Appendix A Discussion of propensity score matching methodologies A.1 Propensity score and matching methods in a binary settingIn the first step of our empirical analysis we aim at estimating the treatment effect τ i for the unit (municipality) i, defined as the difference of the outcome measured between the binary cases of participation Y i (1) and non-participation Y i (0) to the policy.The standard problem of the missing counterfactual is partially solved through the estimation of the propensity score, defined as the probability, for the unit i, of receiving the treatment D i , given the pre-treatment characteristics X i (Rosenbaum and Rubin, 1983):The balancing and unconfoundedness properties of the propensity score (for a formal proof see among others Cerulli et al., 2015;Dehejia and Wahba, 2002) allow us to use the estimated propensity score to quantify the policy effect proxied by the average treatment effect on treated:Given that the propensity score (A2) is a continuous measure that can assume any value between zero and one, we need to define some rules in order to match treated and control units after the estimation of the propensity score itself. A range of different metrics have been definedin literature in order to match treated to control units and compute the ATET, and no method is, ex-ante, better than the others. Since all these methods should asymptotically give the same results, the literature (see among others Caliendo and Kopeinig, 2008

Relational networks and intangible factors are crucial elements for the competitiveness of a territory. Public-Private-Partnerships (PPPs), in particular, allow for the provision of goods and services that favour the exploitation of complementarities between public and private resources. They aim at promoting an increase in the overall efficiency of investment projects through a complex mechanism that distributes risk and revenues among stakeholders. This paper examines the local and territorial determinants of PPPs through an econometric analysis based upon Italian municipal data, grouped at provincial level. Using a tobit model, we analyse the relationship between the realisation of successful PPP initiatives and different sets of factors, including less analysed local and territorial determinants. We stress the role of the local management of infrastructure assets, the administrative efficiency of local authorities and the diffusion of previous local development initiatives. Local management and territorial context factors explain most of the occurrence of successful PPP initiatives in the pre-crisis period while usual determinants (infrastructure endowment and financial distress) display a weaker effect.

No abstract