19
of non-compete enforcement.
31
Interestingly, we find stricter non-compete enforcement to be
associated with both lower wage growth and lower initial wages.
32
The first column of Table 1 shows the percentage change in wages from a one-unit increase in a
non-compete enforceability index, holding constant a number of worker characteristics.
33
It
suggests that a standard deviation in non-compete enforcement reduces wages by about 1.4
percent. Recent work by Starr and coauthors finds broadly similar results to those presented
here.
34
It is possible to refine this approach by focusing more narrowly on populations likely to be
affected by non-competes. Workers with bachelor's degrees are more than 50 percent more
likely to be bound by non-competes than those without, suggesting that one might better
approximate the “eligible” subgroup by restricting the sample to workers with bachelor's
degrees. This is shown in Table 1, column 2. Note that the magnitude of the wage effect of non-
compete enforcement increases for this subgroup, as expected. A slightly more nuanced
approach makes use of the occupational breakdown provided in recent work. Rather than
omitting non-college workers, we instead reweight the sample to be more representative of
workers with non-competes. For example, this will imply placing a higher weight on workers in
the architecture and engineering occupations than in the personal services occupations. Table 1,
column 3 shows results from this reweighted approach. The magnitude of the wage impact is
again above that of column 1, but not dramatically so.
35
31
We use the 2014 merged outgoing rotation groups of the Current Population Survey (CPS), which provide a cross
section of population-representative workers. Merged with this data is the Starr-Bishara index of non-compete
enforceability by state (generously provided by Evan Starr), as well as the fraction of workers with non-competes by
major occupation from Starr, Bishara, and Prescott (2015).
32
Here again, the particular proposed explanation for non-competes is important. For instance, if screening is the
dominant explanation, and workers are fully informed about non-competes, we would expect stricter enforcement to
cause an initial wage premium but slower subsequent wage growth. Workers would only be willing to sign the non-
compete if they were compensated at the time of signing. If, on the other hand, salience is the dominant
explanation, we would expect no initial premium and slower wage growth, as workers are prevented from taking
advantage of outside opportunities or using outside opportunities as leverage for wage growth at the current firm.
33
These controls consist of education, age, gender, marital status, occupation, industry, public sector status, and
union status.
34
See forthcoming work by Balasubramanian, Chang, Sakakibara, Sivadasan, and Starr, as well as Starr, Ganco, and
Campbell.
35
This is perhaps to be expected given the fact that that non-competes are used quite broadly. While non-competes
are more common in particular occupations (e.g., management, computer and mathematical, and architectural and
engineering occupations), they are also found in a wide variety of unexpected occupations and education levels.