The Geography of Self-Reported Innovation: Results from the 2017 Annual Business Survey
Working Papers | NCSES 22-206 | April 20, 2022 |
National Center for Science and Engineering Statistics, National Science Foundation
This working paper provides empirical evidence of the concentration and dispersion of the broad-based definitions of self-reported innovation available in the 2017 Annual Business Survey (ABS). The large sample size of the inaugural ABS provides a unique opportunity to produce relatively precise estimates of innovation rates at the substate level that had not been available in earlier innovation data collections such as those that were part of the 2014 Annual Survey of Entrepreneurs or the Business R&D and Innovation Survey. The ABS is a firm- or enterprise-level survey, so the geography of innovation is only precisely estimated for single-unit firms. The statistics that follow are all based on the single-unit subpopulation that comprises close to 60% of the sample. The correspondence between single-unit firm estimates and published estimates using the complete sample is addressed with respect to state innovation rates below.
Innovation rates at the national level broken out by industry, firm size, metropolitan/nonmetropolitan classification, and settlement size categories provide benchmarks that will be helpful for making sense of the state-level breakouts that follow. These estimates are preceded by an explanation of exactly how innovation is defined in the ABS and how these measures will be used. The working paper concludes by identifying the most innovative labor market areas in the United States, in terms of overall innovation rates and innovation rates that consider the concentration of high-tech, medium-tech, and low-tech industry in the local economy.
Until the 1990s, the measurement of innovation was largely limited to utility patents and expenditures on research and development. Recognition that utility patents are imperfect and incomplete measures of invention (much less of innovation) and that R&D expenditures are inputs to the innovation process rather than measures of innovation output compelled an international effort to develop a more inclusive measure of innovation (Arundel and Smith 2013). The Oslo Manual, first published in 1992 by the Organisation for Economic Co-operation and Development (OECD), provided guidance to national statistical agencies on the collection and interpretation of self-reported innovation data. The third edition of the Oslo Manual, published in 2005, guided the design and development of the 2017 ABS used in this analysis and thus provides the appropriate lens for examining these data (OECD/Eurostat 2005).
The two critical requirements for reporting a product innovation are (1) the innovation has been introduced on the market and (2) the innovation is a new good or service for the firm or an existing good or service that was significantly improved. Market success of an innovation is not, however, a criterion. Thus, whether the innovation is “good” or “bad” as indicated by popularity or sales volume is not considered in classifying a firm as innovative or not. This positive measure of product innovation has the advantage of making data collection and classification easier, along with directly informing the role of innovation in adaptive efficiency of the economy; that is, providing a measure of the number of firms that are introducing new products or significantly improving existing products whether successful or not. The disadvantage of a positive measure is that it is more difficult to connect the innovation measure directly to economic impact. This may be one reason why relatively little research has been done on self-reported innovation measures in the United States to date. Another possible reason is that an innovation measure that combines the introduction of completely new products with the significant improvement or imitation of existing products may be at odds with how many Americans perceive “innovation” (Peric and Galindo-Rueda 2014; Tuttle, Alvarado, and Beck 2019).
The most objective way to limit the measure of self-reported innovation to truly novel products or improvements is to require that innovating firms be the first to introduce such products in their market. Respondents are asked if any of their innovations have been “new to market,” being the first among competitors introducing an innovation. This qualification eliminates the type of innovation that was most problematic for participants in a cognitive interview study: imitation qualifying as innovation (Peric and Galindo-Rueda 2014; Tuttle, Alvarado, and Beck 2019). Many U.S. participants expressed the view that their interpretation of “innovation” was limited to the creation of something unique.
Two innovation rates are thus reported in all the tables that follow: (1) any product innovation comprising new-to-firm and new-to-market innovation and (2) a new-to-market rate of innovation that captures the creation or origination of unique products.
The collection of industries operating in a region is likely to affect the observed rates of any product innovation and new-to-market innovation. Emerging industries that are exploring and developing new uses for foundational technology likely contain firms more prolific in new-to-market innovation. More mature industries with established product lines and uses are likely to demonstrate much lower rates of new-to-market innovation. However, differences in any product innovation between emerging and mature industries is an empirical question because firms in mature industries may be making numerous changes to existing products in order to remain competitive.
Measuring the relative innovation of a region given its collection of emerging and mature industries can be done by first measuring the innovation rate for detailed industries at the national level. The predicted level of innovation for a region can then be computed by assigning the national innovation rate to each firm within a detailed industry. Subtracting the predicted innovation rate for the region from the observed innovation rate would provide an estimate of the region’s innovative shift—the degree to which the collection of industries is more or less innovative than the same collection of industries at the national level.
Regional Innovative Shift = Observed Innovation Rate – Predicted Innovation Rate
Deriving a regional innovative shift estimate that is statistically significant (i.e., the range of estimates does not include zero after adding or subtracting a margin of error) is highly unlikely, given that the error associated with it is the addition of the error rates from the observed and predicted rate estimates. Given this limitation, the analysis that follows will only be able to indicate whether a regional innovative shift is likely to exist, indicated by the observed innovation rate being statistically different from the predicted innovation rate. A positive regional innovative shift with the observed innovation rate larger than the predicted innovation rate would indicate that local businesses, on average, are more innovative than their peers throughout the nation.