Category Archives: Employment

Job growth in the Seventh District in 2016—A post-benchmarking update

Here at the Chicago Fed, we closely track one of the most important regional economic indicators, the U.S. Bureau of Labor Statistics’ (BLS) payroll employment survey (also known as the Current Employment Statistics, or CES). The survey is important because it provides a good picture of the overall state of an economy and its initial results are released quickly (unlike some other regional data that are released with a lag of a month or more). Unfortunately, the relatively quick turnaround also means we must exercise caution. While we get an initial estimate only three weeks after the end of the reference month, the estimate can sometimes be revised substantially later on. The BLS makes minor revisions to the data one month after the initial release and major revisions once a year, when the survey is benchmarked to the unemployment insurance census (released as the Quarterly Census of Employment and Wages, or QCEW).[1]

The BLS recently released newly benchmarked data, so we now have an update on how well the Seventh Federal Reserve District[2] did in 2016. Table 1 shows that, for the most part, the job growth numbers that the BLS had been reporting before the most recent benchmarking held true. Overall, Seventh District employment grew at a pace of about 1 percent in 2016, with the BLS reporting a 0.9% increase before benchmarking and a 1.1% increase after. In addition, the most recent benchmarking further affirmed the division in performance between the eastern and western parts of the District over the past few years. Job growth for Indiana and Michigan in 2016 was revised up, whereas job growth for Illinois, Iowa, and Wisconsin was revised down. So now, the job growth numbers for Michigan and Indiana in 2016 easily exceeded the numbers for Illinois, Iowa, and to a lesser extent, Wisconsin.


The most recent benchmarking by the BLS also gives us the opportunity to see how well we did at predicting the benchmark revisions using a method known as “early benchmarking.”[3] The main idea behind early benchmarking is that it’s possible to use the QCEW data, which are released quarterly, to predict the major annual revisions to the job growth numbers.

This year we found that, unlike last year, the early benchmarked estimates (which we produced in January 2017) were further from the newly released job growth numbers than the March 2016 benchmarked CES estimates were. Our only consolation is that the early benchmarked estimates did get closer to the newly released numbers for the first half of 2016. I’ll explain what happened a little later.

Let’s look first at how the early benchmarking procedure did for the District as a whole. Figure 1 shows three versions of the Seventh District’s CES data, where the solid portion of a line represents data that have been benchmarked using the QCEW and the dashed portion represents data that have not.[4] The black line is the version released in March 2017 (with benchmarked data through September 2016), the red line is the early benchmarked version that we calculated in January 2017 (with benchmarked data through June 2016), and the blue line is the version released in March 2016 (with benchmarked data through September 2015). Surprisingly, in spite of nine months of additional benchmarked data, the March 2016 benchmarked CES appears to track the newly released data better than the early benchmarked CES.


The first row of table 2 summarizes figure 1 in terms of December-to-December job growth for 2016. For this period, the early benchmarking method performed notably worse: It underestimated District job growth by 83,000, while the March 2016 benchmarked data underestimated job growth by only 25,000. The remaining rows of table 2 summarize 2016 job growth for District states (figures plotting the data for District states are at the end of the post). This year, Indiana was the only state for which the early benchmarked growth estimates had a smaller error than the March 2016 benchmarked data, though the errors for Michigan were very close. The early benchmarking method did particularly poorly for Illinois, estimating that employment fell by 13,000 for the year when it actually grew by 19,000.


The poor performance of the early benchmarking method was surprising. We expected the early benchmarking method to do better for the first half of 2016 because it could take full advantage of the available QCEW data. And for the second half of the year, we knew by definition that the method would perform the same as the March 2016 benchmarked data because it used that version’s growth numbers. So, given these half-year results, how could the early benchmarking method do worse for the full year? Table 3 shows how. The early benchmarking method’s errors were negative in both halves of the year, while the errors of the March 2016 benchmarked CES were positive in the first half and negative in the second half. So the early benchmarking method’s errors built on each other, whereas the errors in the March 2016 benchmarked data cancelled each other out.


Even though the early benchmarking method did not do well this year, we still think it is a useful tool for predicting benchmark revisions because it can take full advantage of the available QCEW data. We will see what happens in 2018.






[1] For more information on the BLS’s benchmarking process, go here.

[2] The Seventh District, which is served by the Chicago Fed, comprises all of Iowa and most of Illinois, Indiana, Michigan, and Wisconsin.

[3] An earlier post discussing the method in detail is here.

[4] As is clear in figure 1, the BLS also revises already benchmarked data, though the revisions are typically small.

Our First Look at Job Growth in the Seventh District in 2016—New Estimates Using Early Benchmarking

Last week we received the December 2016 report from the U.S. Bureau of Labor Statistics’ (BLS) state payroll employment survey (also known as the Current Employment Statistics, or CES), so it’s our first opportunity to look at how well the Seventh Federal Reserve District[1] did in 2016. The recent report is not the final word on job growth in 2016 because the data will eventually be benchmarked against more complete data, primarily those from the Quarterly Census of Employment and Wages (QCEW).[2] Data for January through September 2016 will be benchmarked by the BLS in the middle of March of this year, while data for October through December 2016 will not be benchmarked by the BLS until March of next year.

In June 2015, I wrote a blog post detailing a method called early benchmarking, which predicts how the BLS will revise the CES data (the method was first introduced by our colleagues at the Dallas Fed). The BLS rebenchmarks the CES using QCEW data only once a year. However, QCEW data are released quarterly, so it’s possible to use the QCEW data to predict how the BLS will revise the CES (this process is explained in detail in my earlier post). The benchmark revisions to the CES can be quite large, and last year, I found that for the District and most District states, the early benchmarked jobs numbers were closer to the final benchmarked numbers than the non-benchmarked numbers were.

Table 1 shows that for 2016, the early benchmark procedure is predicting employment in the District grew by 106,000 rather than the 164,000 that the BLS’s current estimates indicate. This difference is largely the result of lower job growth numbers for Illinois and Wisconsin, though Iowa’s job growth number is also lower. The early benchmark procedure also suggests that Indiana’s job growth will be revised up, and Michigan’s will be unchanged.


Figures 1 through 6 show the employment series as currently published by the BLS (in blue) and the early benchmarked series (in red) for the District and District states. The dashed portions of the series represent data that have not yet been benchmarked using the QCEW. The lines are identical through September 2015, but then the lines follow different paths because the early benchmarked series use growth rates from the QCEW until June 2016. While their levels differ, both series have the same growth rates starting in July 2016 (again, for more details on the early benchmarking procedure, see my earlier post).

After the BLS releases newly benchmarked data in March, I will review how well the early benchmarking procedure performed at predicting job growth in the District.







[1] The Seventh District comprises all of Iowa and most of Illinois, Indiana, Michigan, and Wisconsin.

[2] For more information on the BLS’s benchmarking process, go here.

Where are the Jobs? Midwest Jobs Recovery by State and Sector

The Great Recession, which lasted from December 2007 through June 2009, had a profound effect on U.S. employment. At its height, U.S. unemployment reached 9.5% and job losses totaled approximately 7.3 million. The current recovery is now entering its eighth year, so it may be a good time to take a look at how many jobs lost during the recession have been recovered as one measure of the nature of the labor recovery. In this blog, we examine the pattern of recovery for all five Seventh District states (Iowa, Illinois, Indiana, Michigan, and Wisconsin) in four industry sectors that were heavily impacted by the recession—construction, finance (finance, insurance and real estate or FIRE), manufacturing, and professional and business services.

Figure 1 provides evidence of the job recovery through September 2016 across industries for the U.S. and the five Seventh District states. The first column for each state shows the number of jobs lost in that sector during the Great Recession (in thousands). The second column shows the number of jobs added during the current recovery (in thousands). And the third column shows the recovery rate (jobs added divided by jobs lost in percent). To provide some perspective on the concentration of each industry’s employment within the state relative to the U.S. total, the fourth column shows an employment location quotient. For example, if you look at Illinois, employment in professional and business services is 1.10 (or 10%) above the U.S. total.


The first thing to notice is that there is considerable variation in job loss and recovery both within sectors and across states. Two industries, health and education and government, actually added jobs during the recession (except in Michigan, where government employment decreased). However, while job gains have been sustained for the health and education sector, government has actually contracted during the recovery. At the other end of the spectrum, job losses were largest in manufacturing, construction (mining, logging, and construction), and transportation (trade, transportation, and utilities). While transportation has seen a strong recovery (143.6% nationally), the job recovery in manufacturing is only 29.7%, and in construction the figure is 48.1%. On a state level, total non-farm employment recovery ranges from a low of 114.2% to 240.9%. Based solely on this measure, the best job recovery rate among Seventh District states would appear to be in Iowa and the worst in Illinois.

Iowa has consistently outperformed the region and the U.S. in job recovery, both in aggregate and by sector, since the end of the Great Recession (figure 1). A clear contributor was the role of agricultural production and rising farm incomes that the state experienced up until 2015. Currently, low crop prices are weighing on farm incomes, so it may be that this pattern of growth is starting to flatten out. As figure 2 shows, over the longer time horizon, Iowa tracked the rate of U.S. job growth coming out of recession but shows a noticeable divergence starting in 2014.


Michigan entered the Great Recession significantly weaker than other states in the region and the U.S. as a whole, having continued to lose jobs and suffer economic decline through the recovery from the 2000–01 recession. However, since the end of the Great Recession, the state’s job recovery has accelerated. Particularly noticeable is that Michigan has almost fully recovered its lost manufacturing jobs (98.2%), far exceeding the U.S. (29.7%) and significantly better than Indiana, the most manufacturing-intensive state in the region (73%). Clearly, the auto recovery has been a boon to the state.

Illinois has clearly underperformed both the region and the nation in post Great Recession job recovery, both in aggregate and across industry sectors. While Illinois has more than recovered the total number of jobs it lost during the recession (114.2%), only two sectors of the state’s economy have significantly gained jobs. The leisure and hospitality sector is up 482.3% and the professional and business services sector is up 191.5%. While wages in professional and business service jobs are relatively good, the gains in leisure and hospitality are likely to be lower paying jobs. Now, we look at job performance since January 2000 in four key industries—construction, finance, manufacturing, and professional and business services.


As we see in figure 3, construction employment in the Seventh District states has underperformed the U.S. with the clear exception of Iowa, which has seen a gain of 40% over the 2000 level. U.S. construction employment has returned to its 2000 level, while the remaining four Seventh District states have yet to reach their January 2000 level, with Michigan particularly hard hit—off roughly 25% over the period. The loss in jobs was particularly sharp in 2009 through 2012. Since then, jobs have gradually but steadily come back.


Finance, Insurance and Real Estate

Like construction, finance experienced a boom running up to the Great Recession, followed by a sharp contraction during the recession. Figure 4 indicates that the employment declines in this sector were most pronounced in Illinois, Indiana, and Michigan.  Recovery rates are now above the 2000 level for the U.S., Wisconsin, Iowa, and Michigan. Where recovery has clearly lagged is in Illinois and Indiana.



Manufacturing has been seen as a relatively good news story during the recent recovery. Particularly in the automobile industry, production levels have been consistently high. In addition, export-led capital goods had strong sales growth until the past couple of years. The question is has the rebound in output been reflected in employment? Figure 5 shows that manufacturing continues to follow a stair-step path when it comes to employment. Since 2000, after a shock, employment steps down; during recovery it tends to level off and not fully recover. Manufacturing continues to see a trend toward higher production with fewer labor inputs. For example, even in Michigan, with record-level auto production, manufacturing employment is still 30% below its 2000 level. Similarly in Indiana, the most manufacturing-intensive state in the Seventh District, employment is still roughly 25% below the 2000 level.


Professional and Business Services

The sturdiest employment performance emerging from the recession has been in professional and business services. Figure 6 shows that all five Seventh District states (and the U.S.) now have professional and business service employment levels above the 2000 level. Admittedly, the sector was not hit as hard by the recession as many others. Indeed, in many cases employment levels remained above the 2000 level during the worst months of the recession. The clear laggard is Michigan. As the figure shows, Michigan was shedding professional service jobs during the expansionary period emerging from the 2000 recession and then faced an even sharper downturn during the Great Recession. As such, it had had more ground to make up, and by early 2016 it had moved back up to its January 2000 level.



There is little doubt that aggregate employment growth has rebounded from the depths of the Great Recession. However, as we see in the data, the distribution of those gains has been uneven across states and industry sectors.

Early Benchmarking the State Payroll Employment Survey—Update

In June of 2015, I wrote a blog post detailing a method called early benchmarking, which predicts how the U.S. Bureau of Labor Statistics (BLS) will revise the state payroll employment survey data when it updates its benchmarks each March (the method was first introduced by our colleagues at the Dallas Fed). The primary source of the revisions for state payroll employment (also known as the Current Employment Statistics, or CES) is the Quarterly Census of Employment and Wages (QCEW). While the BLS only rebenchmarks the CES using new QCEW data yearly, QCEW data are released quarterly, so it’s possible to use new QCEW data to predict how the BLS will revise the CES (this process is explained in detail in my earlier post). Benchmark revisions to the CES can be quite large, and we have found that our early benchmarking procedure typically reduces their size.

The BLS recently released the March 2016 benchmarked data, so we can see how the early benchmarking method performed for the Seventh Federal Reserve District and the District states for 2015. Table 1 summarizes how much employment grew in the Seventh District and District states in 2015 based on when the data were benchmarked using the QCEW. At the District level, the early benchmarking procedure performed marginally better: It underestimated job growth by 32,000, while data benchmarked in March 2015 underestimated job growth by 38,000. Both datasets estimated a 1.1 percent growth rate for 2015.

Early benchmarking is more useful at the state level, because CES sample sizes are smaller. For 2015, the largest error was for Illinois, where the March 2015 benchmarked data estimated that employment declined by 3,000, while the March 2016 benchmarked data indicate that employment actually increased by 51,000. The January 2016 benchmarked data roughly split the difference, estimating job growth of 29,000.

Table 1 - Job Growth

The figures that follow show graphically the differences between the three series summarized in table 1 for the District and District states. The dashed portion of the lines for each series represents data that are not benchmarked using the QCEW. For example, the early benchmarked data are benchmarked using QCEW data through June 2015, while the March 2016 benchmarked data use QCEW data through September 2015. In some figures it is clear that the BLS also revises already-benchmarked data, though these revisions are typically small (Wisconsin is an exception for 2014).

Later this year, we will begin publishing early benchmarked estimates of District and District state employment growth on our website on a monthly basis (coinciding with the release of the Midwest Economy Index). We will be sure to notify our Midwest Economy blog readers when we make the estimates available.







“Early Benchmarking” the State Payroll Employment Survey

One of the most important sources of information we have on current economic conditions is the Bureau of Labor Statistics’ (BLS) Current Employment Statistics (CES) program (also known as the payroll survey). The payroll survey reports monthly estimates of nonfarm payroll employment, hours, and earnings at the national, state, and metro area levels. Because the CES covers only a representative sample of employers, the BLS is able to release the results of the survey in a timely manner (typically three weeks after the reference month at the U.S. level, eight weeks after at the state level, and ten weeks after at the metro area level). At the same time, relying on a survey—instead of a census—means the results are subject to sampling error. The BLS does revise the CES estimates to address sampling error. With each new monthly release, it adjusts the two previous months’ estimates; and once a year in March, it revises the two previous calendar years at the U.S. level and the five previous calendar years at the state and metro area levels.1 In this blog post, I discuss a method pioneered by researchers at the Dallas Fed to predict the once-a-year revisions at the state level as many as nine months in advance. These predictions can provide valuable information, because state-level revisions can be substantial.

For their state CES estimates, the BLS relies primarily (but not entirely) on a sample of about one-third of the universe of employers that participate in states’ unemployment insurance (UI) programs. The information on employment and earnings from all employers participating in UI is eventually reported under the BLS’s Quarterly Survey of Employment and Wages (QCEW) program, with a delay of six months from the reference period. As noted above, in March of each year, the BLS revises state-level CES data for the prior five calendar years, using the previously unavailable QCEW data as a benchmark.2 It is important to note here that QCEW is only a benchmark (not the final answer), because the CES and the QCEW do not cover the same universe of workers. The CES does not count farm workers, but it does count the 3 percent of nonfarm workers who do not receive unemployment insurance. The QCEW only counts workers with unemployment insurance, which includes some farm workers. Because their universes are not exactly the same, the CES and QCEW have different levels. However, they overlap enough that their trends are about the same.

While the BLS rebenchmarks the payroll survey data once a year, the QCEW data are released four times a year (three months at a time). And because the payroll survey uses the QCEW as its benchmark, it is possible to use newly released QCEW data to predict how the BLS will eventually revise the CES data. For example, table 1 shows when the new QCEW data used for the March 2015 rebenchmark became available. Data for the fourth quarter of 2013 were available nine months before they were used to benchmark the CES. And by December 2014, it was possible to have a good idea of how the CES would be revised through the second quarter of 2014, even though the BLS had only revised the CES through the third quarter of 2013.

In what follows, I explain a method for “early benchmarking” the state payroll survey data pioneered by researchers at the Dallas Fed (explained by them here). I then look at how the method performed at predicting the March 2015 revisions for the five states in our Seventh Federal Reserve District.


The method begins with the last month of CES data that are benchmarked to the QCEW. From then on, monthly changes are calculated using growth rates from the (seasonally adjusted) QCEW, not the CES. Once the QCEW data run out, monthly changes are again calculated using the CES. In essence, the method amounts to substituting in growth rates from the QCEW data that are newly available but have not yet been used to benchmark the CES.

Figure 1 shows the early benchmarking method visually for the Seventh District states using the data available in December 2014.3 The blue line is the CES, where the dashed portion of the line represents months that have not been benchmarked by the QCEW. The green line is the QCEW; and it is clear that while the CES and the QCEW have different levels, their trends are quite similar. The red line is the early benchmarked version of the CES. The solid portion of the red line is the path generated by monthly growth rates from the QCEW, and the dashed portion of the red line is the path generated by monthly growth rates in the remaining CES data.



How did the early benchmarking procedure do at predicting the March 2015 CES rebenchmark? Figure 2 shows the revised payroll survey data from March 2015 in black. The QCEW-benchmarked data now go through the third quarter of 2014. Overall, the early benchmarked CES series appears to predict the most recent rebenchmarking of the CES better than the March 2014 rebenchmarked CES series does. This is especially true before the QCEW data run out in June 2014.


One way to measure the accuracy of a forecast is to calculate its root mean squared error. Table 2 shows the root mean squared error for predictions of the March 2015 rebenchmarked CES using the March 2014 rebenchmarked CES and the early benchmarked CES. The early benchmarked CES performs as well or better at both the District and state levels. At the District level, the early benchmarked CES is typically about 31,000 jobs off, while the March 2014 rebenchmarked CES prediction is typically about 51,000 jobs off. In the case of Michigan, the early benchmarked CES performed substantially better. Figures 3 through 7 show the state-level comparison. Taken as a whole, these results demonstrate that early benchmarking the CES provided a tangible improvement in measuring growth in the Seventh District over the last year or so.







  1. For example, in March 2015, the BLS revised data for 2013–14 at the U.S. level and 2009–14 at the state and metro-area levels.
  2. While the new QCEW data only cover the previous year, the revisions go back five years because the BLS also recalculates its seasonal adjustment factors.
  3. I calculate the early benchmark series for the Seventh District by summing the early benchmark series for the District states.

Making Value for America: A study by the National Academy of Engineering

Production technology and the nature of work are changing rapidly, giving rise to job declines in the manufacturing sector. In this context, can the U.S. design policies that support manufacturing while providing greater opportunity for U.S. workers? This was the question asked of a study panel at the National Academy of Engineering, which produced “Making Value for America: Embracing the Future of Manufacturing, Technology and Work.”

On May 4, Nick Donofrio, who chaired the study committee, joined a panel of experts at the Federal Reserve Bank of Chicago to present the findings and implications from the report. Donofrio was the Executive Vice President for Technology at IBM and a member of the National Academy. Joining him were panelists Dan Swinney (Manufacturing Renaissance and study committee member), Chad Syverson (University of Chicago Booth School of Business and study commission member), Haven Allen (World Business Chicago) and Craig Freedman (Freedman Seating).

Setting the stage for the program was Bill Testa, Vice President and Director of Regional Programs at the Chicago Fed. Testa focused on the key role of manufacturing in the Midwest economy. Within the Midwest region, the manufacturing job base remains 53% more concentrated than in the U.S. as a whole. However, dramatic declines in manufacturing jobs have led many policymakers to pursue other economic development targets. Since 1969, when manufacturing accounted for roughly 40% of the regional job base, manufacturing employment has fallen to 13.5%. Testa noted that from 2000 through the recession of 2009, manufacturing jobs in the Seventh District states of Illinois, Indiana, Michigan, Wisconsin, and Iowa dropped by 35% or 1.16 million; and only 224,000 or 21% of those jobs have returned during the recovery to date.

Testa commented that some of this trend can be traced to technology and productivity. What took 1,000 workers to make in 1950 can be done by 200 today. Moreover, increasingly workers in manufacturing jobs are upskilling and are required to have the ability to work with advanced technology—and to do so at pay scales set by global competition. Despite the many challenges, Testa argued that the Midwest still holds some advantages for manufacturers, such as a strong supply chain and the infrastructure to move goods efficiently. Recreating the region’s network of rail, road, air hubs, and ports would be difficult elsewhere. Second, manufacturing directly comprises 20% of the region’s overall output, but more importantly is responsible for almost two-thirds of the research and development in the region. Additionally, the region’s universities are well positioned to create innovation in engineering and technical fields that can lead to new products and processes.

Donofrio presented the findings from Making Value for America. Donofrio emphasized that this was not a typical manufacturing study. He suggested that in some ways manufacturing is a 20th century word and that what matters today is not simply making things but adding value through production. Understanding where value is added is critical; and in many cases, value comes from services related to the final product. Specifically, he defined adding value as the process of using ingenuity to convert resources into a service or process that contributes additional value to a person or a society. In particular, the study was interested in how the nature of work is changing. Donofrio quoted former MIT president Charles Vest who said, “Far too much of our nation is waiting for new ways of work to arrive. We hear lots of rhetoric about how the nature of work will change, as it relates to some unknown distant future. The fact is that it is happening now, and we need a broader recognition of this fact and policies and education that reflects it.”

Donofrio noted that U.S. employment in manufacturing will never regain its historical levels, but growth will be facilitated by policy that focuses on adding value and innovation. Recommendations from the study focused on three areas—education, collaboration, and being inclusive. A goal is to develop a robust innovation ecosystem that includes business; federal, state, and local governments; economic development organizations; educational institutions; and research organizations. A partial list of recommendations appears in Table 1; the complete list is available at


Chad Syverson kicked off the panel discussion by talking about the university’s role in supporting education at the pre K to 12th grade level. He noted that the University of Chicago has been active in urban education, ranging from the creation of charter schools to the creation of the Urban Education Lab. All of these efforts use rigorous social science evaluation to develop best practices for education.

Additionally, the university works directly to support entrepreneurship through the Polsky Center, which serves as a resource for supporting business formation. Syverson noted that the U.S. has had a 30-year downward trend in new business formation and that the dynamics behind this are not well understood. However, it is though business formation that significant innovation occurs, so the trend is disturbing, he said. Complementary services are provided by the Booth School of Business working with the Chicago Innovation Exchange, the 1871 tech incubator, and directly with businesses to disseminate best practices. Syverson noted that large amounts of value are left on the table when firms fail to keep up with best practices and that significant gains for the U.S. economy can be provided by better educating firms.

Haven Allen described the multitude of programs that can be found in Chicago that create manufacturing networks. These include national efforts such as the Digital Manufacturing and Design Innovation Institute (DMDII) and its related Illinois Manufacturing Labs. A clear goal is to connect local firms to this national network of innovative firms. This requires broad partnerships that include businesses, local community groups, and education. Allen stressed that training is key and that recent apprenticeship programs offered by organizations such as the Jane Addams Resource Center, the University of Illinois-Chicago, and Daley College provide models for supporting work force development. Finally, Allen emphasized that successful partnerships will focus on improvements in people, process, and product.

Craig Freedman offered the perspective of a manufacturer. Freedman suggested that the pace of change means that businesses and their workers need to get smarter at all levels. Business needs to link with education in order to make sure that relevant skills are learned at all grade levels. For example, Freedman cited the Manufacturing Connect program at Austin Polytech as a model that deserves replication. The school is located in the Austin community on the west side of Chicago. The program teaches high school students metal-working skills and provides them with certified credentials upon graduation. Local businesses partner with the school and provide apprenticeships and job shadowing opportunities.

Freedman also praised local training organizations such as the Jane Addams Resource Center, Daley College, and the 1000 Jobs Campaign as helping support critical work force development. Finally, Freedman suggested that more work needs to be done to promote the image of manufacturing as a good career path. Current perceptions of manufacturing are based on outdated notions of the industry that need correcting if young people will be attracted to these jobs.

Concluding the discussion was Dan Swinney. He noted that social inclusion should be a goal of the new emphasis of manufacturing rebirth in cities. He also cited the efforts of the Manufacturing Renaissance and the Manufacturing Connect program in the Austin Community in Chicago. Swinney stressed the importance of bringing these types of programs to areas that have suffered disinvestment. The challenges communities like Austin face are clear in the data. While unemployment in Chicago has fallen to 6.4%, in Austin it hovers near 30%. Similarly, manufacturing job loss in Austin is near 90%. However, to address the challenges faced by communities like Austin, Swinney argued, programs like Manufacturing Connect must be scaled up, since they currently only reach a small subset of students.

A Seventh District Labor Market Dashboard

In March 2013, Federal Reserve Board Chair Janet Yellen (then Vice Chair) gave a speech where she discussed the centrality of labor market conditions in the Fed’s decision-making process. The most often discussed indicator of labor market conditions is the unemployment rate, but in this speech and others since,1 Chair Yellen has said that she tracks a number of other labor market indicators as well. This broader set of indicators is now widely referred to as Chair Yellen’s Labor Market Dashboard. The dashboard serves to show that while the unemployment rate has been improving steadily since the depths of the Great Recession, many other labor market indicators have made less progress, and few have returned to where they were prior to the Great Recession.

In this blog post, I look at indicators on Chair Yellen’s dashboard, with a focus on how they measure the Midwest economy. The set of indicators I’ve compiled tell a story quite similar for the Midwest to that for the United States as a whole. While conditions in the Midwest have steadily improved since the depths of the Great Recession, things are not back to normal yet.

Household Survey Measures. Figure 1, panel A shows the unemployment rates for the United States and the Seventh Federal Reserve District2 since 2000. The two rates track each other closely, though there are some notable differences. First, the unemployment rate was somewhat higher in the District before the recession that hit in 2008. Figure 1, panel B shows that the unemployment rate in Michigan was particularly high during this period, largely because of struggles in the auto industry. Second, in terms of unemployment, the District did worse than the nation during the depths of the recession, though it caught up by 2011 as manufacturing recovered. Finally, the District’s unemployment rate again diverged from the nation’s in 2012, but by now it has nearly closed the gap.

1 - Unemployment

A key feature of the unemployment rate is that to be counted as unemployed, an individual without a job must be actively looking for work. Those who report that they are not working and not actively looking for work in the U.S. Bureau of Labor Statistics’ (BLS) Current Population Survey, or the household survey as it’s commonly called, are not considered unemployed. Thus, the unemployment rate can decrease because people give up looking for work, not because they find a job. For this reason (and to provide greater detail), the BLS publishes what it calls alternative measures of labor market underutilization. I take these alternative measures and break them down into five comparable rates.3 Each rate is a percentage of the labor force plus marginally attached workers. The rates are as follows:

  1. Short-term unemployed: people unemployed for fewer than 15 weeks.
  2. Long-term unemployed: people unemployed for 15 weeks or more.
  3. Marginally attached: people who want to work and have looked for work in the last 12 months, but not in the last four weeks.
  4. Discouraged: marginally attached workers who are not looking for work because they believe there are no jobs available or none for which they would qualify.
  5. Part time for economic reasons: workers who are employed part time but would like to be employed full time.

2 - Under_usdi

Figure 2 presents the five measures for the United States and the Seventh District. The measures are 12-month moving averages, so trend movements show up later than in figure 1. The bottom, darkest blue area represents the short-term unemployed. Many people in this group are between jobs or will find work without ever being long-term unemployed. Aside from a blip in 2009, this measure has been steady at around 3.5 percent for the United States and 3.7 percent for the District. The story is much different for the long-term unemployed. There was a substantial increase when the Great Recession hit, and the rate remains elevated today, for both the United States and the District. The same is true for marginal workers (discouraged and not discouraged) and part-time workers who would like to have full-time work.

Figure 3 breaks the District’s unemployment and underutilization rates down by state. Each state has a different baseline. Before the recession, Iowa and Wisconsin had the lowest rates in the District, while Michigan had the highest. Despite the differences in labor market structure highlighted by the different baselines, no state has returned to its pre-recession unemployment and underutilization levels.

3 - Under_dist

The BLS’s household survey provides two additional measures that account for the possibility that the unemployment rate can go down even if labor market conditions don’t improve. The two measures are benchmarked by the overall (working-age) population (specifically, noninstitutional civilians age 16 and over) so that people who leave the labor force are still counted. The disadvantage of benchmarking by the population is that people may not be in the labor force for reasons unrelated to difficulty finding work, such as to attend school, to have a family, or to retire.

The first measure of these two additional BLS measures is the labor force participation rate, which is the number of people in the labor force as a percentage of the population. The second measure is the employment-to-population ratio, which (as you may guess) is the number of employed people as a percentage of the population. Figures 4 and 5 show the measures since 2000 for the United States, the District, and the District states. There is a downward trend in both measures, which largely reflects an aging population. Given the trends, it is possible to spot cyclical movements as well. There is clear upward movement over the housing bubble years and clear downward movement over the recession. It is difficult to discern visually where the measures are relative to trend, but recent research by Chicago Fed economists suggests that they are currently below the trend.4

4&5 - LFP&EtoP

Business Survey Measures. The BLS surveys households to obtain the unemployment rate (and related indicators) and it surveys businesses to obtain their employment levels, openings, turnover, and costs. The most widely discussed indicator from BLS surveys of businesses is payroll employment growth. Figure 6, panel A shows quarterly payroll growth for the United States and the District since the onset of the Great Recession. I’ve scaled the District growth numbers to be as if the District and the nation had the same baseline employment levels.5 Like the unemployment rate, employment growth numbers for the United States and the District are generally in sync, with some notable exceptions. The growth numbers show that the District did worse than the nation during the recession, over 2012, and in the first quarter of 2014, when the District experienced especially harsh winter weather. Figure 6, panel B breaks down employment growth by state. As usual, Michigan is noteworthy. While Michigan was hit hard during the recession, the subsequent strong recovery of the auto industry made important contributions to District employment growth, especially during 2012, when the other states were struggling.

6 - Employment

The BLS also asks businesses about their employment turnover, and figure 7 shows three such measures.6 Panel A is the ratio of the number of unemployed workers to the number of job openings that businesses report; panel B is the number of new hires as a percentage of overall employment; and panel C is the number of quits as a percentage of overall employment. Once again, movements in the United States and Midwest are closely related, though the Midwest was doing somewhat worse before the recent recession. Each measure has a clear cyclical component. Taking the pre-recession period as a baseline, the rate of unemployed per job opening is now close to normal (the Midwest’s rate is slightly closer to normal than the nation’s). The hires and quits rates have not yet fully recovered, but have been growing at a strong pace since the middle of 2013. The quits rate, in particular, tends to be a leading indicator of improvement in labor market conditions, as people are more likely to leave their current positions when their job prospects elsewhere are improving.7

7 - Jolts

The BLS also asks businesses about their employment costs, which are driven in large part by wages. Figure 8 shows two indexes of wage growth for the Midwest states (see note 5) and the United States. Panel A is nominal wage growth, panel B is wage growth adjusted for inflation. Taken together, the panels show that while wage growth has consistently been positive, it has not consistently been above inflation (periods of real wage growth are often followed by periods of real wage decline). This reflects a long trend of slow real wage growth for all but the highest wage earners.

8 - Wages

Figure 9 shows that the quits rate and nominal wage growth move together quite closely in both the United States and the Midwest. Figure 9 also shows that nominal wage growth during the recovery has underperformed a bit relative to the quits rate. This may be a sign that wage growth will pick up in the future.

9 - QtsWge

Taken as a whole, this labor market dashboard indicates that while the U.S. and District labor markets continue to recover from the Great Recession, they are still not back to normal. The employment turnover numbers are the most promising, and they may well lead the other indicators I’ve highlighted. The labor market is clearly headed toward full recovery, but the amount of time it will take to get there remains unclear.

  1. See for example, Chair Yellen’s first press conference and her August 2014 speech at Jackson Hole.
  2. The Seventh District comprises all of Iowa and most of Illinois, Indiana, Michigan, and Wisconsin.
  3. For details on how I calculated the rates, please contact me at
  4. See the Explaining the Decline in the U.S. Labor Force Participation Rate by Aaronson, Davis, and Hu and Estimating the Trend in Employment Growth by Aaronson and Brave.
  5. Specifically, for each quarter, I calculated the ratio of U.S. employment to District employment and multiplied District employment growth by that factor. U.S. employment is roughly eight times District employment.
  6. Note that these measures are not available at the state level; the smallest geographic area for which these measures are available is the Census Region. According to the U.S. Census Bureau, the Midwest region comprises the following states: Illinois, Indiana, Iowa, Kansas, Michigan, Minnesota, Missouri, Nebraska, North Dakota, Ohio, South Dakota, and Wisconsin.
  7. See Labor Market Flows in the Cross Section and Over Time by Davis, Faberman, and Haltiwanger.

Are Seventh District Labor Markets Still Slack?

By Bill Testa and Jacob Berman

There is no question that the U.S. labor market has been gradually but steadily healing after the Great Recession. The national unemployment rate peaked at 10% in October 2009, but it has since fallen to 6.2% (as of July 2014). The nation experienced a net loss of 8.7 million jobs during the downturn, and finally finished making up for those job losses just this past May. So, undeniably, progress has been made in the labor market, but now the questions facing policymakers and other government officials are how much slack capacity in the employable population remains and whether further tightening of labor market conditions will push up wages and prices.

Recently, short-term unemployment—defined as the share of the labor force that has been unemployed for 26 weeks or less (see below)—has fallen to levels that have been historically associated with robust economic conditions. In contrast, despite post-recessionary declines in long-term unemployment (i.e., the share of the labor force that has been unemployed for greater than 26 weeks), its recent levels remain well above the historical norm. Though high long-term unemployment may be a sign of considerable labor market slack, some argue that the vast majority of the long-term unemployed lack the specific skills and other characteristics to be hired or trained. If this proves to be correct, it would imply that the U.S. labor market is nearing its full capacity.

Source: U.S. Bureau of Labor Statistics, Current Population Survey, from Haver Analytics

In order to provide more useful guideposts for macroeconomic policymaking, economists Dan Aaronson and Andrew Jordan recently investigated the relationships between rising wages and indicators of labor market tightness. In their recent Chicago Fed Letter, the authors find a strong correlation between real wage growth and two prominent measures of labor market slack—medium-term unemployment (i.e., the share of labor force unemployed for five to 26 weeks) and the percentage of the labor force reporting they are working part-time involuntarily for economic reasons (such as unfavorable business conditions or seasonal decreases in demand). Partly because both of these measures of labor slack remain elevated today, the authors conclude that real wage growth in June 2014 would have been one-half of a percentage point to one full percentage point higher under the labor market conditions of the 2005-07 U.S. economy.

While we often speak of the labor market as one monolithic term, labor market conditions vary widely by occupation, industry, and location. In their analyses, Aaronson and Jordan identify statistical relationships between real wage growth and labor market conditions by observing individual states. In the chart below, we see the general pace of employee compensation for both the United States and for the East North Central Region, which includes four of the five states of the Seventh Federal Reserve District.[1] In both the nation and the region, recent growth in labor compensation continues to fall short of that in the pre-recessionary period.[2]

Source: U.S. Bureau of Labor Statistics, Current Population Survey, from Haver Analytics

Also, as seen in the next three charts, Seventh District states generally exhibited signs of greater labor market slack in 2013 relative to the pre-recession year of 2007.[3] Long-term unemployment—both in the Seventh District states and in the nation—has stayed high during the economic recovery. In 2013, the long-term unemployment rate in Illinois was the highest among the District states (followed by Michigan). Notably, Michigan’s long-term unemployment rate had been at a high rate already in 2007 as a result of the severe restructuring of the automotive industry in the past decade.

Source: U.S. Bureau of Labor Statistics, Current Population Survey, from Haver Analytics

Medium-term measures also remained elevated among Seventh District states in 2013. However, they suggested that the District’s state labor markets may be less slack than the national one; in particular, Iowa and Wisconsin, where 2013 medium-term unemployment rates had almost returned to their 2007 levels, showed their labor markets may be improving faster than the nation’s.

Source: U.S. Bureau of Labor Statistics, Current Population Survey, from Haver Analytics

According to the measures of the percentage of the labor force who are involuntary part-time workers, there also appeared to be additional work force supply available in both the Seventh District states and the nation in 2013 as compared with 2007. This was the case for all five District states. Moreover, it should be noted that Michigan, Indiana, and Illinois displayed a higher percentage of involuntary part-timers than the nation did in 2013. Involuntary part-time workers are those who would choose to work more hours if it were possible. Typically, such workers have had their hours cut back in their current job, or they are part-time workers who cannot find a full-time job due to poor economic conditions in their occupation.[4]

Source: U.S. Bureau of Labor Statistics, Current Population Survey, from Haver Analytics

As these measures indicate, even while labor markets continue to tighten in the economic recovery, there is significant variation across states. According to the charts above, state labor markets in the Seventh District continue to be somewhat slack. Further, the observed pace of wage and employee compensation increases are still below those of the pre-recessionary period.
[1]The Federal Reserve’s Seventh District comprises major parts of Indiana, Illinois, Michigan, and Wisconsin, as well as the entirety of Iowa. The U.S. Census Bureau’s East North Central Region comprises Ohio and the entirety of the Seventh District states excepting any part of Iowa.(Return to text)

[2]Labor compensation includes both employee wages and benefits.(Return to text)

[3]State averages are reported here, though we acknowledge that local conditions and markets for specific skills and occupations differ. The Chicago Fed’s regional research staff keeps abreast of such conditions and markets through local meetings with labor market participants and businesses, as well as through formal surveys.(Return to text)

[4]For a full discussion see Rob Valletta and Leila Bengalli, “What’s Behind the Increase in Part-time Work?(Return to text)

Job Recovery in the Seventh District

by Bill Testa and Max Lichtenstein

The resumption of growth in the U.S. economy, beginning in mid-2009, has been welcome news. However, the pace of recovery has been disappointing, relative to the severity of the recession. Following a period in which U.S. output shrank 4.1 percent from the fourth quarter of 2007 to the second quarter of 2009, the economy did not regain its former size until the fourth quarter of 2010. The deep recession and relatively slow recovery have left behind startling numbers of unemployed working age adults. According to the Household Survey of employed and unemployed, 13.7 million people reported they were unemployed during the first quarter of 2011, double the number reported during the fourth quarter of 2006.

Growth in employment has resumed, especially in recent months. Outside of the government sector, payroll jobs have grown an average of 138,000 per month over the past year, and 188,000 per month over January, February, and March of this year. We can characterize this performance as a mildly encouraging start toward repairing a very large deficit in employment.

To see the extent of recovery so far, the chart below indexes payroll employment back to the first quarter of 2007, near the peak of employment in the Seventh District states. From that time, payroll jobs declined 7.2 percent in the Seventh District and 6.2 percent in the U.S. From its low point, the U.S. has regained 0.9 percent in payroll employment. The Seventh District states have recovered more strongly, regaining 1.4 percent from the trough.

As of the first quarter of this year, Michigan, which had the largest decline in employment in the District since the start of the recession, now ranks first in the District and fifth in the nation in household employment growth on a year-over-year basis. Illinois, Wisconsin, Indiana, and Iowa rank 15th, 28th, 31st, and 33rd, in the nation, respectively.

The District’s relatively strong recovery is partly explained by its relatively steep descent. Our region’s economy is tilted toward durable goods manufacturing— including autos and machinery, which fall precipitously during U.S. economic downturns. However, on the upside, manufacturing tends to bounce back more rapidly. It has done so again over the recent recovery as retailers and wholesale establishments began rebuilding their inventories in response to revived expectations of sales. Exports of manufactured goods abroad also contributed, as the world economy pulled out ahead of the U.S. recovery. This influence of durable goods can be seen in the chart below; U.S. manufacturing payroll jobs declined steeply during the recession, but have been rising at a healthy clip during the recovery.

Job growth has made a down payment toward lowering unemployment in the Seventh District. Per the chart below, both U.S. and Seventh District unemployment rates have been falling throughout 2010 and into the early part of 2011. Seventh District unemployment has fallen more steeply; it has now converged on the U.S. level, following several years of above-average rates.

Unemployment rates have fallen in all five District states (below), with broad variation among them. Although the automotive sector’s recovery has exerted significant downward pressure on Michigan’s unemployment rate, it remains the highest in the District (10.3 percent). With 6.1 percent unemployment, Iowa’s rate is the lowest in the District, owing to its concentrations in production agriculture, food processing, and export-oriented agricultural machinery.

As a group, the District states of Wisconsin, Illinois, Michigan, and Indiana have experienced steeply declining unemployment rates over the past year (map below). Gains in manufacturing activity, along with related services and transportation, have led employment gains here and eastward throughout the Midwest industrial belt.

Declining unemployment rates are a positive development. However, while “job destruction” levels appear to have abated, “job creation” levels have yet to rebound significantly. Chicago Fed Economist Lisa Barrow finds that the largest factor in recent national unemployment rate declines has been a reduction in the number of workers transitioning from employment to unemployment, rather than job growth. The Chicago Fed Letter reports that, from November 2010 to March 2011, the pace at which employed workers became unemployed slowed markedly. In the Seventh District, this trend is similarly evident from data reporting workers who file initial claims for unemployment insurance—another measure of “job destruction.” As illustrated below, in recent months initial claims for unemployment insurance have been running below year-ago levels and far below the worst months of the recession in 2009.

Despite the labor market progress to date, there is ample room for growth in employment among persons of working age. We can see this if we compare the proportion of the working age population (16 years of age and older) who are currently employed with the 2000 level (see the table below). The employed status of the population lies well below normal. In the first quarter of 2011, fewer than 6 in 10 of those of working age counted themselves as employed.

Labor market indicators suggest that, as economic recovery continues to unfold, employers will increasingly shift toward net hiring. Many employers are reaching the limit of the sales and production gains they can achieve using their existing work forces. In particular, measures of the average hourly workweek continue to tighten in both the District and in the nation so that, barring rapid growth in productivity, employers will need to hire in order to meet heightened demand for goods and services.[1] Data that more closely reflect actual hiring decisions also portend a potential hiring upswing. The national survey of job openings and labor market turnover (JOLTS) reports a strong growth in job openings—nearly 3 million since the trough of the recession.[2]


[1] These data are reported by the Bureau of Labor Statistics (BLS) covering nonproduction nonsupervisory workers in the private nonfarm sector. See (Return to text)

[2] See BLS, (Return to text)

Job Revisions Downward

Over the past few years, the drop in employment has been steep and painful. A recent reassessment of the job count shows that even steeper declines have taken place than initially thought—both across the U.S. as a whole and in the Seventh Federal Reserve District, which covers all of Iowa and most of Illinois, Indiana, Michigan, and Wisconsin.

The U.S. Bureau of Labor Statistics (BLS) releases much awaited monthly estimates of payroll jobs for the nation. These estimates, which exclude self-employed individuals, are derived from a sample of reporting firms. Once each year, these monthly estimates of payroll employment levels are revised in a major way by the BLS. The BLS revisions are reported during the first quarter of each year to reflect an almost complete count of nonfarm payroll jobs that becomes available for the month of March of the previous year[1]. In a separate release , state payroll job estimates are similarly revised.

This year’s BLS revisions were unfavorable to both the U.S and to Seventh District states. On an average monthly basis for the year 2009 in its entirety, the revisions indicated further job decline in the U.S., by more than 1 million payroll jobs, or approximately 0.8 percent (table below). All five Seventh District states also reported downward revisions. Only Indiana’s downward revision of 1.2 percent exceeded the nation’s. Revisions for the states of Wisconsin and Michigan were relatively minor.

These downward revisions constitute more bad news because the original estimates already indicated rather sharp job declines (chart below). In the chart below, yearly decline is measured by the 12-month percent change from December 2008 to December 2009 (red bars)[2]. In examining the annual rates of decline before the revisions, payroll job counts for the Seventh District region were off 4.1 percent. Prior to the revisions, the U.S. as a whole showed a net decline of 3.1 percent of total payroll employment over the same period. The revisions to these data widen the U.S.–Seventh District gap, from 1 percentage point to 1.3 percentage points.

Among individual states, Illinois and Indiana have the sharpest revisions—both at 0.9 percentage points. And the revisions for Wisconsin and Iowa are just behind, at 0.8 percentage points and 0.7 percentage points, respectively. After the revisions, Michigan was actually 0.1 percentage point higher than before—a small bit of good news for a state that lost 840,000 nonfarm payroll jobs since mid-year 2000, an 18 percent decline; by comparison, the U.S. experienced a 1.7 percent decline since that time.

Note: Chenfei Lu and Christian Delgado de Jesús assisted with this essay.


[1]These BLS counts are not of workers but of jobs. A worker may hold one or more jobs. Data in the chart are seasonally adjusted. (Return to text)

[2]The region reported covers the full state boundaries of those states. As mentioned before, the Seventh Federal Reserve District covers only the parts of Illinois, Indiana, Michigan, and Winsconsin. (Return to text)