Short-run trends in united states patent activity

Abstract

We examine a newly created data series consisting of the monthly number of American patents granted since 1853. An initial examination divides the series into four time periods. An analysis of short-run cycles demonstrates that the same Box-Jenkins model is not applicable to all four periods. Differences in nineteenth and twentieth century model parameters may be a result of frequent bureaucratic reinterpretations of America's patent law during the last century, or-as many have claimed-may represent changes in the process of innovation itself over time. Our findings suggest that future researchers discriminate between two periods in their analyses. The first lasts until the late 1870's, during which time there was a very high variability in the number of patents issued. Since the late 1870's there has been a substantial decline in variability, and the amount of variance that can be explained by a simple Box-Jenkins model has increased. Still, not much variation can be explained using short-run cycles, and longer cycles appear to be both time period specific and highly unstable. The dynamics of American inventive activity are complex, and inventive activity appears to be largely driven by exogenous factors such as wars, economic conditions and changes in governmental policy, rather than by its own internal dynamics.