We use confidential micro-data from the Occupational Employment Survey (OES), conducted by the Bureau of Labor Statistics (BLS). This data comes from an annual or biannual survey of individual establishments in the U.S. No establishment is surveyed twice within three years, however, it is common for larger establishments to appear in the data exactly once every three years. The surveyed establishments are selected in a manner to allow for optimal inferences about the US economy as a whole. Aggregated versions of this data are released publicly and used to measure national occupational employment.
For each establishment-year, we observe employment in 800 different occupational categories (represented by 6-digit SOC codes). Within each of these occupations at a given establishment- year, we then observe the count of employment within twelve separate wage bins, where the exact cutoff points for each wage bin changes over time to best reflect changing wage distributions. Furthermore, for each surveyed establishment, we also observe its location (by county), EIN, name, legal name (ultimate owner), industry and a time invariant establishment-identifier which we can use to track establishments which have switched owners over time.
We identify horizontal M&A deals, namely M&As in the same 4-digit NAICS industry, from SDC Platinum. We match those deals to the OES survey over the 2001-2007 period. We start in 2001 as the identifier which we need to link establishments over time is unavailable in earlier years. We end in 2007 to avoid any overlap with the financial crisis which affected both the intensity of M&A activity and firms’ labor market outcomes. We identify a total of 348 horizontal M&A deals in the OES survey that cover 2,141 establishments that had an M&A occurring during the time period the establishment is sampled by OES.2 We create a set of possible control establishments
after excluding all establishments identified to be involved in M&As during our sample period from this group. For each target establishment, we find two control establishments satisfying the following matching criteria:3 i) they operate in the same 4-digit NAICS industry as the target establishment and appear for the first time in the same year in the OES survey, ii) they are sampled for the second time within one year of the treated establishment’s second sampling, iii) they have similar size with the target as measured by number of employees (within 100% of employment distance), iv) they are similar with the target in terms of pre-treatment routine share intensity (within 100% of routine share intensity distance).4 We end up with a sample of 3,081 control
2We use a two-step procedure to match M&A deals to the OES survey. First, we match using EIN and the
target firm’s Compustat provided EIN. However, since firms often report multiple EINs, we also use a name matching procedure. We start with a fuzzy logic algorithm to identify possible candidates, then hand match all likely candidates. A match is only retained if we observe the target establishment strictly before and after the M&A is completed.
3We allow matched establishments to repeat.
4In cases where more than two control establishments satisfy the matching criteria, we keep those establishments
establishments. Both treated and control establishments are observed exactly twice in the 2001- 2007 period.5
We define routine share intensity following Autor and Dorn (2013).6 Routine share intensity
(RSH) of an establishment is defined as the ratio of total employment of routine task intensive occupations over total employment in the establishment. We use a log transformation of one plus the average value of RSH at the establishment level to avoid dropping cases where an establishment has no routine occupations. In the Internet Appendix, we will also present results using instead RTI as our measure.7
We define the share of high-skill employment as total employment identified as high-skill at the establishment-year level as a percent of total employment. We define high-skill employment as managerial occupations in the baseline analysis. We also present robustness using three different definitions of high-skill employment. First, we use data from the 2000 American Commuting Survey (ACS). An occupation is high skill if the percent of workers who have completed some college education is above the 75th percentile of the distribution across all occupations in the ACS
sample. Second, to show our results are robust to alternative cutoffs, we again use the ACS survey but instead define an occupation to be high skill if the percent of workers who have completed some college education is above the 66thpercentile of the distribution across all occupations in the ACS sample. Third, we define high skill occupations following Hecker (2005). These occupations are scientific, engineering and technician occupations.
We define offshorability of a given occupation following Autor and Dorn (2013) and compute
5OES data are imputed when missing. To confirm that our results are not driven by imputation we drop cases
where establishment data is imputed for either one or for both years and re-estimate our baseline regressions. Results are robust.
6Autor and Dorn (2013) define the frequency of “routine” tasks typically performed by employees assigned to a
given occupation. Since occupations involve multiple tasks (routine, abstract, manual) at different frequencies, Autor and Dorn (2013) create an indicator which measures the routine task intensity (RTI) by occupation and define an occupation as routine task intensive if in the top employment-weighted third of routine task-intensity. We merge RTI to occupations in OES by SOC codes using crosswalks from David Dorn’s website. http://www.ddorn.net/data.htm.
78% of establishment-years in our sample have no routine occupations. We find qualitatively similar results if we
an employment weighted average of offshorability at the establishment level.8 We measure wages by taking the occupation-wage bin employment-weighted median within each establishment. All wages are adjusted for inflation and reported in 2001 dollars. We define all variables used in our analysis in the Appendix.
Table 2.1 reports summary statistics for our sample establishments. The average establishment in our sample employs 199 employees. As described earlier, the OES survey over-samples larger establishments. This limits our ability to reach conclusions about the smallest of establishments but ensures that our results are based on a sample of economically important entities. Fifty-three percent of employment at the average establishment is identified as routine occupations and twelve percent as high-skill occupations. Given occupations are coded as routine if they have a routine intensive measure in the top one third of the data, these results suggest that target establishments (and mechanically the matched control establishments) tend to have a disproportionate share of routine employment. Our sample firms have an average (median) wage of $16.5 ($14.2) per hour. This is comparable to the mean (median) hourly US wage in 2001 of $16.35 ($13.0).9 Finally, we
report an average standard deviation of hourly wages equal to 8.4.
We require treated and control establishments to match in terms of pre-treatment employment size and routine share intensity and we report summary statistics pre-treatment for the two groups in columns 4-9, Table 2.1. Our control and treated establishments show economically similar characteristics with the exception of routine share intensity. This may indicate that firms with high routine-share intensity ex-ante are more likely to be M&A targets as the acquirers are aware of the benefits of replacing routine occupations with technology. In untabulated results, we run predictive logit regressions and confirm this to be the case. Our sample of target establishments covers a wide range of industries. About a quarter of our sample M&As take place in the manufacturing sector and 70% in services. The industry distribution is similar across treated and control samples by
8We use SOC codes to merge with the OES sample using crosswalks from David Dorn’s website.
http://www.ddorn.net/data.htm.
definition.
2.2.2 Methodology
To identify the effect of M&As on labor outcomes, we estimate the following difference-in- differences specification at the establishment-year level:
yi,t =αt+αi +γ1·Postt+γ2·Postt·M&Ai+β·Xi,t+i,t (2.1)
whereidenotes establishments andt denotes years. P osttis an indicator set equal to one for
years following M&As—zero otherwise. M&Ai is an indicator equal to one for establishments
targeted by M&As (treated) and zero for the matched set of control establishments.10 Both treated
and control establishments are observed exactly twice in our sample, once prior to the year of the M&A and once after. Xi,t controls for offshorability to alleviate concerns that changes in
establishments’ offshoring potential could affect both the probability of M&As and our measured outcomes. αi is an establishment fixed effect which controls for establishment characteristics that
do not vary over our sample period; andαt is a year fixed effect which absorbs aggregate shocks
affecting all establishments. In all specifications, we report robust standard errors clustered at the firm level.