Innovation and new business formation: the role of innovative large firms

We adopt the measures developed by Kogan et al. (2017). The authors use a firm’s stock price reaction around its patent announcement dates to proxy for the market value of its patents. We hereinafter refer to KPSS with respect to their measurements of patent value. They also measure the scientific value of patents using the number of forward citations those patents receive in the future. We thank Kogan et al. (2017) for making their data on patent value available at https://​github.​com/​KPSS2017/​Technological-Innovation-Resource-Allocation-and-Growth-Extended-Data. The most recent dataset, posted in August 2020, includes patent data from 1926 through 2019.

We recognize the limitation of this approach to treating inventive firms’ headquarters as the locations of patent applicants. Due to the limitation of current database, we are unable to distinguish the locations of applicants and inventors. We hence follow the common practice in the innovation literature and use the locations of inventors as the proxy for applicants’ locations.

CBP data report industry-level business formation by standard industrial classification (SIC) from 1986 through 1997, and by NAICS from 1998. We carefully select the manufacturing sector from CBP data for these two time periods to form a consistent time series of data for the manufacturing sector.

This return is estimated after adjusting for idiosyncratic stock return volatility and aggregate market risk.

Compustat backfills the headquarters information of firms in the database. Although firms do not change their headquarters locations frequently, we track their location information from annual reports to ensure accuracy in defining the local community for firms with patent grants.

USPTO publishes general statistics and miscellaneous reports that describe patenting activity in both corporate and academic sectors.

A core-based statistical area (CBSA) is a geographic area that consists of one or more counties that are closely tied to an urban center based on commuting distance. It is defined by the Office of Management and Budget based on Census 2000 data.

Opponents to the Kaplan–Zingales (KZ) index believe that it uses more common characteristics that also capture firms’ financial distress, while Kaplan and Zingales (2000) argue that financial distress can be considered a variation of financial constraint. The general findings in the literature show that the KZ index captures the distress aspect of financial constraint more accurately than the HP index.

The HP index is a combination of asset size and firm age and is calculated as (− 0.737*Assets + 0.043*Assets² − 0.040*Age).

The WW index is estimated as the linear combination of cash flow to total assets ( −), sales growth ( −), long-term debt to total assets ( +), log of total assets ( −), dividend policy indicator ( −), and the firm’s three-digit SIC industry sales growth ( +).

One example of taking the median value as the aggregate measure is found in Rajan and Zingales (1998). In their study, the median value of firms’ reliance on external finance is used as the proxy to avoid the dominance of large firms when summarizing ratios across firms.

Alternatively, we take the logarithm value of total establishment and repeat the regressions, and the results remain unchanged. We report the results using standardized values to offer a direct interpretation of regression coefficients.

In Table 4, the measures of innovative activities are constructed from manufacturing firms only.

Using evidence from Finland, Forsman (2011) compares the innovations between manufacturing and service industries and finds slight differences between the manufacturing and service industries. Studies tend to address technology intensity within manufacturing sectors, in contrast to knowledge intensity within service sectors (Amara et al., 2009; Miles 2000; Pavitt 1984). In our tests, we focus on the total establishment in manufacturing industries as our main dependent variable and relate it to the innovations generated in the manufacturing industries. Alternatively, we measure innovations from all industries. The results, however, show no significant difference when using the broader measurements taking innovations from all industries, suggesting the dominating role of manufacturing industries in the innovative activities. We hence only report the results using the innovations from the manufacturing industries.

We also scale the variable by total county population and the results are consistent.

Results are not reported here due to space considerations, but are available upon request.

For the growth rate in large size startups, we find that the effects of patent and R&D activities are insignificant.

Tests using the WW index show consistent results and are omitted here due to space considerations.