 2004 survey includes workers displaced in the 2001 recession.

A comparison across surveys reveals that:

• About 40% more workers per year were displaced from “long-tenured” jobs between 2007 and 2011 compared to the 1999-2003 period.
• Workers displaced from “long-tenured” jobs from 2007 to 2011 were 18% less likely to find re-employment within the next 18 months compared to workers displaced from 1999 to 2003.

The next bar chart compares the earnings of displaced workers who were able to find a full-time job after displacement to their previous earnings, based on BLS surveys from January 1996 to January 2012.  The red bar measures the fraction of re-employed full-time workers who are paid at least 20% less than previous earnings.  The green bar measures the fraction of full-time re-employed workers who are paid least as much as they earned previously.

A comparison across surveys reveals that:

• Workers displaced from “long-tenured” jobs between 2007 and 2011, who found another full-time job, were slightly more likely (8.5%) to experience a pay decrease of 20% or more compared to workers displaced from 1999 to 2003.
• Workers displaced from “long-tenured” jobs from 2007 to 2011, who found another full-time job, were no more likely to receive a pay decrease of any kind compared to workers displaced from 1999 to 2003.

The recession of 2007-2009 was the deepest downturn since World War II.  The recovery since 2009 has been tepid.  This is reflected in both the number of workers displaced from jobs they held for at least three years and the low rate at which these workers found jobs during the recovery.  There are much smaller differences in the relative earnings of workers who found full-time work in this recovery compared to displaced workers after the 2001 recession.  The most troubling empirical finding is that between 2007 and 2011 only about half of workers displaced from jobs they held for at least three years were employed within the next 18 months.

Stimulating Reading: How Ezra Klein’s Wonkblog is Helping the Romney Campaign

Earlier this week Ezra Klein’s Wonkblog in the Washington Post published an article titled “The Romney campaign says stimulus doesn’t work.  Here are the studies they left out.”  The article criticizes a Romney campaign paper by Glenn Hubbard, Greg Mankiw, John Taylor and Kevin Hassett for its selective references of empirical studies of the Obama Administration’s American Recovery and Reinvestment Act (ARRA).  The Wonkblog article, posted by Dylan Matthews, provides a list of 15 papers (most of which have not yet been peer reviewed) and claims that 12 of them find that the ARRA “worked.”

I decided to read a few of these studies to see how much empirical support there was, in the academic community, for the idea that the ARRA was effective.  I read four of the papers, and so far, I am underwhelmed by the empirical evidence.

I read three of the papers listed by Wonkblog as providing evidence that the Obama stimulus worked.  The first, by James Feyrer and Bruce Sacerdote estimated the impact of the ARRA on jobs at the state and local level.  They find that the ARRA had some impact on jobs but estimated that the cost of each job created was between $170,000 and $400,000.  In addition, they find that different types of spending have different effects on jobs.  For example they report that spending “to fund teachers and police have if anything a negative impact.”  In general, their results are supportive of the study by John Cogan and John Taylor that found that state and local governments used Federal stimulus funding to reduce borrowing rather than increase hiring.

Feyrer and Sacerdote cleverly control for the statistical problem that stimulus spending is generally larger in places where the economy is worse through a technique called “instrumental variables.”  Their study validates what many ARRA critics have asserted: the ARRA was in large part pork barrel spending.  They find that states where members of Congress have longer tenure in Washington received more stimulus spending.  Their study uses this variation in pork barrel spending to estimate the effectiveness of the ARRA.

A second paper listed by Wonkblog under the “stimulus works” category is “Measuring Tax Multipliers: The Narrative Method in Fiscal VARs” by Carlo Favero and Francesco Giavazzi.  This paper has nothing to say about the effectiveness of the ARRA; it relies on data through the first quarter of 2007.  (The purpose of the article was to reconcile two different methods for estimating the impact of tax cuts or tax increases on economic activity.)

A third paper listed as supportive of the ARRA is “Fiscal Policy Multipliers on Subnational Government Spending” by Jeffrey Clemens and Stephen Miran.  Again this paper has nothing to say about the effectiveness of the ARRA; it relies on data through 2004.  It is fair to say, however, that the authors find low multiplier effects of government spending programs, perhaps due to crowding out of private expenditures.  They state:

“our relatively low multiplier estimates (based on deficit-financed government spending) can be reconciled with the larger estimates in several recent studies (based on windfall-financed government spending) if government debt crowds out current private consumption and investment. We view these contrasting, but not contradictory, results as evidence of an important role for neoclassical considerations.”

Finally “The American Recovery and Reinvestment Act: Public Sector Jobs Saved, Private Sector Jobs Forestalled” by Timothy Conley and Bill Dupor finds that the ARRA had a slight positive effect on public sector employment and a slight negative effect on private sector employment.  They report that the largest declines in private sector employment were in health, education, and private business (HELP) services.  In particular they report that government employment increased by between zero and 900,000 jobs and private HELP service employment decreased by 166,000 to 1,378,000 jobs.  They report that ARRA had a negative, but insignificant, effect on private sector jobs in goods-producing industries.  Taken together, Conley and Dupor find the ARRA had no statistically significant positive impact on overall employment, and may have actually reduced employment.

 Dylan Matthews and the editors at Wonkblog apparently don’t understand what it means for a study to find no significant effect of a policy.  The Wonkblog article states “The biggest problem with the Conley and Dupor study is that their estimates are not statistically significant.”  What is true is that any study of the ARRA will be fairly imprecise and have limited statistical power because of a small sample size and the methods required so that estimates don’t obviously suffer from endogeneity bias.  Given the statistical precision of their study, Conley and Dupor could have detected a statistically significant increase in jobs in service-producing industries as small as 1.5%, if such an increase had occurred.  Conley and Dupor most likely found no positive effect of the ARRA on private sector employment because the stimulus was either ineffective or had such a small effect that it could not reliably be measured.

I have more reading to do, but my initial reaction is: two studies had nothing to say about the effectiveness of ARRA, one study found no positive effect on employment at all, and the other gave mixed results.  Even the study with mixed results found that the ARRA cost hundreds of thousands of dollars per job created and that the stimulus resembled political pork: spending was higher where Representatives had more tenure in Congress.  My primary conclusion is that Wonkblog has helped the Romney campaign by calling more attention to academic studies questioning the efficacy of the Obama Administration’s economic policies.

The Welch Consulting Employment Index Continues Its Sharp Decline

The Welch Consulting Employment Index fell to 93.3 in July 2012.  An index value of 93.3 means that full-time equivalent employment (from the BLS household survey) is 6.7% below its level in the base year of 1997, after adjusting for both population growth and changes in the age distribution of the labor force.  The index is down sharply over the past three months; the index was 94.7 in April 2012.  The index remains below its level three years ago and is only 1.2% above its historic low reached exactly one year ago.  The index remains weak because part-time employment is at historically high levels and employment growth has barely exceeded population growth over the past year.

The Welch Consulting Employment Index, disaggregated by gender, shows that the labor market recovery has been much weaker for women than men since 2009.  The index for men is 91.2 for July 2012, up 1.9% over the past twelve months, and up 1.2% since July 2009.  The index for women is 96.7 for July 2012, up 0.5% over the past twelve months, but down 1.8% over the past three years.  Finally, since President Obama took office in January 2009, the employment indices are down 2.8% for men and down 4.9% for women.

Technical Note: Full-time equivalent employment equals full-time employment plus one half of part-time employment from the BLS household survey.  The Welch index adjusts for the changing age distribution of the population by fixing the age distribution of adults to the 1997 base year.  The Welch Index adjusts for population growth by fixing total population to its 1997 level.  Seasonal effects are removed in a regression framework using monthly indicator variables.

Be Wary of the Seasonal Adjustment in the July Jobs Report

The Bureau of Labor Statistics (BLS) will almost surely report that payroll employment declined by over one million jobs on Friday, but it will all be erased by a procedure known as seasonal adjustment.  The headline number for job creation reports the change in jobs after making statistical adjustments that attempt to eliminate the employment fluctuations due to weather and other seasonal factors.  The July jobs report will have the second biggest seasonal adjustment (after December to January) of the year.  July is typically a much weaker month for payroll employment than June, because of seasonal factors.  Whether the headline number for job creation exceeds expectations or is viewed as disappointing may have more to do with the non-partisan statisticians at the BLS than the Federal Reserve, the Congress, or the Obama Administration.

Over the past decade employment fell between June and July by 1.33 million jobs, on average.  In contrast, seasonally adjusted payroll growth between June and July has been reported as an increase of 300, on average.  (That is 300 not 300 thousand jobs.)  This means the BLS has consistently adjusted a decline of 1.33 million jobs from June to July to be reported as no change at all, seasonally adjusted.  Put somewhat differently, even if there are no raw employment gains in Friday’s report, BLS statisticians will conclude that this is equivalent to the economy creating over 1.33 million jobs in a single month, after seasonal adjustment.

Last July employment was 1.3 million lower than it was in June, but that translated to an increase of 96,000 jobs after applying the BLS seasonal adjustment factor.  Although the BLS allows seasonal adjustment factors to evolve over time to reflect changing economic conditions, the June to July adjustment has been fairly stable over the last decade.  Nonetheless, even the slight difference between using the 2010 and 2011 seasonal adjustment factors for June to July is equivalent to a difference in 111,000 jobs for the headline payroll employment report on Friday.

Over the first three months of 2012 payroll employment was 1.56% higher each month, on average, than the previous year.  Over the past three months payroll employment was 1.34% higher each month, on average, than the previous year.  The labor market recovery has slowed.  Every economist and politician is looking to Friday’s report for an indication of whether the recent trend is likely to continue.

One monthly BLS employment report is noisy, subject to substantial seasonal adjustment, and should be interpreted with caution.  Sophisticated observers of Friday’s July report will be looking at both the seasonally adjusted headline number, as well as seasonally unadjusted figures (compared to July 2011), and revisions to the May and June reports, before reaching any conclusions about the direction of the labor market.

The Unemployment Rate is 8.216%

The BLS announced today that 80,000 jobs were created in the U.S. in June, compared to 77,000 in May and 68,000 in April.  Despite the small apparent increase in payroll employment gains from April to June, there is no meaningful difference in the number of jobs created between April, May and June; their difference is well within the statistical margin of error.  Payroll employment gains are reported to a level of accuracy far beyond our ability to count all employees in the U.S. in real-time.  The difference between an employment gain of 80,000 and a consensus forecast of 125,000 is still far below the margin of error for estimated job growth.  While sophisticated labor market observers recognize this fact, the casual consumer of economic news should be more concerned about longer term labor market trends than a single month’s report.

The BLS reports the number of jobs created to the nearest 1,000, even though the standard error of each monthly estimate is approximately 60,000 jobs.  This means that the margin of error for each month’s payroll employment gain is about 100,000.  For non-statisticians, labor economists are only 90% certain that job growth in June was within an interval from a loss of 20,000 jobs to a gain of 180,000 jobs.

The unemployment rate is reported to the nearest one tenth of a percentage point, while the standard error of the unemployment rate estimate is about .12 percentage points.  In other words the BLS reports the unemployment rate to the nearest 0.1 while the margin of error is approximately 0.2.

If the BLS were to report payroll employment similarly to the unemployment rate, payroll employment would be reported to the nearest 50,000 jobs not the nearest 1,000.  If the BLS reported payroll employment rounded to the nearest 50,000 jobs, those without a statistics background would recognize how noisy monthly employment estimates can be.

Put somewhat differently, if the BLS instead reported the unemployment rate to the same level of inaccuracy as payroll employment, it would report the unemployment rate to the nearest two one-thousandths of one percent.  If that were the case the BLS would have reported that the unemployment rate ticked up from 8.206% in May to 8.216% in June.

The labor market recovery is very weak because over three months, where the margin of error is much smaller, employment gains totaled 225,000 jobs.  If the labor market were healthy, the economy would create more than 225,000 jobs per month not per quarter year.  During the recovery in 1983-1984, there were 15 straight months where job growth exceeded 225,000 per month.  We are on a long slow road back to full employment.

Welch Consulting Employment Index for June 2012

The Welch Consulting Employment Index is 93.8 for June 2012.  The index is down sharply over the past three months, it was 94.9 in March, and fell slightly below its value in June 2009.  The index is up 1.5% from June 2011 (seasonally adjusted).  An index value of 93.8 means that full-time equivalent employment (from the BLS household survey) is 6.2% below its level in the base year of 1997, after adjusting for both population growth and changes in the age distribution of the labor force.  The index is up about 1.8% from its trough in July 2011.  

The Welch Consulting Employment Index, disaggregated by gender, shows that the labor market recovery has been weaker for women than men over the past three years.  Of course, men lost more jobs than women during the first year of the recession and therefore had more ground to make up in the past three years.  The index for men is 91.6 for June 2012, up 1.9% over the past twelve months, and up 1.1% in the past three years.  The index for women is 96.7 for June 2012, up 0.9% over the past twelve months, but down 0.5% over the past three years.  Finally, since President Obama took office in January 2009, the employment indices are down 2.3% for men and down 4.1% for women.

Technical Note: Full-time equivalent employment equals full-time employment plus one half of part-time employment from the BLS household survey.  The Welch index adjusts for the changing age distribution of the population by fixing the age distribution of adults to the 1997 base year.  The Welch Index adjusts for population growth by fixing total population to its 1997 level.  Seasonal effects are removed in a regression framework using monthly indicator variables.

Outlook for June Jobs Report

If tomorrow’s jobs report for June looks like the June 2011 report, we will see payroll job growth of 132,000.  If the labor market report matches the average growth in the decade prior to the 2008-2009 recession payroll employment would increase by 193,000.  Even this more rapid growth would correspond to an annual employment growth rate of 1.75%, barely enough jobs to keep pace with population growth and the rising participation rates of women and older workers.

Labor market observers should pay attention to the growth in full-time employment between May and June (from the household survey) for adults age 20-24, many of whom are just finishing school.  From 1998 to 2007 the full-time employment of workers in this age group increased by 9%, on average, between May and June.  Over the past four years, as new graduates have struggled to find jobs, full-time employment grew by an average of only 6.3% between May and June for adults age 20-24.

56.5% of the Job Losses Since 2009 Occurred for Women

Last month the Romney campaign cited Bureau of Labor Statistics (BLS) data to report that 92% of all job losses since President Obama took office in January of 2009 were suffered by women.  The media reaction to the Romney campaign’s claim was mixed, but most analysts did not question the accuracy of the underlying data.  Some analysts opined that it was expected that most of the job losses in the past three years occurred for women because government employment has contracted slightly since January 2009.  My blog post from April showed that 95% of the jobs eliminated at the U.S. Postal Service were held by women.  The problem with all of these analyses is that they are based on faulty data.

The BLS announced on its website yesterday that:

Estimates of women employees in the U.S. Postal Service and some related series from the Current Employment Statistics survey were temporarily removed from the BLS data-retrieval system on May 14, 2012. BLS staff discovered data-processing errors that occurred during the November 2009–April 2012 period and resulted in an incorrect ratio of women employees to all employees. Correcting these errors will increase the number of women employees but does not affect total employment levels. Series of women employees were removed for the U.S. Postal Service, federal government, government, service-providing, and total nonfarm.

Although I am not sure about how the BLS discovered their error, I believe I have an explanation.  Alan Robinson of the Direct Communications Group (@CEP_Observer) didn’t believe the numbers in my April blog post (three weeks ago) about women’s job losses at the US Postal Service.  After I sent him the data I used and a link to the BLS website, Alan still didn’t believe the data.  His inquiries to the BLS caused their economists and statisticians to take a closer look at the data, which uncovered the apparent errors.

The BLS is assigned an incredibly difficult task, and generally produces extremely reliable and valuable data.  This time they made a mistake, and are working to correct the problem.  Until the establishment employment data for women are updated, our best information on the gender composition of job losses comes from the household survey, also administered by the BLS and the Census Bureau.  The household survey shows that the fraction of the adult population that was employed:

• Declined from 66.2% to 64.3% for men, between January 2009 and April 2012
• Declined from 55.3% to 53.0% for women, between January 2009 and April 2012

Women comprise about 52% of the working age (civilian) population.  The 2.3% decline in employment relative to population for women means that 2.89 million fewer women were employed in April 2012 than would have occurred if the employment to population ratio for women had remained steady since January 2009.  Similarly, there are 2.24 million fewer men employed in April 2012 than would have occurred if the employment to population ratio for men had remained constant since January 2009.

The calculations above indicate that the best estimate is that 56.5% of the relative employment declines since January 2009 were suffered by women.  Next month the BLS will post updated data on the gender composition of employment based on the establishment survey.  There are many reasons why the household and establishment survey data will look somewhat different, but it is likely that the revised establishment numbers will mirror the household data and show that the majority (but far less than 92%) of the jobs lost since January 2009 were jobs held by women.

Durable Goods Manufacturers Struggle to Find Qualified Applicants

The typical job vacancy in durable goods manufacturing remains unfilled longer, on average, than at any time in the past decade.  This suggests that durable goods manufacturers struggle to find qualified applicants for their job openings as the economy slowly recovers from the deep recession of 2008 and 2009.  In contrast the average construction job vacancy remains unfilled half as long as it did during the construction boom because there are many qualified jobless workers for each job opening.  Construction employment fell by almost 30% between 2007 and 2010, and has not grown since January 2010.

The BLS JOLTS data report job openings at the end of each month and the number of persons hired per month, by major sector of the economy.  For example, there have been an average of 194,000 job openings in durable goods manufacturing at the end of each month, and 157,000 workers hired per month over the past three months.  This implies that the average job vacancy remains unfilled for about 38 days.  In contrast, at the depth of the recession in the fall of 2009 the average durable goods job vacancy remained unfilled for less than two weeks.  The following figure shows the average number of days a job vacancy remained unfilled over the past decade in three sectors impacted by the recession: construction, financial services and durable goods manufacturing.

 

The average job vacancy in financial services now remains unfilled for about 40 days compared to 25 days in the fall of 2009.  The financial services labor market had the least slack in the summer of 2006, when the typical job vacancy remained unfilled for about 7 weeks.

It was most difficult to hire a construction worker in the summer of 2007 when the average job vacancy remained unfilled for about 15 days.  In contrast, in the early fall of 2009, the average construction job opening remained unfilled for less than 5 days.

The average job vacancy in financial services now remains unfilled for about 40 days compared to 25 days in the fall of 2009.  The financial services labor market had the least slack in the summer of 2006, when the typical job vacancy remained unfilled for about 7 weeks.

It was most difficult to hire a construction worker in the summer of 2007 when the average job vacancy remained unfilled for about 15 days.  In contrast, in the early fall of 2009, the average construction job opening remained unfilled for less than 5 days.

Welch Consulting Employment Index Up Slightly in Past Year

The Welch Consulting Employment Index is 94.7 for April 2012, up 1.0% from April 2011 (seasonally adjusted).  An index value of 94.7 means that full-time equivalent employment (from the BLS household survey) is 5.3% below its level in the base year of 1997, after adjusting for both population growth and changes in the age distribution of the labor force.  The average annual change in the index over the previous three months was 0.9%.  Thus the increase of 1.0% from April 2011 to April 2012 means there has been virtually no change in the rate of full-time equivalent employment growth over the first four months of 2012.  Employment continues a slow growth trajectory.

In future months Welch Consulting will present more detailed analyses of changes in full-time equivalent employment by age, gender, race and educational attainment in addition to the overall Welch Consulting Employment Index. 

Technical Note: Full-time equivalent employment equals full-time employment plus one half of part-time employment from the BLS household survey.  The Welch Consulting Employment Index adjusts for the changing age distribution of the population by fixing the age distribution of adults to the 1997 base year.  The Index also adjusts for population growth by fixing total population to its 1997 level.  Seasonal effects are removed in a regression framework using monthly indicator variables.