John Haltiwanger: recipient of the 2020 Global Award for Entrepreneurship Research

The Global Award for Entrepreneurship Research was initiated in 1996 and has since become the most prestigious prize in entrepreneurship research. It consists of 100,000 Euros and a statuette designed by the internationally renowned Swedish sculptor Carl Milles.

According to the original statutes, the award should be given to “a person who has produced scientific work of outstanding quality and importance, thereby giving a significant contribution to theory-building concerning entrepreneurship and small business development, the role and importance of new firm formation and the role of SMEs in economic development.” The main aims of the award are (1) to highlight the importance of research produced in the areas of entrepreneurship and small business, (2) to further stimulate and promote research within these fields, and (3) to diffuse state-of-the-art research among scholars, practitioners, and people involved in small business development.

The domain of entrepreneurship research is broad (Carlsson et al., 2013), which means that entrepreneurship research that can be considered for the award is undertaken in several different disciplines, including economics, management, sociology, history, business administration, geography, and psychology. Any aspect of entrepreneurship research is eligible, including the environment and the organizations in which entrepreneurship is conducted, the character of the entrepreneur (personality, cognitive and affective aspects), or the role of the entrepreneur and/or the entrepreneurial function in a wider sense (at the level of the community, region, country, or industry).

One ambition of the Prize Committee is that the award-winning contributions, seen together over a longer time span, reflect the extraordinary breadth of entrepreneurship as a research field in the social sciences. The key criteria for prize-worthy contributions are originality and influence (Braunerhjelm & Henrekson, 2009). It is recognized that contributions can be influential in many ways. A contribution can, for example, be influential because it has had a significant impact on subsequent scientific work, furthered entrepreneurship as a field (through creating important data bases or by starting influential journals, scientific communities, etc.), furthered entrepreneurship education and training at the academic level, and/or influenced policymaking and society more broadly.

When selecting prize-worthy contributions, the prize committee emphasizes the qualitative aspects of the contributions of candidates. Quantitative metrics, such as citation counts and impact factor-adjusted publication volumes, do provide important information about candidates, but they will never replace qualitative judgment. This means that quantity will never substitute for quality, and it is even possible for a scholar to receive the award for a single landmark contribution.

John Couch Haltiwanger was born on 12 December 1955, and is a citizen of the USA. He obtained a B.S. in Applied Mathematics and Economics from Brown University in 1977, and a Ph.D. in Economics from Johns Hopkins University in 1981. He then held academic positions on the faculty of University of California, Los Angeles (UCLA), and Johns Hopkins University, before joining the faculty at the University of Maryland in 1987, where he is currently the Dudley and Louisa Dillard Professor of Economics and Distinguished University Professor of Economics. Haltiwanger is a member of the Federal Economics Statistics Advisory Council, and is a research associate at both the Institute for the Study of Labor (IZA) and the National Bureau of Economic Research (NBER). Haltiwanger is also a Fellow of the Society of Labor Economics and the Econometric Society and has previously served on the Editorial Board of Small Business Economics.

John Haltiwanger’s career is also marked by important contributions through public service. In the late 1990s, Haltiwanger served as Chief Economist of the US Census Bureau. He currently continues his collaboration with the US Census Bureau through his role as Senior Research Fellow at its Center for Economic Studies. In addition, he has been a member of the Committee on National Statistics for the National Academy of Sciences, as well as a consultant for the Conference Board, and also a member of the Brookings Panel on Economic Activity.

Haltiwanger has received a number of honors and awards for his research. His 1996 book with Steven Davis and Scott Schuh on Job Creation and Destruction received widespread critical acclaim and also won the Choice Outstanding Academic Title in 1996. A subsequent book with Clair Brown and Julia Lane (Brown et al., 2006) won the same prize in 2007. He was also awarded the Julius Shiskin Memorial Award in 2013, and the 2014 Roger Herriot Award for Innovation in Federal Statistics (along with his co-authors John Abowd and Julia Lane), in recognition of his work with statistical agencies on the development of statistical and measurement methods to study firm dynamics.

John Haltiwanger has published over 100 academic articles in the world’s most prestigious economics journals such as the American Economic Review, the Quarterly Journal of Economics, the Review of Economic Studies, the Review of Economics and Statistics, and the Journal of Economic Literature. In addition to these journal articles, he has published numerous book chapters and books. These publications have been highly influential in the areas of economics, management, and entrepreneurship.

Haltiwanger’s contributions to our knowledge of job creation and destruction are primarily found in his prize-winning book (Davis et al., 1996a), which summarizes and extends the authors’ previous research on the topic (such as Davis & Haltiwanger, 1992).

Despite the importance of the topic of job creation, and ubiquitous concerns whether the turbulent economy would be capable of creating sufficient jobs, there was little existing evidence on the scale of job creation and destruction. The book, which is empirical in nature, therefore made an important contribution to the gap in the literature. This valuable evidence was obtained thanks to the painstaking efforts of Haltiwanger and colleagues to assemble the necessary database. Collaborating with staff at the Center for Economic Studies, US Bureau of the Census, Haltiwanger and colleagues further developed the Census of Manufactures and the Annual Survey of Manufactures into the Longitudinal Research Database (LRD). This resulted in a database of over 160,000 manufacturing plants and their parent firms. This in itself was a notable achievement: “the assembling of such a database is clearly an immense effort for which the authors and their co-workers deserve unqualified praise” (Haskel, 1998, p. 155). As entrepreneurship research has shifted emphasis from the analysis of cross-sectional data towards longitudinal data (Davidsson & Wiklund, 2000), these data collection efforts are commendable.

The book includes several facts and results on job creation, job destruction, and job reallocation. Main results in the book include:

The book received widespread critical acclaim, heralded by many as a seminal contribution. The book review by Blanchflower (1997, p. 1400) writes that: “This is an important piece of work. Not many books start literatures. This one is likely to. Buy it.” Haskel (1998, p. 156) considers that it is “a book which is a definitive documentation of job creation and destruction in the United States and has already proved to be the starting point for a rich body of work. How many other books can claim to be so influential in their field?” Burgess (1997, p. 1567) writes: “This is a seminal book, one which will make (indeed has already made) a significant change to the way in which macroeconomics and labor economics are thought about. It is a book that any economist would be pleased to have written: scholarly yet not turgid, new and controversial yet measured and reasonable, profound yet accessible.”

One of the prominent findings from the 1996 book was to dispel the myth of the job-creating prowess of small business (see in particular Davis et al., 1996a, Chapter 4). Haltiwanger and colleagues show that the conventional wisdom about the job-creating prowess of small business comes from statistical fallacies and misleading interpretations of the data. This issue was addressed head-on in Davis et al. (1996b). In this article, the authors show that the conventional wisdom of the job-creating prowess of small business is driven by statistical fallacies, analysis of flawed data, confusion between gross and net job creation rates, and misleading interpretations of the data. Instead, they argue that it is large plants and firms that contribute most of the newly created manufacturing jobs. 

This in itself spawned a subsequent re-evaluation in the literature into the relative contribution of small and large employers (Davidsson et al., 1998; Neumark et al., 2011).

More recently, the debate has moved on from comparing the job creation of small vs large employers to the job-creating prowess of small firms vs young firms, thanks to the new impetus provided by a relatively recent article (Haltiwanger et al., 2013). Indeed, most small firms are young firms, but a rigorous statistical analysis can disentangle the two variables. The novelty in Haltiwanger et al. (2013) was not to suggest that firm growth rates may depend on age, because this had already been suggested by others, such as Fizaine (1968). Instead, the novelty was to provide high-quality large-sample evidence, involving data rich enough to zoom in on developments taking place in the first few years of a firm’s existence. This contribution was made at a time when large-sample databases with information on firm age were rare (Decker et al., 2014).

Haltiwanger et al. (2013) turned the conventional view on its head, by showing that it is young firms, rather than small firms, that create the disproportionately large number of jobs. Controlling for age, they find that there is no longer any relationship between size and job creation. This has major implications for policy: job creation should be supported by targeting young firms rather than small firms. Small firms that are old should not be expected to create many jobs.

Haltiwanger et al. (2013) therefore initiated a debate on the relative contribution to job creation by small firms vs young firms. There have been many follow-up papers, including by entrepreneurship scholars, as researchers try to replicate, confirm, and extend his analysis. For example, focusing only on papers published in Small Business Economics that replicate and extend Haltiwanger et al. (2013), we can mention the multi-country studies by Ayyagari et al. (2014), Navaretti et al. (2014), Anyadike-Danes et al. (2015), and Criscuolo et al. (2017), as well as Lawless (2014) on Irish firms and Coad et al. (2018) on Swedish firms.

Another major contribution of Haltiwanger et al. (2013) is that most of the age-related effects of job creation seem to disappear after about 5 years. This interesting finding has raised the bar for empirical investigations into firm age and performance. Previously, entrepreneurship research had considered that firms with an average age of 12, 16, or 18 years were suitable samples for the analysis of the initial conditions of entrepreneurial firms (see the empirical literature that is surveyed in Bamford et al., 2004, Table 1). After Haltiwanger et al. (2013), researchers interested in age-related effects must have representative data on firms observed from their very first years.

Another area of research pioneered by Haltiwanger and colleagues regards the relative performance of entrepreneurial firms in terms of their contribution to productivity growth. More specifically, Haltiwanger and colleagues decompose productivity growth in order to distinguish between the contributions from continuing, entering, and exiting plants or firms. This decomposition technique sheds light on the relative importance of underlying processes of learning, selection, reallocation, and creative destruction.

Foster et al. (1998) put forward the following decomposition of productivity growth: where \(\Delta {PROD}_{t}\) corresponds to the share-weighted growth of productivity, for the period ending at time t. Plants or firms are indexed by i and may be classified as either continuing (C), entering (N), or exiting (X). θ denotes the activity shares attributed to plant or firm i. Bars over variables indicate that the average has been taken over all the plants or firms. Productivity can be measured either in terms of labor productivity (LP) or total factor productivity (TFP).¹

The right-hand side of the equation decomposes productivity growth (\(\Delta {PROD}_{t}\)) into 5 terms, which (respectively) correspond to the within-effect, the between-effect, the covariance term, the contribution of entry, and the contribution of exit.

The contribution of continuing firms can be broken down into within effects, between effects, and a covariance term. Within effects correspond to changes in productivity within a firm, holding constant its market share. Within effects therefore correspond to the productivity growth that occurs within an individual firm, due to learning over time. Between effects correspond to the reallocation that occurs due to changes in the shares of plants. If the between effect is positive, this means that productive capacity is being reallocated from the least efficient towards the more efficient firms, in line with evolutionary theory (e.g., Nelson & Winter, 1982). In contrast, if the between effect is negative, then the least productive firms are growing faster than the more productive firms. The covariance term, or cross term, will be positive if growing firms experience productivity growth as a consequence of their growth (e.g., if growth leads to economies of scale), and negative if firms with productivity increases have decreasing employment shares (e.g., if productivity growth occurs via downsizing).

Foster et al.’s (1998) methodology is considered to be the most common methodology for the decomposition of productivity growth (Holm, 2014). Overall, results of this type of analysis suggest that the largest component of productivity growth is within-plant or within-firm learning. About one-third to one-half of the total productivity growth comes from the within component – according to which continuing plants or firms improve their productivity over time (holding size constant). Hence, an important mechanism for productivity growth is experience and learning about more efficient production techniques and more profitable exploitation of existing market opportunities. The second largest component of productivity growth seems to be the contribution due to net entry, which is due to entrants with above-average productivity, or the exit of firms with below-average productivity.

It is also worth mentioning a few publications by Haltiwanger that focus specifically on estimating the productivity of small firms. Baily et al. (1996) investigate the relationship between downsizing and productivity growth and find that continuing plants that have productivity growth do on average have decreases in employment. Haltiwanger et al. (1999) investigate how productivity relates to firm size and firm age.

Some of Haltiwanger’s more current research documents the decline in the pace of creative destruction and business dynamism in recent decades. This research provides a useful backdrop for contextualizing the trends and developments in contemporary entrepreneurship research and is of clear interest to policymakers. While Haltiwanger and colleagues focused on the USA, other scholars have confirmed these findings for other countries such as Belgium (Bijnens & Konings, 2018).

In Decker et al. (2014), the authors observe that “the share of US employment accounted for by young firms has declined by almost 30 percent over the last 30 years” (Decker et al., 2014, p. 4). This has occurred alongside a decrease in the rate of business startups and a decline in the pace of “business dynamism” – measured as the pace of job creation and job destruction and flows of workers across firms (Decker et al., 2014). Furthermore, there has been deceleration in the activity of high-growth young firms since 2000 (Decker et al., 2016b).

Declining business dynamism is a serious problem because a dynamic economy requires the continuous reallocation of the factors of production towards the more efficient firms. If there is a decline in the rate of reallocation of workers towards higher-productivity opportunities, this will slow down productivity growth at the aggregate level (Decker et al., 2016a). Furthermore, if young firms with growth opportunities are finding it harder to grow fast after entry, this could be an obstacle to innovation and transformative entrepreneurship.

One of the problems facing research into firm growth is the non-trivial decision of how to measure growth. A common approach (e.g., Friedman, 1992) is to measure growth in terms of the change in an indicator of size (such as total sales or number of employees) denoted here as X, from time t to time t + 1, scaled down by initial size, i.e.:

However, a problem with this growth indicator is that it is not symmetric (Tornqvist et al., 1985): a firm growing from 10 employees in t, to 20 employees in t + 1, and then back to 10 employees in t + 2 to end up at its initial size, would be attributed a growth rate of 100% in t + 1 but a growth rate of − 50% in the next period. This growth indicator could therefore exaggerate the growth rates of growing small firms. The problem arises because growth rates are calculated based on initial size X_t.

Figure 1 illustrates how the regression fallacy can exaggerate the job creation prowess of small firms. In the case of growth from time t to time t + 1, the initial size is 450 employees, and the final size is 550. The growth rate would usually be calculated as (550–450)/450 = 0.22. Furthermore, given that initial size was below the size class threshold, this would be classified as a growing SME. In the case of growth over the period (t + 1:t + 2), initial size is 550 and final size is 450. The growth rate would be − 0.18, and this firm would be classified as declining large firm, because its initial size (at t + 1) was above the size threshold of 500. In this example, growth is mistakenly attributed to SMEs, and decline is mistakenly attributed to large firms. The problem therefore arises when continuing firms are assigned to a size category according to base year employment. Furthermore, the growth rate of the “SME” (0.22) is larger in magnitude than the growth rate of the “large” firm (− 0.18), even though the firm is the same size in years t and t + 2.

Haltiwanger and colleagues also explain how other statistical fallacies can be misinterpreted as evidence of the job creation prowess of small business (Davis et al., 1996a, chapter 4). One such example would be the size distribution fallacy, which results from confusion about how to classify the share of employment of firms that migrate across size class boundaries. Another fallacy is the confusion between net job creation and gross job creation: small firms may make a large contribution to net job creation but a small contribution to gross job creation. Nightingale and Coad (2014) contains a further discussion of how other statistical fallacies all err on the side of exaggerating the job creation prowess of small business.

Haltiwanger and colleagues therefore put forward a new growth indicator (e.g., Davis & Haltiwanger, 1992, p. 825):

This growth indicator scales down the growth amount \({X}_{t+1}- {X}_{t}\) by the average size \(\frac{1}{2}\left({X}_{t}+ {X}_{t+1}\right)\). It ranges from a theoretical minimum of − 2 to a maximum of + 2. One statistical feature of this growth rate indicator is that it can take into account entry and exit events. When a firm exits, its final size \({X}_{t+1}\) is equal to zero; therefore, its growth rate is calculated as − 2. Similarly, when a firm enters, its initial size \({X}_{t}\) is equal to zero; therefore, its growth rate is calculated as + 2.²

Applying the DHS growth indicator to the example in Fig. 1, the regression fallacy is avoided, because the growth rates for the periods (t:t + 1) and (t + 1:t + 2) are the same in magnitude, and there is no problem of attributing growth to smaller firms and decline to larger firms.³

Haltiwanger’s research has popularized the term “up-or-out” dynamics to refer to the post-entry struggles of new firms. “Up-or-out” dynamics here are interpreted as the dynamics of young firms who face uncertainty about their abilities (building on work by the 2019 prize winner of the Global Award for Entrepreneurship Research Boyan Jovanovic, Jovanovic, 1982) as well as a shifting business environment and dynamic forces of selection and learning effects. Young firms with superior capabilities and routines move “up” in terms of size and performance (e.g., productivity), while young firms that are revealed to have inferior capabilities are more likely to decline and exit from the market (Huber et al., 2017). Indeed, while the majority of new firms will exit in their first 5 years, the remaining survivors will show considerable average growth conditional on survival (Haltiwanger et al., 2013).

Foster et al. (2016) takes an alternative perspective to the post-entry performance of new establishments. Here, the authors put forward a dynamic model of post-entry growth that is driven by a demand accumulation process (e.g., building a customer base). This serves as a reminder to entrepreneurship scholars that the growth of new firms and new plants is not simply driven by setting up production capacity, but that it crucially depends on the steady accumulation of customers. The authors analyze a sample of US manufacturing plants in industries that are carefully selected because of the commodity-like nature of their products – e.g., ready-mixed concrete, cardboard boxes, and manufactured ice. Even in these sectors, new plants are smaller than established plants, and interestingly, these differences are not the result of supply-side cost differences. New plants are just as technically efficient as older plants, and may even be slightly more efficient, thereby having lower costs. However, young producers may appear to be less productive because they may charge lower prices at the start (Foster et al., 2006). The smaller size of new plants, despite them having a cost advantage, is explained by the slow process of accumulating a customer base (Foster et al., 2016). The results suggest that firms that price their outputs in order to influence future demand can benefit from stronger post-entry growth.