This paper examines the relationships among the site risk, the probability of having accidents, the accident rate itself and the probability of firm survival in construction industry. We first assess safety levels on site using CONSRAT (Forteza et al. 2016). We have examined during 6 years (2004-2009) 502 construction sites of 272 Spanish companies in Mallorca. We built a panel data with these safety assessments, the accident rates from Safety Authorities and the event of firm survival or mortality from SABI database from Bureu van Dijk. Our general hypotheses are that: 1) Risk on site has an effect on the probability of having and accident and also in the current accident rate of a company; and 2) that accident rate affects the probability of a company to survive in the sector. Using probit regression model with panel data, we explore which is the relationship among all those variables in order to test the hypotheses. The results can be useful to highlight the importance of defining policies to control accident rates effectively, since this may be determinant in the sustainability, competitiveness and growth of the construction sector in the economy of a region.