Understanding Business Dynamism
“[T]he overall ability of entrepreneurship to facilitate … economic prosperity depends disproportionately on the … performance of a very small number of new firms.” Guzman and Stern, 2016.
For at least the past 30 years, the rate of U.S. business formation has been falling and the average age of existing firms has been rising. Since 2000, two other things have happened: productivity growth has slowed, while many skilled jobs have disappeared. Startups are thought be a key source of innovation in the economy and of net job creation. At the same time, as Schumpeter’s notion of creative destruction suggests, the death of old firms is a critical part of the renewal process. So, the declining trend of entry and exit has people worried that U.S. business dynamism is ebbing (see our earlier post).
How concerned should we be? To be completely honest, we don’t really know; at least, not yet. But the answer is important, because it can help orient the U.S. economic policy framework to support the creation of successful businesses that generate high-quality jobs. In this post, we summarize some new research aimed at helping us understand what the decline in business formation really means. Does it signal a fall in the number of successful firms that contribute substantially to business value added, productivity, and employment? Or, is it a decline in the formation of firms that never exceed a tiny scale and have little impact on the broader economy?
To make some headway in addressing these questions, we start with a few facts. A key finding of the foundational work by John Haltiwanger and his coauthors is shown in the following chart, where we plot the rates at which firms are born and die. The Business Dynamics Statistics (championed by Haltiwanger when he was chief economist at the Bureau of the Census) shows that, as a share of existing firms, the number of new firms fell from an average of 14 per 100 in the 1980s to less than 10 over the most recent five years of data (to 2014). The rate at which firms die declined by less—from a bit under 12 in the 1980s to roughly 9 in the recent period. Because the economy grew over this period—since 1980, real GDP per capita increased by 70 percent while employment rose by almost 50 percent—we can infer that firms became older and larger.
Rate of firm entry and exit (percent of existing establishments), 1977-2014
Over this same period, productivity growth rose and then fell. The following chart plots the 10-year lagged growth rate of total factor productivity (TFP) growth (adjusted for capacity utilization). The first data point, plotted as 1977, shows that from 1967 to 1977 TFP grew at nearly a 1½% average annual rate. During the decade ending in 2004, TFP growth peaked at almost 2%. Since then, it has plunged below ½%.
Total factor productivity adjusted for capacity utilization (10-year lagged growth rate in percent), 1977-2016
Are these patterns of business formation and productivity growth related? A cursory glance at the two charts does not suggest a close match. That is, it would be difficult to account for the rise and fall of productivity growth using the persistent downward trend in business formation. The following chart shows the cross-correlations of contemporaneous and lagged entry and exit rates with annual TFP growth. What stands out is that, at a lag of five years when the correlation magnitudes are largest, these are negatively correlated (the estimated correlations are -0.18 and -0.20, respectively). As an aside, we note that this says nothing about employment. To the extent that start-ups play a role in job formation, we should still be concerned, independent of whether their rates of birth and death are related to productivity growth.
Correlations of total factor productivity with lagged firm entry and exit rates (annual data), 1977-2016
Source: Authors’ calculations.
Recent work by Guzman, Stern and coauthors starts from here—that is, from the lack of a close relationship between measured business formation and productivity growth. They examine the intriguing notion that the ebbing rate of start-ups represents a decline in the formation of firms that matter relatively little for the economy as a whole—for value added, productivity, and job creation.
The key is to distinguish between two types of firms: those that are likely to stay small and those that are likely to grow. Most new firms—like plumbers and self-employed consultants—have little potential to either generate jobs or contribute to productivity growth. Indeed, an exceedingly small share of firms ever reaches substantial size or stature. Taking a very restrictive view of success, Guzman and Stern (GS) examine the probability that, within six years of being formed, a new firm will either conduct an initial public offering (IPO) or be acquired. In their sample, which covers over 12 million firms in 15 states representing roughly half of U.S. GDP from 1988 to 2014, the likelihood of such a “growth event” is 1 in 3,500. That is, for every 3,500 firms that are started, only one “succeeds”. Put differently, outcomes are extremely skewed, with only a tiny share of firms accounting for virtually all the growth performance.
So, has the observed fall in the aggregate U.S. start-up rate been associated with a decline in the number of firms that are likely to succeed? To answer this question about “entrepreneurial quality”, GS note that the likelihood of success is associated with a limited number of early-stage characteristics. Strikingly, they find that if a firm is both registered in Delaware and has filed for a patent in its first year of existence, then it is nearly 200 times more likely to succeed. That is, such firms succeed at the rate of more than 1 in 20 (rather than 1 in 3,500)! The likelihood of success increases further if a firm’s name is short and does not include the name of the founder, and it is located in a high-tech cluster such as information technology in Silicon Valley or biotechnology in Boston. GS then use this predictive model to estimate the current quality of entrepreneurs: high-quality firms are those that have these propitious characteristics at birth.
The following chart, reproduced from their paper, shows the “Regional Entrepreneurship Cohort Potential Index” for the entire United States. For each cohort (based on the year of founding) the chart depicts the expected number of successful firms (scaled by real GDP) computed from firm characteristics at founding. The pattern is striking: between the dot-com bust in 2000 and the financial crisis in 2007, the quality of entrepreneurial startups was higher than at any time prior to 1998. And, with the exception of the dot-com boom and bust, the “nowcast” measure for 2014 is at the highest level over the period for which they have collected data.
Source: Guzman and Stern, Figure 3.
We draw several inferences from this collection of results. First, while the rate at which people start firms has fallen by something like 40 percent, most of these “missing” start-ups were unlikely to expand in a way that influences the macroeconomy in a notable way. In fact, the chances of being truly successful—defined as issuing publicly traded equity or being acquired—is vanishingly small. Of the 20 million or so firms created in the past 20 years, fewer than 6,000 have “succeeded”. And, importantly, the number of these high-quality firms has not been declining.
But the key lesson from this research is that we still have much to learn about how and under what conditions firms thrive. And, even more importantly, we need to know more about the conditions that lead firms―both new and old―to hire people and pay them well. How can we encourage more startups of the kind that are likely to succeed, enhancing productivity and creating good jobs? To be clear, this is not a question of industrial policy. It is not about some government bureaucrat picking winners and losers, something that has typically proved counterproductive. Rather, it is about establishing rules and conditions that encourage entrepreneurial success. We applaud the researchers who are trying to figure this out, and look forward to analyses that will advance our understanding and guide future policy.