Revolutionizing Economic Measurement At The Census Bureau’s Center For Economic Studies: Pioneering Research On Firm Dynamics, Employment Flows And Transformative Statistical Products

In the late 1980s and early 1990s, the Census Bureau’s Center for Economic Studies (CES) became an epicenter for groundbreaking research that would reshape how economists understand business and labor market dynamics. Mark and John, as visiting researchers, played central roles in this transformation. Their influential work along with collaborators continues to inform economic thought and policy, catalyze countless follow-on studies in the U.S. and abroad, and has been referenced in Nobel Prize citations.

Mark came to CES as an American Statistical Association / National Science Foundation Fellow in 1985 to work with what was then called the Longitudinal Establishment Database (LED). The LED, which linked establishment data from the Annual Survey of Manufactures (ASM) and the Census of Manufactures (CM) for the years 1972 to 1981, was developed just a few years prior as a collaboration between the Census Bureau and Yale University economists Nancy and Richard Ruggles. The LED was a bear to use on the Census Bureau’s UNIVAC mainframe and CES was established in 1982 to facilitate improvements in, and access to, the LED and similar datasets.

Mark and his then graduate student, Tim Dunne, initially intended to use the LED to study capital investment in U.S. manufacturing with their colleague Larry Samuelson. It didn’t take them long to realize that a key strength of the data was the analysis of firm entry and exit and firm growth – something economists and policymakers were keenly interested in, but for which systematic empirical evidence was scarce.

Mark, Tim and Larry restricted their analysis to the quinquennial Census of Manufactures to get the full view of firm and establishment entry and exit. The Census Bureau added the 1982 CM to the LED, but Mark and colleagues also added the 1963 and 1967 CMs. They encountered issues with matching establishments and firms across CM years and many other data problems in files not processed with microdata analysis in mind. Their careful attention to data quality set the standard for CES researchers and influenced core Census Bureau processing moving forward.

All this hard work allowed them to provide stark documentation of firm heterogeneity and business dynamics. One of their landmark papers, “Patterns of Firm Entry and Exit in U.S. Manufacturing Industries” (Dunne, Roberts & Samuelson, 1988, RAND Journal of Economics), analyzed new and exiting firms, challenging the static view of industry structure. They showed that even within narrowly defined industries, new entrants differed in both their initial and post-entry characteristics.

Another key paper, “The Growth and Failure of U.S. Manufacturing Plants” (Dunne, Roberts & Samuelson, 1989, Quarterly Journal of Economics), offered deep insights into plant-level growth and survival, demonstrating the prevalence and economic significance of turnover. In these papers, Mark and team were able to highlight that net changes are much smaller than gross changes in the number of firms. For example, published aggregate statistics might show the number of firms in an industry increasing from 20 in one period to 22 in the next. A reasonable inference, common at the time, would be that two new firms entered. Mark and colleagues demonstrated that it was far more likely that this net change of two was driven by something like eight entering and six exiting firms.

An important insight from their research was that establishment and firm dynamics drive both net and gross employment growth. Like the example above, net employment change in the economy is dwarfed by gross job flows associated with business dynamism.

Arriving at CES shortly after Mark and colleagues, John Haltiwanger started a project examining gross job flows with Steven Davis and Scott Schuh, then a graduate student. Researching job flows and the business cycle, they focused on the quality of the linkages of cross-sectional data from the ASM and CM. As Mark and team found, the Census Bureau processing of the ASM and CM at the time did not prioritize tracking establishments and firms across survey years. This resulted in too many observed entries and exits and biased measures of firm growth. Working closely with CES staff, in particular Tim Dunne and Jim Monahan, as well as other Bureau experts, John and colleagues carefully analyzed and then improved the linkage methodology – taking care to account for ASM sampling strategies - and added additional years leading to the creation of the Longitudinal Research Database (LRD). The LRD would become the workhorse dataset at CES for the next decade.

They produced several highly influential papers systematically documenting the magnitude and variation of gross flows across detailed manufacturing industries. Importantly, they demonstrated that idiosyncratic (micro) factors were more important in explaining the time series of aggregate net employment growth than were aggregate (macro) or even sectoral factors fundamentally changing economists and policymakers understanding of business cycles. The work that would become their authoritative 1996 book, Job Creation and Destruction, informed the discussions at the 1994 G-7 Jobs Summit. This high policy visibility encouraged efforts to construct statistics on job flows and business dynamics around the world.

John and his colleagues did much to nail down the basic methodology and terminology of research on business and employment dynamics that is still prevalent today. Their work was critical in putting forward the view the producer heterogeneity and business dynamism at the microeconomic level were key to understanding business cycles at the macroeconomic level.

The work of Mark, John and their colleagues overwhelmingly demonstrated the value of previously scarcely analyzed micro data at statistical agencies. Similarly, people recognized the important role of organizations like CES that curated such data and made it available to expert researchers.

The empirical revelations delivered by these papers made statistical agencies, especially the Census Bureau, reconsider how published aggregate statistics could obscure the real dynamism of the economy.

When Bob McGuckin, the visionary CES chief who helped facilitate all this work, left in 1996, John stepped in to lead CES as the Census Bureau’s chief economist. At the same time Tim Dunne rejoined CES as research director. During their two-year tenure, work began on two important infrastructure projects that produced better data for researchers as well as transformative public-use data products.

A shortcoming of the LRD was that it was limited only to manufacturing – a sector declining in importance to the broader U.S. economy then as now. At CES, Javier Miranda and I began the work to address that concern by longitudinally linking annual snapshots of the Census Bureau’s Business Register (then known as the Standard Statistical Establishment List). The result was the economy-wide Longitudinal Business Database (LBD) which has become a staple in applied economic research (including the many papers John and I wrote with Javier and other coauthors over the years) and underlies the public-use Business Dynamics Statistics (BDS).

At this same time, John along with John Abowd and Julia Lane led a team at the Census Bureau to assemble linked employer-employee data (as had been done in other countries, including France). The Longitudinal Employer Household Dynamics (LEHD) program used state unemployment insurance data to build key products like the Quarterly Workforce Indicators (QWI), OnTheMap and OnTheMap for Emergency Management and led to several innovations at the Census Bureau in computing, privacy protection and product provision.

The early work of Mark Roberts and John Haltiwanger and collaborators established the Center for Economic Studies as a nexus for empirical innovation and economic measurement. Their publications, and the rich literature they influenced, set the standard for how economists analyze firm and labor market dynamics. Their legacy is visible in both the structure of economic inquiry and the statistical infrastructure underpinning policy analysis.