This paper applies the Inverse Hyperbolic Sine (IHS) transformation to explore the variables that determine a firm’s R&D collaborative expenditure. The IHS specification is used to overcome the inconsistencies deriving from non-normality of error terms which are typical in censored data. This represents a novelty in R&D studies. The sample employed refers to 1231 Italian firms where the dependent variable under investigation is strongly skewed by the zero values and by the extreme observations. The results show that standard errors are smaller in the IHS model than in the more common logarithmic one. The analysis also shows that size and public grants are effective in determining the level of cooperative R&D expenditure. Absorptive capacity, outsourcing inputs or services externally and the industry, also play an important role.