spatial analysis of start-up health and pharmaceutical firm survival

The presence of knowledge spillovers and shared human capital is at the heart of the Marhall-Arrow-Romer externalities (MAR) hypothesis (see Marshall, 1890, Arrow, 1962, Romer, 1986) which represented the fount of a flood of scientific contributions produced in the last decades in the field of firm formation, agglomeration, growth and survival. According to the MAR hypothesis, similar firms located close by increase the chance of human interaction, labor mobility and knowledge exchange which in turn has an effect on firm creation, development and survival. Most of the earlier empirical contributions on knowledge externalities, mainly due to data limitation, considered data aggregated at a regional level leading to contrasting empirical results (Mansfield, 1995, Henderson, 2003; Rosenthal and Strange, 2003). In particular, the role of agglomeration economies has been considered to explain firm entry at a regional level and its effects on the growth of regional employment and regional production (Glaeser et al., 1992; Henderson, 2003). In most of this literature we can identify two major shortcomings. The first pertains the statistical sphere and concerns the way in which the agglomeration theories are empirically tested, the second concerns the main focus of these studies. More specifically the first limitation refers to the fact that, with data aggregated at a regional level, conclusions are based on the arbitrary definition of jurisdictional spatial units so that aggregating them in a different way we can obtain different (and often, contrasting) evidences: this is the essence of the so-called Modifiable Areal Unit Problem (Arbia, 1989). Furthermore theoretical models of new firm formation and firm exit are generally grounded on the behavior of the individual economic agent (see Hopenhayn, 1992; Krueger, 2003; Lazear, 2005) and they can be tested empirically on regional aggregates only under the unrealistic assumption of a homogeneous firm behavior within the region. The second limitation is constituted by the fact that, somewhat surprisingly, while concentrating on the effects of agglomeration on firm creation and growth, the literature has, conversely, largely ignored its effects on firm survival. These are the issues our paper seeks to contribute. The search for a statistical solution to the MAUP has lead, in recent years, to an increasing interest towards the use of micro-data and establishment-level information thus eradicating at its deep roots the problem of defining a-priori the level of geographical partition and thus providing more robust evidences. This approach allows the use firm-level variability in spatial concentration in order to test if industrial localization and concentration influences firm demography and, also, to reconcile the contrasting empirical evidences found at an aggregated level. In this respect, a series of recent papers that introduced the use of point pattern analysis methods (generated mainly in the ecological and epidemiological literature, see e. g. Ripley, 1977; Diggle, 2003) are of tremendous potential impact in this area in order to measure geographical patterns of industries and their effect on firm creation, growth and survival (e. g. Arbia et al., 2010; 2012; 2013; Marcon and Puech, 2010). The use of establishment-level data, however, until now, has been traditionally concentrated on firm creation (see Helfat and Lieberman, 2002 for a review) while only few papers are devoted to study industrial localization effects on firm exit, survival and bankruptcy. Some remarkable examples are those reported in Staber (2001), Folta et al. (2006), Sahver and Flyer (2000) and De Silva and McComb (2012). The present paper aims at contributing to the existing literature by answering to some of the open methodological questions reconciling the literature of Cox proportional hazard with that on point pattern and thus capturing the true nature of spatial information. In particular our interest is in modeling the effects of spatial concentration and interaction on the probability of firm survival by incorporating both geographical variables and spatial interaction effects. We present a methodological advance with respect to the current literature (e. g. De Silva and McComb, 2012) in that we suggest to model the probabilities of firm exit of individual firms (with explicit consideration of their location), with a Cox proportional hazards model with micro-founded spatial covariates which takes into account both the spatial interactions among firms and the potential effects deriving from agglomeration. We also present some empirical results based on a recently released database on Italian firm demography managed by the Italian National Institute of Statistics (ISTAT), created in accordance with the procedures suggested by OECD and Eurostat. This database overcomes a series of inaccuracies due to non-demographic events (such as changes of activity, mergers, break-ups, split-off, take-over and restructuring) and to obtain a more realistic picture of firm demography with respect to that obtained examining traditional microdata drawn from Business Register. In view of achieving our aim the present paper is divided into 3 more sections. Section 2 will be devoted to a brief summary of the state-of-the-art empirical methods to analyze survival data. Section 3 presents the methodology and shows the results of an empirical application to the case of start-up firms in the health and pharmaceutical sector during the years 2004-20008 in Italy. Finally Section 4 contains some concluding remark