Knowledge connectedness within and across home country borders: Spatial heterogeneity and the technological scope of firm innovations

Abstract

We explore how knowledge-based connections to domestic and foreign locations affect the technological scope of firm innovations. Inspired by a blend of Economic Geography and International Business perspectives, we propose a theoretical framework that distinguishes between domestic subnational differences and cross-national spatial heterogeneity. Further, we combine the Penrosean view of managerial capabilities with the attention-based theory of the firm. Analyzing a sample of US-based firms between 1990 and 2006, we show that both domestic and international knowledge connectedness affect the technological scope of firm innovations, but their effects are different. The breadth of international knowledge connectedness appears to be positively associated with the technological scope of firm innovations. However, the breadth of domestic knowledge connectedness positively contributes to the technological scope of firm innovations up to a certain point, beyond which the bounded rationality of managers constrains firms’ ability to further leverage subnational heterogeneity. Thus, domestic search is more likely to be challenged by limited managerial bandwidth. Lastly, domestic and international knowledge connectedness significantly interact with each other to explain the technological scope of firm innovations.

Resume

Nous étudions comment les connexions fondées sur les connaissances aux localisations domestiques et étrangères influencent la portée technologique des innovations de la firme. Inspirés par la combinaison des perspectives de la géographie économique et de l'international business, nous proposons un modèle théorique qui distingue les différences infranationales domestiques et l'hétérogénéité spatiale transfrontalière. Par ailleurs, nous combinons la perspective de Penrose sur les capacités managériales avec la théorie de la firme fondée sur l'attention. Analysant un échantillon de firmes basées aux Etats-Unis entre 1990 et 2006, nous montrons que les connectivités domestique et internationale des connaissances influencent la portée technologique des innovations de la firme; mais leurs effets sont différents. L'ampleur de la connectivité internationale des connaissances semble positivement associée à la portée technologique des innovations de la firme. Toutefois, l'ampleur de la connectivité domestique des connaissances contribue de manière positive à la portée technologique des innovations de la firme à un certain point ; point au-delà duquel la rationalité limitée des dirigeants gêne la capacité de la firme à tirer davantage profit de l'hétérogénéité infranationale. Par conséquent, la recherche domestique est davantage susceptible d'être défiée par une bande passante managériale limitée. Enfin, les connectivités domestique et internationale des connaissances interagissent de manière significative l'une avec l'autre pour expliquer la portée technologique des innovations de la firme.

Resumen

Exploramos cómo las conexiones basadas en el conocimiento a sedes domésticas y extranjeras afectan el alcance tecnológico de las innovaciones de la empresa. Inspirados en una mezcla de perspectivas de la Geografía Económica y de los Negocios Internacionales, proponemos un marco teórico que distingue entre las diferencias sub-nacionales domésticas y la heterogeneidad espacial. Además, combinamos la perspectiva penroseana de las capacidades gerenciales con la atención en la teoría de la empresa. Analizando una muestra de empresas ubicadas en los Estados Unidos entre 1990 y el 2006, mostramos que tanto la conectividad al conocimiento interno como el internacional afecta el alcance de las innovaciones de la empresa, pero sus efectos son diferente. La amplitud de la conectividad del conocimiento doméstico contribuye al alcance tecnológico de las innovaciones de la empresa hasta cierto punto, más allá del cual la racionalidad limitada de los gerentes restringe la habilidad de las empresas para beneficiarse de la heterogeneidad sub-nacional. Por tanto, la búsqueda doméstica es más probable que sea desafiada por el limitado ancho de banda gerencial. Por último, la conectividad de conocimiento doméstico e internacional interactúan significativamente entre sí para explicar el alcance tecnológico de las innovaciones de la empresa.

Resumo

Nós exploramos como conexões baseadas no conhecimento com locais domésticos e estrangeiros afetam o escopo tecnológico das inovações da firma. Inspirado por uma mistura de perspectivas de Geografia Econômica e Negócios Internacionais, nós propomos um modelo teórico que distingue entre diferenças domésticas subnacionais e a heterogeneidade espacial transnacional. Além disso, nós combinamos a visão de Penrose de capacidades gerenciais com a teoria baseada na atenção da empresa. Analisando uma amostra de empresas com sede nos EUA entre 1990 e 2006, mostramos que tanto a conexão de conhecimento doméstico quanto a internacional afetam o escopo tecnológico das inovações da firma, mas seus efeitos são distintos. A amplitude da conexão do conhecimento internacional aparenta ser positivamente associada com o escopo tecnológico das inovações da firma. No entanto, a amplitude da conexão de conhecimento doméstico contribui positivamente para o escopo tecnológico das inovações da firma até certo ponto, depois do qual a racionalidade limitada de gerentes restringe a capacidade das empresas de se beneficiar da heterogeneidade subnacional. Assim, é mais provável que a busca doméstica seja desafiada pela limitada largura de banda gerencial. Por fim, a conexão de conhecimento doméstico e internacional interagem significativamente entre si para explicar o escopo tecnológico das inovações da firma.

概要

我们探索基于知识的国内外位置的联系如何影响公司创新的技术范围。受经济地理学与国际商务观点融合的 启发, 我们提出了一个理论框架, 将国内次国家差 异与跨国空间异质性区分开。进一步, 我们将彭罗斯 管理能力观与基于注意力的公司理论相结 合。通过分析1990至2006年间的美国公司的样本, 我们显示, 国内和国际的知识联系对公司创新的技术范围产生影 响, 但效果不同。国 际知识联系的广度似乎与公司创新的技术范围正相关。然而, 国内知识联系的广 度有助于公司创新的技术范围到一定程度, 在这之后管理者的有限理性限 制了公司进一步发挥次国家 级异质性的能力。因此, 国内搜索更有可能受到有限的管理带宽的挑战。 最后, 国内外知识联系明显相互影响, 从而解释公司创新的技术范围。

Appendix

Appendix A. Description of the Robustness Checks

We applied the IV procedure on Model 3 presented in Table 2 of the main text. We employed four instruments for our three allegedly endogenous variables (i.e., Breadth of Domestic Connectedness, Breadth of Domestic Connectedness Squared and Breadth of International Connectedness). Therefore, the application of the IV procedure consisted in four regressions jointly estimated. In the first three regressions, each potentially endogenous variable is the dependent variable in a regression which contains the instrumental variables and all other control variables. In the fourth regression, the dependent variable is Technological Scope and regressors are the estimated values of the potentially endogenous variables, along with all other control variables.

The first instrument employed is the knowledge endowment of the CBSA in which the firm is located, and it is computed as the number of granted USPTO patents measured at the CBSA level in year t − 1. To identify the CBSA in which the firm is located, we gathered information about the address of each firm from Compustat North America. Then, following previous studies (e.g., Agrawal et al., 2014), we retrieved all USPTO patents applied for (and subsequently granted) between 1989 and 2005, and featuring at least one US-based inventor. We used the ZIP codes of the patents’ inventors to identify their precise location (and corresponding CBSA). If a patent had at least one inventor from a particular CBSA, we assigned such patent to the relative CBSA. This measure intends to proxy the amount of technological knowledge available locally (i.e., within the firm’s CBSA) to feed the firm’s innovation processes, but it does not account for the technological composition of such knowledge and, in turn, the technological specialization/diversity of the CBSA. We expect that the local knowledge endowment is correlated with the firm ability and propensity to connect with other domestic or foreign locations (Perri et al., 2017), but it is not directly correlated with firm ability to produce innovations with higher (lower) technological scope as the instrument does not include information about the technological distribution of the locally produced knowledge. The second instrument is the number of USPTO patents measured at CBSA level in year t − 1 featuring multiple assignees, i.e., patents involving formal cross-organization collaborations. This variable is expected to be correlated with the firm’s ability and propensity to establish future knowledge-sourcing collaborations, but to have no direct influence on the technological scope of firm innovations (Fritsch & Lukas, 2001). The third instrument is the number of international flight departures from airports based within the US State in which the firm is located and computed in year t − 1, using the US International Air Passenger and Freight data published by the US Department of Transportation. This variable is expected to be correlated with the ability and propensity of the firm and its inventors to establish geographically dispersed knowledge-based connections by facilitating individuals to travel from a specific location (Ejermo & Karlsson, 2006), but to have no direct influence on the technological scope of firm innovations. Finally, we have also included the squared term of the number of flight departures.

Table A1 displays the results of the IV estimates, which are consistent with our main findings presented in Table 2 of the main text. The sample used for the IV estimates has fewer observations than the sample used for the main analysis (7409 instead of 7432) due to missing values related to the instrumental variables.

Table A1 Results of the two-stage least-squares within-group estimator (with heteroskedastic-robust standard errors)

Several key tests confirm the validity and goodness of the instruments. First, the F-tests of the first-stage regressions reject the null hypothesis that the exclusion restrictions are jointly null (p < 0.001), reassuring the reader on the goodness of the selected instruments. Second, we can reject the null hypothesis that the system of equations is under-identified (p < 0.001), concluding that the model is well identified. Third, the Hansen test does not reject the validity of our instruments (p > 0.4), proving that they are not correlated with the error term and that the excluded instruments are allegedly excluded from the main equation.

We also checked the sensitivity of our results to firm size. We split the sample by firm size and tested our full model specification on two subsamples, using as sample-splitting criterion the threshold of the 75th percentile of the Firm Size distribution. By doing so, we separated the subsample of the small and medium enterprises (SMEs) with fewer than 500 employees (in line with the definition provided by the Statistics of US Businesses, SUSB), from the subsample of large firms. The subsample analysis provided results that are very much consistent with the main analysis, showing similar effects as regards to the relationship between the independent and the dependent variables. Second, we tested our full model specification on a reduced sample that comprises only observations with values of Firm Size within the 5th and the 95th percentiles of the distribution, in order to exclude very large and very small companies from our sample. We also repeated the same test using as exclusion criteria the 2nd and the 98th percentiles of the variable distribution. These additional tests confirmed the results obtained in our main findings (results are available upon request from the authors).