Scaling invariant distributions of firms’ exit in OECD countries  Corrado Di Guilmi a Mauro Gallegati a Paul Ormerod b a Department of Economics, University of Ancona, Piaz.le Martelli 8, I-62100 Ancona, Italy b Volterra Consulting, 121 Mortlake High Street, London SW14 8SN, UK Abstract Self-similar models are largely used to describe the extinction’s rate of biological species. In this paper we analyse the extinction’s rate of firms in 8 OECD countries. Firms are classified by industrial sectors and sizes: we find that while a power law distribution with exponent close to 2 fits very well the extinction rate by sector, a Weibull distribution is more appropriate if one analyses firms’ size. 1. Introduction Power laws are definitely not new in economics. 1 Only recently scaling plot techniques have been applied to research on firms’ extinction rate, the main examples being [9] (henceforth, CO), who present evidence of power law scaling for the demise of US firm firms in the long run. In particular, CO show that the exit rate follows approximately a power law distribution with exponent close to 2. This value is very much in line with the literature on the Raup-Sepkoski’s kill curve according to which biological extinction events “can be reasonably well fitted to a power law with exponent between 1 and 3” ([10] p.165). This suggests that the mainstream economic model (i.e. the General Economic Equilibrium approach) may be incomplete or inadequate, since it assumes that the distribution of observations depends on the scale. 2 Here we aim at extending and controlling for robustness the CO findings. Differently from CO, who limit their empirical analysis to the US data, we use an OECD data set of eight industrialized countries. 3 Using a larger country data set allows us to get rid of the country-specific evidence which may affect the findings. Second, rather than taking an historical perspective by using data spanning well over one century we use a very short term perspective, because data are much reliable. The paper is organised as follows. In the second section we classify firms by industries obtaining a total of 5051 observations and treat them as a sum of daily uncorrelated events: their distribution follows a power law with exponent close to 2. In section 3 we classify firms by size: In this case  G.Giulioni and A.Palestrini provided very useful comments. 1 Topics include income distribution [1], returns on financial assets [2], firm sizes [3], city sizes [4], revenues in the motion picture industry [5] and business cycles [6-8]. 2 [11] emphasizes that scale-free behaviour are connected to agents’ interaction, while [12] points out the failure of neoclassical theory of firms in explaining at what level agents behaviour can be analysed autonomously from the market. 3 [13]: the countries are Denmark, Finland, France, Italy, Netherlands, Portugal, United Kingdom and United States.