Industrial clusters are geographical concentrations of related industries. They foster innovation, job creation and business formation. Previous studies find that firms in clusters on average are more innovative than firms outside. They interpret this as evidence that clusters encourage firms to innovate. This interpretation is misleading because two different mechanisms can lead to the same result. On the one hand, firms in clusters improve innovativeness through knowledge spillovers and network building. On the other, less innovative firms are forced out of clusters by tough competition. Most studies fail to differentiate these two mechanisms. I separate these mechanisms and examine their variations across industries and establishments. I also search for the optimal spatial scale of industrial clusters to maximize their effect on innovation.

In this dissertation, I match establishment data with patent data for the state of Maryland from 2004 to 2013. I improve the methodology of quantifying the causal relationship between clusters and innovation, and apply this method to employment centers. Employment centers on average encourage establishments to file for 8% to 11% more patents. This effect is maximized within a one- to two-mile radius region. I also compare how much clusters encourage innovation across different industries, and find significant heterogeneity. In Metalworking Technology, the effect of clusters peaks at a three-mile radius region and increases patent applications by 18%. In contrast, in Business Services, the effect is essentially zero, even when it is maximized in a one-mile radius region. These differences can be explained by industrial characteristics, such as the different level of reliance on tacit knowledge. Finally, I examine how industrial clusters shape the originality of small versus large establishments. I find that small in-cluster establishments improve innovation numerically more than large establishments, but their differences are statistically insignificant.

This dissertation can provide guidance to the design of industrial policies. It helps to more precisely evaluate the benefit of cluster policies. Policymakers can also implement cluster policies targeting at the most beneficiary industries and the optimal spatial scales.