A Tarnished Silver Lining: The Great Recession and Productivity

Bookmark and Share

Written by:
Lucia Foster and Cheryl Grim – Center for Economic Studies, U.S. Census Bureau
John Haltiwanger – University of Maryland and U.S. Census Bureau

Recessions are costly in terms of high unemployment, lost output, and lost consumption. Still there is long running debate about whether recessions may have a silver lining, a long term “productivity-enhancing” effect.

This productivity-enhancing effect can occur through the movement or “reallocation” of resources (e.g., workers) across firms. In a well-functioning economy, resource reallocation typically directs resources from less productive firms to firms that are more productive.  One strand of economic theory hypothesizes that the pace of this resource reallocation increases during recessions thus increasing productivity for the aggregate economy.

Alternatively, it is possible recessions will not be productivity enhancing due to market distortions. For example, even a high productivity firm might not be able to get credit needed to survive and grow if a recession distorts credit markets.

Previous research found evidence that recessions prior to the 2007-2009 “Great Recession” were productivity enhancing. However, the severity and persistence of the Great Recession and its close connection to the financial crisis suggest it may have been different. Did the Great Recession differ in its impact on the reallocation of resources across firms and the efficiency gains associated with such reallocation? To find out, we study recessions in the U.S. economy over the last 30 years.

The pace of job reallocation was low during the Great Recession

We start by looking at job reallocation, which is the sum of job creation and job destruction. The U.S. economy typically features a high pace of job reallocation across firms: the annual average job creation rate for the U.S. private sector over the last 30 years is close to 18 percent while the analogous job destruction rate is 16 percent.  These combine to yield an overall annual average job reallocation rate of 34 percent.


Figure 1. U.S. Job Flows and the Business Cycle, 1981-2011
Source: Authors’ calculations based on the Business Dynamics Statistics and the Current Population Survey.
Notes: Cycle is the change in the unemployment rate. Changes are from March of year t-1 to March of year t. Shaded areas are NBER recessions.

Figure 1 shows job creation was as low during the Great Recession as during any period in the last 30 years. By 2011, job creation had still not fully recovered to pre-recession levels.

Further, job reallocation was also at a historical low during the Great Recession. When job destruction peaked in 2009, job reallocation was 28 percent, in contrast with a 35 percent reallocation rate in 1983. Lower than usual (for a recession) reallocation suggests the productivity-enhancing effect may have been weakened in the Great Recession.

Reallocation in the Great Recession was less productivity enhancing than in other recent recessions

Figure 2. Differences in Employment Growth Rates Between High and Low Productivity Establishments Over the Business Cycle


Source: Authors’ calculations based on the Annual Survey of Manufactures, Census of Manufactures and Longitudinal Business Database.
Notes: The vertical bars in Figure 2 show the predicted difference in employment growth rates between high and low productivity establishments at five different points in the business cycle. A high [low] productivity establishment is one standard deviation above [below] industry-year mean productivity. Normal is zero change in state-level unemployment; mild contraction is a one percentage point increase in state-level unemployment; sharp contraction is a three percentage point increase in state-level unemployment; Great Recession is for the period 2007-2009.

Figure 2 shows the predicted difference in employment growth rates between high and low productivity establishments at different points in the business cycle. In all cases, high productivity establishments are more likely to grow than low productivity establishments (all of the differences are positive).

The productivity-enhancing effect of recessions prior to the Great Recession is evident in the first three bars: the difference between growth rates increases with the severity of the recession. In contrast, the difference between growth rates decreases with severity during in the Great Recession (compare the last two bars).

In sum, we find the Great Recession is less productivity enhancing than earlier recessions because its reallocation is less intense and less efficient.  The Great Recession was costlier than earlier recessions due to not only higher unemployment, more lost output, and more lost consumption, but also because there was less productivity-enhancing reallocation.

Our work here is about the “what” rather than the “why” of the Great Recession. In future work, we plan to examine why the Great Recession is less productivity enhancing than other recent recessions. One particular area of interest for future work is the role of the financial crisis in the Great Recession.

Want to know more? See our working paper “Reallocation in the Great Recession: Cleansing or Not?

Posted in Uncategorized | Leave a comment

Preliminary Results from America’s Churning Races: Race and Ethnic Response Changes Between Census 2000 and the 2010 Census

Bookmark and Share

Written By: Sonya Rastogi, PhD, Senior Researcher, Center for Administrative Records Research and Applications (CARRA), Carolyn Liebler, PhD, Assistant Professor of Sociology, University of Minnesota, Leticia Fernandez, PhD, Researcher, CARRA, and James Noon, Researcher, CARRA

At the 2014 Population Association of America meetings (May 1-3), we presented preliminary results of research (authored by Carolyn Liebler, Sonya Rastogi, Leticia Fernandez, James Noon, and Sharon Ennis) examining patterns of race and Hispanic origin response change at the individual level between Census 2000 and the 2010 Census.

Our research, America’s Churning Races: Race and Ethnic Response Changes Between Census 2000 and the 2010 Census, measures the extent of these changes.

Our preliminary findings have generated a lot of interest, and we are taking this opportunity to clarify what we have found so far.

Previous research (see references in abstract) suggests that responses to questions of race and Hispanic origin can change over time or across contexts for a variety of reasons, including life experiences, changing social forces, a change in who is reporting the race/Hispanic origin, or questionnaire design. The race and Hispanic origin response changes that we observe from 2000 to 2010 may be influenced by many of these factors, and some groups may be differentially impacted by questionnaire design changes from Census 2000 to the 2010 Census.

Our research paper documents the extent of the response changes and gives information about their patterns. After these patterns are well-documented, further research can be done about why responses are changing.

Our research used census responses that were linked to the same anonymized individual in both 2000 and 2010.  While this yields a rich and unique dataset, it is not representative of the total United States population.

Like many other demographers, sociologists, and census researchers, we found that some race and Hispanic origin responses can and do change between censuses.  In preliminary results, we found that race and Hispanic origin responses changed for about 10 million people (or 6 percent) out of the 168 million individuals in our linked dataset from Census 2000 to the 2010 Census.

Race response change was concentrated in some groups, specifically American Indians and Alaska Natives, Native Hawaiians and Other Pacific Islanders, people who reported multiple races, and Hispanics who reported a race. Responses were generally stable among single race non-Hispanic whites, blacks, and Asians in our data.

The most common response change was from Some Other Race (SOR) to single race white among those who identified (or were identified by someone in the household) as Hispanic in both Census 2000 and the 2010 Census.  The second most common response change was the inverse — from single race white to SOR for those who reported (or were reported as) Hispanic in 2000 and 2010.

Across most types of response change, we note that “inflows” to each race/Hispanic group are similar in size to the “outflows” from the same group, indicating that cross-sectional data might show a small net change. We also note that race and/or Hispanic origin response change occurs in a wide variety of ways. These included changes from multiple-race to single-race, from single-race to multiple-race, from one single race to another, or adding (or dropping) a Hispanic response.

Again, these findings are preliminary and cannot be considered final. We are still analyzing the data and plan to release more detailed results in a working paper in the coming months. We caution readers not to reach conclusions based solely on the preliminary research.  We look forward to talking about the final results when the process is complete.

Posted in Uncategorized | 2 Comments

Making Sense of Big Data to Enhance Official Statistics

Bookmark and Share

Written by: Nancy Bates, Senior Researcher for Survey Methodology, Research and Methodology Directorate

A recent story in the Washington Post,Want to take economy’s pulse? Try digging into Google, Twitter,” compared big data to traditional sources for statistics.

As the statistical agency responsible for many of the nation’s demographic and economic indicators, we are researching the use of big data from publicly available sources to enhance our traditional methods of gathering official statistics. In 2011, the National Science Foundation and the Census Bureau came together to fund discovery and innovation concerning measurement in the social and economic sciences – discoveries meant to match the rapid changes in today’s population and economy. This program is known as the NSF-Census Research Network or NCRN.

NCRN-funded research by Matthew Shapiro of the University of Michigan has shown that information from Twitter tracks job loss trends in real time. As the Washington Post reported, Shapiro and his colleagues “scoured more than 19 billion tweets over two years for references to unemployment, hunting for phrases such as “axed,” “pink slip” and “downsized.”

The NCRN program is just one example of the Census Bureau’s commitment to leverage advances in technology and social media to reduce cost to the public while maintaining high standards of data accuracy and quality. As the costs of traditional forms of data collection continue to rise, encouraging research that makes sense of big data sources is one avenue the Census Bureau is pursuing.

Research like the University of Michigan NCRN node may present new directions for statistical agencies by tapping into information that individuals, households and businesses create in the normal course of the day. Ultimately, it may be possible to collect more timely data at higher frequency, greater precision and lower costs.

Posted in Uncategorized | Leave a comment

Increasing our Understanding of Business Dynamics through the First-Ever Census Bureau Business Management Survey

Bookmark and Share

Lucia Foster, Chief, Center for Economic Studies
Ron Jarmin, Assistant Director for Research and Methodology

How much does management matter for the success of a business? Some light can be shed on this much debated topic using results from the Census Bureau’s first-ever business management survey, the Management and Organizational Practices Survey (MOPS). This supplement to the 2010 Annual Survey of Manufactures (ASM) collected data from more than 30,000 of the approximately 50,000 manufacturing plants surveyed in the ASM. With the addition of the MOPS, the Census Bureau now provides information on manufacturing plants’ inputs, outputs, and management and organizational practices. The MOPS will be an important tool in assessing plant performance, worker productivity, and the ability of plants to succeed in a highly competitive global economy.

The MOPS represents a collaboration between Census Bureau staff and researchers from Stanford University (Nick Bloom and Itay Saporta-Eksten), the Massachusetts Institute of Technology (Erik Brynjolfsson) and the London School of Economics (John Van Reenen). The National Science Foundation also contributed to the collaboration via a grant to the academic research team.

Preliminary results from the first research paper using the MOPS, Management in America, were presented at the annual American Economic Association meetings in early January. This first paper focuses on the management component of the MOPS and provides summary statistics from the new survey and empirical exercises intended to provide validation of the survey results. The authors constructed a management score that summarizes plants’ intensity of use of structured management practices based upon the responses to the 16 management questions. Structured management practices are those that are more specific, formal, frequent or explicit.

The authors found that use of structured management practices varies across U.S. manufacturing establishments: 18 percent of establishments adopt at least 75 percent of structured management practices related to performance monitoring, targets and incentives, while 27 percent of establishments adopt less than 50 percent of these practices.

The authors also linked the survey results to other Census Bureau (and outside) data sources in order to focus on how differences in management practices are related to variations in plant performance. Plant performance metrics include productivity, profits, output growth, and exports  and measures of innovation (research and development expenditures and patents, both per employee, calculated from the Business R&D and Innovation Survey).

The figure below shows the correlations between these measures of plant performance and structured management practices. For each panel, the establishments are grouped into deciles according to their management index scores (where 1 is the decile with the lowest scores and 10 is the decile with the highest scores).  The performance measures are on the vertical axis and are the median value for each management index decile. The figures show that all of the performance measures are rising across the deciles.

Plant Performance Measures and Structured Management Practices

Plant Performance Measures and Structured Management Practices

This analysis is refined in a regression framework controlling for firm and establishment characteristics (such as capital intensity, size, education of workforce, and industry). This basic positive correlation persists even with these controls. The results provide support that the MOPS is systematically capturing meaningful content about management practices rather than just statistical noise. Future research will address causality, thus shedding light on how much management matters.

Users will be able to access MOPS results from a website that will host summary tables, a benchmarking tool where businesses can compare themselves to their peers and public-use version of the data (subject to Census Bureau approval). In addition, the full data set will be available to qualified researchers on approved projects via the Census Bureau’s secure Research Data Centers.

More on the Management and Organizational Practices Survey is available at <www.census.gov/mcd/mops>.

Management in America by Bloom, Brynjolfsson, Foster, Jarmin, Saporta-Eksten, and Van Reenen is available at <http://www2.census.gov/ces/wp/2013/CES-WP-13-01.pdf>.

Posted in Uncategorized | 2 Comments

Statistics Matter

Bookmark and Share

Written By: Thomas A. Louis, PhD, Associate Director for Research and Methodology

Statistics touch every part of daily life and provide measures of most everything – the rise and fall of the tides, the size of communities and of the U.S. economy, the probability of storms, the balance of trade, the prevalence of disease, the financial cost of hurricanes, commuting patterns and time use, the effectiveness of medical treatments, performance in sporting events, the health effects of environmental exposures; plus thousands of other aspects of human behavior and natural phenomena.  The substantial societal benefits of these measures depend on their quality and relevance.

International Year of Statistics
To draw attention to the value statistics play in our lives and the importance of our profession, the statistical community has designated 2013 “The International Year of Statistics.”  Statistics is the science of learning from data, and of measuring, controlling and communicating uncertainty. It provides the navigation essential for directing the course of scientific and societal advances.

Statistics and statisticians will play increasingly important roles as complex “big data” inform and empower our future. How will society mine the haystacks of information on social networks, time-use, economic, and other activities to benefit science and business? The answer is sound statistical practice.

Statistics inform public policy
A few examples: Each year billions of dollars are allocated to school districts based on the Census Bureau’s county-specific estimates of income and poverty, produced by combining information from the most recent decennial census, from the Current Population Survey, from the American Community Survey and administrative records.  Municipalities use these and other data sources to make decisions on transportation infrastructure.  The nation uses Economic Census statistics in setting the industry benchmarks that shape the Gross Domestic Product, our best indicator of economic health.

Growing demand
Academia, business, government, and individual stakeholders increasingly rely on data-driven decisions.  As Marie Davidian (current president of the American Statistical Association) and I highlighted in our editorial in Science (Vol. 336, April 6, 2012), substantially more statisticians and other data scientists are needed to meet the burgeoning demand to develop valid information and make sense of the data tsunami.   Success will depend on novel statistical designs and analyses, and on innovative communication strategies.

Our data-rich future demands that scientists, policy-makers, and the public be able to interpret increasingly complex information and recognize both the benefits and potential pitfalls of statistical information.  Consequently, it is a good sign that there is a strong push to promote statistics as a key component in precollege education.  We must encourage students to develop skills in describing data, developing statistical models, making inferences, evaluating the consequences of decisions, and asking questions that help calibrate quality.  These are skills that students will use throughout life, whatever their careers.  A data-driven future awaits, and statisticians must lead the way.

In my role as Associate Director for Research and Methodology at the U.S. Census Bureau, I will support our talented researchers in developing new approaches that ensure the Census Bureau remains a world leader in achieving the highest attainable quality of our statistical products. Through substantive collaboration we identify the highest priority issues, develop and evaluate approaches, then transfer the best to practice, thereby ensuring that Census statistics continue to support the public good. Visit Research@Census to learn more.

Posted in Uncategorized | 1 Comment

Trends in Health Insurance Premiums for Public and Private Employers

Bookmark and Share

Written by: Alice Zawacki, Senior Economist, Center for Economic Studies

Undoubtedly, you have seen headlines or heard reports in the media comparing employee benefits between the public and private sectors.   An important component of employee benefits is health insurance.  In a current project, Tom Buchmueller (University of Michigan), Jessica Vistnes (Agency for Healthcare Research and Quality), and I are using data from the Medical Expenditure Panel Survey-Insurance Component (MEPS-IC)  to analyze recent trends in health insurance premiums and benefits for public and private sector employers.

Looking at publicly available estimates based on the MEPS-IC, we found that the gap between premiums for the public sector (state and local governments) and private employers grew dramatically from 7.5 percent in 2000 to 20.5 percent just nine years later.  The figure below shows this growing gap in premium costs for enrollees.  In 2009, the single premium per employee enrolled in state and local government health plans was $5,627 versus $4,669 for plans offered by employers in the private sector.  A more detailed analysis (not shown) indicates that the higher growth in premiums in the public sector was driven by rising premiums for local government establishments.

One possible explanation for this divergence is that private sector employers responded more to increases in health care costs and the financial pressures brought on by the Great Recession.  In our ongoing work using the MEPS-IC microdata, we will examine whether private sector employers were more likely to alter benefits in order to “buy down” health insurance premiums.  In particular, we will test the extent to which benefit changes can explain the growing gap illustrated in the figure.  Look for future postings in the Research Matters blog with more details on our results.

Average Total Single Premium per Employee Enrolled

Average Total Single Premium per Employee Enrolled

Source: Agency for Healthcare Research and Quality. 1996-2009 Medical Expenditure Panel Survey-Insurance Component Internet tables I.C.1. and III.C.1. http://meps.ahrq.gov/mepsweb/data_stats/quick_tables_search.jsp?component=2&subcomponent=1

The Census Bureau collects the MEPS-IC under sponsorship of the Agency for Healthcare Research and Quality.  The Census Bureau sponsors or co-sponsors the collection of data on health insurance in other surveys, including the American Community Survey (ACS), the Current Population Survey Annual Social and Economic Supplement (CPS ASEC), and the Survey of Income and Program Participation (SIPP). For more information, see the Census Bureau’s Health Insurance webpage.

Posted in Uncategorized | Leave a comment

Falling House Prices and Labor Mobility

Bookmark and Share

Written by:  Christopher Goetz, Economist, Center for Economic Studies

Has the recent housing bust, which left approximately a third of households with negative equity, “locked” workers into their current home and unable to move for new jobs? America has long been known as a place where people are willing to relocate for new opportunities. However, obstacles to labor mobility, such as house lock, may discourage workers from changing jobs and prevent the unemployed from finding a job, possibly prolonging the slow economic recovery.

Previous studies addressing this issue lack the scope and detail to distinguish between the impact of house prices and the general effects of the recession, which also reduce migration.  My research attempted to explain this.

Using a data source with a large sample size, such as the American Community Survey (ACS), can help get around this problem, but the ACS is a snapshot that doesn’t allow us to observe people moving during one time period to the next. However, by merging employment information from the Longitudinal Employer-Household Dynamics (LEHD) jobs database, I can determine if ACS respondents later begin a new job in a different city.

Using this matched dataset, I can observe many people located in a particular metropolitan area during the same time period. This enables me to compare homeowners to renters, and see how changes in their home price differentially affect their probabilities of relocating to another city for a new job in the LEHD data. Because these workers are all exposed to the same local economic conditions, if the mobility of owners appears to decline more compared to renters when the value of their home has fallen, then we can infer that the difference is due to the changes in the owners’ home equity.

To estimate whether a respondent has negative equity in their home, I used historical price information from a real estate company to see if the value of the individual’s house has declined since the date that they moved into it.  Renters effectively serve as the control group in this setup because their migration behavior should not be directly impacted if the value of the house they are renting falls.

Results from statistical analyses using this strategy on data from 2002 to 2010 show that a homeowner with negative equity was about 20 percent less likely to move for a new job. The impact is similar if we look only at the unemployed.  For context , the attached  figure shows  that the migration rate of homeowners fell by about 60 percent over the studied time period (from 1.7 percent per quarter  to 0.7 percent ), while that of renters fell by less than half.

Owners and Renters

Note also that the migration rate fell for both owners and renters starting in 2005. During the depths of the housing crash in 2009-2010, the patterns for the two groups really began to diverge widely. This means that while migration fell in large part because of the general economic decline, the housing crash put additional downward pressure on the mobility of homeowners and could continue to do so until the market recovers.

Posted in Uncategorized | Leave a comment

Fewer Children are in Private Schools, More in Charters and We’re Looking at Possible Links

Bookmark and Share

Written by: Stephanie Ewert

The majority of  U.S. schoolchildren have always attended public schools, but private schools have also educated significant numbers of children. Parents send their children to private schools for a variety of reasons, including the availability of advanced programs and extracurricular activities, religious reasons, dissatisfaction with local public schools, class size and student-teacher ratios. But as tuition costs have risen, and public charter and magnet schools have emerged, has enrollment in private schools continued to grow? Recent research suggests the answer might be “no.”

Data from several surveys show that, while overall school enrollment has been increasing, a decline in private school enrollment took place in the last decade (Figure 1). Based on the Current Population Survey, the number of students enrolled in private school, kindergarten to grade 12, went from 5.4 million in 2002 to 4.5 million in 2010.  The decline in private school enrollment occurred at all school levels but was concentrated among schools that were larger, religiously affiliated, and in cities and suburbs.

Students age 3 and older enrolled in private school, 1989-2010

Limited data make it difficult to uncover the causes of the decline in private school enrollment. However preliminary analysis suggests that growth in charter schools may be a related factor.

We compared data on private school enrollment from the Census Bureau’s American Community Survey with data on charter schools from the National Alliance for Public Charter Schools and found that the majority of states in the U.S. with a decline in private school enrollment also experienced an increase in charter school enrollment. These preliminary results call for additional research on the relationship between private and charter school enrollment.

Current data limitations prevent us from evaluating whether growth in home schooling is a factor, and the data do not suggest that the recession beginning in December of 2007 precipitated the decline. We will collect information on private, charter, and home school enrollment in the Re-engineered Survey of Income and Program Participation and perhaps this will allow us to answer some of those questions.

For more details on trends in private school enrollment and possible factors related to the observed decline, see our working paper “The Decline in Private School Enrollment.”


Posted in Uncategorized | 2 Comments

Collecting Data on Governments – Innovation at Work!

Bookmark and Share

Written by: Carma Hogue, Assistant Division Chief, Governments Division

Today government finance, public pensions, education spending, and taxes are hot issues and in the information age – where information is readily available and more easily monitored and measured – statistics tell the stories. 

2009 State and Local Government Expenditures

The U.S. Census Bureau’s Governments Division collects data on federal, state and local government and constantly researches new ways to make data collection more efficient and the data more precise.

On March 15, 2012, the Council of Professional Associations on Federal Statistics held a workshop on censuses and surveys of governments.  Attendees at the conference included representatives from academia, the private sector, several federal statistical agencies, and members of a 2007 Committee on National Statistics panel on government statistics.

Governments Division staff presented their research, as well as planned research, on a host of topics.  We believe many readers will find this research to be of interest:

For more information on the papers, see http://www.census.gov/govs/pubs/research_reports.html

Posted in Uncategorized | Leave a comment

Using Historical Census Data to Reveal Migration Patterns of the Young, Single, and College Educated

Bookmark and Share

Written by: James Fitzsimmons, Assistant Division Chief, Population Division

Between 1965 and 2000, the young, single, and college-educated population in the United States—the “YSCE” population—migrated in patterns that were often at odds with those of other segments of the nation’s population.

In general, larger metropolitan statistical areas were more likely to have consistent net in-migration of the YSCE population, while smaller metros, micropolitan statistical areas, and areas outside of metros and micros were more likely to experience YSCE net out-migration.  These findings were often opposite those for the total population. Within metro areas, migration to principal cities also was a hallmark of the YSCE population.

Other findings reported in the recently released Population Division working paper Historical Migration of the Young, Single, and College Educated: 1965 to 2000, authored by Justyna Goworowska and Todd Gardner, included the fact that less than one-fifth of states saw consistent net in-migration of the YSCE population during that period.  About half of states, on the other hand, experienced consistent net out-migration of the group.

The working paper’s focus on migration of the YSCE population, a group with outsized human capital and potential impact on population growth, was possible thanks to the Census Bureau’s Historical Census Files Project. That project has recovered all available microdata from the 1960, 1970, and 1980 censuses, and it is in the process of harmonizing these files with ones from the 1990 and 2000 censuses.

The central outcome of the historical files project is a time series of anonymized historical decennial census microdata files available to researchers within the Census Bureau as well as to those with approved projects through the Census Bureau’s national network of secure Research Data Centers.

An eventual project goal is to extend the historical microdata holdings to earlier censuses, but at present the full range of data gathered from the “long form” of five consecutive censuses, along with documentation, is at hand for researchers with approved projects.  In its analysis of migration patterns of the YSCE population, the working paper has shed light on only one of a long list of potential subjects that would lend themselves to further study with the historical microdata series.

Posted in Uncategorized | Leave a comment