The US economy has been one of the most dynamic economies in the world but recent research suggests that US dynamism is in decline. The startup, job creation, and job destruction rates have all declined over the past three decades with a possible increase in the rate of decline in the past decade. The dynamism decline is robust, appearing in a variety of data. Moreover because startups and the movement of resources from low to high productivity firms are closely associated with improvements in productivity, the decline of dynamism may reduce real wages and the standard of living.
Could regulation be increasing barriers to entry, raising the costs of reallocation, and slowing the diffusion of productivity innovations? To test the hypothesis that regulation is reducing dynamism we combined data on dynamism with an industry level measure of regulation. Our measure of regulation is produced by an innovative technique that combs the Code of Federal Regulations (CFR) for restrictive terms or phrases such as “shall,” “must,” “may not,” “prohibited,” and “required”. The count of restrictive words in each section is then associated to industries via a machine learning algorithm that recognizes similarities between the language in that CFR section and industry language (e.g. a section of the text with words such as “pipeline” would be associated with the oil and gas industry). In this way, we can associate each industry with an index of regulation derived from the entire CFR.
The following figure shows the startup rate against the regulatory stringency index (both averaged by industry over the period 1999-2011). Contrary to expectation, there is a slight positive relationship; industries with greater regulatory stringency have higher startup rates. We find a similar relationship with job creation rates.
Of course, it could be the case that more dynamic industries attract greater regulation so the apparent positive relationship in our graph would not reflect a causal connection and could even be masking a negative causal connection. Thus, to further test the relationship, we statistically test whether increased regulatory stringency is associated with reduced dynamism within an industry over time (we give each industry a “fixed effect”). After subjecting the data to a number of different tests we find no statistically significant relationships between dynamism and regulatory stringency (see the paper for details).
One simple test divides manufacturing industries into those that experienced a large increase in regulation (+50% or more) during our time period and those where regulation hardly changed at all (+10%-to -10%). If regulation were the cause of changes in dynamism we would expect to see big differences between these groups. The figure below, however, shows that startup rates, for example, track similarly across the two types of groups suggesting that regulation is not a primary cause of declining startup rates (the same is true for job creation and destruction rates).
It’s important to note that regulation could have large negative (or positive) effects without having a big effect on dynamism. A tax, for example, could reduce the size of the industry without have a big effect on the startup rate or how well the industry responds to shocks by reallocating labor from low to high productivity firms. In short, regulation can have significant effects on levels without necessarily having large effects on growth rates.
If regulation is not responsible for the decline in dynamism then what is? We offer some suggestive hypotheses in another paper. First, it could be the case that we are mis-measuring entrepreneurship. If entrepreneurship is measured as new firm creation, for example, we miss the entrepreneurship inherent in rebuilding and revitalizing larger and older firms. Since most workers work for larger and older firms, revitalizing these firms may be a more important use of entrepreneurship than starting new firms. In an increasingly global economy we may also miss some of the outsourcing of dynamism that has occurred in recent decades. Apple, for example, is measured in US data as a relatively stable firm but the Apple ecosystem from which Apple sources its product is a maelstrom of entry and exit as Apple hires and fires new firms with each new iteration of the iPhone.
Even if dynamism has declined is this necessarily a bad thing? We should not let word associations influence our evaluations of underlying realities. Dynamism as measured by, for example, job reallocation rates might equally well be called churn. Declining churn doesn’t sound as bad as declining dynamism. Moreover, combining the last two points perhaps the reason for some of the declining dynamism as measured in the US statistics is that we have outsourced some of our churn. A very different way of describing the same data.
More generally, information technology may allow us to reduce churn while still allowing adaption and innovation. Creative destruction is necessary for a growing economy but if we can boost the ratio of creative to destruction that counts as an improvement in welfare.
The remarkable reallocation of labor and capital across sectors is an important force driving the American economy forward. We don’t fully understand, however, what the causes of declining dynamism are or exactly how our measures of dynamism relate to entrepreneurship, growth and improvements in the standard of living.
The preceding post comes to us from Alexander Tabarrok and Nathan Goldschlag, George Mason University – Economics Department.