This dissertation investigates the effects of innovation in the U.S. metropolitan statistical areas (MSA) on survival chances of standalone non-patenting firms. Extensive literature argues that knowledge spillovers are likely to be present in agglomerated regions with greater accumulated stock of knowledge. According to this view, firms exposed to knowledge spillovers should become more innovative and productive. Empirical research consistently finds a negative relationship between innovation and productivity, on the one hand, and the probability of exit, on the other. An alternative argument, going back to Schumpeterian `creative destruction', contends that greater innovation leads to increased competition and forces less innovative firms to leave the market. The dissertation tests these contesting hypotheses by estimating the hazard rates faced by firms in two high-technology sectors, namely computer and electronic product manufacturing and healthcare services.In the two sectors, all firms established within the continental U.S. MSAs in 1991 are identified and traced until a firm exits or year 2008, whichever happens first. Firms that are non-independent, have at least one patent, or exit via merger or acqusition are not included in the analysis. Two variables approximate the level of innovation in a metropolitan area, the population-adjusted number of patents, and the population-adjusted number of patents in technological classes related to the sectors of interest. Results of the non-parametric analysis suggest that in more innovative MSAs, computer and electronic product manufacturing firms tend to exit sooner, while healthcare firms enjoy greater likelihood of survival. In most dense metropolitan areas, however, relationship between innovation and firm survival becomes positive for manufacturing firms, and switches from positive to negative after about 12 years of operation for healthcare firms.After controlling for a number of regional and firm-level characteristics in the semi-parametric and parametric analyses, this study implies that innovation in a MSA is either not significant, or forces non-patenting firms out of business in high-technology manufacturing, while it is not a significant predictor of firm longevity in healthcare. These results stay the same in the separate analyses by density groups or NAICS4 industries. In general, the dissertation presents support for the Schumpeterian hypothesis, while the knowledge spillover view is not supported.