The Atlanta Fed's macroblog provides commentary on economic topics including monetary policy, macroeconomic developments, financial issues and Southeast regional trends.
- BLS Handbook of Methods
- Bureau of Economic Analysis
- Bureau of Labor Statistics
- Congressional Budget Office
- Economic Data - FRED® II, St. Louis Fed
- Office of Management and Budget
- Statistics: Releases and Historical Data, Board of Governors
- U.S. Census Bureau Economic Programs
- White House Economic Statistics Briefing Room
November 24, 2014
And the Winner Is...Full-Time Jobs!
Each month, the U.S. Census Bureau for the U.S. Bureau of Labor Statistics (BLS) surveys about 60,000 households and asks people 15 years and older whether they are employed and, if so, if they are working full-time or part-time. The BLS defines full-time employment as working at least 35 hours per week. This survey, referred to as both the Current Population Survey and the Household Survey, is what produces the monthly unemployment rate, labor force participation rate, and other statistics related to activities and characteristics of the U.S. population.
For many months after the official end of the Great Recession in June 2009, the Household Survey produced less-than-happy news about the labor market. The unemployment rate didn't start to decline until October 2009, and nonfarm payroll job growth didn't emerge confidently from negative territory until October 2010. Now that the unemployment rate has fallen to 5.8 percent—much faster than most would have expected even a year ago—the attention has turned to the quality, rather than quantity, of jobs. This scrutiny is driven by a stubbornly high rate of people employed part-time "for economic reasons" (PTER). These are folks who are working part-time but would like a full-time job. Several of my colleagues here at the Atlanta Fed have looked at this phenomenon from many angles (here, here, here, here, and here).
The elevated share of PTER has left some to conclude that, yes, the economy is creating a significant number of jobs (an average of more than 228,000 nonfarm payroll jobs each month in 2014), but these are low-quality, part-time jobs. Several headlines have popped up over the past year or so claiming that "...most new jobs have been part-time since Obamacare became law," "Most 2013 job growth is in part-time work," "75 Percent Of Jobs Created This Year  Were Part-Time," "Part-time jobs account for 97% of 2013 job growth," and as recently as July of this year, "...Jobs Report Is Great for Part-time Workers, Not So Much for Full-Time."
However, a more careful look at the postrecession data illustrates that since October 2010, with the exception of four months (November 2010 and May–July 2011), the growth in the number of people employed full-time has dominated growth in the number of people employed part-time. Of the additional 8.2 million people employed since October 2010, 7.8 million (95 percent) are employed full-time (see the charts).
The pair of charts illustrates the contribution of the growth in part-time and full-time jobs to the year-over-year change in total employment between January 2000 and October 2014. By zooming in, we can see the same thing from October 2010 (when payroll job growth entered consistently positive territory) to October 2014. Job growth from one month to the next, even using seasonally adjusted data, is very volatile.
To get a better idea of the underlying stable trends in the data, it is useful to compare outcomes in the same month from one year to the next, which is the comparison that the charts make. The black line depicts the change in the number of people employed each month compared to the number employed in the same month the previous year. The green bars show the change in the number of full-time employed, and the purple bars show the change in the number of part-time employed.
During the Great Recession (until about October 2010), the growth in part-time employment clearly exceeded growth in full-time employment, which was deep in negative territory. The current high level of PTER employment is likely to reflect this extended period of time in which growth in part-time employment exceeded that of full-time employment. But in every month since August 2011, the increase in the number of full-time employed from the year before has far exceeded the increase in the number of part-time employed. This phenomenon includes all of the months of 2013, in spite of what some of the headlines above would have you believe.
So, in the post-Great Recession era, the growth in full-employment is, without a doubt, way out ahead.
Author's note: The data used in this post, which are the same data used to generate the headlines linked above, reflect either full-time or part-time employment (total hours of work at least or less than 35 per week, respectively). They do not necessarily reflect employment in a single job.
November 20, 2014
For Middle-Skill Occupations, Where Have All the Workers Gone?
Considerable discussion in recent years has concerned the “hollowing out of the middle class.” Part of that story revolves around the loss of the types of jobs that traditionally have been the core of the U.S. economy: so-called middle-skill jobs.
These jobs, based on the methodology of David Autor, consist of office and administrative occupations; sales jobs; operators, fabricators, and laborers; and production, craft, and repair personnel (many of whom work in the manufacturing industry). In this post, we don't examine why the decline in middle-skill jobs has occurred, just how those workers have weathered the most recent recession. But our Atlanta Fed colleague Federico Mandelman offers an explanation of why this has occurred.
So how have workers in middle-skill occupations fared during the last recession and recovery? Let's examine a few facts from the Current Population Survey from the U.S. Bureau of Labor Statistics.
Only employment in middle-skill occupations remains below prerecession levels
Chart 1 shows employment levels by skill category (using 12-month moving averages to smooth out the seasonal variation). From the end of 2007 to the end of 2009, the overall number of people working declined by more than 8 million. Middle-skill jobs were hit the hardest, declining about 10 percent from 2007 to 2009. As of September 2014, the level was still about 9 percent below the 2007 level. In contrast, employment in low-skill occupations is 7 percent above prerecession levels, and employment in high-skill occupations is about 8 percent higher than before the recession.
For full-time workers (working at least 35 hours a week at all jobs) the decline in middle-skill occupations is even more dramatic. From 2007 to 2009, the number of full-time workers whose main job was a middle-skill occupation fell more than 15 percent from 2007 to 2009 and is still about 11 percent below the level at the end of 2007.
Those in middle-skilled occupations were most likely to become unemployed
In the 2001 recession, the chances of being unemployed after one year were similar for those working full-time in middle- and low-skill occupations. During the most recent recession, the likelihood of becoming unemployed rose sharply for everyone, but much more sharply for those working in middle-skill occupations. At the recession's trough, almost 6 percent of people who were employed in middle-skill occupations one year earlier were unemployed, compared with about 3 percent of workers in high-skill occupations and 3.5 percent of workers in lower-skill occupations (see chart 2).
Underemployment has improved only slowly at all skill levels
The share of people who are working part-time involuntarily about doubled for workers in low-, middle-, and high-skill occupations. For middle-skill occupations, the share rose from around 1.7 percent to 4.3 percent and is currently around 2.4 percent. For low-skill occupations, involuntary part-time employment increased from 2.4 percent to 5 percent and was still 3.8 percent as of September 2014. And for those in high-skill occupations, the chances of becoming involuntarily part-time rose from 0.8 percent to 1.8 percent and are now back to about 1 percent (see chart 3).
Ready for some good news?
Those who held middle-skill jobs are more likely to obtain high-skill jobs than before the recession
Currently, of those in middle-skill occupations who remain in a full-time job, about 83 percent are still working in a middle-skill job one year later (see chart 4). What types of jobs are the other 17 percent getting? Mostly high-skill jobs; and that transition rate has been rising. The percent going from a middle-skill job to a high-skill job is close to 13 percent: up about 1 percent relative to before the recession. The percent transitioning into low-skill positions is lower: about 3.4 percent, up about 0.3 percentage point compared to before the recession. This transition to a high-skill occupation tends to translate to an average wage increase of about 27 percent (compared to those who stayed in middle-skill jobs). In contrast, those who transition into lower-skill occupations earned an average of around 24 percent less.
In summary, the number of middle-skill jobs declined substantially during the last recession, and that decline has been persistent—especially for full-time workers. Many of the workers leaving full-time, middle-skill jobs became unemployed, and some of that decline is the result of an increase in part-time employment. But others gained full-time work in other types of occupations. In particular, they are more likely than in the past to transition to higher-skill occupations. Further, the transition rate to high-skill occupations has gradually risen and doesn't appear directly tied to the last recession.
Authors' note: The middle-skill category of jobs consists of office and administrative occupations; sales; operators, fabricators, and laborers; and production, craft, and repair personnel. The other two broad categories of occupations are labeled high-skill and low-skill. High-skill occupations consist of managers, technicians, and professionals. Low-skill occupations are defined as those involving food preparation, building and grounds cleaning, personal care and personal services, and protective services.
November 13, 2014
A Closer Look at Employment and Social Insurance
The Atlanta Fed's Center for Human Capital Studies hosted its annual employment conference on October 2–3, 2014, organized once again by Richard Rogerson of Princeton University, Robert Shimer of the University of Chicago, and the Atlanta Fed's Melinda Pitts. This macroblog post summarizes some of the discussions.
Social insurance programs in the United States and other developed countries represent a large and growing share of expenditures relative to gross domestic product (GDP). Assessing the costs and benefits of the diverse programs that make up the U.S. social insurance system is a key input into the design and implementation of effective programs. This conference featured seven papers that dealt with various aspects of this assessment. Although each program is designed to address specific issues and hence needs to be studied in the context of those issues, many of the same basic economic questions arise in each context. For example, what is the rationale for social insurance programs? Do they address inefficiencies, or are they mainly designed to redistribute from one group to another? Who benefits from specific programs? How do programs designed to achieve specific objectives distort economic outcomes? These are the questions that featured prominently in the conference.
A classic question in economics concerns the extent to which markets cannot achieve efficient outcomes without government intervention. It is well known that the so-called "invisible hand" can achieve efficient outcomes in a wide range of standard settings, but do these results extend to situations in which information asymmetries exist? In 1976, Michael Rothschild and Joseph Stiglitz's article "Equilibrium in Competitive Insurance Markets" suggested that in the presence of certain kinds of private information, insurance markets could not achieve efficient allocations. In fact, they argued that competitive equilibrium might not even exist in these settings. In "Adverse Selection Is Not a Justification for Social Insurance," Ed Prescott challenges this result and shows that competitive equilibrium exists and achieves efficient allocations in settings that include information problems such as Rothschild and Stiglitz's adverse selection problem. Key to this result is the presence of mutual insurance companies, and how this presence influences the contracts offered by insurance companies in equilibrium. In the Rothschild and Stiglitz environment, insurance companies were effectively agents with deep pockets that were outside the model.
Providing insurance to individuals in situations in which they face bad outcomes may distort individual behavior and lead to negative outcomes that outweigh the benefits of the insurance. This basic issue was addressed by three of the papers at the conference in three separate contexts. Jason Abaluck, Jonathan Gruber, and Ashley Swanson examined how prescription drug coverage through Medicare influences prescription drug usage; Hamish Low and Luigi Pistaferri studied the disability insurance (DI) system; and Bradley Heim, Ithai Lurie, and Kosali Simon examined whether the extension of health benefits to young adults as mandated by the Affordable Care Act (ACA) influenced the behavior of young adults.
In "Prescription Drug Use Under Medicare Part D: A Linear Model of Non-linear Budget Sets," Jason Abaluck, Jonathan Gruber, and Ashley Swanson study how prescription drug use responds to price changes associated with social insurance through Medicare. At the conference, Gruber discussed one key objective of their analysis: uncovering the elasticity of prescription drug use with respect to price. A large elasticity implies that providing insurance in the form of lower prices will distort behavior and lead to much higher drug use, and some recent papers have argued that this elasticity may be quite large. Their basic strategy is to study how changes in the details of Medicare coverage over time influenced individual choices. A novel feature of the estimation strategy is to take advantage of the fact that the marginal price people face depends on their overall annual expenditure on prescriptions, so that individuals can be sorted into groups based on histories of usage, interacted with changes in the details of coverage. A first key finding of this paper is that the elasticity is relatively small. A second key set of findings concerns the extent to which individual choices (in terms of plan selection and yearly expenditure conditional on plan choice) reflect departures from rationality, such as myopia or salience. The paper finds an important role for both of these effects.
Disability insurance (DI) represents a clear and classic example of the tension between insurance provision and insurance. While one would like to provide insurance to individuals who are unable to work, it can be difficult to assess the true ability of an individual to work, thereby creating the opportunity for people who are not disabled to also collect. Luigi Pistaferri addressed this issue in the paper he coauthored with Hamish Low, "Disability Insurance and the Dynamics of the Incentive-Insurance Tradeoff." This paper builds and estimates a structural model that incorporates labor supply, health shocks, earnings shocks, and the key details of the DI application process. The authors conduct various counterfactuals and assess the tension between insurance and incentives in the context of the U.S. DI program. Several results emerge. First, making the review process less strict would enhance welfare despite worsening incentives for people to misreport their health status. This is because the current system denies too many truly disabled individuals from collecting. But decreasing generosity would also increase overall welfare by decreasing the incentives for false collection.
One of the first measures of the Affordable Care Act (ACA) to be enacted was the provision that allowed dependent individuals to remain covered by their parents' healthcare plans until the age of 26. The paper by Bradley Heim, Ithai Lurie, and Kosali Simon, "The Impact of the Affordable Care Act Young Adult Mandate: Evidence from Tax Data," aims to assess the extent to which this provision has affected outcomes for young adults in terms of employment, wages, schooling, and marriage. As Simon described it at the conference, the novel aspect of this analysis is that it tracks outcomes using administrative IRS data, which affords a large sample size. The main empirical strategy is to compare the change in outcomes from before and after the provision was enacted for individuals below the age threshold with the change in outcomes for individuals just above the age threshold. The paper also reports estimates based on triple differencing that uses information on parental health insurance status. The main message from the analysis is that one cannot find robust, statistically significant effects of this ACA provision on outcomes for young individuals. One important qualification is that despite the large sample size, standard errors are still quite large, so that the analysis cannot rule out the possibility of economically significant effects.
Naoki Aizawa and Hanming Fang also considered the effects of the ACA in their paper "Equilibrium Labor Market Search and Health Insurance Reform." However, in contrast to the above papers that focus on how a particular program feature might influence individual choices, this paper focuses on how the creation of health insurance exchanges and the individual insurance mandate would affect the overall equilibrium in the labor market, taking into account the firms' decisions on whether to offer insurance and the wages that they offer to workers. In his presentation, Fang discussed building a structural equilibrium model of the labor market and estimating it using a variety of data sets. The authors find that the ACA will reduce the uninsured rate from about 20 percent to about 7 percent. But interestingly, the paper finds that the uninsured rate would drop even further if the employer mandate were dropped from the ACA. General equilibrium responses are key to understanding this result, illustrating the importance of studying these effects.
One of the rapidly growing social insurance programs is Medicaid. Mariacristina De Nardi, Eric French, and John Bailey Jones assess the benefits of this program in their paper "Medicaid Insurance in Old Age." As French described at the conference, this paper uses a structural approach to assess the extent to which households with different income and health status benefit from Medicaid. The analysis focuses on individuals from age 70 and forward using data from the Health and Retirement Study, emphasizing the risks that individuals face as a result of health shocks. Medicaid offers partial insurance against these shocks, particularly the large expenditures associated with nursing home care, and the paper assesses the value of this insurance for individuals in different positions in the wealth distribution at age 70. The paper has two main findings. First, the insurance value of Medicaid is substantial, and decreasing the size of the program would entail large welfare costs in excess of one dollar for every dollar of reduced spending. Second, expanding the size of the program would offer significant insurance value only to wealthy households. The authors conclude that in terms of managing the risks of the elderly, the current scope of Medicaid seems appropriate.
As the above discussion emphasizes, a critical input into the design and assessment of social insurance programs are data that allow us to reliably document the outcomes and groups that the insurance program wishes to help, as well as measure the efficacy of existing programs in achieving desirable outcomes. In the paper "Welfare Programs and Survey Misreporting: Implications for Income, Poverty and Disconnectedness," Bruce Meyer and Nikolas Mittag documented the serious shortcomings of several standard publicly available data sets when it comes to measuring the resources available to the poorer segments of the population. Meyer presented the paper at the conference, and it uses administrative data from New York State that allow them to link income and transfer data, both cash and in-kind, and compare the measures obtained using these administrative data with the measures obtained using data from the Current Population Survey (CPS), which is a standard source for publicly available data on the income distribution. The results are striking. Relative to analysis based on data from the CPS, analysis using administrative data shows better outcomes in terms of inequality and disconnectedness and yield larger effects from existing programs in terms of their ability to affect these outcomes.
Full papers or presentations for most of these papers are available on the Atlanta Fed's website.
By Melinda Pitts, director of the Atlanta Fed's Center for Human Capital Studies, Richard Rogerson of Princeton University, and Robert Shimer of the University of Chicago
November 10, 2014
Wage Growth of Part-Time versus Full-Time Workers: Evidence from the CPS
Last week, our Atlanta Fed colleagues Lei Fang and Pedro Silos highlighted the wage growth trends of full-time and part-time workers in recent years. Using data from the U.S. Census Bureau's Survey of Income and Program Participation (SIPP), they showed relatively weak growth in hourly wages of part-time workers between 2011 and 2013. The Current Population Survey (CPS)—administered jointly by the Census Bureau and the U.S. Bureau of Labor Statistics—also contains wage information and has data through September 2014. We thought it would be interesting to see if the CPS data revealed a similar post-recession pattern, and if the more recent data show any sign of improvement. The short answer is that they do.
The following chart displays the median year-over-year growth in hourly earnings of wage and salary earners (shown as quarterly averages). The wage data are constructed using a similar methodology to that outlined in this paper by our San Francisco Fed colleagues Mary Daly and Bart Hobijn. The orange line is the median year-over-year growth in the hourly wages of all workers. The green line is the median wage growth of workers who worked full-time in both the current month and 12 months earlier (it is close to the orange line because most workers work full-time hours). The blue line is the median wage growth of workers who were part-time in both periods. Note that the median part-time wage growth is less precisely estimated (and thus demonstrates relatively more quarter-to-quarter variation) than its full-time counterpart because the CPS's sample size of wages for part-time workers is much smaller than for full-time workers.
Despite the noisy nature of the part-time wage data, it seems clear that the median wage growth of people usually working part-time fell dramatically behind that of full-time workers between 2011 and 2013. This finding is consistent with that of Fang and Silos. Interestingly, the other period when median part-time wage growth slipped behind was during the sluggish labor market recovery following the 2001 recession, albeit much less dramatically than the recent episode.
The SIPP data used by Fang and Silos ended in mid-2013. The more recent CPS data suggest that overall wage growth has picked up during the last year and that the wage growth gap has closed a bit, which are encouraging findings. But the wage growth of part-time workers, as a group, continues to lag well behind that of full-time workers. The relatively low wage growth of part-time workers heightens the importance of the fact that the number of people working part-time—especially involuntarily part-time—remains elevated.
November 06, 2014
Wage Growth of Part-Time versus Full-Time Workers: Evidence from the SIPP
Debates about the sluggish recovery in output, the low growth in labor productivity, and the actual level of slack in the U.S. economy are common within policy circles (see, for example, this speech by Fed Chair Janet Yellen and previous macroblog posts—here and here). One of the defining features of the recovery from the Great Recession has been the rise in the number of people employed part-time. As reported by the U.S. Bureau of Labor Statistics, roughly 10 percent more people are working part-time in September 2014 than before the recession. Part-time workers generally earn less per hour than full-time workers, so lower hours and lower per-hour earnings both contribute to their lower incomes. Despite those differences in wage levels, less is known about wage growth of part-time relative to full-time workers. Has wage growth been different? Has wage inequality increased across the two groups of workers?
To find out, we employ data from the Survey of Income and Program Participation (SIPP) to analyze the wage growth of part-time and full-time workers. The SIPP is a longitudinal survey designed to be representative of the U.S. labor force. It is constructed as a sequence of panels of households who are interviewed for three to five years. Designed and maintained by the U.S. Census Bureau, the first panel began in 1984, and the most recent panel started in 2008. Households are interviewed every four months during the time they remain in the sample, providing information on work experience (employment, hours, earnings, occupation, and industry, among other variables) for the months between interviews.
The 2008 SIPP panel data that we use cover the period from August 2008 to April 2013. We restrict the analysis to hourly workers, a group representing roughly half of all employed in the 2008 panel. The reason we focus on this group is that they provide the cleanest measure of the price of labor: a wage rate for each hour they work. The remainder of workers—those compensated with a monthly or annual salary—do not report such a measure, and it needs to be inferred from their responses about total earnings and total hours worked. Because hours reported in the SIPP include much missing data and are sometimes inaccurate, we discard salaried workers. We also exclude anyone whose wages or hours information was allocated or imputed and anyone at the top or bottom of the wage distribution.
We divide the sample into two groups: those whose usual hours are fewer than 35 hours a week (part-time workers) and those who usually work 35 hours or more per week (full-time workers). We then compare the distribution of wage growth for each group and compute the median wage growth rate. To eliminate short-term fluctuations and seasonal effects, we compute median hourly wage growth rates over a three year period, expressed as an annual rate. Since the data start from August 2008, our series for the wage growth rate starts from August 2011.
Chart 1 shows the median wage growth rate of individuals over time. During the recovery, the median growth rate of full-time workers has been higher than that of part-time workers. In particular, wage declines were more common among part-time workers.
To further analyze the wage growth pattern of full-time and part-time workers, we subdivide the sample by education. Chart 2 plots the median wage growth rates for those with at least a bachelor's degree and those with some college or less. The median wage growth rates for full-time workers are larger than for part-time workers within each education group and highest for college graduates working full-time. Also apparent is that the weak wage growth of part-time workers is significantly influenced by the sluggish wage growth among those with less than a bachelor's degree.
Overall, we find that part-time workers as a group appear to experiencing a lower average wage growth rate than full-time workers during the recovery from the Great Recession. Education matters for wage growth, but the pattern of lower wage growth for part-time workers persists for people with broadly similar educational attainment.
By Lei Fang, research economist and assistant policy adviser, and
Pedro Silos, research economist and associate policy adviser, both in the Atlanta Fed's research department
November 04, 2014
Data Dependence and Liftoff in the Federal Funds Rate
When asked "at which upcoming meeting do you think the FOMC [Federal Open Market Committee] will FIRST HIKE its target for the federal funds rate," 46 percent of the October Blue Chip Financial Forecasts panelists predicted that "liftoff" would occur at the June 2015 meeting, and 83 percent chose liftoff at one of the four scheduled meetings in the second and third quarters of next year.
Of course, this result does not imply that there is an 83 percent chance of liftoff occurring in the middle two quarters of next year. Respondents to the New York Fed's most recent Primary Dealer Survey put this liftoff probability for the middle two quarters of 2015 at only 51 percent. This more relatively certain forecast horizon for mid-2015 is consistent with the "data-dependence principle" that Chair Yellen mentioned at her September 17 press conference. The idea of data dependence is captured in this excerpt from the statement following the October 28–29 FOMC meeting:
[I]f incoming information indicates faster progress toward the Committee's employment and inflation objectives than the Committee now expects, then increases in the target range for the federal funds rate are likely to occur sooner than currently anticipated. Conversely, if progress proves slower than expected, then increases in the target range are likely to occur later than currently anticipated.
If the timing of liftoff is indeed data dependent, a natural extension is to gauge the likely "liftoff reaction function." In the current zero-lower bound (ZLB) environment, researchers at the University of North Carolina and the St. Louis Fed have analyzed monetary policy using shadow fed funds rates, shown in figure 1 below, estimated by Wu and Xia (2014) and Leo Krippner.
Unlike the standard fed funds rate, a shadow rate can be negative at the ZLB. The researchers found that the shadow rates, particularly Krippner's, act as fairly good proxies for monetary policy in the post-2008 ZLB period. Krippner also produces an expected time to liftoff, estimated from his model, shown in figure 1 above. His model's liftoff of December 2015 is six months after the most likely liftoff month identified by the aforementioned Blue Chip survey.
I included Krippner's shadow rate (spliced with the standard fed funds rate prior to December 2008) in a monthly Bayesian vector autoregression alongside the six other variables shown in figure 2 below.
The model assumes that the Fed cannot see contemporaneous values of the variables when setting the spliced policy—that is, the fed funds/shadow rate. This assumption is plausible given the approximately one-month lag in economic release dates. The baseline path assumes (and mechanically generates) liftoff in June 2015 with outcomes for the other variables, shown by the black lines, that roughly coincide with professional forecasts.
The alternative scenarios span the range of eight possible outcomes for low inflation/baseline inflation/high inflation and low growth/baseline growth/high growth in the figures above. For example, in figure 2 above, the high growth/low inflation scenario coincides with the green lines in the top three charts and the red lines in the bottom three charts. Forecasts for the spliced policy rate are conditional on the various growth/inflation scenarios, and "liftoff" in each scenario occurs when the spliced policy rate rises above the midpoint of the current target range for the funds rate (12.5 basis points).
The outcomes are shown in figure 3 below. At one extreme—high growth/high inflation—liftoff occurs in March 2015. At the other—low growth/low inflation—liftoff occurs beyond December 2015.
One should not interpret these projections too literally; the model uses a much narrower set of variables than the FOMC considers. Nonetheless, these scenarios illustrate that the model's forecasted liftoffs in the spliced policy rate are indeed consistent with the data-dependence principle.
By Pat Higgins, senior economist in the Atlanta Fed's research department
October 15, 2014
What's behind Declining Labor Force Participation? Test Your Hypothesis with Our New Data Tool
The share of people (age 16 and over) participating in the labor market—that is, either working or looking for work—declined significantly during the recession. As many researchers have noted (see our list of supplemental reading under "More Information"), there is clearly a cyclical component to the decline. When labor market opportunities dry up, it influences decisions to pursue activities other than work, such as schooling, taking care of family, or retiring. However, much of the decrease in the overall labor force participation rate (LFPR) could be the result of the continuation of longer-term behavioral and demographic trends.
While the U.S. Bureau of Labor Statistics (BLS) produces aggregate statistics on LFPR by various demographic measures, the published data tables don't detail the reasons people give for not participating in the labor market. However, we have cut and coded the micro monthly data from BLS's Current Population Survey so that you can explore your own questions.
For example: Are millennials less likely to participate in the labor market than earlier cohorts? Are people retiring sooner? Are women less likely to stay at home than in the past? You can answer these and other types of questions on our new Labor Force Participation Dynamics page (click on "Interact and Download Data").
In addition to allowing you to create your own charts and download the chart data, the website also guides you through some of the major factors we found that contributed to the decline in LFPR from 2007 to mid-2014, as well as a picture of the trend in those factors before the recession began.
The chart below (also in the Executive Summary) provides an overview of the major factors that we noted in our analysis of the data. Each bar shows the contribution to the 3 percentage point change in the overall LFPR from 2007 to mid-2014.
What the chart doesn't show is whether the trends were occurring before the Great Recession. For a deeper dive into any of the factors in the chart, see the "Long-Term Behavioral and Demographic Trends" section. The most influential factor has been the changing distribution of the population (see "Aging Population"). The fact that a greater portion of Americans are retirement age now than in 2007 accounts for about 1.7 percentage points of the decline. At the same time, older Americans are more likely to be working than in the past, a trend that has been putting upward pressure on LFPR for some time. All else being equal, if those older than 60 were just as likely to retire as they were in 2007, LFPR would be about 1.0 percentage point lower than it is today.
Other factors bringing down the overall LFPR include an increased incidence of people saying they are unable to work as a result of disability or illness (click on "Health Problems"), increased school attendance among the young (click on "Rising Education"), and decreased participation among individuals 25–54—the age group with the greatest attachment to the labor force (click on "Focus on Prime Working-Age Individuals").
These are the factors we found to be the most significant drivers of changes in LFPR, but you can also explore many other questions with these data. Check out the interactive data tools and read our take on the data and let us know what you think.
September 29, 2014
On Bogs and Dots
Consider this scenario. You travel out of town to meet up with an old friend. Your hotel is walking distance to the appointed meeting place, across a large grassy field with which you are unfamiliar.
With good conditions, the walk is about 30 minutes but, to you, the quality of the terrain is not so certain. Though nobody seems to be able to tell you for sure, you believe that there is a 50-50 chance that the field is a bog, intermittently dotted with somewhat treacherous swampy traps. Though you believe you can reach your destination in about 30 minutes, the better part of wisdom is to go it slow. You accordingly allot double the time for traversing the field to your destination.
During your travels, of course, you will learn something about the nature of the field, and this discovery may alter your calculation about your arrival time. If you discover that you are indeed crossing a bog, you will correspondingly slow your gait and increase the estimated time to the other side. Or you may find that you are in fact on quite solid ground and consequently move up your estimated arrival time. Knowing all of this, you tell your friend to keep his cellphone on, as your final meeting time is going to be data dependent.
Which brings us to the infamous “dots,” ably described by several of our colleagues writing on the New York Fed’s Liberty Street Economics blog:
In January 2012, the FOMC began reporting participants’ FFR [federal funds rate] projections in the Summary of Economic Projections (SEP). Market participants colloquially refer to these projections as “the dots” (see the second chart on page 3 of the September 2014 SEP for an example). In particular, the dispersion of the dots represents disagreement among FOMC [Federal Open Market Committee] members about the future path of the policy rate.
The Liberty Street discussion focuses on why the policy rate paths differ among FOMC participants and across a central tendency of the SEPs and market participants. Quite correctly, in my view, the blog post’s authors draw attention to differences of opinion about the likely course of future economic conditions:
The most apparent reason is that each participant can have a different assessment of economic conditions that might call for different prescriptions for current and future monetary policy.
The Liberty Street post is a good piece, and I endorse every word of it. But there is another type of dispersion in the dots that seems to be the source of some confusion. This question, for example, is from Howard Schneider of Reuters, posed at the press conference held by Chair Yellen following the last FOMC meeting:
So if you would help us, I mean, square the circle a little bit—because having kept the guidance the same, having referred to significant underutilization of labor, having actually pushed GDP projections down a little bit, yet the rate path gets steeper and seems to be consolidating higher—so if it’s data dependent, what accounts for the faster projections on rate increases if the data aren’t moving in that direction?
The Chair’s response emphasized the modest nature of the changes, and how they might reflect modest improvements in certain aspects of the data. That response is certainly correct, but there is another point worth emphasizing: It is completely possible, and completely coherent, for the same individual to submit a “dot” with an earlier (or later) liftoff date of the policy rate, or a steeper (or flatter) path of the rate after liftoff, even though their submitted forecasts for GDP growth, inflation, and the unemployment rate have not changed at all.
This claim goes beyond the mere possibility that GDP, inflation, and unemployment (as officially defined) may not be sufficiently complete summaries of the economic conditions a policymaker might be concerned with.
The explanation lies in the metaphor of the bog. The estimated time of arrival to a destination—policy liftoff, for example—depends critically on the certainty with which the policymaker can assess the economic landscape. An adjustment to policy can, and should, proceed more quickly if the ground underfoot feels relatively solid. But if the terrain remains unfamiliar, and the possibility of falling into the swamp can’t be ruled out with any degree of confidence...well, a wise person moves just a bit more slowly.
Of course, as noted, once you begin to travel across the field and gain confidence that you are actually on terra firma, you can pick up the pace and adjust the estimated time of arrival accordingly.
To put all of this a bit more formally, an individual FOMC participant’s “reaction function”—the implicit rule that connects policy decisions to economic conditions—may not depend on just the numbers that that individual writes down for inflation, unemployment, or whatever. It might well—and in the case of our thinking here at the Atlanta Fed, it does—depend on the confidence with which those numbers are held.
For us, anyway, that confidence is growing. Don’t take that from me. Take it from Atlanta Fed President Lockhart, who said in a recent speech:
I'll close with this thought: there are always risks around a projection of any path forward. There is always considerable uncertainty. Given what I see today, I'm pretty confident in a medium-term outlook of continued moderate growth around 3 percent per annum accompanied by a substantial closing of the employment and inflation gaps. In general, I'm more confident today than a year ago.
Viewed in this light, the puzzle of moving dots without moving point estimates for economic conditions really shouldn’t be much of a puzzle at all.
By Dave Altig, executive vice president and research director of the Atlanta Fed
September 15, 2014
The Changing State of States' Economies
Timely data on the economic health of individual states recently came from the U.S. Bureau of Economic Analysis (BEA). The new quarterly state-level gross domestic product (GDP) series begins in 2005 and runs through the fourth quarter of 2013. The map below offers a look at how states have fared since 2005 relative to the economic performance of the nation as a whole.
It’s interesting to see the map depict an uneven expansion between the second quarter of 2005 and the peak of the cycle in the fourth quarter of 2007. By the fourth quarter of 2008, most parts of the country were experiencing declines in GDP.
The U.S. economy hit a trough during the second quarter of 2009, according to the National Bureau of Economic Research, but 20 states and the District of Columbia recovered more quickly than the rest. The continued progress is easy to see, as is the far-reaching impact of the tsunami that hit Japan on March 11, 2011, which disrupted economic activity in many U.S. states. By the fourth quarter of 2013, only two states—Mississippi and Minnesota—experienced negative GDP.
The map shows that not all states are growing even when overall GDP is growing, and not all states are shrinking even when overall GDP is shrinking. But if we want to know more about which states are driving the change in overall GDP growth, then the geographic size of the state might not be so important.
Depicting states scaled to the size of their respective economies provides another perspective, because it’s the relative size of a state’s economy that matters when considering the contribution of state-level GDP growth to the national economy. The following chart uses bubbles (sized by the size of the state’s economy) to depict changes in states’ real GDP from the second quarter of 2005 through the fourth quarter of 2013.
This chart shows how the economies of larger states such as California, New York, Texas, Florida, and Illinois have an outsize influence on the national economy, despite some having a smaller geographic footprint. (Conversely, changes in the relatively small economy of a geographically large state like Montana have a correspondingly small impact on changes in the national economy.)
Overall GDP is now well above its prerecession peak. But have all states also fully recovered their GDP losses? The chart below depicts the cumulative GDP growth in each state from the end of 2007 to the end of 2013. The size of the circle represents the magnitude of the change in the level of real GDP between the end of 2007 and 2013. Most states have fully recovered in terms of GDP. (North Dakota’s spectacular growth stands out, thanks to its boom in the oil and gas industry.) However, Florida, Nevada, Connecticut, Arizona, New Jersey, and Michigan had not returned to their prerecession spending levels as of the end of 2013. For Florida, Nevada, and Arizona, the depth of the collapse in those states’ booming housing sectors is almost certainly responsible for the relative shortfall in performance since 2007.
The next release of the state-level GDP data, scheduled for September 26, will provide insight into the relative performance of state economies during the first quarter of 2014 at a time when overall GDP shrank by more than 2 percent (annualized rate). Some analysts have suggested that weather disruptions were a leading cause for that decline. The state-level GDP data will help tell the story.
By Whitney Mancuso, a senior economic analyst in the the Atlanta Fed's research department
August 25, 2014
What Kind of Job for Part-Time Pat?
As anyone who follows macroblog knows, we have been devoting a lot of attention recently to the issue of people working part-time for economic reasons (PTER), which means people who want full-time work but have not yet been able to find it. As of July 2014, the number of people working PTER stood at around 7.5 million. This level is down from a peak of almost 9 million in 2011 but is still more than 3 million higher than before the Great Recession. That doesn’t mean they won’t ever find full-time work in the future, but their chances are a lot lower than in the past.
Consider Pat, for example. Pat was working PTER at some point during a given year and was also employed 12 months later. At the later date, Pat is either working full-time, still working PTER, or is working part-time but is OK with it (which means Pat is part-time for noneconomic reasons). How much luck has Pat had in finding full-time work?
As the chart below shows, there is a reasonable chance that after a year, Pat is happily working full-time. But it has become much less likely than it was before the recession. In 2007, an average of 61 percent of the 2006 Pats transitioned into full-time work. The situation got a lot worse during the recession, and has not improved. In 2013, only 49 percent of the 2012 cohort of Pats had found a full-time job. The decline in finding full-time work is largely accounted for by the rise in the share of Pats who are stuck working PTER. In 2007, 18 percent of the Pats were still PTER after a year, rising to around 30 percent by 2011, where it has essentially remained.
Now, our hypothetical Pats are a pretty heterogeneous bunch. For example, they are different ages, different genders, different educational backgrounds, and in different industries. Do such differences matter when it comes to the chances of Pat finding a full-time job? For example, let’s look at Pats working in goods-producing industries versus services-producing ones. In goods-producing industries, the chance is greater that Pat will find full-time work (more jobs in goods-producing industries are full-time), and there is a bit more of a recovery in full-time job finding for goods-producing industries than for services-producing ones. But overall, the dynamics are similar across the broad industry types, as the charts below show:
As another example, the next four charts show the average 12-month full-time and PTER job-finding rates for all of our hypothetical Pats by gender and education. The full-time/PTER finding rates display broadly similar patterns across gender and education, albeit at different levels. (The same holds true across age groups but is not shown.)
People who find themselves working part-time involuntarily are having more difficulty getting full-time work than in the past, even if they stay employed. But it doesn’t seem that much of this can be attributed to any particular demographic or industry characteristic of the worker. The phenomenon is pretty widespread, suggesting that the problem is a general shortage of full-time jobs rather than a change in the characteristics of workers looking for full-time jobs.
By John Robertson, a vice president and senior economist, and
Ellyn Terry, an economic policy analysis specialist, both of the Atlanta Fed's research department
- Signs of Improvement in Prime-Age Labor Force Participation
- Could Reduced Drilling Also Reduce GDP Growth?
- Are Shifts in Industry Composition Holding Back Wage Growth?
- Are Oil Prices "Passing Through"?
- Business as Usual?
- What's (Not) Up with Wage Growth?
- Are We Becoming a Part-Time Economy?
- Contrasting the Financing Needs of Different Types of Firms: Evidence From a New Small Business Survey
- Gauging Inflation Expectations with Surveys, Part 3: Do Firms Know What They Don’t Know?
- Gauging Inflation Expectations with Surveys, Part 2: The Question You Ask MattersA Lot
- March 2015
- February 2015
- January 2015
- December 2014
- November 2014
- October 2014
- September 2014
- August 2014
- July 2014
- June 2014
- Business Cycles
- Business Inflation Expectations
- Capital and Investment
- Capital Markets
- Data Releases
- Economic conditions
- Economic Growth and Development
- Exchange Rates and the Dollar
- Fed Funds Futures
- Federal Debt and Deficits
- Federal Reserve and Monetary Policy
- Financial System
- Fiscal Policy
- Health Care
- Inflation Expectations
- Interest Rates
- Labor Markets
- Latin America/South America
- Monetary Policy
- Money Markets
- Real Estate
- Saving, Capital, and Investment
- Small Business
- Social Security
- This, That, and the Other
- Trade Deficit