Section 3:

Entrepreneurship Trends


The papers in this section consider trends in entrepreneurial activity, both in the United States and internationally. Chiara Criscuolo, Patrick Blanchenay, and Flavio Calvino examine OECD research to offer an international perspective on economic dynamism in their paper, “Business Dynamics and Public Policies: Cross-County Evidence from New Data.” They find that, across many countries, new, young, and high-growth companies are major contributors to job creation and productivity, although there is some variation according to institutional environments. Like the United States, many countries recently have experienced a fall in economic dynamism. In response to these circumstances, the authors suggest that the public policy environment must be favorably oriented toward experimentation and a high level of creative destruction, with both high entry and high exit rates; that there needs to be reduction in the costs of entry and post-entry growth, the cost of uncertainty, and the cost of exit; that competition must be encouraged and preserved; and that labor market mobility must be increased.

Focusing exclusively on the United States, John Haltiwanger’s paper, “Top Ten Signs of Declining Business Dynamism and Entrepreneurship in the United States,” offers evidence for ten facts related to the change in the dynamics of U.S. businesses and workers in the last several decades. Across a variety of indicators and databases, he concludes, economic dynamism in the United States has been declining for many years, predating the Great Recession. Historically, he explains, startups and young firms have contributed to job creation and innovation, and high rates of firm entry and exit and worker movement between companies and jobs were the core driver of productivity. The recent decline in dynamism, then, could have negative implications for growth and innovation. But Haltiwanger also explores potential causes for the decline and suggests that declining dynamism could reflect benign and even positive economic changes, such as more efficient labor market matching. It’s possible that a decline in the employment-at-will doctrine across states has raised barriers to labor market mobility, and an increase in the burden of occupational licensing could have had a similar effect. Globalization and the effects of the IT revolution could also be causes, as they may be shifting the relative structural advantages of the United States. The causes and impact of the declining trends in dynamism and fluidity, he concludes, is an open question that should be a high priority for future research.

Petr Sedláček also explores the causes and consequences of this decline in U.S. entrepreneurship in his essay, “Startups and Young Firms in the Economy.” Not only has the startup rate fallen over the past few decades in the United States, he explains, but the rate of business creation scaled to different measures of population also has fallen. He suggests that demographic change—namely, slower growth in population and labor supply—could explain most of the decline in the startup rate. Like Haltiwanger, he calls for more work to understand other potential causes of the decline, especially for the small fraction of young, high-growth firms that contribute most of the net new jobs.

Business Dynamics and Public Policies: Cross-Country Evidence from New Data

By Patrick Blanchenay Directorate for Science, Technology and Innovation Organisation for Economic Co-operation and Development
Chiara Criscuolo Directorate for Science, Technology and Innovation Organisation for Economic Co-operation and Development
Flavio Calvino OECD Directorate for Science, Technology and Innovation Paris School of Economics – University Paris 1 and Scuola Superiore Sant’Anna
Patrick Blanchenay
Patrick Blanchenay
Chiara Criscuolo
Chiara Criscuolo
Flavio Calvino
Flavio Calvino


1. Introduction

The dynamism of employment varies significantly across countries. Recent cross-country evidence suggests that these differences exist not only in terms of the entry and exit patterns of firms, but also in the size of firms at entry, and in their post-entry growth performance (Bartelsman, Scarpetta and Schivardi, 2003; Calvino, Criscuolo and Menon, 2015). For example:

  • The size of entering and exiting firms tends to be smaller in the United States than in Europe.
  • Successful, young firms tend to expand relatively more quickly in the United States than elsewhere (Bartelsman, Haltiwanger and Scarpetta, 2013).
  • Europe has a higher share of slow-growing and stagnant firms relative to the United States (Bravo-Biosca, Criscuolo and Menon, 2013).

The gap between the United States and Europe suggests that differences in institutional factors — which shape differences in the cost of reallocating resources — may explain the relative sluggishness of some European countries in capitalising on the information and communications technology revolution, (Conway et al. 2006; Bartelsman, Gautier and de Wind, 2010) and realising the potential for growth embodied in knowledge-based capital. Accordingly, this paper reviews evidence from recent Organization for Economic Co-operation and Development (OECD) research that examines cross-country differences in business dynamism, especially across sectors and firm types. A number of key findings emerge that carry important implications for public policy. First, as noted in the U.S. by Decker et al. (2014), a decline in business dynamism has been observed in several, but not all, OECD economies; we document the heterogeneity of the decline across sectors and countries. Second, as suggested by recent U.S. evidence (Haltiwanger, Jarmin and Miranda, 2013), not all small businesses are net job creators. Rather, estimates show that only young businesses — which are predominantly small — are the primary drivers of job creation. While this seems an empirical regularity across countries, we will show significant cross-country differences in the extent to which young businesses contribute to aggregate employment growth. The paper proceeds as follows:
  • Section 2 briefly describes the data used in the underlying analysis.
  • Section 3 documents some stylised facts about the patterns of business dynamism across countries and sectors during the past decade, and shows that the extent of the decline varies significantly across sectors.
  • Sections 4 and 5 focus on young businesses’ contributions to employment creation and destruction, and their employment growth dynamics respectively.
  • Section 6 draws policy implications and offers some concluding thoughts.

2.  Data

The analysis underlying the evidence reported in this paper is based mainly on micro-aggregated data from the OECD DynEmp project. We will describe the DynEmp Express and the DynEmp v.2 briefly, in turn. The Data Appendix contains a more detailed description of the DynEmp v.2 database.

The DynEmp data used in this work are the outcome of the ongoing DynEmp project,1 which is led by the OECD Directorate for Science, Technology and Innovation, with the support of national delegates and national experts in member and nonmember economies (see also Criscuolo, Gal and Menon, 2014a; Calvino, Criscuolo and Menon, 2015).

The DynEmp project is based on a distributed data-collection exercise aimed at creating a harmonised, cross-country, micro-aggregated database about employment dynamics from confidential microlevel data. National business registers are the primary sources of firm and establishment data. The project is supported by a network of national experts who run common Stata routines developed centrally by the OECD DynEmp team (see Criscuolo, Gal and Menon, 2014b for details), using the confidential microdata to which they have access. The experts also implement country-specific disclosure procedures to ensure confidentiality is respected.

The purpose of the DynEmp database is to collect cross-country evidence from countries’ business registers to identify the sources of job creation across countries and time. The project aims to quantify the extent to which firms that differ in terms of age, size and sector contribute to job creation and job destruction, and to see how firm entry, growth and exit shape employment dynamics across countries and time. The resulting statistics also provide insights about how business dynamics behaved during the recent global financial crisis.

As emphasized by Criscuolo, Gal and Menon (2014a), the advantages of using harmonised, micro-aggregated data from business registers for the study of business employment dynamics are manifold. First of all, the different channels of employment variation can be identified separately, distinguishing between gross job creation and job destruction, and between the extensive (firm entry and exit) and intensive margin (post-entry growth). Furthermore, the role of firm age and size can be examined. Finally, each of these elements can be compared across countries, sectors, and time.

Measuring entrepreneurship and its economic effects in terms of job creation is not an easy task. Appropriate data, taking into account the age — not only the size — of businesses, are necessary. Furthermore, very few databases allow the researcher to follow different cohorts of units of analysis over time, despite the wide recognition that this is crucial when studying business dynamics, especially in the case of entrants (see, for instance, Bartelsman, Haltiwanger and Scarpetta, 2009). Even fewer cross-country databases combine a cohort approach with a detailed industry disaggregation. The DynEmp database provides an ideal framework for this type of investigation.

  • Mergers and acquisitions could be accounted for in only a few countries.
  • Differences in the source data might persist, for example, in the minimum threshold above which a business is captured in the relevant country’s business register.

Owing to methodological differences in constructing these indicators, and the efforts made to harmonise the underlying data, DynEmp statistics may deviate from official statistics published by national statistical offices.

3.  A General Picture of Business Dynamism Trends

To document the cross-country heterogeneity in the existence and severity of a decline in business dynamism, we examine two measures of economic dynamism. The first is the entry rate of firms, which is calculated as the number of entering units in the sector compared with total number of units in that sector,2 thus indicating how many firms are created during a given year in a given sector. The second measure we consider is sectors’ churning rate as an indicator of their ability to allocate resources, specifically labour.

Figure 1 . Measures of Firm Dynamics for the Period 2002-2010

Source: Blanchenay et al. (forthcoming), based on the OECD DynEmp v.2 preliminary data, June 2015.

Figure 1 shows that, during the 2002-2010 period, entry rates consistently fell on average across the countries in our sample (Austria, Belgium, Denmark, Finland, Italy, Hungary, Luxembourg, New Zealand, Sweden).3 The decline seems to have been fairly steady during the period, and exacerbated by the 2007 recession. In parallel, the average size of firms (both incumbents and new entrants) increased only by 6 percent. Thus, if there was a process of concentration, it seems only a partial explanation to the decline in entry rates. Interestingly, a reduced entry rate combined with an increasing size of entrants suggests that barriers to entry may have become more salient with the 2007 recession, allowing only bigger firms — perhaps with more access to resources — to enter the market.

Figure 2. Startup Rates Across Countries

Source: Criscuolo et al. (2014a).

Data collected in the DynEmp Express (see Data Appendix) show that while most economies observed declining startup rates during the 2000s, some exceptions remain. For example, as shown in Figure 2, some countries saw a rise in their entry rates (e.g., the United Kingdom and Finland, until the crisis) or relatively stable entry rates (e.g., Belgium and Portugal until the crisis, and Sweden). Our ongoing research is trying to relate differences in policies to these diverging patterns.

As a second measure of the economy’s dynamism, we consider sectors’ churning rates as an indicator of their ability to allocate resources, more specifically labour. Following Davis and Haltiwanger (1999), we define churning rate of sector j in country i at time t:

L stands for employment, and GJC and GJD stand for gross job creation and gross job destruction, respectively. A higher churning rate indicates that more jobs are reallocated within the sector and reveals the intensity of the Schumpeterian process of creative destruction (see Schumpeter, 1942).

Figure 3 . Measures of Job Reallocation, 2002-2010

Source: Blanchenay et al. (forthcoming), based on the OECD DynEmp v.2 preliminary data, June 2015.

Figure 3 depicts the intensity of job reallocation over time, and disaggregates churning into its components: job-creation and job-destruction rates. Over time, the churning rates profile seems rather flat, with a small decline toward the end of the period. However, this apparently stable pattern hides important differences in the periods before and after the crisis. After 2007, overall job destruction rose sharply, while job creation fell across the board. In a well-functioning market, the Schumpetarian process of creative destruction should reallocate jobs from firms destroying jobs to firms creating jobs, so we should observe job destruction and job creation moving together. This is largely the case in our dataset before the crisis. However, after 2007, the crisis created a sharp disconnect between job creation and job destruction: Between 2007 and 2009, the job destruction rate sharply rose by 31 percent while the job creation rate fell by almost the same amount, 30 percent, indicating that the crisis strongly affected the process of job reallocation.

Finally, it is important to notice that the decline in business dynamism, as captured by firm-entry rate and job-churning rate, has been uneven across sectors. While on average in our dataset, the entry rate fell during the 2002-2010 period, the amount by which it declined differs greatly across sectors. In a number of sectors (e.g. telecommunications) entry rates fell sharply during the period. On the other hand, entry rates have been relatively stable in other sectors (e.g., textiles). This heterogeneity is even more pronounced when comparing how sectors fared after the crisis. During the 2007-2010 period, entry rates experienced annual changes ranging from -17.5 percent in the scientific research and development sector, to a yearly increase of 0.1 percent in the textile sector.4 The heterogeneity observed across rates seems to suggest that entry rates might have been more resilient in sectors in which demand is more inelastic, such as food products, accommodation and textiles.

Figure 4. Three Different Effects of the Crisis on Entry Rates

Source: Blanchenay et al. (forthcoming), based on the OECD DynEmp v.2 preliminary data, June 2015.

Figure 4 shows how entry rates in three different sectors were affected by the crisis. In the textile industry, entry rates only mildly declined during the crisis. The legal and accounting sector experienced a steady decline in entry rates during the period, particularly after 2004, and this downward trend continued during the crisis. In the information-technology services sector, the crisis led to a sharper decrease following the downward trend already evident since 2002.

In line with the preliminary evidence that sectors experienced very different changes during the 2002-2010 period, we further investigate how much of the observed change in the economy-wide firm-entry rate comes from a change of each sector’s entry rate (“within”), or from a change in the relative weights of sectors in the economy (“between sectors”). Following Decker et al. (2014), we implement a shift-share decomposition by decomposing, in a given country, the change in entry rate during the period 2002-2010 as follows:

The first term, “within,” represents the total change of entry rate within sectors holding their shares in the economy constant. The second term, “between,” represents the change in sectors’ weights holding their entry rate constant. The last term, “cross,” is a covariance term, which represents the joint change of weights and entry rate (a positive term means that sectors with increased entry rates also became more important in the economy).

Figure 5. Decomposition of Change in Entry Rates (2002-2010)

Source: Blanchenay et al. (forthcoming), based on the OECD DynEmp v.2 preliminary data, June 2015.

Consistent with Decker et al. (2014), the visual representation of each component of the decomposition, given in Figure 5 above, highlights that, in most countries in our dataset, the change in entry rates during the period is driven by within-sector changes, rather than by changes in relative size of the sectors in the economy. The same holds for job-churning rates.5 This seems to suggest, in particular, that changes in the composition of the economy, with a decline in the share of manufacturing and the rise of new technologies of information and communication (NTIC) sectors, is not sufficient to explain the decline of firm-entry rates in all countries in our dataset. As we show below, it is crucial to understand the source of this decline, because young firms account for a disproportionate part of job creation and growth potential in the economy.

4. Firm Age is a Key Characteristic

In addition to a great heterogeneity across countries and across sectors, data from the DynEmp project also can help shed light on firms’ dynamism, depending on their age. Indeed, one of the main contributions of the DynEmp project has been to allow, for the first time, close investigation of the role of age — not just size — in the growth dynamics of firms across eighteen OECD countries covering the period 2001-11. This is particularly important because policies generally have focused on targeting small- and medium-sized enterprises, which typically represent a large proportion of economic activity — firms with fewer than fifty employees represent more than 95 percent of all businesses in an economy (Figure 6) and between 30 and 65 percent of total employment (Figure 7). However, these small firms vary enormously in their age, and the age distribution of small firms varies significantly across countries (Figure 8). The fact that, in some economies (e.g., Italy), most firms are small and old might reflect a lack of dynamism in the economy, as evident also from looking at entry rates and churning rates in the country. Conversely, the large proportion of small firms in Brazil reflects the high rate of new entrants, typical of an emerging economy.

Figure 6. Share of Firms of Different Size by Country

Source: Criscuolo et al. (2014a)
Notes: The period covered is 2001-11 for Belgium, Canada, Finland, Hungary, The Netherlands, the United Kingdom and the United States; 2001-10 for Austria, Brazil, Spain, Italy, Luxembourg, Norway and Sweden; 2001-09 for Japan and New Zealand; 2001-07 for France; and 2006-11 for Portugal. Sectors covered are manufacturing, construction and nonfinancial business services. Owing to methodological differences, figures may deviate from officially published national statistics. For Japan, data are at the establishment level; for other countries, at the firm level. Average across all available years.


Figure 7. Share of Employment by Different Firm Size and by Country

Source: Criscuolo et al. (2014a)
Notes: see notes to Figure 6.

Figure 8. Age Composition of Small Businesses
Average over time, firms with fewer than fifty employees 

Source: Criscuolo et al. 2014a 4.1.
Notes: The figure shows the share of firms by different age groups in the total number of micro and small firms (fewer than fifty employees) in each economy on average during the available years. The period covered is 2001-11 for Belgium, Canada, Finland, Hungary, The Netherlands, the United Kingdom and the United States; 2001-10 for Austria, Brazil, Spain, Italy, Luxembourg, Norway and Sweden; 2001-09 for Japan and New Zealand; 2001-07 for France; and 2006-11 for Portugal. Sectors covered are manufacturing, construction and nonfinancial business services. Owing to methodological differences, figures may deviate from officially published national statistics. For Japan, data are at the establishment level; for other countries, at the firm level. Data for Canada abstract from merger and acquisition activity.

4.1. Young Firms are More Dynamic

The cross-country, aggregated microdata also show that, across all countries in the sample, young firms are more dynamic than old firms, both in terms of jobs created and destroyed. However, young firms systematically create more jobs than they destroy, independent of their size (Figure 9). In particular, young firms represent only around 20 percent of total employment, but they account for almost 50 percent of total job creation in the economy; their share in job destruction is around 25 percent. These patterns also hold at the sector level (Figure 10). Small/young firms account for almost 45 percent of job creation in services, and just more than 30 percent in manufacturing. Even during the global financial crisis, the majority of jobs destroyed in most countries reflected the downsizing of mature businesses, while net job growth in young firms (less than five years of age) remained positive. Therefore, the age dimension is particularly important for the design of policy, especially for policies aimed at small firms.

Figure 9. Contribution to Employment, Job Destruction and Job Creation, Nonfinancial Businesses
By firm age and size, average across eighteen countries

Source: Criscuolo et al. (2014a), based on the OECD DynEmp data collection, March 2014. Contribution to Employment, Gross Job Creation and Gross Job Destruction, Manufacturing and Services
See notes to Figure 6.

Figure 10. Contribution to Employment, Gross Job Creation and Gross Job Destruction, Manufacturing and Services
By firm age and size, average across eighteen countries

Source: Criscuolo et al. (2014a), based on the OECD DynEmp data collection, March 2014
See notes to Figure 6.

Differences in the magnitude of this phenomenon across countries points to the importance of national policies and business environments in fostering the birth and growth of new firms. In some countries — for example, Brazil, New Zealand, and Spain — young firms account for more than half of the economy’s total gross job creation. In others — such as Japan and Finland — young firms account for less than 30 percent of jobs created (Figure 11).

Figure 11. Contribution of Young Firms to Employment, Gross Job Creation and Gross Job Destruction

Source: Criscuolo et al. (2014a), based on the OECD DynEmp data collection, March 2014
See notes to Figure 6.

4.2. Deconstructing Job Creation by Entrants

Given the importance of young firms for the firm dynamics, the contribution of startups to the creation of new jobs can be further unpacked into four different elements6 (Calvino, Criscuolo and Menon 2015):

  • Start-up ratio.7
  • Average size at entry.
  • Survival rate.
  • Average post-entry growth of survivors.

These four levers help investigate the extent to which entrants contribute to aggregate job creation, with potentially very different policy implications. Indeed, even across economies with similar aggregate startup contributions, these elements interplay in different ways, highlighting the lack of a one-size-fits-all pattern. Figure 12 separately illustrates the four elements described above, focusing on the first three years of activity of surviving entrants. It demonstrates that the startup ratio, average size at entry, and post-entry growth of survivors exhibit substantial cross-country heterogeneity, while the survival rate appears more homogeneous. For instance, the startup rate is very low in Belgium, but the post-entry growth rate of survivors in that country is the highest in the sample. Conversely, in New Zealand and Turkey, the startup rate is higher but the average post-entry growth is significantly lower. Further analysis reveals that, in most countries, startups’ likelihood of exiting is highest at year two and decreases (linearly) beyond that age (Calvino, Criscuolo and Menon, 2015).

This evidence confirms that entrepreneurship is a highly complex phenomenon, and that similar outcomes in terms of net job creation might mask very different startup dynamics across countries. Decomposing net job creation by surviving entrants might help fine-tune policy interventions in the light of the relative weight that each of these four elements has in a specific economy.

Figure 12. Four-Way Decomposition of Job Creation by Surviving Entrants
Panel A. Startup Ratio

Panel B. Survival Rate (after three years)

Panel C. Average Size at Entry

Panel D. Average Post-Entry Growth

Source: Calvino et al.(2015) based on OECD DynEmp v.2 database. Data for some countries are still preliminary.
Notes: The graph illustrates the four elements of the growth decomposition. Panel A: Startup ratio expressed as total number of entering units (entrants) across total employment (in thousands). Panel B: Survival share of entrants expressed as number of entering units surviving across total number of entrants percent. Panel C: Average size of surviving entrants expressed as total employment of surviving entrants across number of surviving entrants. Panel D: Ratio between total employment at t + 3 across total employment of surviving entrants. Figures report the average for different time periods t = 2001, 2004 and 2007, conditional on their availability. Sectors covered are manufacturing, construction, and nonfinancial business services. Owing to methodological differences, figures may deviate from officially published national statistics.

4.3   The Dynamics of Micro Startups

The contribution of young firms to aggregate employment growth reflects a process of creative destruction in which success and failure go hand in hand. Young-firm dynamics are characterised by a so-called “up or out” pattern (Haltiwanger, Jarmin and Miranda, 2013). A significant proportion of startups do not survive beyond the first two years, but those that do survive contribute disproportionately to job creation.

Differences in the extent to which young firms grow are first shown in Figure 13. We infer the potential growth of young firms by comparing the average size of startups with the average size of old businesses (more than ten years old). Figure 13 points to some differences in the size of startups across countries, although these are not striking. France, Finland and The Netherlands have the largest new firms.8 The picture is much more heterogeneous when examining the size of older businesses. The average size of old firms in the U.S. is by far the largest — around eighty employees in manufacturing and forty in services. This is even more striking since the average size of startups in the French manufacturing sector is more than double the average size of U.S. startups. This confirms the previous results of Bartelsman, Scarpetta and Schivardi (2003), who found that seven-year-old U.S. firms are, on average, 60 percent larger than their size at entry; in European countries, the figure ranged between 5 and 35 percent. With the current data, The Netherlands shows one of the lowest ratios between the average size of old firms and startups across all countries, with a particularly low ratio of 1.7 for manufacturing firms. In the U.S., this ratio is greater than five for both manufacturing and services firms.

13. Average Size of Startup and Old Firms across Industries and across Countries

Source: Criscuolo et al. (2014a), based on the OECD DynEmp data collection, March 2014
See notes to Figure 6

These findings are consistent with anecdotal and survey evidence that, in some countries, entrants might be able to start off at a smaller size because it is less costly to experiment. Moreover, they can exit more easily if they are not successful. This, in turn, might contribute to stronger growth prospects for very productive and successful businesses by freeing up scarce resources, such as skilled labour. It also suggests that barriers to growth (e.g., access to markets, burdensome regulation on starting businesses, and lack of competition) might hinder the growth potential of young businesses in some countries.

To better understand the growth dynamics of startups, the OECD DynEmp database follows cohorts of entrants (firms aged zero to two at the beginning of the period) with fewer than ten employees for three, five and seven years.

5.  Post-Entry Growth

Previous work already has highlighted that the average growth rate of startups — although generally positive overall — entails a substantial degree of heterogeneity within cohorts of otherwise similar entrants, with the large majority of small startups growing very slowly, and a tiny proportion of them experiencing very fast growth (Criscuolo, Gal, and Menon, 2014c; Anyadike-Danes et al., 2014). It is, therefore, important to further explore this firm-level heterogeneity in the growth performance of startups in order to draw useful policy implications.

Figure 14 analyses the post-entry dynamics of micro entrants (entrants with zero to nine employees) in the nonfinancial business sector, classifying them according to their size class five years later, or in the exit group if they did not survive. From Panel A, which presents the figures in terms of number of units, it is evident that most micro startups either remain stable (i.e., at the end of the period they are in the same size class as at the beginning of the period) or exit the market. In every country, the number of micro startups moving to a higher size class at the end of the period is extremely small — on average around 3 percent and never more than 8 percent. The graph also shows that, in all economies but Austria, Brazil, and Turkey, the number of surviving micro startups is higher than the number of exiting ones.

Figure 14. Focus on Micro Entrants: Stable vs. Growing vs. Exiting
Panel A. Size class of micro entrants (zero to nine employees) five years after entry 

Panel B. Contribution to total net job creation

Source: Calvino et al.(2015), based on the OECD DynEmp v.2 database. Data for some countries are still preliminary. Notes: Panel A represents the share (in terms of number of units) of micro (zero to nine employees) entrants by size class at time t + 5. Panel B represents the contribution to net job creation (defined as net job creation by the group across total net job creation of micro entrants) for micro (zero-nine employees) entrants by size class at time t + 5. Size classes are aggregated as follows: zero-nine (stable), ten to nineteen and twenty or more (growing), exit (shrinking) and units for which the size class at time t + 5 is missing. Figures report the average for different time periods t = 2001, 2004 and 2007, conditional on their availability. Sectors covered are manufacturing, construction, and nonfinancial business services. Owing to methodological differences, figures may deviate from officially published national statistics.

The pattern is, however, completely different if, rather than looking at firms, one looks at jobs. Panel B in Figure 14 illustrates post-entry dynamics of micro entrants in terms of their net job creation (the difference between employment at the beginning and at the end of the five-year period, respectively). The very few micro-entrants with more than twenty employees at the end of the horizon are responsible for most job creation of micro startups in all countries — on average, 37 percent with a maximum of 52 percent in Sweden. Furthermore, in most countries, gross job creation by surviving micro startups more than compensates for gross job destruction by exiting ones.9

Three main features are worth noting:

  • Very few micro startups – between 2 and 9 percent – grow to include more than ten employees, but their contributions to employment change ranges from 19 to 54 percent.
  • Most firms remain within the same size class (i.e., they still employ fewer than ten employees after three years), and while they still create a reasonable number of jobs, their contribution is less than proportional to their weight in terms of number of firms. For some cohorts and countries, these small startups actually contribute negatively to net job creation.
  • The extent to which micro startups survive is very different across countries.

Understanding the role that differences in policies play in explaining these features is crucial for evidence-based policymaking, as discussed in the next section.

6. The Role of Public Policy: Discussion and Future Research

The paper has shown some empirical regularities about the important role of startups for net job creation and about the up-or-out dynamics that characterise their growth profiles. Nevertheless, the new DynEmp database also points to some important cross-country differences in the magnitude of young firms’ contributions to net job creation, and in the extent to which they can upscale. Both the empirical regularities and the cross-country differences convey important evidence for policymakers.

In particular, the “empirical facts” that young businesses are net job creators across the board, while they also are characterised by an up-or-out dynamic, suggest that often-heard statements such as “SMEs are net job creators” should be qualified as “young SMEs are net job creators.” At the same time, the facts speak against using “picking winners” policies for this group of firms, given the high turbulence characterising startups. Rather, both of these facts characterising startups indicate that an environment that favours experimentation, and allows for a healthy process of creative destruction with high entry and exit rates, and the realisation of high growth potential, are key for both employment and productivity growth. The existence of cross-country differences in the growth contribution from startups, and in the possibility for them to upscale points to the importance of national structural features and institutions. In particular, it highlights the importance of designing public policies such that entrepreneurs can enter and operate in an experimentation-friendly and growth-supportive environment. This entails policies that can reduce the costs of entry and post-entry growth (regulations affecting product, labour and financial markets), uncertainty (contract enforcement and judicial efficiency) and exit (employment-protection legislation and bankruptcy laws).

To that end, ongoing research using the DynEmp v.2 database has looked into policies that relate to entry and exit rates, size at entry, and post-entry growth performance. Preliminary results of this research point to the importance of competition in product markets with tight product-market regulations and barriers to foreign competition being associated with reduced firm entry. Slower growth of firms, in line with existing evidence that links the lack of competition with a less efficient allocation of resources, is associated with slower growth of innovative firms and lower exit rates of inefficient firms.

Secondly, regulation of the labour market is also significantly related to the growth performance of young businesses. This is in line with previous OECD research (Bravo-Biosca, Criscuolo and Menon, 2013, and Andrews, Criscuolo and Menon, 2014) showing that less stringent employment-protection legislation is associated with higher increases in the ability of innovative firms to attract resources that are required to implement and commercialize new ideas, especially for young businesses. It is consistent with strict employment-protection legislation slowing down the reallocation process, and making it less attractive for firms to experiment with highly uncertain technologies by raising the costs of exit in case of business failures

Third, well-developed financial and judicial systems are also an important policy lever that can explain differences in growth performance of entrants. In particular, availability of seed and early-stage venture capital, clear enforcement of contracts and efficient judicial systems are key.

Finally, given the high risk of failing when starting a business, bankruptcy legislation that does not excessively penalise failure is associated with both higher entry rates and faster post-entry growth. By contrast, if the time it takes to wind down a business is particularly lengthy, risky entrepreneurial ventures might not be brought to the market to avoid incurring high exit costs in case of failure.

From a methodological perspective, the evidence presented in this paper and the ongoing research agenda at the OECD point to the importance of accessing, harmonizing and using microdata for constructing the evidence base needed for policymaking. The DynEmp project has been an important step in bringing this to the attention of policymakers in the area of employment dynamics.

Using the same methodology as the DynEmp project, we also are working on providing new cross-country evidence about the microdrivers of productivity, thanks to a new project called MultiProd. The MultiProd project aims at building a collection of statistics at the detailed sectoral level for different firm characteristics, such as wages, profits, investment and, in particular, for various measures of labour and multifactor productivity. The objective is to better understand the heterogeneity in the Schumpeterian process of creative destruction across countries and sectors. In particular, it will focus on four issues:

First, MultiProd gauges whether resources are efficiently allocated through the analysis of the entire firm-level productivity distribution, with further refinements by size, age, and ownership categories.

Second, the project identifies firms at the “frontier”—the best performers— and understands how they differ across countries, what drives their performance, and how much they contribute to aggregate productivity growth.

Third, it investigates the cross-country differences in firm-level productivity performance and in allocative efficiency before, during and after the financial crisis. Finally, it looks at the effect of productivity dispersion on wage inequality.

Thanks to this new collection of data, the project will be able to examine the effectiveness of various policy frameworks aimed at shaping firm productivity and enhancing resource allocation to more productive firms. The MultiProd project also will complement the evidence gathered by the DynEmp project because it will be useful to understand whether young firms, which disproportionately contribute to job creation in the economy, also contribute to the growth of aggregate productivity. The project will allow comparing the characteristics of young firms to old ones, as well as comparing the age of firms in the top and bottom of the distribution of productivity growth.

About the Authors
Patrick Blanchenay, Flavio Calvino and Chiara Criscuolo are with the Directorate for Science, Technology and Innovation, of the Organization for Economic Co-operation and Development (OECD).

Authors’ Acknowledgements

The views expressed in this paper do not reflect the view of the OECD or of its member countries. Errors remain our own. We are grateful to Isabelle Desnoyers-James for statistical assistance.


  2. Note that the focus of the analysis is on units with positive employment.
  3. Note that the trends illustrated in Figure 1 are weighted country averages. Larger countries, therefore, count more in driving the dynamics. A certain degree of cross-country heterogeneity remains present, as also shown in Figure 2.
  4. Note that changes in entry rate are expressed in compound annual changes, so small differences in percentages can amount to large discrepancies when compounding during the 2002-2010 period.
  5. See Blanchenay et al. (forthcoming).
  6. See Annex A in Calvino, Criscuolo and Menon (2015) for the exact formula
  7. Measured as number of startups across total employment in the economy
  8. These differences might be partly influenced by the intensity of merger and acquisition (M&A) activity across countries, and whether the new businesses that result from M&A deals appear as “entrants” in a country’s business register.
  9. A caveat is, however, necessary for this analysis. Due to confidentiality requirements and the relatively small size of their economy, the partial blanking of the result datasets for Denmark, Finland, New Zealand, and Netherlands may lead to underestimating the share of micro startups moving to a higher size class, as well as their contributions to net job creation.
  10. More information:
  11. Data relative to 2010-2011 in Norway have been excluded from the analysis due to ongoing checks on unusual dynamics in the underlying data; data relative to 2006 in The Netherlands have been also excluded due to the redesign of the Business Register in that year.


M. Anyadike-Danes, C.M. Bjuggren, S. Gottschalk, W. Hölzl, D. Johansson, M. Maliranta,  A.G. Myrann, “Accounting for job growth: disentangling size and age effects in an international cohort comparison,” IFN Working paper (2014)

D. Andrews, C. Criscuolo, C. Menon, “Do Resources Flow to Patenting Firms? Cross-country Evidence from Firm Level Data,” OECD Economics Department Working Papers No 1127 (2014)

E. Bartelsman, J. Haltiwanger, S. Scarpetta, “Measuring and Analysing Cross-Country Differences in Firm Dynamics,” Producer Dynamics: New Evidence from Micro Data (University of Chicago Press) pp. 15-76 (2009)

E. Bartelsman, J. Haltiwanger, S. Scarpetta, “Cross-Country Differences in Productivity: The Role of Allocation and Selection,” The American Economic Review 103 (1): 305–34 (2013)

E. Bartelsman, S. Scarpetta, F. Schivardi, “Comparative Analysis of Firm Demographics and Survival: Micro-Level Evidence for the OECD Countries,” OECD Economics Department Working Papers, No. 348 (OECD, Paris) 2003

E. Bartelsman, P.A. Gautier, J. De Wind, “Employment Protection, Technology Choice, and Worker Allocation,” IZA Discussion Paper No. 4895 (2010)

A. Bravo-Biosca, C. Criscuolo, C. Menon, “What Drives the Dynamics of Business Growth?” OECD Science, Technology and Industry Policy Papers, No 1, OECD Publishing (2013)

P. Blanchenay, F. Calvino, C. Criscuolo, (forthcoming), “Cross-Country Evidence on the Decline of Business Dynamism”

F. Calvino, C. Criscuolo, C. Menon, “Cross-country Evidence on Start-up Dynamics,” OECD Science, Technology and Industry Working Papers, 2015/06 (OECD Publishing, Paris) 2015

P. Conway, D. de Rosa, G. Nicoletti, F. Steiner, “Product Market Regulation and Productivity Convergence,” OECD Economic Studies, Vol. 2006(2), pp. 39-76, (OECD Publishing, Paris) 2006

C. Criscuolo, P. Gal, C. Menon, “The Dynamics of Employment Growth: New Evidence from 18 Countries,” OECD Science, Technology and Industry Policy Papers, No. 14, OECD Publishing, (2014a)

C. Criscuolo, P. Gal, C. Menon, “DynEmp: A Stata® Routine for Distributed Micro-data Analysis of Business Dynamics,” OECD Science, Technology and Industry Working Papers, 2014/2, OECD Publishing (2014b)

C. Criscuolo, P. Gal, C. Menon, “Do Micro Startups Fuel Job Creation? Cross-country Evidence from the DynEmp Express Database,” OECD Science, Technology and Industry Policy Papers, forthcoming 

S.J. Davis, J. Haltiwanger, “Gross job flows,” Handbook of Labor Economics, 3, 2711-2805 (1999)

R. A. Decker, J. Haltiwanger, R.S. Jarmin, J. Miranda, “The Secular Decline in Business Dynamism in the U.S.” 2014

J. Haltiwanger, J.R. Jarmin, J. Miranda, “Who Creates Jobs? Small vs. Large vs. Young,” Review of Economics and Statistics, 95(2), 347-361 (2013)

The Sources of Economic Growth in OECD Countries (OECD Publishing, Paris, 2003)

J. Schumpeter, Capitalism, Socialism, and Democracy (New York, Harper & Bros., 1942)

Data Appendix: The DynEmp v.2 database

This paper uses data collected by the OECD for a database called DynEmp. A first phase of the project (DynEmp Express) has collected nonconfidential, comparable statistics about employment, and gross job creation and destruction by firm age, size and macrosectors for eighteen countries: Austria, Belgium, Brazil, Canada, Finland, France, Hungary, Italy, Japan, Luxembourg, The Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, the United Kingdom, and the U.S. The second phase of the project is ongoing. It aims to collect detailed data at the two-digit, industry-code level, and follow cohorts of businesses through three, five and seven years.

The data used in this report are the intermediate outcome of the ongoing second round of data collection within the DynEmp project, which is led by the OECD Directorate for Science, Technology and Innovation, with the support of national delegates and national experts in member and nonmember economies.

The DynEmp project is based on a distributed data-collection exercise aimed at creating a harmonised cross-country micro-aggregated database about employment dynamics from confidential microlevel data for which the primary sources of firm and establishment data are national business registers. The main building blocks of the data produced by the DynEmp v.2 routine can be summarized as: i) “flow datasets,” ii) “transition matrices,” and iii) “distributed regressions”.

The flow datasets contain annual statistics about gross job flows (such as gross job creation and gross job destruction, defined as the total job variation of growing and shrinking units, respectively) and about several statistical indicators of unit-level employment growth (mean, median, and standard deviation).

The transition matrices summarize the growth trajectories of cohorts of units from years 2001, 2004, and 2007, observed after  three-, five- and seven-year periods. The matrices contain a number of statistics for different combinations of age classes and size classes at time t and t + j, plus a focus on the dynamics of high-growth units.

The DynEmp v.2 network also is collecting three sets of distributed regression outputs. The first set of OLS regressions focuses on growth-rate dynamics. The second set of regressions investigates (by means of a Linear Probability Model) the units’ exit probabilities. These regressions control for two-digit industries specificities and for different age-size effects, including a focus on pre- vs. post-crisis dynamics. The third set of regressions is aimed at describing the firm or establishment growth distribution, and at identifying potentially significant discontinuities in these distributions, possibly due to the institutional and regulatory environment.

At the time of writing, sixteen countries have been successfully included in the DynEmp v.2 database (namely, Austria, Belgium, Brazil, Costa Rica, Denmark, Finland, Hungary, Italy, Luxembourg, The Netherlands, Norway, New Zealand, Portugal, Spain, Sweden, and Turkey). Data at firm level are available for all the above-mentioned countries. For most countries, the time period between 2002 and 2011 is covered. For Costa Rica and Portugal, only fewer years are available, while the time horizon for Austria, Denmark, Norway and Sweden is longer. For Costa Rica, no transition matrix is available due to the limited-time extension of the source data. Details about temporal coverage by country are summarized in Table 1.11

Table 1. Temporal Coverage DynEmp v.2 Through Time by Country

Source: OECD DynEmp v.2 database. Data for some countries are preliminary.

Note: Temporal coverage by country of the database used for the analysis. Years for which annual flow data are available are coloured. For Costa Rica, no transition matrix is available due to the limited-time extension of the source data. Texture boxes correspond to years that have been excluded from the analysis due to data issues.

For section 3 of this work, we focus on data ranging from 2002 to 2010 for Austria, Belgium, Denmark, Finland, Italy, Hungary, Luxembourg, New Zealand, and Sweden.

Top Ten Signs of Declining Business Dynamism and Entrepreneurship in the United States

By John Haltiwanger Distinguished University Professor in the Department of Economics University of Maryland and National Bureau of Economic Research
John Haltiwanger
John Haltiwanger


The United States has exhibited a substantial and pervasive decline in measures of business dynamism, entrepreneurship, and labor market fluidity in the last several decades. We have learned this through the relatively recent development of comprehensive longitudinal business databases tracking the U.S. private, non-farm sector. Numerous studies have documented the decline and explored its causes and consequences.1 In this short synopsis, the basic facts of this decline are summarized by highlighting the top ten signs of the decline.

Before discussing the top ten facts, it is useful to provide some brief remarks about whether these trends have adverse consequences for U.S. economic growth. A hallmark of the U.S. economy has been a high pace of job and worker reallocation with a high pace of business entry and exit. The evidence shows that this dynamism and flexibility of the U.S. labor market has been important for productivity growth, earnings growth, and job creation. The reallocation of jobs across businesses historically has reflected moving resources from less-productive to more-productive businesses. The churning of workers across jobs has reflected workers finding better matches and moving up the job ladder. Better matches yield both higher productivity and higher wages. Startups and high-growth young firms have been a critical component of this dynamism. Startups and young firms have disproportionately contributed to job creation and been key sources of innovation and experimentation. From this perspective, the declines in dynamism and fluidity would appear to have adverse consequences. However, it could be that shifts in the business model and changes in the ways workers build careers imply that job creation, productivity, and earnings growth can be achieved without such a high pace of worker and job reallocation. These alternate views are at the core of the open question about the impact of the declining trends in dynamism and fluidity. Further discussion of the ramifications of the declining dynamism is provided in the concluding remarks, but, first, the core facts are stated and discussed.

Fact 1: The decline in business dynamism accelerated in the post-2000 period.

Using measures such as the pace of job reallocation and the dispersion of growth rates across businesses and within business volatility, the decline in business dynamism for the U.S. private, non-farm sector dates at least to the mid-1980s (see Davis, Haltiwanger, Jarmin, and Miranda 2007, Davis and Haltiwanger 2014, Decker, Haltiwanger, Jarmin, and Miranda 2014a,b). Figure 1 shows the annual pace of job reallocation from the Business Dynamic Statistics (BDS) for the U.S. private, non-farm sector from 1980–2012. Figure 1a shows the quarterly pace of job reallocation from the Business Employment Dynamics (BED). In each figure, the actual and trend (Hodrick-Prescott Filtered) series are depicted. The trend decline is apparent, as is a notable acceleration in the decline in trend in the post-2000 period. The acceleration of the decline in the trend is associated with change in the character in the decline in the post-2000 period, as discussed in the next fact.

Figure 1: Job Reallocation Rate, U.S. Private, Non-Farm, Quarterly 1990:2–2014:3

Source: BED

Figure 1a: Job Reallocation Rate, U.S. Private, Non-Farm, Annual 1980–2012

Source: BDS

Fact 2: The decline in business dynamism has seen a shift in the industries and types of firms impacted.

Prior to 2000, the decline in indicators of business dynamism were dominated by sectors such as retail trade and services. In the retail trade sector, the decline in entrepreneurship and dynamism arguably has been driven by benign factors reflecting a shift in the retail trade business model. The shift has been away from single-unit establishment firms (“Mom and Pop” firms) to large national and multi-national chains. The latter have taken advantage of IT and globalization to build efficient distribution and supply chain networks. Establishments in retail trade that are part of large, national firms are both more productive and more stable.2

In contrast, prior to 2000, key sectors like high tech as well as publicly traded firms exhibited a rise in indicators of dynamism.3 The evidence shows that in high tech, high-growth young firms play an especially critical role in job creation and productivity growth. Likewise, newly listed publicly traded firms (IPOs) are high-growth young firms critical for job creation and productivity growth. In the robust period of aggregate productivity and job growth in the 1990s, the high-tech sector and newly listed public companies (there is considerable overlap here) exhibited increases in indicators in dynamism and entrepreneurship. However, since 2000, the high-tech sector and publicly traded firms have exhibited a decline in dynamism. The number of IPOs has fallen in the post-2000 period, and those that have entered have not exhibited the same rapid growth as earlier cohorts.

Fact 3: The decline in dynamism has been attenuated by the changing composition of U.S. industries.

The shift away from goods-producing industries in the United States should have implied an increase in measures of business dynamism. Sectors such as retail trade and services have higher average rates of dynamism than the manufacturing sector, even taking into account the especially large declines in dynamism in retail trade and services discussed in Fact 2.4 This implies that the puzzle of declining dynamism is even larger once one controls for industry composition. That is, the within-industry decline is even larger than the overall decline. As discussed in Fact 2, the patterns of within-industry declines in dynamism vary substantially over time.

Fact 4: The decline in dynamism encompasses a decline in the startup rate.

Figure 2 shows the startup rate of firms along with the exit rate of firms from the BDS. These startup and exit rates abstract from M&A activity and reflect organic entry and exit.5 The startup rate decline has been ongoing for the last few decades, although, consistent with Fact 2, its character has changed over time. Prior to 2000, the startup rate decline was dominated by sectors like retail trade. Since 2000, there has been a sharp decline in the startup rate in key innovative sectors like high tech. The decline has been substantial enough that the net entry rate has turned negative in the last several years.

Figure 2a shows the employment-weighted startup rate over the same period (from the BDS), which shows the same general patterns. The employment-weighted startup rate declined from 2006 to 2009 by one percentage point, which implies more than one million fewer jobs created by startups in 2009 compared to 2006. Figure 2a also shows the employment-weighted establishment annual openings rate from the BDS from 1981 to 2012 and the analogous series from the BED from 1992 to 2015. All series are on a March-to-March basis, with the BED series depicted representing four times the average of the quarterly openings rate for the relevant quarters of the year.6 Interestingly, in the overlapping years, the BDS and the BED series show quite similar rates in terms of magnitudes and trends. The establishment openings rate is about twice that of firm startup rate, reflecting the fact that many new establishments are for existing firms. Indeed, Haltiwanger, Jarmin, and Miranda (2013) show that most new establishments of existing firms are for large, older firms. This implies that appropriate caution must be used in drawing inferences from the establishment openings series. Still, for both the BDS and the BED, the downward trend in the establishment openings series largely mimics the downward trend in the startup rate series. This is important because it provides a proxy for the likely pattern of the startup rate series through 2015. With these reservations in mind, it is striking that the establishment openings rate exhibits a continued downward trend through this period.

The decline in these indicators of entrepreneurship is a core component of the overall decline in dynamism discussed in Fact 1. Young businesses are the most volatile businesses, so a decline in startups implies a decline in the share of activity accounted for by young businesses. Decker, Haltiwanger, Jarmin, and Miranda (2014) show that this shift away from young firms accounts for about 30 percent of the decline in dynamism.

Figure 2: Startup and Exit Rates for Firms in the U.S. Non-Farm Sector, 1981–2012

Source: BDS

Figure 2a: Annual Employment-Weighted Startup Rate of Firms and Establishments, U.S. Private, Non-Farm Sector, BDS (1981–2012), BED (1992–2015).

Source: BDS and BED


Figures 2 and 2a also show a sharp decline in startup activity in the Great Recession, with little or no recovery. For the startup rate itself, we observe this only through 2012, but the weak pattern for establishment openings through 2015 suggests a continued weak recovery of startups. One way to see that the recovery for young and small businesses has been especially anemic is to compare the recovery in employment for young and small businesses from the Great Recession to the early 1980s recession. Figure 3 shows that it is especially young (less than five years old) and small/medium (less than 500 employees) businesses that have experienced a weak recovery.

Figure 3: Employment Relative to Trough, by Firm Size and Firm Age

Source: BDS. This figure draws on the analysis in Davis and Haltiwanger (2015).

Fact 5: The typical worker is increasingly more likely to work at a large, mature, national (or global) firm.

Figure 4 shows the share of employment at large (500+), mature (five years plus) firms; small/medium (less than 500 employees), mature firms; and small/medium, young (less than five years old) firms. The share of employment for the latter has dropped in half over the last several decades, while the share of employment at large, mature firms has risen from about 40 to 50 percent of employment. Large, mature firms account for less than 1 percent of total firms (see Haltiwanger, Jarmin, and Miranda 2013), but, obviously, a very large and increasing fraction of activity. Decker, Haltiwanger, Jarmin, and Miranda (2014, 2015) show that this shift to large, mature firms is also accompanied by a shift toward national firms (with activity in all regions of the country). The shift in the age and size structure of firms is directly related to the decline in startups documented in Fact 4.

Figure 4: Share of Employment by Broad Firm Age and Firm Size Classes, U.S. Non-Farm, 1981–2012

Source: BDS

Fact 6: Worker reallocation and worker churn has declined.

Worker reallocation is about twice the pace of job reallocation. Put differently, hires are about twice the pace of job creation, and separations (mostly quits plus layoffs) are about twice the pace of job destruction (see Davis and Haltiwanger 2014). Figure 5 shows the pace of worker reallocation, job reallocation, and worker churn (the difference between worker and job reallocation). Worker churn is the amount of worker reallocation over and above that needed to accommodate the shifting of jobs across establishments.7

Worker churn actually was rising over the 1990s, but then fell in the post-2000 period. Given the decline in job reallocation through the entire period that accelerated in the post-2000 period, worker reallocation fell sharply in the post-2000 period.

Figure 5: Worker Reallocation, Job Reallocation and Churn, U.S. Private Sector, 1990:2–2015:1

Source: BED and Job Openings and Labor Turnover Survey (JOLTS). This uses an extended version of the series developed by Davis, Faberman, and Haltiwanger (2012). This figure is an extended version of a figure in Davis and Haltiwanger (2014). The worker reallocation series extends to 2015:1. The job reallocation and churn series stops in 2014:3, which is the most recent quarter of the BED.

Fact 7: The decline in labor market fluidity has induced a decline in the employment-to-population ratio.

States and demographic groups with the largest declines in labor market fluidity also experienced the largest declines in employment-to-population ratios (see Davis and Haltiwanger 2014). Specifically, young, less-educated males experienced the largest declines in both labor market fluidity and employment-to-population ratios. Economic theory suggests this relationship may be causal. A fluid labor market is especially important for young workers as they try to build careers. A fluid labor market implies a high likelihood of finding a job upon entry into the labor market and an ability to job hop to find the best match. For marginally attached young workers, a decline in fluidity makes participating in the labor market less attractive. Using instrumental variable estimation procedures, Davis and Haltiwanger (2014) find empirical support for this labor fluidity hypothesis.

Fact 8: The recent recovery in net job creation is driven more by the continuing decline in job destruction and layoffs than by an increase in job creation and hiring.

The United States has had a sustained, albeit slow, recovery from the Great Recession with twenty quarters of consecutive positive net employment growth since the first quarter of 2010. Figures 6 and 6a show that the recovery is driven more by a decline in job destruction and layoffs than by a recovery in job creation and hires. The latter remain substantially below the 2006 levels and far below the robust rates in the late 1990s. Hires and quits have picked up in the fourth quarter of 2014 and first quarter of 2015. This looks to be an increase in worker churn and not an increase in job reallocation (we will not know for sure until the BED data for those quarters are released). Worker churn by construction does not contribute to net job creation.

Figure 6: Hires and Job Creation for U.S. Private, Non-Farm, Quarterly, 1990:2–2015:1

Source: BED and JOLTS. This uses an extended version of the series developed by Davis, Faberman and Haltiwanger (2012). This figure is an extended version of a figure in Davis and Haltiwanger (2014). The hires series extends to 2015:1. The job creation series stops in 2014:3.

Figure 6a: Layoffs, Quits, and Job Destruction for U.S. Private, Non-Farm, Quarterly, 1990:2–2015:1

Source: BED and JOLTS. This uses an extended version of the series developed by Davis, Faberman, and Haltiwanger (2012). This figure is an extended version of a figure in Davis and Haltiwanger (2014). The layoffs and quits series extend to 2015:1. The job destruction series stops in 2014:3

Fact 9: The decline in dynamism is not due to a decline in the volatility of shocks but, rather, in the responsiveness to those shocks.

One possible potentially benign explanation for the decline in dynamism is that the structure of the U.S. economy and the business model have changed so that there is less need for so much dynamism. According to economic theory, a high pace of dynamism is part of a healthy economy if that dynamism reflects moving resources away from less-productive to more-productive uses. If the dispersion of productivity differences decreased between firms, then this would be one indicator of change in structure, implying less need for reallocation. Evidence from the manufacturing sector shows a rise in the dispersion of TFP differences between plants (see Decker, Haltiwanger, Jarmin, and Miranda 2014b). This holds overall in manufacturing, but also in the critical high-tech components of manufacturing (e.g., computers and semi-conductors). This rise in TFP dispersion is consistent with a rising dispersion in the volatility of idiosyncratic productivity shocks impacting businesses. This implies that there should be a rise in dynamism, as there are increased incentives for reallocation. To reconcile the decline in dynamism with a rise in firm-level volatility of shocks, Decker, Haltiwanger, Jarmin, and Miranda (2014b) show there has been a decline in the responsiveness of manufacturing plants in terms of growth and survival to TFP shocks.

Fact 10: The declining responsiveness to idiosyncratic productivity shocks implies the contribution of reallocation to productivity growth has declined.

High-productivity firms expand while low-productivity firms contract and exit; thus, reallocation contributes to productivity growth. This finding is one of the most ubiquitous findings in the firm dynamics literature (Syverson 2011). Fact 9 implies that the contribution of reallocation to aggregate productivity growth has declined. A counterfactual exercise by Decker, Haltiwanger, Jarmin, and Miranda (2014b) suggests that the difference in the annual contribution of reallocation to productivity growth between the 1980s and the post-2000 period is about one percentage point. This is a large quantitative effect.

Concluding Remarks and Discussion

The ten facts presented above establish that something has changed in the dynamics of U.S. businesses and workers. U.S. policymakers in the 1990s highlighted the dynamism of U.S. businesses and the flexibility of U.S. labor markets as underlying the robust economic growth over that period. Such claims cannot be made for the post-2000 period and, especially, in the post-Great Recession recovery. The factors that underlie these fundamental structural changes to the U.S. economy are not yet understood. In some sectors, like retail trade, a case can be made that the declining dynamism largely has been benign, reflecting a shift in the business model to large, national chains that are both more productive and more stable. However, in other sectors, such as high tech, such arguments are less persuasive. In this respect, it is worth noting that the acceleration in the decline in dynamism overall and, in particular, the decline in dynamism in key innovative sectors like high tech in the post-2000 period coincides with a decline in U.S. economic growth, whether measured by productivity or job growth.

Economic theory implies that one possible explanation is increased barriers to adjustment for firms and workers. In practice, it is not apparent that there was a single major change in the U.S. business climate over this period that increased adjustment frictions. It may be that this is more like “a death by a thousand cuts” with many small changes in the regulatory and institutional environment contributing to the decline in dynamism and flexibility. Davis and Haltiwanger (2014) show that the decline in the employment-at-will doctrine contributes non-trivially to the decline in business dynamism. Likewise, they speculate that the rapid increase in occupational licensing requirements may be an important factor. Hsieh and Moretti (2015) have argued that, in the idea economy, it is even more important than ever for economic activity clustering to occur, and they argue that zoning and other restrictions on local economic growth may be stifling U.S. economic growth. In this case, it may not be so much a change in the regulatory environment, but a change in the impact of existing regulations as the economy has changed structure.

Related concerns correlate to the combined influence of IT and globalization. IT facilitates global linkages in the production process within and across firms. It may be that the United States has become relatively less attractive for the type of dynamic, entrepreneurial activity that has been its hallmark. Some have speculated that the United States is still the place to develop and design new products and processes, but it is relatively less attractive as the place to produce such new products.8 In considering these issues, this is not an argument against globalization, because it brings greater economic efficiency and considerable gains to consumers. Likewise, further IT innovations will raise productivity and yield consumer gains. But it may be that the United States is less well positioned to deal with the inevitable disruptions that occur when there are structural changes, such as IT and globalization.

Another possible explanation is that there has been a shift in the willingness to take on the risks that are inherent in a highly entrepreneurial economy. Such a shift may come from several sources. Potential entrepreneurs may be less willing to undertake the risky process of starting new businesses. Credit markets may have become less willing to finance risky startups and young firms. Why might there be less willingness to take on risk? One difficult-to-assess hypothesis is that more recent generations are more risk averse. Perhaps more easy to assess is the hypothesis that something in the economic environment has changed that makes taking on risk less attractive. A plausible candidate hypothesis here is the rise in various indicators of economic and policy uncertainty that have been documented by Baker, Bloom, Canes-Wrone, Davis, and Rodden (2014).

Much of this discussion is speculative. Figuring out what has driven these fundamental changes in these indicators of dynamism and fluidity and, in turn, exploring the consequences of these changes for U.S. economic growth should be a high priority.

About the Author
John C. Haltiwanger, is a Distinguished University Professor in the Department of Economics at the University of Maryland. He also is the first recipient of the Dudley and Louisa Dillard Professorship in 2013. He is a research associate of the National Bureau of Economic Research, a fellow of the Society of Labor Economics and a senior research fellow at the Center for Economic Studies at the U.S. Census Bureau.


  1. See e.g., Davis, Haltiwanger, Jarmin, and Miranda (2007); Hyatt and Spletzer (2013); Decker, Haltiwanger, Jarmin, and Miranda (2014a, 2014b, 2015).
  2. See Foster et al. (2006) and Jarmin et al. (2009) for further discussion.
  3. See Comin and Philippon (2003), Davis, Haltiwanger, Jarmin, and Miranda (2006), Haltiwanger, Hathaway, and Miranda (2014), Decker, Haltiwanger, Jarmin, and Miranda (2015).
  4. But an interesting feature of the within-industry declines is that they have been larger for the sectors with the larger initial levels. This implies there has been convergence in reallocation rates across sectors.
  5. See Haltiwanger, Jarmin, and Miranda (2013) and Decker, Haltiwanger, Jarmin, and Miranda (2014a, 2015).
  6. There may be time aggregation issues with some establishments opening and closing within the March-to-March period captured in the BED but not included in the BDS. Building on the approach of using four times the average quarterly rates for the year, the preliminary estimate for 2015 is four times the average of the second- and third- quarter rates. This estimate should be used with appropriate caution.
  7. Figure 5 is an extended version of the series developed in Davis, Faberman, and Haltiwanger (2012).
  8. See, for example, the remarks of Laura Tyson at the Hamilton Project Event on the “The Future of Work in the Age of the Machine,” February 2015.


Baker, Scott, Nicholas Bloom, Brandace Canes-Wrone, Steven Davis, and Jonathan Rodden. 2014. “Why Has U.S. Policy Uncertainty Risen Since 1960?” NBER Working Paper No. 19826.

Comin, Diego A., and Thomas Philippon. 2005. “The Rise in Firm-Level Volatility: Causes and Consequences.” Chap. 3 in NBER Macroeconomics Annual 2005, edited by Mark Gertler and Kenneth Rogoff, vol. 20, pp. 167–228. Cambridge, MA: National Bureau of Economic Research, MIT Press.

Davis, Steven J., and John Haltiwanger. 2014. “Labor Market Fluidity and Economic Performance.” NBER Working Paper No. 20479 (published in the Federal Reserve Bank of Kansas City Jackson Hole Conference Symposium).

Davis, Steven J., and John Haltiwanger. 2015. “Dynamism Diminished: The Role of Credit Conditions,” in process.

Davis, Steven J., R. Jason Faberman, and John Haltiwanger. 2012. “Labor Market Flows in the Cross Section and Over Time.” Journal of Monetary Economics 59(1): 1–18.

Davis, Steven J., John Haltiwanger, Ron Jarmin, and Javier Miranda. 2007. “Volatility and Dispersion in Business Growth Rates: Publicly Traded versus Privately Held Firms.” Chap. 2 in NBER Macroeconomics Annual 2006, edited by Daron Acemoglu, Kenneth Rogoff, and Michael Woodford, vol. 21, pp. 107–180. Cambridge, MA: MIT Press.

Decker, Ryan, John Haltiwanger, Ron Jarmin, and Javier Miranda. 2014a. “The Role of Entrepreneurship in U.S. Job Creation and Economic Dynamism.” Journal of Economic Perspectives 28(3): 3–24.

Decker, Ryan, John Haltiwanger, Ron Jarmin, and Javier Miranda. 2014b. “The Secular Decline in Dynamism in the U.S.” mimeo.

Decker, Ryan, John Haltiwanger, Ron Jarmin, and Javier Miranda. 2015. “Where Has All the Skewness Gone? The Decline in High-Growth (Young) Firms in the U.S.” mimeo.

Foster, Lucia, John Haltiwanger, and C.J. Krizan. 2006. “Market Selection, Reallocation, and Restructuring in the U.S. Retail Trade Sector in the 1990s.” The Review of Economics and Statistics 88(4): 748–758.

Haltiwanger, John, Ian Hathaway, and Javier Miranda. 2014. “Declining Business Dynamism in the U.S. High-Technology Sector.” The Kauffman Foundation.

Haltiwanger, John, Ron S. Jarmin, and Javier Miranda. 2013. “Who Creates Jobs? Small vs. Large vs. Young.” Review of Economics and Statistics 95(2): 347–361.

Hathaway, Ian, and Robert Litan. 2014. “Declining Business Dynamism in the United States: A Look at States and Metro,” Brookings Institution.

Hsieh, Chiang-Tai, and Enrico Moretti. 2015. “Why Do Cities Matter? Local Growth and Aggregate Growth.” NBER Working Paper No. 21154.

Hyatt, Henry R., and James Spletzer. 2013. “The Recent Decline in Employment Dynamics.” IZA Journal of Labor Economics 2(3): 1–21.

Jarmin, Ron S., Shawn Klimek, and Javier Miranda (2009), ‘The Role of Retail Chains: National, Regional, and Industry Results,’ in Timothy J. Dunne, J. Bradford Jensen, and Mark J. Roberts (eds.), Producer Dynamics, Chicago, IL: University of Chicago Press, pp.237–262.

Syverson, Chad. 2011. “What Determines Productivity?” Journal of Economic Literature 49(2): 326–365.

Startups and Young Firms in the Economy

Trends, the Great Recession, and a look ahead

By Petr Sedláček Assistant Professor, Department of Economics University of Bonn
Petr Sedláček
Petr Sedláček


1. Some reasons for concern

The Great Recession was officially deemed to be over in 2009. Although many economic indicators have since turned toward brighter days, developments along certain dimensions related to startups and young businesses remain gloomy. For example, 2009 was also the year that witnessed, for the first time in recorded U.S. history, more businesses shutting down than starting up.1 This pattern has not yet recovered. Even in 2012, three years after the official end of the crisis, the number of startups was still 31 percent below its pre-crisis level. The persistently lower number of startups also spilled over to a smaller share of young firms in the economy: in 2006, firms not older than five years accounted for 40 percent of all businesses; in 2012, it was 33 percent.

The above patterns are reasons for concern, because young firms are known to be the “engines of growth,” in the sense that they account for a disproportionately large chunk of aggregate job creation (see, e.g., Haltiwanger, Jarmin, and Miranda 2013) and that a fast pace of entry and exit is associated with productivity-enhancing creative destruction (see, e.g., Foster, Haltiwanger, and Krizan 2006). This report discusses potential sources and consequences of the declining share of startups and young firms in the economy. The last section is devoted to an attempt at pointing out missing links in our knowledge about the functioning of the economy. Further research in such areas will, I hope, improve our position for providing policy recommendations.

2. The secular decline in the startup rate

While the Great Recession was characterized by large drops in the number (and rate) of startups, it was not a historical exception. In fact, there is a clear secular decline in the startup rate accompanied by a relatively stable firm exit rate (left panel of Figure 1).2 Together, these patterns imply that the firm population is aging. While in 1982 half of the universe of firms was younger than six years, in 2012 only one-third of firms were that young. This pattern is also reflected in a drop in the employment share of young firms from 20 percent to 10 percent (right panel of Figure 1).

Figure 1. Entry and Exit Rates (left panel), Firm and Employment Shares of Young Firms (right panel)

Note: The startup rate is defined as the number of firms younger than one year over the total number of firms in the economy. The exit rate is the number of firms shutting down over the total number of firms. Young firms are those younger than six years. Source: Business Dynamics Statistics

What kind of startup rate?

The last three decades were characterized also by strong trends in the population of individuals. Therefore, perhaps a more meaningful measure of whether or not the economy has lost some of its ability to create new businesses is one in which the number of startups is scaled by the population, labor force, or employment level. However, even these alternative measures display a downward trend (left panel of Figure 2).

The demographic changes mentioned above also include an aging worker population. A common misperception is that new businesses are founded by young individuals. Interestingly, according to the Kauffman Firm Survey, the median age of business founders is forty-five. Therefore, an aging population (of individuals) might lose its ability to create startups simply because the pool of potential entrepreneurs shrinks. Measuring the startup rate in terms of the number of economically active individuals aged forty to forty-five, however, also reveals a strong downward trend (right panel of Figure 2).

Figure 2. Alternative Measures of the Entry Rate

Note: Startup rates are defined as number the of firms younger than one year over the civilian “population,” “labor force,” and “employed” (taken from the BDS) and over the labor force aged thirty-five to forty-five and forty to forty-five. Source: Business Dynamics Statistics and Bureau of Labor Statistics

What are the consequences?

Pugsley and Sahin (2014) investigate the consequences of an aging firm population for the aggregate economy. They find that once we take the developing firm demographics into account, incumbent firms have not changed their behavior over the past three decades. In other words, conditionally on age, firms’ survival rates, employment growth rates, and business cycle sensitivity have remained virtually identical to what it was thirty years ago.3

In an accounting sense, the shift away from young firms has a direct negative effect on aggregate trend employment growth. Moreover, because older businesses are less sensitive to business cycle movements, these secular patterns also dampen the cyclical component of aggregate employment growth. The trend decline in firm entry exacerbates the dampening during booms and masks it during recessions. Put together, these forces help explain the emergence of jobless recoveries when aggregate employment recovers slowly relative to output.

What are the potential sources?

While there is ample evidence documenting the presence of the secular decline in the startup rate (see also, e.g., Davis and Haltiwanger 2014; Decker, Haltiwanger, Jarmin, and Miranda 2014; Reedy and Litan 2011), its sources are not yet well understood. Identifying the driving forces is crucial for policy: are we observing an efficient response of the economy to technological or demographic shifts, or is the declining rate of business formation the result of increasing misallocation due to, for instance, rising costs of starting a business?

While increasing costs of starting a business may be part of the story, they would have to leave incumbent firm behavior unchanged (as documented by Pugsley and Sahin 2014).4 Moreover, a declining startup rate also may arise “naturally” in a growing economy. At this point, it is useful to revert back to a simple equilibrium condition, namely, that labor supply is equal to labor demand. On the one hand, labor supply in the economy depends on the size of the population (L) and the employment rate (E/L). On the other hand, the number of workers demanded by firms in the economy is determined by the number of firms (N) and their average size (S). We can then rewrite this equilibrium condition in terms of growth rates

d ln L + d ln E/L = d ln N + d ln S
(1.2%)   (0.4%)    (1.1%)    (0.5%)

which simply states that labor supply growth (either because of a growing population or a rising employment rate) can be “accommodated” by the economy through an increase in the number of firms and/or by firm size growth. The numbers in brackets show the average values for the U.S. economy in the past thirty years.

What effect can this have on the firm entry rate? Keeping the relative contributions to growth fixed, a back-of-the-envelope calculation would suggest that population growth alone would result in a 0.8 percent annual growth in the number of firms.5 This, together with a roughly constant inflow of the number of startups (as in the BDS data), implies a drop in the startup rate amounting to about 70 percent of the observed secular decline (Figure 3).6 Karahan, Pugsley, and Sahin (2015) come to a similar conclusion using cross-state variation in demographics and startup rates to identify the impact of a slowdown in population growth on the rate of firm entry.7

In other words, and perhaps surprisingly, the majority of the startup rate decline seems to be explained by demographic changes. Missing from the above simple exercise, however, are feedback effects from the lack of startups to incumbent firms, an analysis of reasons why the economy adjusts both on the extensive (number of firms) and intensive margins (firm size), and, most importantly, what factors determine the relative strength of these adjustments.

Figure 3. Startup Rate, Trend, and Trend Implied by Population Growth

Note: Level and (linear) trend of actual startup rate and trend startup rate implied by observed population growth. Source: Business Dynamics Statistics and author’s calculations

1. Has the Great Recession left a persistent scar?

The latest numbers of firm startups not only are the result of the trend decline discussed above, but also stem from a particularly strong recession. Of all the recessions since the late ’70s, the Great Recession gave rise to the largest decline in startups, even when we take into account the severity of the downturn (either by looking at output growth or the increase in the unemployment rate). Can this unprecedented cyclical decline in the number of firms have persistently negative effects on the aggregate economy in future years?

A lost generation of firms

A decline in firm entry not only has a direct effect through a lack of job creation by startups, but also creates a ripple effect in later years as the smaller cohort of entrants grows older. In fact, the pro-cyclical nature of firm entry spills over to young firms, resulting in pro-cyclical movements in the share of young businesses.8 Moreover, a simple variance decomposition of the variation in aggregate employment reveals that young firms account for 40 percent of its cyclical fluctuations even though they employ only 16 percent of the workforce. This reiterates the findings of Haltiwanger, Jarmin, and Miranda (2013) about the disproportionately large contribution of young firms to aggregate job creation on average, but this time at business cycle frequencies. Together, these facts raise concerns about the medium- to long-run impact of the recently lost generation of firms in the Great Recession.

We can obtain a rough estimate of the potential impact of the lost generation of firms by simulating a drop in the entry rate in an economy that has survival and employment growth rates of incumbent firms fixed to their sample averages. The unemployment rate impact of a drop in the number of startups (of the magnitude seen in the Great Recession) is quite alarming: even ten years after the shock subsides, the unemployment rate remains more than one percentage point above its pre-crisis level (see Sedláček 2015).

However, this type of calculation is somewhat misleading and Sedláček (2015) shows that the behavior of incumbent firms is instrumental for understanding the impact a lack of startups may have on the economy. Using an estimated structural model, the paper shows that, even if the number of startups had remained at its pre-crisis level up until 2012, the effect on the aggregate economy is limited (in subsequent years, the unemployment rate would have been, at most, 0.5 percentage points lower). The reason is that incumbent firms take advantage of the lack of startups (and, thus, job creation), and they hire and retain workers more easily. Moreover, depressed wages increase firm profits, and this promotes higher firm entry in future years.

The take-away message from the above analysis is that a lower number of startups per se does little to the aggregate economy unless there are reasons preventing incumbents from taking advantage of the slack in the labor market. The paragraphs below discuss two such reasons.

The composition of startups

One can imagine that the composition of startups is different in recessions as compared to booms. On the one hand, perhaps only the relatively more productive startups are good enough to survive in the downturn. On the other hand, it may be that new firms starting up in recessions are dominated by “necessity” entrepreneurs trying to escape unemployment. Sedláček and Sterk (2014) analyze the cyclical changes in the composition of new firms by investigating the growth potential of cohorts of startups several years after they enter the economy. They find that cohort-level employment is extremely persistent and that the majority of its changes are driven by differences in firm sizes, rather than the number of firms across cohorts. In other words, cohorts of startups that enter small (typically in recessions) turn out to remain small even years down the track.

Moreover, using an estimated structural model, they find that composition effects are important not only at the cohort level, but also for aggregate fluctuations. In particular, changes in “birth” conditions of firms help shape the trend in aggregate employment. Their results therefore suggest that (all else equal) the unprecedented drop in the number of startups in the Great Recession may be followed by a prolonged period of lower aggregate employment due to negative selection effects among firms born during the crisis.

Figure 4. Startups per Number of Young in the Labor Force

Note: Number of startups per the number of young (twenty to twenty-five and twenty-five to thirty-five) in the labor force. Source: Business Dynamics Statistics and Bureau of Labor Statistics

Who do startups and young firms hire?

Another reason incumbent firms may not be able (or indeed willing) to compensate for the lack of job creation due to a lower number of startups is that new and young firms hire in “different markets.” Ouimet and Zarutskie (2013) document that young businesses are more likely to employ (and hire) younger workers. In fact, they show that a regional increase in the supply of young workers is followed by a rise in the startup rate in those regions. Interestingly, this pattern is apparent also at the aggregate level, where the number of startups per the number of economically active young is essentially flat (Figure 4).

A cyclical decline in the number of startups could then give rise to mismatch unemployment as the pool of job seekers becomes disproportionately occupied by workers who are not appropriate candidates for incumbent firms. That said, current research seems to assign only a limited role to mismatch for determining unemployment rate dynamics (see, e.g., Sahin, Song, Topa, and Violante 2013).9

4. A look ahead: what would we like to know?

The currently low share of new and young firms in the economy is a combination of a thirty-year secular decline and a particularly strong cyclical downturn following the Great Recession. Recent research shows that such changes in the firm age distribution have implications for the aggregate economy and may lead to a persistent drag on employment in years to come.

However, what is not yet well understood are the sources of these changes. To provide effective policy advice, we need to understand why (the slowdown of) population growth affects both the number of firms (i.e., startups and firm closures) and the average firm size, as well as what determines the extent of adjustment along these two margins. In particular, what is the role of (possibly changing) costs of starting up and running a business in this adjustment process?

Moreover, existing evidence suggests that most young firms either fail or do not grow. The prowess of young firms in terms of creating jobs therefore comes from a relatively small fraction of very successful firms. To better understand the consequences of periods of low firm entry, we need to know more about these success stories: what do they do, who do they hire, and what does this imply for aggregate productivity and employment growth?

About the Author
Petr Sedláček is an assistant professor of Economics at the University of Bonn in Germany and a member of the Hausdorff Center for Mathematics. His main areas of interest include macroeconomics of entrepreneurship and firm dynamics and, in particular, their impact on the labor market. Sedláček obtained his PhD at the University of Amsterdam in 2011. Prior to this, he worked as an economic analyst at the Czech Statistical Office in Prague.


Davis, S., and J. Haltiwanger. 2014. “Labor Market Fluidity and Economic Performance.” NBER Working Paper 20479.

Davis, S., J. Haltiwanger, R. Jarmin, and J. Miranda. 2006. “Volatility and Dispersion in Business Growth Rates: Publicly Traded versus Privately Held Firms.” NBER Macroeconomics Annual.

Dawson, J., and J. Seater. 2013. “Federal Regulation and Aggregate Economic Growth.” Journal of Economic Growth, 18(2), 137–177.

Decker, R., J. Haltiwanger, R. Jarmin, and J. Miranda. 2014. “The Role of Entrepreneurship in US Job Creation and Economic Dynamism.” The Journal of Economic Perspectives, 28(3), 3–24.

Foster, L., J. Haltiwanger, and C. J. Krizan. 2006. “Market Selection, Reallocation, and Restructuring in the US Retail Trade Sector in the 1990s.” Review of Economics and Statistics, 88(4), 748–758.

Haltiwanger, J., R. Jarmin, and J. Miranda. 2013. “Who Creates Jobs? Small versus Large versus Young.” Review of Economics and Statistics, 95(2), 347–361.

Karahan, F., B. Pugsley, and A. Sahin. 2015. “Understanding the 30-Year Decline in Business Dynamism: A General Equilibrium Approach.” mimeo.

Ouimet, P., and R. Zarutskie. 2014. “Who Works for Startups? The Relation between Firm Age, Employee Age, and Growth.” Journal of Financial Economics, 112(3), 386–407.

Pugsley, B., and A. Sahin. 2014. “Grown-Up Business Cycles.” Federal Reserve Bank of New York, Staff Report 707.

Reedy, E.J., and R. Litan. 2011. “Starting Smaller; Staying Smaller: America’s Slow Leak in Job Creation.” Kauffman Foundation Research Series: Firm Foundation and Economic Growth.

Sahin, A., J. Song, G. Topa, and G. Violante. 2014. “Mismatch Unemployment.” American Economic Review, 104(11), 3529–3564.

Sedláček, P. 2014. “The Aggregate Matching Function and Job Search from Employment and Out of the Labor Force.” mimeo.

Sedláček, P. 2015. “The Lost Generation of Firms and Aggregate Labor Market Dynamics.” mimeo.

Sedláček, P., and V. Sterk. 2014. “The Growth Potential of Startups over the Business Cycle.” mimeo.

Stangler, D., and P. Kedorsky. 2010. “Exploring Firm Formation: Why is the Number of New Firms Constant?” Kauffman Foundation Research Series: Firm Formation and Economic Growth.


  1. All referenced data about firms and establishments are for the U.S. economy and come from the Business Dynamics Statistics (BDS), unless stated otherwise. The BDS covers almost the entire universe of private employment and gives annual information on the number and employment behavior of firms and establishments from 1977 to 2012.
  2. On the contrary, the number of startups has been relatively stable in the past three decades (with the exception of the latest downturn, from which the number of startups has not yet recovered).
  3. Davis, Haltiwanger, Jarmin, and Miranda (2006) show, however, that business dynamism-measured (employment weighted) average volatility of firm growth rates has declined by more than 40 percent since 1982. Only about one-third of this decline can be traced back to a shift toward older businesses.
  4. Nevertheless, there is some evidence that increased U.S. federal regulation has negatively impacted aggregate economic activity (see, e.g., Dawson and Seater 2013).
  5. Population growth accounts for three-fourths of the labor supply increases. Three-fourths of the 1.1 percent growth in the number of firms gives an annual growth rate of 0.8 percent.
  6. See Stangler and Kedorsky (2010) for an analysis of the constancy of the number of startups.
  7. In their analysis, between 50 percent and 75 percent of the observed secular decline in the startup rate is due to a slowdown in population growth.
  8. Cyclical changes in firm survival rates turn out to explain only about 1 percent of the fluctuations in aggregate employment.
  9. Sedláček (2014) shows, however, that the severity of mismatch may be underestimated when job seekers from outside unemployment are not taken into account.

Measuring Entrepreneurship


In nearly any conversation on entrepreneurship, talk eventually turns to data and measurement. Entrepreneurship is a slippery thing to define, let alone measure. And, while enormous strides have been made in improving data on entrepreneurship, there is still a great deal of nuance and variation that is hard to capture. One conference contributor, in arguing that the United States will not return to the robust GDP growth of the 1990s, highlighted a fundamental difference in high-tech firms between 1990 and today. In the 1990s, we saw a small fraction of high-growth, young firms create an enormous number of jobs. Young high-tech firms, in particular, that were high in the productivity distribution took off and saw tremendous job growth. While data for more recent years still show huge dispersions in productivity, we are not seeing those firms achieve job growth and output growth.

Furthermore, he explained that, when we look at the last 100 years of data, across time and sectors, robust periods of economic growth (both productivity and jobs) have been accompanied by high degrees of entrepreneurship, within-sector disruption, and fluidity (workers moving across jobs). These indicators, he said, are consistent with models of economic growth, and they mean that the lower levels of entrepreneurship we are seeing today are cause for concern. This is not, moreover, solely an American phenomenon: the OECD has documented similar declines in economic dynamism across other countries. He challenged the group to consider how we can achieve the higher levels of entrepreneurship, creative destruction, and dynamism we need for high levels of economic growth in the future.

Responding to the trends in data that were presented, another participant affirmed that there is a solid empirical project in this measurement of business dynamism, and the decline in dynamism it shows tells us something real and important about jobs. He suggested, however, that there may now be a disconnect between high growth and jobs, in which firms are doing well, but not creating jobs in the United States, perhaps because the jobs are growing internationally. And, going beyond this question to the larger issue of measurement, he explained that these trends don’t reflect the economy people who aren’t empirical economists perceive because we hear so much about entrepreneurial firms in Silicon Valley and we see many students in MBA programs seeking to become entrepreneurs. Entrepreneurship, it appears, is even taking off in places we didn’t used to see it, like New York City. While no one is denying the jobs story, he reiterated, there appears to be a significant amount of entrepreneurial activity.

The contributor suggested that we need to understand how both of these observations can be true at the same time. The key to this understanding, he proposed, may be that there are different types of entrepreneurs. The vast majority of new businesses have no growth aspirations and see success only as reaching five employees. On the other end of the spectrum, the highest-quality entrepreneurial firms are different from the outset because they have such high growth goals. The number of these firms is quite small, however, and they often fail because the process is so risky. While we are captivated by what we see happening in this top group of firms, the data primarily reflect the vast majority of new firms in the country and trends within this larger group.

The participant cited the data from the dot-com boom as evidence of this measurement challenge. While the large-scale data from the Census Business Dynamics Statistics didn’t register any boom in new business registrations during these years and even showed a decline in new entrances and GDP during that time, the period is known for the small number of high-growth firms that saw tremendous growth and captured our attention. In fact, 1996 was the single best year to start a business; the cohort that began in that year was five times more likely to be successful than those that began in the 2000s.

Yet another participant agreed, suggesting that we need to distinguish between different types of entrepreneurs when we consider larger trends and when we establish policies that affect new businesses. The new businesses started by Harvard and MIT graduates, for example, are not representative of all new businesses, and 1996, he pointed out, was only the best year to start a business for those in a very small subset of businesses positioned for the Internet. Furthermore, he suggested that there may be a third distinct type of entrepreneur, those like app developers, who are qualitatively different from the small businesses with little growth ambition (like new hair salons) and the entrepreneurs with high growth goals (such as venture capital-backed companies). Like the high-growth entrepreneurs, this group is very small. Perhaps, he suggested, businesses that intend to remain small represent 95 percent of all firms; another 4 percent are entrepreneurs like these app developers; and the last 1 percent represents the firms with the ambition and potential to grow. While we may move beyond a binary understanding of entrepreneurship conceptually, he noted that it is difficult to distinguish between these three types and identify the high-growth businesses in data analysis. Other participants agreed, noting that we need higher resolution data sources to give us a better sense of the characteristics of entrepreneurs and the dimensions of entrepreneurial outcomes.

The other contributor responded that we can identify the high-growth businesses by creating an algorithm to look at all business registrants and determine if they have growth ambitions and potential. The business name, for example, is an important indicator. Something with “pizza” in the name, for example, is unlikely to be a growth firm, but “Thermonuclear Biotic” likely has growth ambitions, even if it may fail. Using this algorithm, we can map firms that grow and identify startup characteristics to learn about factors associated with high-growth firms.

While Jorge Guzman and Scott Stern acknowledge the overall decline in entrepreneurship in their essay, “Measuring Growth Entrepreneurship and the Performance of Entrepreneurial Ecosystems: Evidence from Five U.S. States, 1988–2014,” they also point to a simultaneous explosion of entrepreneurship and investing, especially in places like Silicon Valley. Census data, for example, show a steady downward trend in firm formation, but measurement of “growth events” displays a different pattern of spike, fall, and stabilization. The authors develop new measures of entrepreneurial quality and regional entrepreneurship potential and regional entrepreneurial acceleration, finding a more cyclical pattern that contrasts with the secular decline of dynamism. While there is significant variation across time and place in these measures, the authors conclude that the United States has experienced rising entrepreneurial quality and potential, but a breakdown in growth conversions. Policies, they argue, should be oriented toward helping the growth and performance of existing entrepreneurs, rather than focused on increasing the supply of quality entrepreneurship.

In his essay, “Where is Innovation Falling Short? Using Labor Market Indicators to Map the Successful Innovation Frontier,” Michael Mandel calls for new measures of the American innovation economy and presents an analysis with new labor market indicators. He finds that these indicators show a rapid pace of innovation in areas such as information technology and robotics and oil and gas, but a lower and slower rate of innovation in areas that include the biosciences and material sciences. This “uneven innovation,” he suggests, may be holding down overall productivity growth, and he suggests that further mapping of the innovation frontier with new data and measurement tools should be undertaken as a way to spur more entrepreneurship and innovation.

Kyle Handley examines the impact of globalization on entrepreneurship in his paper, “Entrepreneurship and the Challenge of Globalization.” The offshoring of tasks and global unbundling of production, he explains, have mixed effects for entrepreneurship: new firm growth may be unevenly distributed, and inequality within firms may rise because of differential returns to skill. Increasing global trade and integration expose more firms to more competition, which will have a different impact on different sectors. He suggests that, because firms that export and import are larger, more productive, and pay higher wages, exposing more entrepreneurial firms to global trade could be a net positive. The costs and economic risks involved in entering international trade, however, can be very high, especially for young firms. Policy uncertainty can compound these costs and risks, he explains, but institutions such as the World Trade Organization or free trade zones like the European Union can reduce policy uncertainty for new exporters and, thus, raise the economic payoff.

Finally, Rajshree Agarwal and Ben Campbell’s paper, “Entrepreneurship and Intrapreneurship: The Role of Human Capital and Complementary Assets,” suggests that the decline in U.S. entrepreneurship may be, in part, caused by an increase in intrapreneurship. If this is the case, they explain, the implications of the decline in entrepreneurship for dynamism are more benign, and innovation is continuing to occur. The decision to be an entrepreneur or intrapreneur, they explain, is shaped by the availability and cost of complementary assets, and it is possible that large companies are doing a better job of providing internal opportunities for their employees who seek to pursue new ideas. The authors suggest that we need to look beyond entrepreneurs and look to the connection between innovation, complementary assets, and organizational forms, measuring innovation wherever it occurs. Policymakers, in particular, need to understand the behavior of a more broad category of innovators, because policies make an impact at the individual level, changing individuals’ incentives and constraints.

A New View of the Skew:

A Quantitative Assessment of the Quality of American Entrepreneurship

By Catherine Fazio Managing Director, Laboratory for Innovation Science and Policy The MIT Innovation Initiative
Jorge Guzman Doctoral Student MIT Sloan School of Management
Fiona E. Murray Associate Dean For Innovation, Co-Director MIT Innovation Initiative MIT Sloan School of Management
Scott Stern David Sarnoff Professor of Management and Chair of the Technological Innovation, Entrepreneurship, and Strategic Management Group MIT Sloan School of Management
Catherine Fazio
Catherine Fazio
Jorge Guzman
Jorge Guzman
Fiona E. Murray
Fiona E. Murray
Scott Stern
Scott Stern



The state of American entrepreneurship shapes the outlook for the American economy. High-growth startups contribute disproportionately to net job creation and to impactful innovation, laying a crucial foundation for economic dynamism and prosperity. Fostering these firms is a strategic priority.

Among all new businesses, however, only a very small fraction experience the explosive growth (in terms of jobs, revenue, or valuation) that propels the economy. The state of American entrepreneurship—and its potential to fuel economic dynamism and prosperity—therefore depends more on whether there are enough startups being founded with the potential to realize this outsized performance than on the quantity of new business starts. As Robert Litan, former vice president of research and policy at the Kauffman Foundation, noted in the wake of the Great Recession, “America’s great challenge is to … bring about a substantial increase in the numbers of highly successful new companies … Nothing less than the future welfare of America and its citizens is at stake.”1 From the perspective of a policymaker, a central difficulty in assessing the state of American entrepreneurship is being able to systematically account for “the skew:” the fact that the overall ability of entrepreneurship to facilitate American economic prosperity depends disproportionately on the realized performance of a very small number of new firms. But how do we identify whether the economy at a given point in time is nurturing the types of startups that have the potential for exponential growth?

Accounting for the skew requires confronting a measurement quandary: at the time a company is founded, one cannot observe whether that particular firm will experience explosive growth (or not). On the one hand, this challenge is fundamental, since by its nature entrepreneurship involves a high level of uncertainty and luck. And, some outsized successes certainly result from unlikely origins. Ben & Jerry’s, for example, was founded with the intention to be a one-store, homemade ice-cream shop.2"> On the other hand, many startups aspire to a specific level of performance and then achieve it, including startups that we refer to as innovation-driven enterprises (IDEs) and more traditional small and medium-size enterprises (SMEs).3 Across all new business starts, firms span a wide gamut in terms of their founders’ ambitions and potential for growth. A very large number of new businesses aim to offer successful local services (such as a neighborhood handyman striving to build a steady book of regular clients) in the traditions of the SME phenomenon.4 Others aspire to be regional players or to grow over time in an incremental manner. Still others have aspirations to be the next Google or Facebook (classic IDEs).5 To the extent that the new firms that ultimately contribute to the skew are disproportionately drawn from those IDEs with significant growth ambitions and underlying potential at the time of founding, mapping the skew in a systematic way requires accounting for these differences at an early stage in the entrepreneurial process.

Traditional approaches towards measuring entrepreneurship have by and large abstracted away from firms’ initial differences in growth potential—tracking the rate of entrepreneurship by either counting new firms (considering all firms within a given sector to be equal) or selecting on achieving a performance outcome (such as the receipt of venture funding). Though these quantity-based and performance-based approaches are both instructive, neither provides a clear view of the skew. The first cannot discern whether changes in the quantity of entrepreneurship within a sector reflect changes in the founding rate of firms whose underlying potential for growth is modest versus those with the potential for exponential growth. The second conflates the analysis of startup potential at the moment of founding with other factors influencing later success (such as the relative supply of risk capital, regional ecosystem effects, or luck).

More significantly, these different measurement approaches have led to divergent perspectives about the state of American entrepreneurship and fueled a polarized policy debate.

  • Quantity-based measures document a troubling, three-decade-long decline in the U.S. rate of entrepreneurship and business dynamism (the pace at which the economy reallocates economic activity), with only a very modest leveling off and increase in high-tech, confined to the late 1990s.6, 7, 8 These findings have prompted urgent calls to jumpstart the creation rate of new firms. As the Chairman and CEO of Gallup, Jim Clifton, cautioned: “We are behind in starting new firms per capita, and this is our single most serious economic problem. … This economy is never truly coming back unless we reverse the birth and death trends of American businesses.”9
  • Conversely, outcome-based measures indicate that the rate of entrepreneurship is rising. Early-stage angel and venture capital financing of new ventures has been on a significant upswing over the past several years. In addition, a recent report co-authored by one of us (Fiona Murray) documents a striking shift in the propensity for MIT undergraduates to join startup firms at graduation.10 Some leading entrepreneurs and financiers in ecosystems such as Silicon Valley fear not that there are too few startups, but that we are in the midst of an entrepreneurship bubble!11

As Marc Andreessen succinctly put it: “There’s too much entrepreneurship: Disruption running wild!” “There's too little entrepreneurship: Economy stalling out!”12

This policy brief builds on research conducted by two of us (Guzman and Stern, “The State of American Entrepreneurship: New Estimates of the Quantity and Quality of Entrepreneurship for 15 U.S. States, 1988–2014,”13) that aims to break through this impasse. This work calculates consistent estimates of the underlying growth potential of startups using a combination of comprehensive business registries and predictive analytics and drawing on startup characteristics observable at or near the time of founding. These new metrics allow for the evaluation of the state of entrepreneurship across time (and place) and yield several new findings. Contrary to the secular decline in the rate of net firm births observed with quantity-based measures, the rate at which high-potential growth startups are founded follows a cyclical pattern that is sensitive to the capital market and overall market conditions. Among the key new findings are (1) a sharp, upward swing in the number of expected growth outcomes starting in 2010; and (2) strong differences across place and time in the likelihood of startups (for a given quality level) to realize their potential and scale. These findings demonstrate the importance of accounting for quality when measuring the quantity of entrepreneurship and evaluating its potential impact on future economic growth.

The state of American entrepreneurship looks quite different when one has a clear view of the skew. Startups with the ambition and potential for exponential growth have strikingly different patterns of creation than SMEs. Further, traditional measures of the overall rate of entrepreneurship do not effectively capture the likely potential of these firms to scale. Finally, to the extent that the current state of American entrepreneurship is facing a crisis, it is not in the rate of creation of high-growth-potential startups or even in the initial funding of those firms, but, instead, in the potential of those firms to scale in a meaningful way over time.

These findings set the table for a new conversation about the direction of innovation and entrepreneurship policy—one that calls for reconsideration of whether efforts to jumpstart entrepreneurial quantity independent of quality can effectively lever economy-wide growth and prosperity. As emphasized by former Small Business Administrator Karen Mills, we can do better for both SMEs and IDEs by designing policies more directly tailored to the acceleration of each.14

Traditional Measures Do Not Effectively Capture Entrepreneurial Potential

Broadly speaking, academics traditionally have measured the rate of entrepreneurship in two basic ways: (1) quantity-based, population-level statistics tracking firm birth and exit rates that abstract away from any variation in growth potential or ambition; and (2) performance-based measures that account for heterogeneity in retrospect based on outcomes. Both are highly informative about different aspects of entrepreneurship and regularly used by policymakers to guide decisions aimed at catalyzing high-growth outcomes. But do these measures provide a good signal of entrepreneurial potential to realize explosive growth? We conclude that the signal each offers of the skew is weak, at best. This section provides a brief background of leading examples of each of these types of measures and then highlights important disconnects with other indicators and/or drivers of performance.

Two leading examples of quantity-based measures of the state of American entrepreneurship are the Business Dynamics Statistics series from the U.S. Census (“BDS”) and the Kauffman Index of Startup Activity. BDS measures the overall quantity of new business starts, specifically emphasizing the number of new firm births relative to firm exits. Except for sector differences, the BDS considers firm potential to be equal at the time of founding. In a series of important and insightful papers using the BDS, John Haltiwanger and co-authors document a troubling three-decade-long secular decline in the U.S. rate of entrepreneurship, with only a very modest leveling off and increase in high-tech, confined to the late 1990s.15, 16, 17 They find that this decline in the overall rate of entrepreneurship appears to be linked to a decline in business dynamism—the pace at which the economy reallocates economic activity. While this drop is most pronounced in industries such as retail trade, the overall pattern of decline also is present in other sectors, including high-tech. The foremost index tracking startup activity in the United States, states and metropolitan areas—the Kauffman Index: Startup Activity18—is a quantity-based measure of new self-employment rates using the Current Population Survey.19 This careful tracking of all new entrepreneurs allows one to evaluate differences across regions and time in the rate at which individuals become entrepreneurs (of any type).20

By construction, neither the BDS measures nor the Kauffman Index: Startup Activity evaluate whether changes in the quantity of entrepreneurship reflect changes in the founding rate of firms or founders whose underlying potential is modest versus a change in the founding rate of firms or founders with the potential for exponential growth. The methodology used does not account for differences in initial potential for growth (outside of those that might generally exist across sectors). The question remains whether the trends observed provide a good signal of the high-potential growth skew.

There are at least four disconnects that lead us to conclude they do not:

Disconnect 1: Quantity-Based Measures of Entrepreneurship Have Little Relationship to GDP Growth. Yearly fluctuations in counts of firm births appear to hold little relationship to medium-term measures of economic performance. The Business Dynamics Statistics series from the U.S. Census find that young firms produce the most employment growth.21, 22, 23, 24 Accordingly, we would expect to see the fluctuations in the founding of new firms to roughly track economic boom and bust cycles (or for those cycles to follow the trajectory of new firm starts with a lag). Instead, we see new firms on a long-term secular decline largely independent of the economic cycle.
Figure 1 (Source: Guzman and Stern, 2016)
  • Disconnect 2: Quantity-Based Measures Hold Little Relationship to Equity Growth. The steady decline in entrepreneurship shown in quantity-based measures does not track the more cyclical nature of high-value startup exits. If quantity-based measures were an effective signal of entrepreneurial growth potential, we would expect the opposite. Put differently, more firm births should mean more “shots on goal” and a higher rate of growth firms emerging. Conversely, fewer firm births should lead to fewer shots on goal and fewer growth firms emerging. Yet, when we compare the original “birth dates” of firms within Census Business Dynamic Statistics that achieved successful exits (defined as an IPO or acquisition at a multiple of the firm’s valuation within six years) relative to overall firm births, again, we find no apparent relationship.
Figure 2 (Source: Guzman and Stern, 2016)
  • Disconnect 3: Quantity-Based Measures Cannot Find Silicon Valley. Though directly informative about the rate of self-employment, perhaps the most well-known regional index tracking startup activity across the United States—the Kauffman Index: Startup Activity25—regularly finds the rate of startup activity to be higher in Montana and Alaska than in California.26 Indeed, both the 2014 and 2015 Startup Activity ranking found Montana to be first in the nation in number of startups founded. Kauffman’s 2015 ranking finds more startup activity in Miami, Florida, than in either San Francisco or San Jose, California. The Index likewise ranks Miami, as well as Columbus, Ohio, and Phoenix, Arizona, above Boston, Massachusetts. This mismatch between Index rankings and top hotspots of entrepreneurial activity (like Silicon Valley and Kendall Square) signal strongly that, to the extent that trends in entrepreneurial growth potential are being captured, they have been swamped by the effects of more local or regional businesses.

Alternatively, two leading examples of performance-based measures of entrepreneurship include the 2015 IPO Report by Wilmer Hale and the “PwC/NVCA MoneyTree™ Report.” Instead of counting the number of new firms or founders, both track the number of startups that have achieved certain performance outcomes. The 2015 IPO Report tracks the number of IPOs and dollar volumes by year and finds: “The number of IPOs has seen a steady annual increase in all but one of the past six years, and the last seven quarters have each produced fifty or more IPOs—a level of consistently high activity not seen since 2000.”27 Similarly, the PWC/NVCA MoneyTree Report shows a significant increase in annual venture capital investment dollars following the Great Recession in 2009.28 Though instructive indicators of whether surviving startups have been able to scale, these performance-based measures also fail to measure startup potential for growth by virtue of how they are constructed.

  • Disconnect 4: Performance-Based Measures Put the Cart Before the Horse. Selecting on performance after it has occurred makes it difficult, if not impossible, to disentangle the different effects that might have contributed to or detracted from that outcome. The number of IPOs or employment growth experienced by startups, for example, could have happened for any number of reasons (including luck, market dynamics, ecosystem effects, and the underlying potential of the new firms that realized performance). Rates of performance in periods measured could reflect more about the period in question than about the underlying potential of new firms for growth. Past performance may not be a valid indicator of future rates of success.

Thus, it is not safe to assume that the secular decline in the net births of newfirms mirrors trends for high-potential-growth startups. Nor can we assume that current rates of employment growth or equity outcomes are a good proxy for the present growth ambitions or potential of new firms. Instead, performance measures may be reflecting other issues—such as the effect of an ecosystem on a startup’s ability to realize its growth potential.

To effectively evaluate the state of American entrepreneurship, we need a new approach—one that prospectively accounts for the differences in the potential for growth at the time of founding and recognizes that all new firms have at least some growth potential.

A Quantitative Approach to the Measurement of Entrepreneurial Quality

The State of American Entrepreneurship: New Estimates of the Quantity and Quality of Entrepreneurship for 15 U.S. States, 1988–2014”29 introduces a new lens through which to evaluate startup trends over time—entrepreneurial quality. Its findings complement and enrich quantity and performance-based measures, offering reliable estimates that predict new firms’ average potential for growth, the number of growth outcomes expected, and whether the firms’ potential will be helped or hampered by the ecosystem where they are located.

The quantitative estimation of entrepreneurial quality builds on three interrelated insights. First, a practical requirement for any growth-oriented entrepreneur is business registration (as a corporation, partnership, or LLC).30 State-level, public business registration documents therefore represent a robust sample of entrepreneurs at a similar and foundational stage of their entrepreneurial process (and a viable alternative to firm births in the Business Dynamics Statistics). Second, beyond counts of business registrants, characteristics noted within business registration filings (made at or close to the time of firm founding) are good endogenous signals for growth ambition or potential (what we call “entrepreneurial quality”). These “startup characteristics” include how a firm is organized (e.g., corporation vs. partnership), whether it is registered in Delaware, and how it is named (e.g., after its founder vs. a type of technology, and long vs. short). The paper verifies that early firm name choices are correlated with the founders’ intentions with respect to growth.31 For example, businesses named after an individual (e.g., Florentino’s Handyman Services) or with terms like “café” or “realtor” are more likely to be local SMEs. Firms with short names or tied to specific high-tech sectors (like Stemcentrx in biotech) are likely positioning themselves as innovation-driven enterprises and signaling their intention and aspiration for growth. In addition, other “digital signatures” of early-stage milestones (e.g., filing for a patent or trademark close to their founding date) can help to identify high-potential firms.32 Third, meaningful growth outcomes for startups (IPOs or high-value acquisitions within six years of founding) are observable with a lag and can be mapped back to startup characteristics to estimate the relationships between them. This mapping enables the estimation of entrepreneurial quality for any business registrant within the sample (even where outcomes have not yet materialized).33 Entrepreneurial quality is thus measured as the estimated probability of achieving a growth outcome given startup characteristics at founding. characteristics at founding.34

Table 1 reports the core empirical relationship, which is based on a logit regression model that allows one to examine how the presence or absence of a startup characteristic correlates with the probability of growth (conditioning on the presence or absence of other startup characteristics). Before examining specific results, it is useful to highlight an important broad finding: there is an extremely strong (and robust) correlation between startup characteristics and the probability of growth. Substantial changes in the predicted likelihood of a growth outcome are associated with characteristics observable in real time from business registration records (“nowcasting”) as well as characteristics observable with a lag (e.g., patent and trademark applications). At the time of founding, for example, corporations are four times more likely to grow, firms with short names are 2.5 times more likely to grow, and eponymous firms are 70 percent less likely to grow. The likelihood of growth is five times higher for firms with trademarks and thirty-five times higher for firms that apply for patents. Firms registering in Delaware have an even bigger disparity: they are forty-five times more likely to grow. Firms that both register in Delaware and apply for patents have an outsized 196X boost in their probability of growth. It is very important to emphasize that these startup characteristics are not the causal drivers of growth, but instead are “digital signatures” that allow us to distinguish firms in terms of their entrepreneurial quality. Registering in Delaware or filing for a patent will not guarantee a growth outcome for a new business, but the firms that historically have engaged in those activities have been associated with skewed growth outcomes.35, 36 Finally, these changes in predicted probabilities are multiplicative in nature: a Delaware firm with both a patent and trademark is 984 times more likely to grow than a firm that only registers in its home state and does not apply for intellectual property protection.

These findings can be used to construct, for every registered firm, its underlying probability of growth at founding. The probability of growth for an average firm is very low (on the order of one in 3,500). However, for those firms with multiple startup characteristics that positively predict growth, the probability of growth is dramatically higher (the top 1 percent of firms have a better than one in 100 chance of achieving growth outcomes). These estimates of entrepreneurial quality at the firm level can, in turn, be used to develop new economic statistics illustrating the state of American entrepreneurship over time (and across locations within the United States). We focus on three new indices that simultaneously account for both the quantity and the quality of entrepreneurship:

  • EQI—the Entrepreneurial Quality Index—the average growth potential (or “quality”) of any given group of new firms.
  • RECPI—the Regional Entrepreneurship Cohort Potential Index—the number of startups within a particular location or region expected to later achieve a significant growth outcome.
  • REAI—the Regional Entrepreneurship Acceleration Index—the ability of a region to convert entrepreneurial potential into realized growth.

Each index calculates a different quantitative measure that incorporates the quality of entrepreneurship. The EQI, RECPI, and REAI indexes give a better indication than possible under traditional methods about just how skewed the distributions of growth potential and likely growth outcomes are (and whether and to what extent a greater number of small to medium-sized businesses could be expected to catalyze the same growth outcomes as a high-potential growth firm).37 Additionally, REAI systematically quantifies the ratio of realized to expected growth events for a given cohort of new firms, providing an indication of whether the ecosystem in which a cohort of new firms is located is conducive to growth (or not). As such, these indexes offer policymakers and stakeholders a better view of whether and to what extent their regions are attracting/generating startups with high-growth potential vs. helping/hampering these firms’ efforts to realize their potential.

The State of American Entrepreneurship

Looking at EQI, RECPI, and REAI on an annual basis from 1988–2014 for fifteen states (representing close to 51 percent of U.S. GDP), presents a different and deeper view into the state of American entrepreneurship. Figure 3 highlights several interrelated patterns:

  • The expected number of growth outcomes (think successful startups) in the United States (RECPI relative to GDP or “U.S. RECPI”) has followed a cyclical pattern that appears sensitive to the capital market environment and overall market conditions.
  • U.S. RECPI reflects broad and well-known changes in the environment for startups, such as the dotcom boom and bust of the late 1990s and early 2000s.38
  • While the expected number of high-growth startups peaked in 2000 and then fell dramatically with the dot-com bust, starting in 2010 there is a sharp, upward swing in the expected number of successful startups formed and the accumulation of entrepreneurial potential for growth (even after controlling for the change in the overall size of the economy).
  • Notwithstanding the cyclical nature of U.S. RECPI trends, U.S. RECPI has exhibited an overarching upward trend across the full time-series of our sample (Figure 3). The rate of expected successful startups fell to its lowest point in 1991, at a level that has not been approached again. U.S. RECPI downturns in the wake of the dot-com burst (from 2000–2004) and Great Recession (from 2007–2009) ebbed at levels significantly above its 1991 nadir. U.S. RECPI thus provides a strong signal that the state of American entrepreneurship is not imperiled by a lack of formation of high-growth-potential startups, but, instead, by other dynamics or ecosystem effects that may be inhibiting the ability of startups to realize their growth potential.
  • Finally, relative to quantity-based measures,39, 40, 41 regional variation in the number of expected high-growth startups (RECPI) holds a much stronger relationship to economic growth. Whereas increases in RECPI successfully predict economic growth six years later, quantity-based measures do not (see Guzman and Stern, 2016, Table 5).
­Figure 3 (Source: Guzman and Stern, 2016)

While variation across cohorts’ average growth potential has a clear relationship to later performance, there remain striking differences across place and time in the likelihood of startups (for a given quality level—EQI) to realize their potential (REAI). Figure 4 presents the overall pattern of REAI from 1988–2012:

  • The cohorts of new firms with the greatest average growth potential did not end up being the most successful (in terms of realized growth outcomes). While the 1996 cohort of new firms turned out to be the most successful, the 1998 and 1999 cohorts exhibited the highest level of average growth potential. This may suggest that the “financial guillotine” 42, 43 unleashed after the dot-com crash may have significantly reduced the ability of startups to realize their potential.
  • REAI—the likelihood of startups to reach their potential—declined sharply in the late 1990s and did not recover through at least 2008 (Figure 4). During this time period (which preceded the Great Recession), the American ecosystem for entrepreneurship was not conducive to startup growth. For example, conditional on the same estimated potential, a 1996 startup was four times more likely to achieve a growth event in six years than was a startup founded in 2005.
  • These estimates highlight a potential improvement in the United States’ ability to catalyze startup growth between 2009 and 2011 in parallel with the increased availability of venture capital during that time. But 2009–2011 REAI preliminary estimates still remain markedly lower than REAI levels observed during the 1990s. Put another way, the U.S. entrepreneurial ecosystem is still dramatically less conducive for growth than it was in the dot-com era.
  • There is striking variation in entrepreneurial potential for growth (EQI) across regions and over time. Figure 5 shows a map of the United States where each point represents the EQI of the corresponding zip code. The size of the point reflects the quantity of entrepreneurship, while the color reflects the quality. The darker the shade of red, the greater the average probability of growth. The brighter the shade of white, the greater the number of new firms with lower potential/ambition for growth.44As illustrated in Figure 5, consistent with practitioner perceptions, there are extremely high and persistent levels of entrepreneurial quality in Silicon Valley and Route 128 (in Massachusetts), as well as a pronounced rise in startup activity in the two urban cores closest to both hotspots (San Francisco and Cambridge/Boston). At the same time, however, there are also regions where the “startup nation” has yet to take off, despite some of the highest levels of self-employment per capita in the nation, such as Miami, Florida.
Figure 4 (Source: Guzman and Stern, 2016)

Figure 5 (Source: Guzman and Stern, 2016)

These findings bring into sharp relief the importance of accounting for differences in quality when considering the state of American entrepreneurship. Not every newly founded company has the ambition and potential for significant growth, and those startups that do differ in important ways from those that do not. Thus, policies that aim to increase “shots on goal” and implicitly treat all firms as equally likely candidates for growth are likely to expect “too much” from the vast majority of new businesses, by focusing on a lever—new firm formation—that is only weakly related to economic growth.
While the overall decline in business dynamism observed in quantity-based measures does raise cause for concern,45 that concern, with respect to high-growth-potential startups, may be misplaced. U.S. RECPI does not register a long-term decline, but, instead, shows more cyclical boom and bust cycles with a general upward trend. There has been a sustained increase in U.S. RECPI starting in 2010.
Much more worrisome than the rate of creation of high-growth-potential firms is the decline in the United States’ ability to accelerate the growth of new businesses conditional on initial quality—the REAI—which has been falling since the late 1990s and only recently, and mildly, began to recover. Even as the number of new ideas and potential for innovation is increasing, there seems to be a reduction in startups’ ability to scale in a meaningful and systematic way. Whether this is primarily a challenge for capital markets or reflects systematic reductions in various aspects of ecosystem efficiency remains an important challenge for both future research and policy intervention.
Finally, the regional variation found in startup performance reflects very significant differences in both the underlying quality of ventures started there and the ability of different ecosystems within the United States to nurture startups in order to realize growth. Systematic and real-time metrics for the measurement of entrepreneurial quality and ecosystem performance can serve as powerful tools for policymakers and stakeholders seeking to accelerate and reinforce the impact of entrepreneurship on economic and social progress within their communities.

Time for a New Policy Conversation

The distribution and skewness of entrepreneurial quality empirically demonstrates the need to frame the policy conversation around American entrepreneurship from a different vantage point. Policymakers should account for quality when mapping the rate and trajectory of new firms founded and set objectives for enabling high-growth-potential IDEs that are different from (though coordinated with) their programs and objectives for SMEs.

Though more research is necessary to confirm and deepen these findings, the increasing rate of creation of high-growth-potential startups implies that policy dialogue may benefit from a heightened focus on improving the scaling capability of regional ecosystems. Given the striking variation in entrepreneurial potential for growth (EQI) across regions and over time, tailored analysis of each region’s innovation and entrepreneurial capacity is needed to find where gaps in a region’s ability to accelerate entrepreneurial potential may lie and to develop tailored strategies for policy intervention. Experimental approaches may be needed to collect evidence regarding the effectiveness of proposed interventions.

With respect to other types of firms, including small to medium-sized local businesses whose relative decline is accurately reflected in indices that focus on quantity, different solutions are required. Programs should specifically target the needs of this category of young firms, without expecting that they will necessarily fuel economy-wide growth.

Finally, the mix of support and programs offered should reflect the makeup of new businesses and high-potential-growth startups found in a given region and should be tailored to their specific needs. Where strong innovation-driven entrepreneurial ecosystems (e.g., Silicon Valley, Greater Boston/Kendall Square) may be producing enough startups, the question becomes what types of other businesses are needed. Where regions, like Miami, have high rates of new self-employment but register a low score for entrepreneurial quality, there may be a case for considering dedicated investments in building the foundations for a more robust innovation-driven entrepreneurial ecosystem that also leverages local comparative advantage.

Accurately diagnosing the challenges facing specific ecosystems is likely to be a challenge. As the former administrator of the U.S. Small Business Administration and current Senior Fellow at Harvard Business School, Karen Mills, noted: “there is no one-size-fits-all package to help small businesses, precisely because each of the different types of small businesses has different needs. The SME business owner needs a different kind of capital from the high-tech entrepreneur. For each city or region, the right mix of programs depends on what outcomes the leadership of that area is trying to achieve.”46

At MIT, we have had a chance to put elements of such a playbook into action through the Regional Entrepreneurial Acceleration Program, which works with stakeholder teams from around the world to not only undertake systematic analysis of their innovation-driven entrepreneurial ecosystems, but also to put their insights into action through the development and implementation of a regional entrepreneurship acceleration strategy.47


“To tackle our biggest societal challenges, we need an innovation pipeline that delivers every drop.”48 Quality-based measures of entrepreneurship enrich our understanding of the state of American entrepreneurship and better inform where policy and program interventions in support of startups should be focused. Changes in both entrepreneurial potential and ecosystem effects are important for understanding the state of American entrepreneurship. While the supply of new high-potential-growth startups appears to be growing, the ability of U.S. high-growth-potential startups to commercialize and scale seems to be facing continuing stagnation. Policy interventions to enhance the process of scale-up may be more impactful than those that simply aim to increase shots on goal. Ultimately, we may be able to do better for both fledgling small to medium-sized enterprises and innovation-driven enterprises that aspire to exponential growth by accounting directly for the differences between them. Allowing each to make their distinct contributions to U.S. performance requires a new conversation.

Author Acknowledgements
This paper builds on the MIT Innovation Initiative Laboratory for Innovation Science and Policy working paper “The State of American Entrepreneurship: New Estimates of the Quality and Quantity of Entrepreneurship for 15 U.S. States, 1988–2014,” by Jorge Guzman and Scott Stern and research presented at the 2015 Kauffman Foundation New Entrepreneurial Growth conference. We thank Erik Brynjolffson, John Haltiwanger, Joshua Gans, and Phil Budden for helpful suggestions. We are also grateful to Tetyana Pecherska for copy editing assistance. We acknowledge and thank the Jean Hammond (1986) and Michael Krasner (1974) Entrepreneurship Fund and the Edward B. Roberts (1957) Entrepreneurship Fund at MIT for financial support. All errors and omissions are, of course, our own.

About the Authors
Catherine Fazio is the Managing Director of the Laboratory for Innovation Science and Policy at the MIT Innovation Initiative. Catherine serves on the Development Committee for the Atrium School in Watertown, Massachusetts. She received a J.D. from Stanford University, an M.B.A. from the Sloan Fellows Program in Innovation and Global Leadership at MIT, and a B.A. from the University of California, Berkeley.

Jorge Guzman, Doctoral Student at MIT Sloan School of Management, Doctor of Philosophy (Ph.D.), Entrepreneurship and Strategic Management.

Professor Fiona Murray is the Associate Dean of Innovation at the MIT Sloan School of Management, Alvin J. Siteman (1948) Professor of Entrepreneurship and the Faculty Director of the Martin Trust Center for MIT Entrepreneurship.  She is the Co-Director of MIT’s Initiative for Innovation.  She is also an associate of the National Bureau of Economic Research. Murray holds an MA in chemistry from Merton College, University of Oxford, and an MS and PhD in engineering and applied sciences from Harvard University.

Scott Stern is the David Sarnoff Professor of Management and Chair of the Technological Innovation, Entrepreneurship, and Strategic Management Group at the MIT Sloan School of Management. Stern started his career at MIT, where he taught from 1995 to 2001. Stern is the director and co-founder of the Innovation Policy Working Group at the National Bureau of Economic Research. In 2005, he was awarded the Kauffman Prize Medal for Distinguished Research in Entrepreneurship. Stern holds a BA in economics from New York University and a PhD in economics from Stanford University.


  1. Litan, Robert E. “Inventive Billion Dollar Firms: A Faster Way to Grow.” SSRN Working Paper #1721608 (2010).
  2. As Ben Cohen of Ben & Jerry’s fondly recalls: “[W]e took a $5 correspondence course in ice-cream technology and started making ice cream in our kitchen … When we first started, it was just a lark. We never expected to have anything more than that one homemade ice-cream shop …” How We Met: Ben Cohen And Jerry Greenfield. Interviews by Ronna Greenstreet, INDEPENDENT, May 27, 1995.
  3. Aulet, William, and Fiona Murray. “A Tale of Two Entrepreneurs: Understanding Differences in the Types of Entrepreneurship in the Economy.” Available at SSRN 2259740 (2013).
  4. Hurst, Erik, and Benjamin Wild Pugsley. “What do small businesses do?” Brookings Papers on Economic Activity (2011): 73–143.
  5. Aulet, William, and Fiona Murray. “A Tale of Two Entrepreneurs: Understanding differences in the Types of Entrepreneurship in the Economy.” Available at SSRN 2259740 (2013).
  6. Decker, Ryan, John Haltiwanger, Ron Jarmin, and Javier Miranda. “The role of entrepreneurship in U.S. job creation and economic dynamism.” Journal of Economic Perspectives 28.3 (2014): 3–24.
  7. Decker, Ryan, John Haltiwanger, Ron Jarmin, and Javier Miranda. “Where has all the skewness gone? The decline in high-growth (young) firms in the U.S.”NBER Working Paper # 21776. National Bureau of Economic Research, (2015).
  8. Haltiwanger, John. “Top Ten Signs of Declining Business Dynamism and Entrepreneurship in the U.S.” Working Paper (2015). Web. Accessed on Feb. 8, 2016.
  9. Clifton, Jim, Chairman and CEO of Gallup. “American Entrepreneurship: Dead or Alive?” Business Journal, Jan. 13, 2015.
  10. Roberts, Edward B., Fiona Murray, and J. Daniel Kim. “Entrepreneurship and Innovation at MIT: Continuing Global Growth and Impact.” MIT Innovation Initiative. (2015).
  11. “San Francisco, Silicon Valley, and the strip of land that runs along the shore of the Bay between them have had a tremendous decade … Every year new ideas grow from specks to spectacular. Startups are so commonplace that in San Francisco’s Mission district you can buy greeting cards that say, ‘Congratulations on closing your first round.’ Uber, a six-year-old taxi-hailing company, is valued at $41 billion; Airbnb, a seven-year-old firm through which people turn their homes into hotels, is valued at $26 billion. … And at the same time, you hear the worry that the boom … cannot last …”
    “To fly, to fall, to fly again.” The Economist. Print Edition. July 24, 2015.
  12. Andreesen, Marc (pmarca). ‘“There’s too much entrepreneurship: Disruption running wild!” “There’s too little entrepreneurship: Economy stalling out!”’ Jan. 2, 2015. 9:11 PM. Tweet.
  13. Guzman, Jorge, and Scott Stern. “The State of American Entrepreneurship: New Estimates of the Quantity and Quality of Entrepreneurship for 15 U.S. States, 1988–2014.” MIT Innovation Initiative Laboratory for Innovation Science and Policy Working Paper (2016).
  14. Mills, Karen. “A Playbook for Making America More Entrepreneurial.” Harvard Business Review. May 27, 2015.
  15. Decker, Ryan, et al. (2014) Op. cit.
  16. Decker, Ryan, et al. (2015) Op. cit.
  17. Haltiwanger, John. “Top Ten Signs of Declining Business Dynamism and Entrepreneurship in the U.S.” Working Paper (2015). Web. Accessed on Feb. 8, 2016.
  18. Fairlie, Robert W., E. J. Reedy, Arnobio Morelix, and Joshua Russell. “Kauffman Index: Startup Activity. National Trends” (2015). Web. Accessed on Feb. 8, 2016.
  19. The Current Population Survey (CPS), sponsored jointly by the U.S. Census Bureau and the U.S. Bureau of Labor Statistics (BLS), is the primary source of labor force statistics for the population of the United States.
  20. See, e.g.,Kauffman Startup Indexes for 2013, 2014, and 2015.
  21. For more than two decades, an important line of research has highlighted the role of young, rather than small, firms in employment growth and economic performance in the United States (see [14–16]).
  22. Davis, Steven J., and John Haltiwanger. “Gross Job Creation, Gross Job Destruction, and Employment Reallocation.” Quarterly Journal of Economics (1992): 819–863.
  23. Davis, Steven J., John C. Haltiwanger, and Scott Schuh. “Job creation and destruction.” MIT Press Books (1998).
  24. Haltiwanger, John, Ron S. Jarmin, and Javier Miranda. “Who creates jobs? Small versus large versus young.” Review of Economics and Statistics 95.2 (2013): 347–361.
  25. Fairlie, Robert W., E. J. Reedy, Arnobio Morelix, and Joshua Russell. “Kauffman Index: Startup Activity. National Trends” (2015. Web. Accessed on Feb. 8, 2016. 
  26. See, e.g., Kauffman Startup Indexes for 2013, 2014, and 2015.
  27. The “2015 IPO Report” by WilmerHale.
  28. Annual figures drawn from The “PwC/NVCA MoneyTree™" Report. 
  29. Guzman, Jorge, and Scott Stern. “The State of American Entrepreneurship: New Estimates of the Quantity and Quality of Entrepreneurship for 15 U.S. States, 1988–2014.” MIT Innovation Initiative Laboratory for Innovation Science and Policy Working Paper (2016).
  30. This is true whether the new firm aims to grow in a linear fashion (as most small to medium-sized businesses do) or in an exponential fashion (as most startups do).
  31. Belenzon, Sharon, Aaron K. Chatterji, and Brendan Daley. “Eponymous entrepreneurs.” Working paper (2014).
  32. While the startup characteristics identified and used to estimate entrepreneurial quality have proven an informative starting point, alternative predictors are, of course, possible, and we consider this a fruitful area for research going forward.
  33. The paper separately tests whether the process by which startup characteristics map to growth outcomes remains stable over time to ensure that the underlying correlations relied upon remain valid across the full time frame of the sample.
  34. As discussed in full detail in Guzman and Stern (2016), the paper defines a significant growth event (an initial public offering or acquisition within six years) as the outcome of interest. The paper then draws from the full population of new businesses registered with a state as a corporation, partnership, or LLC, and builds a dataset of characteristics for each new firm observable at or near the time of founding (a set of measures referred to as “startup characteristics”). The regression model is based on a probit model of the relationship between the growth event and observable startup characteristics. Finally, the model’s predictive capability is tested using the standard ten-fold cross-validation approach, which compares the predicted values generated by the model to observed growth across ten different random holdout (testing) samples.
  35. It is important to note that this equity growth outcome is different from employment growth, which has been the focus of many studies, including those by Davis, Decker, Haltiwanger, Jarmin, and Miranda (see references 8–10, 19, 24–26). While we hypothesize that those startups that achieve equity growth outcomes under this approach are highly correlated with those that disproportionately contribute to employment growth, the precise relationship between these two groups remains an open research question.
  36. To test the model’s predictive capability and evaluate the skewness of entrepreneurial growth potential, the paper compares estimated probabilities of growth outcomes to realized growth outcomes in a ten-fold cross validation. Sixty-nine percent of realized growth events fall within the top 5 percent of the models’ estimated entrepreneurial quality distribution, and more than 50 percent of the realized growth outcomes fall in the top 1 percent. Thus, not only is the model highly predictive, but the distribution of realized growth outcomes (conditional on initial entrepreneurial quality) also is likewise highly skewed.
  37. The level of skewness of entrepreneurial quality is highly informative. It indicates how much more likely a startup at the high end of the entrepreneurial quality distribution is to grow than an average firm. If skewness were low, then adding several average firms could have as much regional impact as one high-growth-potential firm. But, if skewness is high (as the findings indicate), then a much larger number of firms with average growth potential is needed to generate the expected impact of one high-potential firm. Given the level of skewness observed, almost 4,000 local limited liability companies (average firm) are needed to generate the same potential as only one new Delaware corporation with an early patent and trademark. Put another way, initial ambition/potential for growth is a key dimension of heterogeneity across new firms. The subset of high-potential-growth startups is very small and fundamentally different than the vast majority of new firms.
  38. RECPI also offers an important comparison relative to simple measures of “success” (e.g., counting the number of high-value exits in a given year or cohort). As described above, success-oriented measures, though informative, necessarily conflate the founding of growth-oriented firms and the process by which the growth potential is realized. RECPI, on the other hand, provides a direct measure of the expected number of high-growth startups independent of other factors or conditions that may impact the realization of growth outcomes.
  39. Fairlie, et al. Op. cit.
  40. Haltiwanger, John, Ron Jarmin, and Javier Miranda. “Business Dynamics Statistics: An Overview.” The Kauffman Foundation. (2009).
  41. Singer, Slavica, José Ernesto Amorós, and D. Moska Arreola. “Global entrepreneurship monitor: 2014 global report.” Global Entrepreneurship Research Association (2015): 1–116.
  42. Nanda, Ramana, and Matthew Rhodes-Kropf. “Investment cycles and startup innovation.” Journal of Financial Economics 110.2 (2013): 403–418.
  43. Nanda, Ramana, and Matthew Rhodes-Kropf. “Financing risk and innovation.” Management Science (Forthcoming).
  44. In the states of New York and Washington, firm registrations do not contain zip code information. Accordingly, estimations of entrepreneurial quality can only be calculated at a state level. The average entrepreneurial quality of new firms is shown at a state level in a single shade of color.
  45. Decker, et al. Op. cit.
  46. Mills, Karen. “A Playbook for Making America More Entrepreneurial.” Harvard Business Review. May 27, 2015.
  47. For more information on the MIT Regional Acceleration Program, see
  48. Reif, Rafael. “A better way to deliver innovation to the world.” The Washington Post. Opinions. May 22, 2015.

Where is Innovation Falling Short?:

Using Labor Market Indicators to Map the Successful Innovation Frontier

By Michael Mandel Chief Economic Strategist Progressive Policy Institute in Washington
Michael Mandel
Michael Mandel


There is a great deal of debate these days about how innovative the “innovation economy” really is. On the pro-innovation side of the ledger is the apparently endless flow of new products and services enabled by the Internet and wireless communications, and the equally endless list of industries, jobs, and existing business models being disrupted or potentially disrupted by new technologies such as ecommerce, mobile apps, robotics, drones, and self-driving vehicles. In 2011, Marc Andreesson explained “Why Software Is Eating the World,” and so far that seems to be true.1

Yet, the economies of the United States and other developed countries do not appear to be performing as if we are in a period of successful innovation. Productivity growth has been slowing rather than accelerating, with the ten-year growth rate of multifactor productivity only 0.6 percent annually from 2004 to 2014, compared to a 1.5 percent annual rate in the 1994 to 2004 period. Private-sector nonresidential investment, rather than booming to take advantage of new innovative opportunities, stood at only a mediocre 12.9 percent of GDP in 2014, roughly about the average of the past 45 years. The long-term real interest rate, as reported by the Federal Reserve, was below 1 percent as of mid-July 2015.2 Furthermore, the decline in the rate of new business formation, as documented by Ian Hathaway and Robert Litan, is seemingly inconsistent with an innovation boom, since new technologies historically generate opportunities to start new businesses.

Finally, the decline in the real wages of young college and graduate school graduates is difficult to reconcile with an economy where innovation is proceeding rapidly across a wide range of areas. From 2000 to 2014, real earnings of full-time male workers, aged twenty-five to thirty-four, with a bachelor’s degree only, dropped by 16 percent. For women with bachelor’s degrees, the comparable figure was an 8 percent decline in real earnings. This decline is disturbing, because new college graduates are the very definition of mobile putty: they have no sunk costs in stagnant skills, sectors, or regions, and are theoretically able to gravitate to those leading-edge companies and vibrant regions enjoying the benefits of innovation.

Pessimists such as Robert Gordon look at this evidence and conclude that the current wave of innovation is relatively weak compared to past industrial revolutions.3 Optimists argue that it simply takes time for the benefits of innovation to diffuse within the economy. Indeed, a new study from the OECD argues that “global frontier” firms still are enjoying rapid productivity growth, and the real issue is the “breakdown of the diffusion machine.”4

The “innovation economy” debate is highly relevant for government policy and for the understanding of new business formation. If the optimists are right, we have a fundamentally fast-potential-growth economy that is being held back by a combination of weak demand and supply-side factors such as excess regulation and slow diffusion of new technologies. In that case, the appropriate demand-side policy measures might include redistribution to spur demand, along with increased government spending. Appropriate supply-side measures might include regulatory reform and government efforts to speed up the diffusion of new technologies to more companies. 

If the pessimists are right, then the United States and other developed countries are stuck with slow-potential-growth economies that can be tweaked but not substantially accelerated in the short or even medium run. In that case, developed economies will have great difficulty paying the benefits that their aging populations expect without creating great hardship for the younger generation, or without opening up their doors to a large number of immigrants.   

Slow-potential-growth economies have tough policy decisions to make in terms of sharing the pain. Will the aging generations have to absorb cuts to their long-promised medical and pension benefits in order to pay for the education, infrastructure and services that principally benefit younger generations? Will younger households have to pay a higher share of their incomes in taxes to support the older members of their community? Or will immigration policies have to be substantially loosened in order to increase the pool of young workers.

An Alternative Narrative

In this paper we will argue for an alternative narrative that includes elements of both the optimistic and pessimistic scenarios.  We call this alternative scenario “uneven innovation.” We argue in areas such as IT, robotics, and oil/gas exploration and production, the innovation frontier is expanding very rapidly. Meanwhile in other critical areas, such as biosciences and materials sciences, the innovation frontier has been expanding slower than expected up to now. 

To support this argument, we use a new approach based on labor market indicators—specifically, occupational employment from government surveys and online help-wanted ads from aggregators—to map out areas of successful innovation.  For example, the sharp rise in the demand for petroleum engineers in recent years is directly related to technological innovation in oil/gas exploration will show that in certain key areas—specifically, IT, robotics, and oil/gas exploration and production—labor market indicators show a rapid pace of successful innovation.

By comparison, labor market indicators show a relative lack of successful innovation in the biosciences and material sciences. These are areas where scientific progress is being made, but successful commercialization is happening at a much slower pace.

We conclude that today’s economy is best described as “unevenly innovative.” We briefly explore policy implications and directions for future research.

Using Labor Market Indicators to Map Successful Innovations

Note that past waves of innovation have been built on multiple major innovations across major categories. For example, the period 1920–1940 saw major advances in transportation (air travel), medicine (antibiotics), materials (plastics and artificial fibers), and information processing and communications (radio, television, computers). Other major innovations in energy (electricity) and transportation (automobiles) diffused to broader populations over the same period. 

Each of these major advances created a corresponding new set of occupations in the workforce to both develop the innovation and produce/deliver it. For example, the advent of commercial air travel brought about the creation of a new specialty of aeronautical engineers, as well as the birth of the commercial airlines and airplane production industries, with all their attendant subspecialties, such as pilots, flight attendants, and aircraft mechanics.5 Similarly, the implementation of electricity created the new specialty of electrical engineers, as well as power station operators, linemen, and the entire electric utility industry.6 More recently, the Internet Revolution has created a vast array of job titles, such as web developer and social media manager, which did not exist before.

Based on historical evidence and common sense, it seems inevitable and non-controversial that the successful introduction of a major innovation into the economy should create new occupations, which then expand rapidly as the innovation diffuses more broadly. Indeed, the creation of such new occupations might be part of the definition of a successful introduction of a major innovation.

So, the creation and growth of new occupations is a sign of a successful major innovation. Conversely, the lack of creation and growth of new occupations is a sign that the corresponding innovation is relatively less successful. (Note that we are not making the much stronger claim that innovation creates net positive job growth across the economy.)

To track the creation and growth of new occupations, we have two potential sources of data. First, in the United States, the Bureau of Labor Statistics produces monthly and annual figures on detailed occupational employment using the Current Population Survey.7 These surveys are well benchmarked and reliable, within the limits of the survey.

However, government surveys have important limitations when it comes to mapping successful innovation. First, there is a long lag before new occupations are added to the survey. For example, ‘web developer’ was only added in the 2010 Standard Occupational Classification, even though it was a common term at least ten years earlier.

Second, there are necessarily a limited number of occupational categories in government surveys. With too many categories, the surveys become both unwieldy and statistically unsound. So, the 2010 SOC has only one category for chemists, even though some types of chemists may be working on innovative new materials while others may be stuck in stagnant parts of the industry.

However, we have available another source of information: online help-wanted ads (or job postings, as they are sometimes called). Earlier work has shown how to use online help-wanted ads to track innovative jobs.8 Online want ads are collected and indexed in real time by job search engines such as Indeed.

As a data source, want ads have several positive and negatives. The positives:

  • Want ads (or job postings) have to contain information about needed skills and location.
  • The want ad databases are searchable by detailed Boolean searches.
  • Job search engines typically collect a very large share of available job postings.
  • Some job search engines cover more than one country.
  • Job search engines have an interest in providing good search results for jobseekers and employers. That means the company develops location algorithms, which are as accurate as possible, while eliminating stale and duplicated ads.
  • Some job search engines, including Indeed, produce some types of longitudinal data as well.

The minuses:

  • There is not a one-to-one relationship between want ads and actual job openings. The same want ad may correspond to multiple jobs. One job opening may correspond to multiple ads, even though the aggregators try to eliminate duplication.
  • There is no guarantee that the data will be consistent over time. Job posting behavior by firms may change, as may the algorithm used by the job search engine.

Note that counts of want ads, by themselves, are not a reliable short-term macroeconomic indicator. They can vary over time for reasons that have nothing to do with the underlying economic situation. In particular, the opening or closing of a job board can immediately change the counts without any connection to the actual labor market.

Despite these limitations, want ads provide a very useful way of determining longer-term trends in the creation and growth of innovative skills. In particular, we will use the ‘trend’ data from Indeed, which gives the percentage of want ads that contain the given search terms.

Government Survey Data

We will use government survey data to get a broad overview of several key areas of potential innovation. In particular, we will look at the growth of occupations connected with information technology, petroleum exploration and production, biosciences, and material sciences.

Figure 1: Occupational Employment Growth and Successful Innovation

*Through November 2015. Data: Current Population Survey

We start with oil/gas exploration and production innovation. Hydraulic fracturing, or fracking for short, combined with horizontal drilling, has dramatically increased the available reserves of oil and gas in the United States in just a few short years and helped sharply drive down the cost of oil and gas.9

The innovation in oil/gas exploration and production is reflected in the soaring employment for mining, geological, and petroleum engineers. Between 2006–2007 and 2014–2015, the number of mining, geological, and petroleum engineers rose by 93.4 percent.10 Today, taking advantage of the oil production opportunities offered by hydraulic fracturing requires skilled petroleum and mining engineers.

Similarly, the rapid pace of innovation in information technology—particularly Internet and mobile-related businesses—has led to a 29.9 percent rise in the number of people employed in computer and mathematical occupations between 2006–2007 and 2014–2015 (That figure includes computer hardware engineers).11

Now let’s consider what these labor market indicators tell us about the pace of successful innovation in the biosciences. Over the past decade, the public and private sectors together have spent roughly $1 trillion on research and development in the biosciences.12 The result has been a series of scientific breakthroughs, including the sequencing of the human genome.

However, there’s debate about the rate at which scientific advances in the biosciences are being transformed into marketable innovations. On the one hand, undeniable advances in gene sequencing have produced some gains in diagnostics and treatments approved by the FDA, but much fewer than expected.13 In particular no human gene therapy product has been approved for sale in the United States.14 Indeed, to deal with this problem, in 2012 the Obama Administration authorized a new National Institutes of Health division called the National Center for Advancing Translational Sciences, with the avowed goal to “reduce, remove, or bypass costly and time-consuming bottlenecks … to speed the delivery of new drugs, diagnostics, and medical devices to patients.”15

On the other hand, pharma and biotech companies have scored some huge innovative gains lately, including to a cure for Hepatitis C, the most common blood borne infection. At the same time, stock market investors in recent years have shown a great belief that the translational drought is over for biotech companies. The Nasdaq Biotech Index, for example, almost quadrupled over the past five years, going from 970 on January 1,2011 to 3525 on December 28, 2015.

To shed light on this debate, we look at the number of employed biological and medical scientists and biomedical engineers. This is an important indicator of successful innovation, because taking advantage of market opportunities created by advances in biosciences requires scientists, engineers, and technicians who are trained in the relevant techniques.

What we find is that employment of biological and medical scientists and biomedical engineers is down slightly between 2006–2007 and 2014-15. That lack of job growth is consistent with a more negative story about the slow pace at which scientific advances in biosciences are being transformed into marketable innovations, perhaps because of regulatory barriers.

Very importantly, we note that employment is likely a coincident indicator of innovation, not a leading indicator. Flat employment of biological and medical scientists and biomedical engineers does not predict, one way or another, whether successful innovation in the biosciences is likely to accelerate.

Finally, what does government employment data tell us about successful innovation in new materials? Historically, the invention and mass use of new materials such as steel, plastics, and artificial fibers have been important contributors to industrial revolutions. More recently, a series of Nobel Prizes were awarded for work in material sciences, and we hear about breakthrough materials such as carbon nanotubes, nanoparticles, and high-temperature superconductors on a regular basis.

However, a closer look at the marketplace suggests that these materials have not yet become commercially viable. For example, the Nobel Prize for high-temperature superconductors was awarded in 1987. However, companies making high-temperature superconductors still have low revenues.16

Similarly, Figure 1 also shows negative growth in the number of employed chemical and material engineers and scientists. This surprising result is corroborated by a report from the American Chemical Society, which found that out of students who graduated during the 2013 academic year with degrees in chemistry and related fields, “14.9 percent were looking for work as of October 2013.”17 That unemployment rate fell slightly to a still high12.4% in 2014, but the ACS noted that “the number of graduates who took on part-time or temporary work grew from 17.7% in 2013 to 18.8% in 2014.”18

Online Want-Ad Data

One criticism that can be levied against the analysis of the previous sector is that the occupational categories shown in Figure 1, based on government surveys, are clearly too broad to pinpoint innovative areas. For example, a fall in the overall number of chemists and material scientists could conceal a more complicated story, in which jobs in advanced materials were booming at the same time that jobs using legacy materials were collapsing.

To answer this critique, we can use online want-ad data as a microscope to study occupational growth in detail. Our data comes from Indeed, which is one of the world’s largest job search engines.19 To give an example of its coverage, Indeed lists almost 500,000 want ads that contain the word ‘software’ in the United States, subdivided by region, state, and locality.

For the purposes of this paper, what’s important is that Indeed also offers a ‘trend’ analysis, which gives the share of want ads containing a key word or key words over time. For example, Figure 2a shows the want-ad trend data for the search term ‘Android,’ which is the name of the Google mobile operating system. From the chart, we see that the share of want ads that contain the term ‘Android’ soars starting in 2009, peaks in 2012, and since then has remained relatively constant.

In this case, the initial surge of ‘Android’ want ads is an indicator of successful innovation, as the introduction of the iPhone and Android-driven smartphones generated the need for software developers to create Android apps. Similarly, Figure 2b shows a rising share of want-ads for the search term ‘robotics.’

Labor market indicators also can be used to assess successful innovation in energy production and distribution. Recall that ‘fracking’—short for ‘hydraulic fracturing’—is a key innovation in the production of oil and gas. This successful innovation is reflected in the labor market, as the share of want ads that contain the term ‘fracking’ soars between 2011 and 2014 (the decline subsequent to 2014 is the result of falling oil and gas prices). 

What about innovation in the biosciences? The previous section showed that employment of biological and medical scientists, and biomedical engineers has fallen in recent years. The question, though, is whether this aggregate pattern of innovation weakness holds for more detailed occupations.

The logical place to start is with want ads containing the word ‘gene,’ because the most important scientific innovation in biosciences over the past ten years is the sequencing of the human genome, followed by the dramatic drop in the cost of sequencing genes. Figure 3a shows the share of want ads containing the term ‘gene’ has been trending down. As of July 2015, roughly 2,000 want ads include the term ‘gene,’ compared to 500,000 that contain the term ‘software.’

Moreover, Figures 3b and 3c show recent downward trends for want ads containing the terms ‘biologist’ and ‘bioinformatics’ (where bioinformatics is the use of information technology to process biological data). This result appears to support the proposition that successful innovation in biosciences is currently relatively weak.

The results for innovation in the material sciences are more mixed. The share of want ads containing the term ‘composite materials’ has been falling, according to Figure 4a. Figure 4b shows that the share of want ads that include the term ‘nanotechnology’ are roughly flat, excepting a blip in 2010 and 2011. And the share of want ads containing the term ‘biomaterials’ has been rising. That covers advanced materials that interact with the body.

Implications and Further Research

This analysis is intended as indicative rather than conclusive. It suggests an uneven pattern of technological innovation, where tech and, now, energy production are leaping ahead, while other areas are stagnating in terms of delivering breakthrough innovations to the marketplace.

To the extent that innovation drives job growth and new business formation, the uneven pattern of innovation may help reconcile divergent views of the current economic situation. Rapid innovation in tech and communications may be creating jobs in those fields, with ambiguous effects on job growth in other industries. Conversely, the lack of commercially important technological breakthroughs in critical areas such as material sciences and biosciences is hampering job growth and business creation in related areas. 

Moreover, the uneven innovation may be holding back productivity and wage growth on the macro level. Past industrial revolutions have included innovations across a wide range of disciplines, including energy, transportation, materials, medicine, and information technology and communication. The relatively limited scope of innovation today may help explain the weak economic performance of the developed economies. 

Moreover, the uneven nature of innovation may help explain some recent trends in income inequality. Advances in information technology have tended so far tended to boost job prospects for workers with a college education. But advances in other areas—say, a new material that spurs construction—could create more inclusive opportunities for a broader set of workers.

From the perspective of policy, there are two separate issues. First, we need to systematically map out the innovation laggards and leaders in the economy, using a search terms across a wide range of disciplines. Such a mapping of the innovative frontier would be helpful for students, entrepreneurs, and policymakers. College students would be able to use the mapping of the innovation frontier to assess which majors or fields of study would be most productive in terms of getting jobs. Entrepreneurs would be able to identify areas of potential opportunity.

And perhaps most importantly, policymakers would have an objective measure for assessing the true extent of the uneven innovation. If they can see where the innovation breakdowns are occurring, they can focus resources on getting roadblocks out of the way. In some cases, as in the case of biosciences, this may require a close look at the relationship between regulation and innovation. In other areas, such as advanced materials, there may be insufficient public sector funding.

The goal in the end is to broaden successful innovation beyond just the information technology and communication sector, and thus boost entrepreneurial vigor, productivity and economic growth across the entire economy.  The best answer for a stagnant economy is more inclusive innovation. 

Figure 2: Selected Labor Market Indicators for Successful Innovation in Tech and Energy Production (share of matching job postings)

Figure 2a: Want-ad trend data for ‘Android’

Figure 2b: Want-ad trend data for ‘robotics’

Figure 2c: Want-ad trend data for ‘fracking’

Data: Indeed (chart downloaded from website June 2015)

Figure 3: Selected Labor Market Indicators for Successful Innovation in Biosciences (share of matching job postings)

Figure 3a: Want-ad trend data for ‘gene’

Figure 3b: Want-ad trend data for ‘biologist’

Figure 3c: Want-ad trend data for ‘bioinformatics’

Data: Indeed (chart downloaded from website June 2015)

Figure 4: Selected Labor Market Indicators for Successful Innovation in Material Science (share of matching job postings)

Figure 4a: Want-ad trend data for ‘composite materials’

Figure 4b: Want-ad trend data for ‘nanotechnology’

Figure 4d: Want-ad trend data for ‘biomaterials’

Data: Indeed (chart downloaded from website June 2015)

About the Author
Dr. Michael Mandel is chief economic strategist at the Progressive Policy Institute in Washington. He was co-principal investigator for a Sloan Foundation grant and testified before Congress on impact of regulation on innovation. Mandel also holds an appointment as senior fellow at Wharton’s Mack Institute for Innovation Management at the University of Pennsylvania, and serves as president and founder of South Mountain Economics LLC. Mandel received a Ph.D. in economics from Harvard University and formerly served as chief economist at BusinessWeek, where he directed the magazine’s coverage of the domestic and global economies.


  1. Marc Andreessen, “Why Software Is Eating the World,” Wall Street Journal, August 20, 2011.
  3. Robert Gordon, “The Demise of U. S. Economic Growth: Restatement, Rebuttal, and Reflections,” Northwestern University, January 2014.
  4. OECD, The Future of Productivity, OECD, Paris, 2015.
  5. MIT offered the first U.S. course in aeronautical engineering in 1914 The occupational category of airplane pilots and navigators was first reported in the 1920 Census.
  6. The American Institute of Electrical Engineers was founded in 1884. The occupational category of electrical engineers was first reported in the 1910 Census. At that time there were 15,000 electrical engineers in the United States. That number rose to 110,000 by 1950. The occupational category of power station operators was also first reported in the 1910 Census.
  7. The BLS also products occupational employment figures through the Occupational Employment Survey, but that survey is not designed to yield longitudinal estimates.
  8. See Michael Mandel, “Where the Jobs Are: The App Economy,” South Mountain Economics, February 2012.
    Michael Mandel and Judith Scherer, “A Low-Cost and Flexible Approach For Tracking Jobs and Economic Activity Related To Innovative Technologies,” South Mountain Economics, Nesta Working Paper No. 15/11, June 2015.
    National Research Council, 2014 (edited by Robert E. Litan, Andrew W. Wyckoff, and Kaye Husbands Fealing). Capturing Change in Science, Technology, and Innovation: Improving Indicators to Inform Policy. National Academies Press, 2014. Available:
  9. Energy Information Administration, “U.S. Crude Oil and Natural Gas Proved Reserves,” December 2014..
  10. All data for 2015 through November of that year.
  11. See also Michael Mandel and Diana Carew, April 2015. “Tech Opportunity for Minorities and Women: A Good News, Bad News Story,” Progressive Policy Institute.
  12. Calculations by author.
  13. Michael Mandel, “Hacking the Regulatory State”.
  16. One market research firm pegged the entire global market for superconducting materials—both high and low temperature—at only $427 million in 2013
  19. We thank Indeed for use of its job-posting data. Indeed bears no responsibility for the analysis or conclusions of this paper.

Entrepreneurship and the Challenge of Globalization

By Kyle Handley Assistant Professor of Business Economics and Public Policy University of Michigan, Ross School of Business
Kyle Handley
Kyle Handley


The range and volume of customers and markets has grown dramatically in the past two decades. This has created new opportunities for entrepreneurial growth and new challenges. As new global markets are integrated by either policy changes or technological advancement, the reach and potential for entrepreneurial growth has expanded. At the same time, however, greater global integration increases the competitive pressures from overseas business and entrepreneurs and exposure to new economic and policy risks. This paper discusses three of the challenges and opportunities for entrepreneurial growth in this environment.

Challenge 1: Task offshoring and the global unbundling of production

The impact of offshoring and trade has become a huge economic and political topic in the United States. With the continuing surge in both globalization and inequality, it will remain center stage politically and economically. The reason is that increasing global trade in both goods and services is playing an important role in labor-market polarization and the decline of manufacturing (Autor, Dorn, and Hanson 2013; Charles, Hurst, and Notowidigdo 2013; Pierce and Schott 2013).

Our understanding and measurement of the extent of offshoring and its impact is limited. For example, if GM closes down an auto parts factory in Michigan, does that mean we know those jobs were offshored? Maybe GM just decided to stop producing a particular auto part, maybe it decided to outsource production to another U.S. producer, or maybe it offshored the factory to China. Likewise, when boutique upstart Shinola opens a watch assembly factory in Detroit, does it matter that the watch parts are imported from Switzerland? Should we value the retail and distribution jobs at Shinola more or less than the manufacturing jobs it replaces?

Two factors are at work behind the aggregate trends. First, the cost of automating routine tasks has fallen with the adoption of information technology, shifting low-skilled workers toward the bottom of the wage distribution (Autor and Dorn 2013). Even if many jobs ultimately are not moved overseas, competitive pressures can have direct effects on wages for tasks and routines that are more easily offshored (Blinder and Krueger 2013). Second, increased import competition is both a cause and consequence of increased offshoring. Greater low-wage import competition, especially from China, reduces wages and employment (cf. Autor, Dorn, and Hanson 2013; Pierce and Schott 2013). It also changes the nature of the employer-employee relationship and wage bargaining (Bertrand 2004).

This presents a challenge for entrepreneurial growth, because new jobs are not likely to replace traditional manufacturing jobs that have been lost. Entrepreneurs can take advantage of new technologies to completely unbundle the stages of production, from designing, to building, to distributing a new product. A new idea can be designed by an engineer in India, with parts manufactured in Thailand or Japan and assembled in China, and the product finally distributed and sold by wholesale and retail operations in the United States. But despite the breadth of new opportunities, much of the investment and job creation will occur overseas in firms that can produce the inputs and final goods at lower cost, higher quality, or both.

Many of these new enterprises are “factoryless” goods producers and are not even classified in the manufacturing sector. Bernard and Fort (2013) assess the size and scope of factoryless goods producers in the U.S. economy in 2002 and 2007. They note prominent examples such as Apple, Inc., which owns no production facilities in the United States, and the vacuum and appliance producer Dyson, which conducts management and R and D in England but outsources most production to Malaysia. They estimate that as many as 431,000 to 1.9 million workers could be counted as manufacturing employment if factoryless goods producers were re-classified from the wholesale to manufacturing sector. This is large and likely to grow as continued advances in technology further fragment the tasks within production, design, and distribution. The implications for wages, employment, productivity, and inequality are not well understood. But it seems likely that some share of future new entrepreneurial and innovative activity will take this form.

Whether it is a new product or service from a successful startup or from within established firms, the new growth from entrepreneurial activity in a more integrated world may be unevenly distributed. It is well known that the gains from international trade are unequally distributed, but recent work by Ma (2014) connects this to another well-known fact: CEO-to-worker pay ratios are large and have been growing in recent decades. Increased global integration may increase compensation for the founders and managers of globally engaged firms much more than for individual workers. By matching executive compensation to firm-level U.S. Census data, Ma (2014) finds that greater global market access increases wage inequality within the firm and increases the CEO-to-worker pay ratio in firms that engage in exporting or foreign direct investment. This provides a link between globalization and inequality through the channel of within-firm inequality. Heterogeneity and diversity in the human capital of CEO and founders translates into higher CEO-to-worker pay ratios (i.e., higher inequality). As a result, further global integration, via policy or technological advancement, that increases the payoff to entrepreneurial growth may end up amplifying inequality.

Challenge 2: Import competition and export market opportunities

The study of international trade has changed tremendously in the past decade to focus on firms that are engaged in trade or affected by trade. In part, this was driven by new firm-level datasets linking import and export transactions directly to firm-level measures of wages, employment, and productivity. At the firm level for the United States, there are approximately 47 million import transactions and 24 million export transactions annually, each of which are tracked at the border and assigned to one of more than 8,000 product categories. Firms that import, export, or both are typically larger and more productive, and pay higher wages than domestic firms do (cf. Bernard, Redding, and Schott 2007). Changes to trade policy and technology affect which firms engage in trade, where they import and export, and the winners and losers. With that in mind, it’s important to remember that firms are competing for market share both domestically and internationally. Nevertheless, it’s not clear cut that imports always compete with domestic firms or that exports are the path to growth.

First, consider some very aggregate data on the import and employment levels of the largest multinational firms. Multinationals employed 28.6 million U.S. workers in 2011,1 accounting for 25 percent of total U.S. wage income. But multinationals and trade have an impact far beyond their direct employees—for example, Blinder (2009) estimates about 25 percent of all U.S. jobs could be offshored, particularly the lower-paid and lower-skilled jobs. As one illustration, Figure 1 plots the cumulative employment changes at the domestic and foreign operations of U.S. multinationals.2 During most of the 2000s, overseas employment at majority-owned affiliates of U.S. multinationals was growing, while domestic employment at multinationals was flat or falling rapidly. Contemporaneously, annual imports from foreign-owned affiliates were increasing. This is strongly suggestive of the type of employment offshoring that is alarming to policymakers.

But, does the increase in imports really substitute for labor that would have been employed to produce the inputs within the firm? Returning to the GM factory example, this need not be the case. First, the imported intermediates may have been imported at arm’s length by subcontracting from a foreign supplier. While this may substitute for domestic employment, the importing firm is not offshoring or physically relocating employment within the firm. Alternatively, the Shinola watch assembly jobs in Detroit might not even exist if precision parts and casings were not easily imported from Switzerland. Second, firms might have been outsourcing production domestically to begin with and simply switched to foreign suppliers. Third, firms might increase intermediate imports in response to supply and demand shocks—for example, to meet excess demand for auto parts after a regulatory shock (like changes in emissions standards) or productivity shocks (like advances in injection molding technology). Moreover, this flexible ebb and flow of imported inputs might be essential to managing domestic overhead labor costs that would result in costly shutdown or layoffs if all stages of production were performed domestically.

On the other hand, one U.S. entrepreneur’s foreign-sourced intermediate input could be another firm’s final product. In general, greater global integration will expose firms and potential entrepreneurs to more competition in their own final goods markets, either domestic or overseas. Recall that exporters and importers are larger and more productive, and pay higher wages. This is the result of competitive reallocation toward more productive firms, whereby the least-productive firms remain small and either don’t participate in international trade or may shutdown entirely.

Entrepreneurism plays an important role in trade-induced reallocation. Products and processes typically have lifecycles and must be replaced by new and innovative products and ideas. A larger potential market can increase innovation, even if many ideas are not successful. This is one reason why new goods make up a large share of trade growth following trade liberalizations (Kehoe and Ruhl 2013). While many firms start small when entering new export markets, those that survive expand employment and sales rapidly (cf. Easton, Eslava, Kugler, and Tybout 2007; Rauch and Watson 2003). It’s important for policymakers to provide these opportunities for new entrepreneurial growth by promoting greater openness and trade. But policymakers also must bear in mind that the challenge of international competition could hurt one industry or product market while contemporaneously increasing employment and productivity in another.

Challenge 3: Barriers to global entrepreneurial growth—trade costs, economic, and policy risks

This final section is concerned primarily with new sources of risk and complexity that face entrepreneurs in a more integrated world. Many of the risks are large and, for new entrants and startups, some risks are non-diversifiable.

I’ll cover three areas I think are important: trade costs and credit, economic risks, and policy uncertainty.

Trade Costs and Trade Credit

The cost of international trade can be very high. In addition to simple freight costs, a variety of other costs differ with the value or frequency of shipments. Firms must be bonded, they need to insure their shipments, they have to provide customs documentation that varies by country of origin or destination, and, in many cases, they have to pay tariffs or other regulatory fees at the border. Moreover, most of these transactions are facilitated and secured by trade credit. Credit constraints can be particularly severe for small entrepreneurs, whereas a large corporation like Boeing can finance its own transactions. As a case in point, credit constraints were an important factor in the dramatic international trade collapse during the Great Recession (Chor and Manova 2012). This suggests that government-sponsored organs such as the U.S. Export-Import bank perform an important function, especially for small firms. In light of this recent experience, the failure of Congress to renew the Export-Import bank charter has unknown potential to disrupt U.S. export activity.

In addition to these variable costs, entering new markets requires large, upfront, and irreversible sunk costs, which can be very high and prohibitive. They cover, for example, the costs of tailoring products to customer tastes abroad, complying with safety and regulatory measures, and establishing distribution and retail outlets abroad. In structural estimates for Colombia and the United States, they often exceed $1 million and are country specific (Das, Roberts, and Tybout 2007; McCallum 2013). Because these costs are high and irreversible, they clearly will lead to a first order reduction in trade participation by both small entrepreneurs and large firms. But choosing to export or import is not a one-shot game. Entrepreneurs also can choose when to invest. When large sunk costs are combined with uncertainty over exchange rates, demand, trade policy, etc., entrepreneurs will delay investment and be less responsive even to good news about demand. I’ll now discuss two sources of uncertainty that potentially could reduce the dynamism of U.S. trade participation.

Economic Risk

For all the benefits of participating in international trade mentioned above, it also comes with additional risk. The entrepreneur is exposed to demand and supply shocks in foreign, as well as domestic, markets. This can occur through economic channels like the global financial crisis or exchange rate fluctuations. But risk (and opportunity) also can be transmitted through natural disaster shocks, such as the eruption of a volcano in Iceland or a tsunami in Japan. In work by Boehm, Flaaen, and Nayar (2015) linking U.S. firms to their affiliates in Japan, they find the Tohuku earthquake, by shutting down a portion of the supply chain, nearly shut down production and sales of affected U.S. plants. Even firms that don’t trade directly may find international events disrupt business by the global activities of upstream and downstream partners.

Policy Barriers and Uncertainty

Policy barriers and uncertainty over those barriers can play a major role in firm-level decisions to enter and invest in new export markets. Regulatory and tax barriers frequently are cited obstacles to startups and new investment. Policies such as investment tax credits for solar or wind energy intended to promote activity can have little effect, or even backfire, if they are subject to uncertain renewal, frequently adjusted, or otherwise not permanent.

Policy uncertainty over trade barriers, namely tariffs, is at least as important as the level of trade barriers in reducing export entry and investment. I’ve quantified this by linking a firm’s investment decision to trade policy uncertainty measures constructed from observable data. I show how institutions like the World Trade Organization (WTO) reduce uncertainty for new exporters in Handley (2014) and how free trade areas like the European Union do the same in Handley and Limão (2014). These effects can be very large. In Handley and Limão (2013), we show the reduction of Chinese exports to the United States from policy uncertainty prior to China’s WTO accession was equivalent to nearly doubling the applied tariff faced by the average firm.

About the Author
Assistant Professor of Business Economics and Public Policy at the University of Michigan Ross School of Business. Ph.D. Economics, University of Maryland, College Park, 2011, M.Sc. Economics, London School of Economics, 2006, B.S. Economics and Mathematics (with distinction), University of Wisconsin, Madison, 2000.


  1. Bureau of Economic Analysis. Summary Estimates for Multinational Companies: Employment, Sales, and Capital Expenditures for 2011, April 18, 2013. (accessed May 19, 2014).
  2. I end this figure in 2008 because there is a break in the BEA data series starting in 2009.


Autor, David H., and David Dorn. 2013. “The Growth of Low-Skill Service Jobs and the Polarization of the U.S. Labor Market.” American Economic Review, vol. 103(5), pp. 1553–1597, August.

Autor, David H., David Dorn, and Gordon H. Hanson. 2013. “The China Syndrome: Local Labor Market Effects of Import Competition in the United States.” American Economic Review, vol. 103(6), pp. 2121–2168, October.

Bernard, Andrew B., and Teresa C. Fort. 2013. “Factoryless Goods Producers in the U.S.” NBER Working Paper 19396, National Bureau of Economic Research, Inc.

Bernard, Andrew B., J. Bradford Jensen, Stephen J. Redding, and Peter K. Schott. 2007.

“Firms in International Trade.” Journal of Economic Perspectives, American Economic Association, vol. 21(3), pp. 105–130, Summer.

Bertrand, Marianne. 2004. “From the Invisible Handshake to the Invisible Hand? How Import Competition Changes the Employment Relationship.” Journal of Labor Economics, vol. 22(4), pp. 723–766, October.

Blinder, Alan S. 2009. “How Many US Jobs Might be Offshorable?” World Economics, vol. 10(2), pp. 41–78, April.

Blinder, Alan S., and Alan B. Krueger. 2013. “Alternative Measures of Offshorability: A Survey Approach.” Journal of Labor Economics, vol. 31(S1), pp. S97–S128.

Bloom, Nicholas, Mirko Draca, and John Van Reenen. 2011. “Trade Induced Technical Change? The Impact of Chinese Imports on Innovation, IT, and Productivity.” NBER Working Paper 16717, National Bureau of Economic Research, Inc.

Boehm, Christoph, Aaron Flaaen, and Nita Nayar. 2015. “Input Linkages and the transmission of shocks. Firm-level Evidence from the 2011 Tohoku Earthquake,” Michigan mimeo.

Charles, Kerwin, Erik Hurst, and Matthew Notowidigdo. 2013. “Manufacturing decline, housing booms and non-employment,” Chicago mimeo.

Chor, Davin, and Kalina Manova. 2012. “Off the cliff and back? Credit conditions and international trade during the global financial crisis.” Journal of International Economics, Elsevier, vol. 87(1), pp. 117–133.

Das, Sanghamitra, Mark J. Roberts, and James R. Tybout. 2007. “Market Entry Costs, Producer Heterogeneity, and Export Dynamics.” Econometrica, vol. 75(3), pp. 837–873.

Eaton, Jonathan, Marcela Eslava, Maurice Kugler, and James Tybout. 2007. “Export Dynamics in Colombia: Firm-Level Evidence.” NBER Working Paper 13531, National Bureau of Economic Research, Inc.

Handley, Kyle. 2014. “Exporting under trade policy uncertainty: Theory and evidence.” Journal of International Economics, Elsevier, vol. 94(1), pp. 50–66.

Handley, Kyle, and Nuno Limão. 2015. “Trade and Investment under Policy Uncertainty: Theory and Firm Evidence.” American Economic Journal: Economic Policy, vol. 7(4).

Handley, Kyle, and Nuno Limão. 2013. “Policy Uncertainty, Trade and Welfare: Theory and Evidence for China and the U.S.” NBER Working Paper 19376, National Bureau of Economic Research, Inc.

Ma, Lin. 2014. “Globalization and Top Income Shares.” Working Paper 14-07, Center for Economic Studies, U.S. Census Bureau.

McCallum, Andrew H. 2013. “The Structure of Export Entry Costs.” Federal Reserve Board, mimeo.

Rauch, James E., and Joel Watson. 2003.”Starting small in an unfamiliar environment,” International Journal of Industrial Organization. Elsevier, vol. 21(7), pp. 1021–1042, September.

Entrepreneurship and Intrapreneurship:

The Role of Human Capital and Complementary Assets

By Rajshree Agarwal Rudolph P. Lamone Chair and Professor in Entrepreneurship University of Marylandand
Benjamin Campbell Associate Professor of Management and Human Resources Max M. Fisher College of Business at the Ohio State University
Rajshree Agarwal
Rajshree Agarwal
Benjamin Campbell
Benjamin Campbell

“The ultimate resource is people—skilled, spirited, and hopeful people—who will exert their wills and imaginations for their own benefit as well as … social concern.” — (Simon, 1998: p xxxviii)
Control over nonhuman assets leads to control over human assets— Hart, 1995: 58

Entrepreneurship and innovation are the primary drivers of value creation. They shape technological change, the creation of new firms and industries, and the process of creative construction for the economic progress of societies. While Schumpeter (1911) referred to entrepreneurship and innovation in a synonymous manner, there is value in examining the causes and consequences of their divergence, particularly as it relates to human capital dynamics in new and established firms.

Specifically, despite positive trends in the overall economy, there are increasing concerns about the U.S. economy’s reduced dynamism as measured by a reduction in entrepreneurship. The recently documented statistics of decreased startup rates and a reduced number of people working at new ventures may well be a leading indicator of a decrease in innovation, inasmuch as entrepreneurship and innovation are positively correlated.

However, an open question is whether entrepreneurship is being substituted by intrapreneurship. This is an important issue because, if established firms are increasingly more able to retain and facilitate the innovative potential of their human capital, it would result in reduced entrepreneurship without a concomitant loss in overall innovation and economic dynamism.

In this paper, we propose that the appropriate level of analysis for measuring dynamism and addressing the open question above is not the firm level alone, but the union of individual and firm levels of analysis. This is because firms provide the necessary context within which innovation occurs, but innovation and entrepreneurship are fundamentally related to individual action and agency. Specifically, innovative individuals seek complementary assets that are specialized to increase the value of their ideas, while existing firms facilitate the contractual relationships between innovators and the owners of other assets jointly required for value creation (Alchian and Demsetz, 1972). When existing firms do not provide the optimal configuration of complementary assets, innovators can venture out on their own, recreating and transferring the necessary complementary assets in a new venture (Campbell et al., 2012). While dynamism and growth have been linked to the latter phenomenon, its relationship to intrapreneurship of individuals within established firms has not been a subject of systematic inquiry. Our main thesis, therefore, is that deeper attention to human capital dynamics—with a primacy on tracing career histories of innovative personnel within and across firm boundaries—may provide a richer measure of economic dynamism than a singular focus on new-venture creation and jobs associated with startups.

In the rest of the paper, we will first, briefly describe our conceptual framework, defining key concepts and their relationship with each other. Then, we turn our attention to factors that impact the cost of accessing complementary assets within new versus established firms, and factors that impact the importance of complementary assets for value creation. Changes over time in the underlying factors have implications for whether individuals harness their innovative ideas through intrapreneurship or entrepreneurship. In the last section, we identify potential implications at the individual- and economy-wide levels, setting out a potential research agenda for developing more holistic and integrative measures of economic dynamism than the traditional focus on new firm creation alone.

Entrepreneurship, Intrapreneurship and Complementary Assets

Focusing on individuals who generate and seek to pursue their innovative ideas, we define “entrepreneurship” as the pursuit of a new opportunity in the context of a new venture, and “intrapreneurship” as the pursuit of a new opportunity within an existing firm.

Related to entrepreneurship, prior research has documented three critical knowledge contexts that shape new venture formation (see review in Agarwal and Shah, 2014). First, employee-created spinouts are new ventures of a firm in the same industry as their “parent.” Gordon Moore and Bill Noyce, for instance, created Intel when they left Fairchild Semiconductor. Second, academic spinoffs are new ventures created by individuals who gained their knowledge in a university context. Google, for example, was founded by Sergey Brin and Larry Page based on ideas they generated while at Stanford University. Third, user entrepreneurship occurs when individuals tinker and find solutions to satisfy their unmet needs, and then found firms based on this knowledge. Apple is an iconic example. The involvement of Steve Jobs and Steve Wozniack in the Amateur Computer Users Group (also called the Homebrew Computer Club) shaped their creation of the personal computer.

Intrapreneurship, a term coined to represent “intra-corporate entrepreneurship,” focuses on individuals who capitalize on their visions and ideas to introduce new products and services within existing organizations (Pinchot, 1987). Intrapreneurs, also referred to as “product champions,” have been associated with higher firm performance because they enhance overall new product development, including product line extensions, and both incremental and radical innovations (Markham and Griffin, 1998). As an example, Apple’s turnaround strategy is deeply intertwined with its introduction of the iPod and the associated business model. What is perhaps less known is that Tony Fadell approached Apple with the idea, first starting as an independent contractor and then becoming an Apple employee who championed the product from development to successful launch (Coff, 2010).

When innovators have novel business ideas they wish to pursue, they face the choice of whether to become entrepreneurs or intrapreneurs. We explore this decision and its implications through the lens of complementary assets. Building on Teece (1986) and Campbell et al. (2012), we highlight that the importance and transferability of complementary assets shapes innovators’ decisions to innovate inside or outside of the boundaries of their employers. Based on Campbell et al. (2012), complementary assets consist of “organizational knowledge (e.g., codified routines, knowledge embodied in products and processes, and intellectual property rights), nonhuman complementary assets (e.g., physical capital, contractual relationships with buyers/suppliers, brand equity, and reputation), and human complementary assets (e.g., tacit knowledge embodied in other employees)” (p. 67).

In Figure 1, we depict our conceptual framework, which ties intrapreneurship and entrepreneurship to the importance of complementary assets, and to the relative costs of assembling them in established firms versus in new ventures. We note that the depiction focuses only on innovations that are sufficiently valuable relative to the costs of justifying their pursuit.

Holding the intrinsic value of the innovative idea constant, Figure 1 focuses on the role of complementary assets in enhancing value creation:

  • The horizontal axis represents the importance of complementary assets to the focal innovator’s human capital for value creation.
  • The vertical axis represents the innovator’s cost of building these complementary assets in an established firm, relative to the cost of creating them in a new venture.

The decision to pursue entrepreneurship or intrapreneurship is then a function of whether complementary assets are important for value creation, and of the relative costs of building these assets inside an existing organization or at a new venture. Accordingly, we differentiate three areas in Figure 1.

If complementary assets are important to value creation, and if the cost of reproducing them inside an incumbent organization is low relative to a new venture, intrapreneurship is the likely organizational form. On the other hand, even when an established firm possesses the necessary complementary assets, an innovator who can easily recreate or transfer those assets to a new venture context will pursue entrepreneurship. The middle area represents a situation in which both organizational forms are likely. In that case, established and entrepreneurial firms compete in the same space, and innovators may choose either option.

As suggested by the opening quotes, the framework gives primacy to innovators and their human capital, which includes not only their knowledge and skills, but also their “will and imaginations” (Simon, 1998). However, the importance, availability, and transferability of complementary assets determine corporate “control” (Hart, 1995).

We hasten to add that we do not intend to imply that the innovator necessarily relinquishes control. Rather, the framework is meant to distinguish between the managerial role of innovators as founders of new ventures, or as product champions in existing establishments.

Importantly, doing so also implies a separation of the innovation stage of the process (identification of an idea that can be commercialized) from the intra/entrepreneurial stage of the process (assembling the complementary assets necessary to profit from the idea). If the necessary complementary assets are more easily available within an existing organization than at a new venture, then the innovative activity will be pursued in the context of intrapreneurship. If the necessary complementary assets are more easily constructed within a new venture than in an existing organization, then the innovative activity will be pursued in the context of entrepreneurship. In other words, it is the ability to build and control complementary assets that determines where innovators will commercialize their ideas.

To examine the relationship between complementary assets and the entrepreneurship-versus-intrapreneurship decision, we first describe several factors that shape the importance of complementary assets to an innovative idea. We then describe several factors that shape the relative costs of accessing complementary assets in an existing organization versus in a new venture. After establishing the sources that determine each factor, we explore the consequences of a variety of potential trends that could change each factor.

Factors that Impact the Importance of Complementary Assets

Focusing first on the horizontal axis of Figure 1, Teece (1986) identifies several key factors that determine the importance of complementary assets to the profitability of an innovation, including the:

  • Maturity of the industry.
  • Size of rivals in the space the firm is entering.
  • Dominant appropriability regime of the industry.
  • Complexity of the innovation.

As industries mature, the minimum efficient scale of production and the importance of industry-specific knowledge and specialized complementary assets tend to increase (Gort and Klepper, 1982; Teece, 1986). Size is important because industries differ in their minimum efficient scale of production, and larger capital outlays may be more important for some industries than others. Size also correlates with a broader portfolio of complementary assets, and with an associated increase in economies of scope. Thus, industry structure dictates whether startups, which typically are small, will be able to compete successfully with larger, established rivals.

Second, if legal protection of an industry’s intellectual property is weak, then innovators must construct other barriers to imitation. Typically, the construction of barriers to entry requires increased investments in complementary assets, such as brand recognition, manufacturing, and distribution. For example, a firm in a weak appropriability regime has a greater imperative to separate itself from potential imitators through product differentiation, increasing the importance of marketing capabilities compared with a firm in a regime that affords strong protection for the innovation.

Third, the complexity of an innovation increases the importance of complementary assets. A more complex innovation, by definition, requires interaction with more assets and knowledge. In the words of Teece (1986): “To produce a personal computer, for instance, a company needs access to expertise in semiconductor technology, display technology, disk-drive technology, networking technology, keyboard technology, and several others” (p. 293).

Factors that Impact Costs of Accessing Complementary Assets in Established Firms Versus New Ventures

Figure 1. Complementary assets drive the intrapreneurship/entrepreneurship decision 

[Image Caption Text]

Turning our attention to the vertical axis of Figure 1, we use two categories to group factors that impact the relative cost of accessing complementary assets in established versus new ventures. The first relates to the infrastructural facilitators and impediments of nonhuman complementary assets, and the second to those that impact human complementary assets.

The differential costs of accessing nonhuman assets in an established firm versus a new venture are shaped by the financial constraints of the innovator and by the functionality of the underlying resource markets. If an innovator is financially constrained, the cost of building the necessary complementary assets is high in terms of both monetary and opportunity costs. If these costs are sufficiently high, an innovator is better off pursuing intrapreneurship, in which an established firm bears many of these costs. Similarly, if there are market frictions that impede contractual solutions to access complementary assets, a new venture will be exposed to appropriation risks from which an established firm—which already controls the necessary complementary assets—is insulated. The transaction costs, not only in product but also in resource markets, thus dictate the ease with which an innovator can configure complementary assets from scratch, as is required for the creation of a new venture.

Access to complementary human assets is shaped by the same dynamics as access to nonhuman assets but, because humans have agency, additional factors come into play. The ability of innovators to rally teammates to join them in risky new ventures impacts the costs of acquiring these complementary human assets. The quality of a founder, his/her leadership skills, and what the founder gives up to pursue the venture all influence the relative costs of hiring the right team (Agarwal et al., 2015). Thus, those factors contribute to the position on the vertical axis of Figure 1.

Finally, access to both human and nonhuman complementary assets is shaped by inertial forces and power dynamics within established firms. Entrepreneurship is often triggered by innovators when they feel frustrated in pursuing their ideas within an existing firm (Agarwal et al., 2004; Klepper, 2016). Inertia caused by lack of cognitive attention to new opportunities (Tripsas and Gavetti, 2000), or by power dynamics and politics within existing firms (Eisenhardt and Bourgeois, 1988) implies that, even though their existing firms have the required complementary assets, innovators’ ability to access those assets is impeded  (Moore and Davis, 2004).

Potential Dynamics in Underlying Factors and the Impact on Intra/Entrepreneurship

Given the theoretical framework described above, we now can examine how economic dynamism may be shaped by entrepreneurship and intrapreneurship. In particular, we discuss examples that may impact this choice, rather than examples that increase or decrease the incidence of both intra and entrepreneurship. While we rely heavily on anecdotal evidence here, these examples may set the stage for a research agenda that explores the extent to which these changes are more pervasive across organizations, leading to implications for economic dynamism.

Increased Access to Complementary Assets Within Existing Firms
Propelled by concerns about employee entrepreneurship due to perceived inertia and lack of internal support for innovative ideas, firms such as Google and IBM have initiated policies and set structures that enable intrapreneurship.

IBM’s emerging business opportunities, for example, recognize that some of the most successful new products are based on innovation within IBM, but were exploited by IBM employees who ventured out. Creating an “ambidextrous” organization that simultaneously can explore and exploit new ideas (Tushman and O’Reilly, 2006) helps IBM retain its most innovative employees.

Similarly, prodded by concerns about the xoogler (ex-Googler) phenomenon that resulted in several Google spinouts, Google is proactively developing sophisticated search algorithms to identify employees at risk of becoming entrepreneurs (Fost, 2008). The company then provides them the support and infrastructure necessary for success within its own boundaries. This has enabled Google to retain more innovators, and provides a complement to its innovative environment for creating new ideas (in which, for example, employees are encouraged to spend twenty percent of their time on self-defined projects that they believe will be valuable to Google).

Corporate venture capital (CVC) is another tool employed by established organizations to provide complementary assets to innovators. Providing venture capital gives the established firm a real option on innovative ideas that may be riskier or more distant to core information. These represent win-win outcomes because they enable innovators to develop ideas outside the inertial constraints of an existing organization, but allow them access to complementary assets provided by the CVC partner.

As exemplified by Cisco, CVC also can be leveraged for “spin-ins.” Cisco invested $135 million in Insieme Networks, a startup led by three of Cisco’s top engineers. After Insieme Networks was successful, Cisco purchased it for $863 million. As Burrows (2012) notes: “The spin-in structure lets Cisco ensure the allegiance of Mazzola and his team, rather than risk competing with them or losing them to a rival. Mazzola’s startups also help Cisco tap top engineering talent that might otherwise avoid working for a large public company.”

A related strategy is the acqui-hiring of startups by established firms (Selby and Mayer, 2013). In an acqui-hire, an established firm acquires a startup primarily to gain access to the underlying team of human capital. Relative to traditional, individual hiring practices—which break up productive teams and increase the number of competitors in a space—acquiring an entire startup preserves the routines, tacit knowledge, and other complementary assets embedded in the team structure. It also constrains the flow of knowledge to competitors.

While acqui-hires look like startup exits, bringing a startup into an existing organization may enhance the innovative potential of the team. As Tony Fadell wrote in his blog when Google acquired Nest Technologies (after Google funded it through several rounds of corporate venture capital): “Google has the business resources, global scale and platform reach to accelerate Nest growth across hardware, software and services for the home globally. And our company visions are well aligned—we both believe in letting technology do the hard work behind the scenes so people can get on with the things that matter in life (Fadell, 2014).” He recognizes that access to Google’s complementary assets and its complementary culture will accelerate Nest’s growth and innovative output.

As these factors become more common, they represent a trend downward along the vertical axis of Figure 1, which results in an increased likelihood of intrapreneurship relative to entrepreneurship. Accordingly, proliferation of these tools may explain the reduced entrepreneurial activity and increased exits of entrepreneurial firms observed in recent data. However, from the innovator’s perspective, these represent higher-growth options for ideas because they increase access to complementary assets within the established firm.

In summary, if economy-wide trends are making intrapreneurship more attractive than entrepreneurship, the data would demonstrate less entrepreneurial activity, but this does not necessarily indicate that innovative activity has stagnated.

Increased Use of Labor-Market Constraints
While the above strategies relate to what established firms may do to enable access to complementary assets, defensive strategies in the use of labor-market constraints may limit an innovator’s exit options.

Of great importance in this context are firms’ threats to enforce non-compete clauses (Starr, 2015) and intellectual property ownership (Ganco, Ziedonis, and Agarwal, 2015). Both discourage innovators from taking their ideas outside the boundaries of the firm and from poaching coworkers. If innovators are uncertain about the costs and legal implications of taking an idea and/or their coworkers outside of the boundaries of a firm, they are less likely to leave the established firm to form a new venture.

However, attempting to limit employee mobility has a mixed impact on employee entrepreneurship—some constraints limit it while others actually enhance It. For example, in the Silicon Valley wage-collusion case, leading technology firms agreed not to compete for each other’s workers, thus preventing employees’ threats of mobility form bidding up wages. The artificially suppressed wages at established firms made the alternative career path of becoming an entrepreneur relatively more attractive. For example, these collusion practices did not prevent innovators, such as Tony Fadell, from leaving Apple to found Nest Technologies.

While labor-market constraints may decrease the ability of innovators to transfer complementary assets to new ventures (represented by a downward movement along the vertical axis of Figure 1), they might strengthen the incentives to do so in some contexts. As a consequence, limiting employee mobility could, in some contexts, enhance entrepreneurship.

New Technologies that Allow Access to Complementary Assets through Markets
Advances in technology are facilitating market-based access to complementary assets and, in so doing, make specialization easier for potential entrepreneurs. According to the Lazear “Jack-of-all-Trades” view of entrepreneurship (2005), entrepreneurs (or founding teams) need to be generalists because they must perform the very broad range of tasks necessary in a new venture. However, new technologies now allow entrepreneurs to outsource many tasks, thus accessing complementary assets through markets. Instead of learning to do the minimum level of necessary accounting, entrepreneurs can buy simple software. Instead of learning know how to write a press release, develop a logo, and build a website, entrepreneurs can outsource those tasks to a task-based, online marketplace. Enhanced access to complementary assets would be reflected in a leftward movement along the horizontal access of Figure 1.

While the availability of these tools potentially makes it easier for individuals to become entrepreneurs by accessing some complementary assets, there may be mixed effects both for incidence of entrepreneurship and share of employment in new ventures.

First, it may be that firms’ competitive advantage has shifted to complementary assets that are not transacted easily, such as brand recognition and scale economies. Consider, for example, the effect of the Internet on retail trade: Companies such as Amazon and eBay have simultaneously encouraged individuals to trade as independent owners, and reduced the number of single “mom and pop” establishments that “go it all alone.” The importance of the Amazon and eBay brands and related complementary assets (e.g., distribution, financing, customer reviews) implies that individuals may choose to partner with established firms in their retail endeavors. Overall, many of the new technologies may diminish the incidence of entrepreneurs who create new ventures, and increase the incidence of self-employment.

Second, entrepreneurs who have access to more of these tools can execute a greater share of founding activities themselves and, thus, do not need to build as large a team. The implication is that founding teams will be smaller and, ultimately, fewer people will be employed in new ventures.   

Finally, it is unclear how the new technologies impact the balance between time costs and financial resources. For example, the increased ability to outsource relaxes time constraints, but tightens financial constraints. If financial constraints are more binding than time constraints, new technologies may not be effective at enabling entrepreneurship.

Increased Regulatory Burdens and their Impact on Innovators’ Intra/Entrepreneurship Decisions
Complementary resources and assets directed toward compliance with federal and state laws result in bureaucratic and regulatory burdens of operation. The need for these complementary resources has increased over time. In fact, the Reg Stats published by George Washington University depict a sharply positive slope across multiple indicators of regulatory burdens, including increases in newly proposed rules, number of pages of the code of federal regulations, fiscal budget devoted to regulatory activity, and number of economically significant regulations issued in a year.1 Importantly, Crain and Crain (2014) show that there are economies of scale in compliance costs, and estimate that small manufacturing firms incur two and a half times the cost per employee compared with large manufacturing firms. Iliev (2010) similarly notes that the Sarbanes Oxley Act resulted in significantly high costs for small firms. Popular press accounts (Chisholm, 2013; Shane, 2014) also highlight increasing subjectivity and complexity in rules that are stifling new venture creation and small business activity.

In our context, in addition to the overall dampening of economic activity, increased regulatory burden also impacts the intrapreneurship versus entrepreneurship tradeoff. In Figure 1, increasing regulatory burdens imply a rightward movement along the horizontal axis (importance of complementary assets), and a downward movement along the vertical axis (relative costs of accessing complementary assets in established firms). Each of these changes increases the relative attractiveness of intrapreneurship.

Intra/Entrepreneurship as Career Decisions for innovators: A Research and Policy Agenda

Many of the mechanisms we highlight above are illustrated by anecdotal evidence, and we hope that both our framework and the issues it identifies help set future research and policy agendas.

First and foremost, we need research—both qualitative and quantitative—that examines whether these anecdotes are representative of broader, meaningful trends. Such research would build a valuable foundation for understanding antecedents and consequences of changes in economic dynamism. Additionally, our proposed framework highlights a general toolkit for thinking about other factors that may drive innovation and intra/entrepreneurship. We now briefly explore several specific areas in which our framework has important implications. 

Implications for Measurement
It is very challenging to measure novel productive activity within an economy. Startup rates and startup employment are only loose proxies of novel productive activity. These measures do not account for the innovation that occurs within incumbent organizations, within self-employment, and within alternative structures.

An alternate measurement approach would focus directly on innovative activities, and the careers of innovators across all organizations. This approach would separate measurement of innovation and value creation from the choice of organizational form. As innovative individuals build their careers, they continuously make choices between intrapreneurship, entrepreneurship (including serial instances), or exiting from entrepreneurship into intrapreneurship —either alone, as part of a small independent team, or as part of a full company acquired by an established firm.

This alternative measurement is motivated by the premise that the vibrancy of a knowledge economy is predicated on the enterprising action and agency of knowledge workers. This approach helps link micro-level determinants of individual intra/entrepreneurship to macro-level outcomes related to value creation and diffusion of innovation.

By measuring innovators’ careers, we would better understand the source of innovative activity, the long-term incentives facing innovators, and the outcomes of innovation on individuals. Accordingly, innovation policy to promote knowledge creation and diffusion can be informed by examining factors that facilitate or create frictions for these choices.

Implications for Human Capital
If changes in the importance and availability of complementary assets are shifting innovative activity to occur within the context of an incumbent firm instead of in a new venture, this has important effects on human capital development.

Experience in an entrepreneurial environment leads to the development (or revelation) of latent human capital (Campbell, 2013). In other words, experience in a new venture requires workers to learn new and different skills. It also allows workers to reveal existing skills that were not on display at an established firm. This enhanced human capital is valuable to workers at future jobs. It is an open question if intrapreneurial experience provides the same benefit to individuals.

Similarly, if entrepreneurs need a broad skill set, trends toward increased specialization of jobs within organizations limit the pool of potential entrepreneurs. The lack of generalist jobs in existing firms means that workers will not develop a broad range of skills on the job. It also weakens incentives for students to develop broad skills, which would be valuable if students become entrepreneurs but would limit their employment options in the traditional labor market.

Together, these trends suggest a positive feedback loop, whereas firms specialize more, and intrapreneurship becomes more attractive. Therefore, more innovative activity moves toward intrapreneurship and away from entrepreneurship, and in turn, the attractiveness of intrapreneurship increases.

Such a feedback loop would impair development of the future stock of potential entrepreneurs.

Implications for Inequality
Growing evidence suggests that increased inequality in the U.S. is not driven by within-establishment changes, but instead primarily is shaped by increased dispersion of mean wages across establishments.

This between-establishment growth in dispersion appears to be robust across industries (Song et al., 2015). In other words, in any industry, average wages paid at top firms are increasing relative to the bottom firms. If high-wage firms are better at retaining innovators—perhaps because of greater access to complementary assets, or because workers at high-wage firms have greater opportunity cost associated with leaving the firm—then the entrepreneurship/intrapreneurship decision promotes the trend of between-firm and within-industry inequality. The best-performing firms are more likely to retain and benefit from their innovative employees, while lower-performing firms are more likely to lose their best employees (either to entrepreneurship or to intrapreneurship within the boundaries of a higher-performing firm). This inter-industry mobility and sorting enhances inequality, and is at least partially shaped by the importance and availability of complementary assets.

Implications for Policy
Understanding entrepreneurship requires looking beyond entrepreneurs in isolation, and understanding more deeply the connection between innovation, complementary assets, and organizational forms. As noted previously, an important recommendation is to better measure innovation wherever it occurs, and to better measure the careers of innovators across all organizations.

Understanding the behavior of innovators broadly defined is particularly important for policymakers because policies make an impact at the individual level. Policies change individuals’ incentives and constraints, so a better understanding of how individual innovators behave will allow the development of more effective programs. Further, given that successful innovation and access to complementary assets are linked, understanding how policies affect the importance and availability of complementary assets to individual innovators is an important component of designing policy that leads to the desired outcomes.


Recent trends in economic dynamism—as measured by entry and exit of new ventures, and by associated job creation and destruction—have suggested a potential decline in the U.S. economy. An implicit assumption when creating links between entrepreneurship rates and economic dynamism is that new ventures alone are key to innovation.

We have argued for moving beyond the firm level of analysis, for incorporating the confluence between individuals and firms, and for focusing on innovators’ career decisions as they choose between organizational forms (established versus new firms). Our framework highlights the importance of complementary assets to the human capital of innovators, and the relative costs of accessing these assets across organizational forms. We provide a brief description of how recent trends may impact these factors, and hope that future research will shed light on whether there may be an optimistic silver lining to the observed trends.

If entrepreneurship is being substituted by intrapreneurship, the implications for dynamism are more benign—pointing to a robust economy in which innovation is not necessarily stifled. However, if entrepreneurship and intrapreneurship are both declining—particularly due to factors such as increased regulatory burdens that dampen innovation and individual initiative—there should be an even greater sense of urgency and call for action to restore America’s dynamism. After all, it is enterprising individuals in both new and established firms who make for an enterprising economy.

About the Authors
Rajshree Agarwal is the Rudolph P. Lamone Chair and Professor in Entrepreneurship at the University of Maryland and director of the Ed Snider Center for Enterprise and Markets. Agarwal’s research interests focus on the implications of entrepreneurship and innovation for industry and firm evolution.

Benjamin Campbell is an Associate Professor of Management and Human Resources at the Max M. Fisher College of Business at the Ohio State University. He received a PhD in economics from UC-Berkeley and has an undergraduate degree in mathematics from the Ohio State University.


  1. The George Washington University, Reg Stats (Sept, 14, 2015)


Agarwal, R., B.A. Campbell, A. Franco, and M. Ganco M. 2015. "What Do I Take With Me?: The Mediating Effect of Spinout Team Size and Tenure on the Founder-Firm Performance Relationship," Academy of Management Journal.

Agarwal, R., R. Echambadi, A.M. Franco, and M.B. Sarkar. 2004. "Knowledge Transfer Through Inheritance: Spinout Generation, Development, and Survival," Academy of Management Journal 47(4), 501-522.

Agarwal, R., and S. K. Shah. 2014. "Knowledge sources of entrepreneurship: Firm formation by academic, user and employee innovators," Research Policy 43(7), 1109-1133.

Alchian, A.A., and H. Demsetz. 1972. "Production, Information Costs, and Economic Organization," The American Economic Review 62(5), 777-795.

Burrows, P. 2012. "Insieme: Cisco’s Latest ‘Spin-in’," BloombergView.

Campbell, B.A. 2013. "Earnings Effects of Entrepreneurial Experience: Evidence from the Semiconductor Industry," Management Science 59(2), 286-304.

Campbell, B.A., M. Ganco, A.M. Franco, and R. Agarwal. 2012. "Who leaves, where to, and why worry? Employee mobility, entrepreneurship and effects on source firm performance," Strategic Management Journal 33(1), 65-87.

Chisholm, J. 2013. "Six ways to save U.S. startups and jobs from death by regulation," Forbes.

Coff, R.W. 2010. "The coevolution of rent appropriation and capability development," Strategic Management Journal 31(7), 711-73.

Crain, M.W., and N.V. Crain. 2014. "The Cost of Federal Regulation to the US Economy, Manufacturing and Small Business," National Association of Manufacturers.

Eisenhardt, K.M., and L.J. Bourgeois. 1988. "Politics of Strategic Decision Making in High-Velocity Environments: Toward a Midrange Theory," Academy of Management Journal 31(4), 737-770.

Fadell, T. 2014. "Welcome home," Nest, June 9, 2015.

Fost, D. 2008. "Keeping It All in the Google Family," The New York Times..

Ganco, M., R.H. Ziedonis, and R. Agarwal. 2015. "More stars stay, but the brightest ones still leave: Job hopping in the shadow of patent enforcement," Strategic Management Journal 36(5), 659-685.

Gort, M., and S. Klepper. 1982. "Time Paths in the Diffusion of Product Innovations," The Economic Journal 92(367), 630-65.

Hart, O. 1995. Firms, Contracts, and Financial Structure (Clarendon Press).

Iliev, P. 2010. "The Effect of SOX Section 404: Costs, Earnings Quality, and Stock Prices," The Journal of Finance 65(3), 1163-1196.

Klepper, S. 2016. Experimental Capitalism: The Nanoeconomics of American High-Tech Industries (Economics Books, Princeton University Press).

Lazea, E.P. 2005. "Entrepreneurship," Journal of Labor Economics 23(4), 649-680.

Markham, S. K., and A. Griffin. 1998. "The Breakfast of Champions: Associations Between Champions and Product Development Environments, Practices and Performance," Journal of Product Innovation Management 15(5), 436-45.

Moore, G., and K. Davis. 2004. "Learning the Silicon Valley way," in Building high-tech clusters: Silicon Valley and beyond, 7-39.

Pinchot, G. 1987. "Innovation Through Intrapreneuring," SSRN Scholarly Paper, Social Science Research Network, Rochester, NY.

Schumpeter, J.A. 1911. The Theory of Economic Development: An Inquiry into Profits, Capital, Credit, Interest, and the Business Cycle (Transaction Publishers).

Selby, J., and K.J. Mayer. 2013. "Startup Firm Acquisitions as a Human Resource Strategy for Innovation: The Acqi-hire Phenomenon," Academy of Management Proceedings 2013(1), 7109.

Shane, S. 2014. "To help small business, cut regulation," Entrepreneur, January.

Simon, J.L. 1998. The Ultimate Resource 2 (Princeton University Press).

Song, J., D.J. Price, F. Guvenen, and N. Bloom. "Firming Up Inequality," working paper, National Bureau of Economic Research.

Starr, E.P. 2015. "Training the enemy? Firm-Sponsored Training and the Enforcement of Covenants Not to Compete," January.

Teece, D.J. 1986. "Profiting from technological innovation: Implications for integration, collaboration, licensing and public policy," Research Policy 15(6), 285-305.

Tripsas, M., and G. Gavetti. 2000. "Capabilities, cognition, and inertia: Evidence from digital imaging," Strategic Management Journal 21(10-11), 1147-1161.

Tushman, M.L., and C.A. O’Reilly III. 2006. "Ambidextrous organizations: Managing evolutionary and revolutionary change," Managing innovation and change, 170.

Session Summary:


New Entrepreneurial Growth Conference: Entrepreneurs and polarization

Economic polarization and income inequality within the American population have been hot topics in the United States over the past few years. At Kauffman’s New Entrepreneurial Growth conference, many participants talked about polarization and inequality among American businesses, and how that could affect entrepreneurship. Some participants suggested that advances in technology will yield rapid growth of only a small fraction of firms, creating an increasingly bimodal distribution of businesses. Large companies, they said, will be stronger, dynamism will decrease, and there will be a growing space between the top- and bottom-performing firms. Middle-sized businesses, they predicted, will be rare, as these firms won’t have access to the capital they need for growth or the reputational capital that ensures consumer trust. Technology, one participant stated, has a centralizing effect, and the market of the future may have only two primary platforms, with other companies serving small niches and very little of the competition that is necessary for a healthy market economy.

Other participants, however, disagreed that the role of new and small companies will diminish, indicating that big companies are sluggish, lack creativity, and almost always make mistakes that give small and nimble emerging firms the opportunities they need. Some participants also anticipated that a decline in costs and increase in the accessibility of knowledge will help entrepreneurs and that infrastructural technologies that enable advances in certain fields always will provide opportunities for entrepreneurship on a worldwide basis.

Citing the data from John Haltiwanger’s paper, one more contributor suggested that these two positions are not incompatible. While we are seeing more, larger firms, even in high-tech and software publishing, the polarization of businesses, he said, is not new. The economy has consistently seen this skewed distribution of productivity in all sectors, with enormous dispersion and only a few firms on the right tail. This polarization, however, does not mean that the role of entrepreneurship in the economy is diminished. In every generation of technology, entrepreneurs play an important part in innovation, but only a very small fraction of them grow. The digital age, he contended, will not change dispersion within industries.

Other participants agreed, also emphasizing that a focus on the U.S. economy is too narrow. The Internet, they said, allows businesses to become instantly international, accessing a much larger market and competing on a very large scale. Entrepreneurs need this broader market. For niche businesses that would be unsustainable in a small community, technology offers them access to the customers they need to survive. And entrepreneurs more generally, they said, need international markets to capture the benefits of their innovations. As rapid business cycles make it more difficult for innovators to achieve profits, international markets are essential to growth.