Identification and estimation of treatment effects in the presence of neighbourhood interactions

This paper presents a counter-factual model identifying Average Treatment Effects (ATEs) by Conditional Mean Independence when externality (or neighbourhood) effects are incorporated within the traditional potential outcome model. As such, it tries to generalize the usual approach, widely used in program evaluation, when SUTVA (i.e. Stable Unit Treatment Value Assumption) is relaxed. This new approach is applied to a locally-based policy. More specifically, we focus on the Garda lake area that is one of the 14 touristic districts in Trentino ? an Alpine province in north-east Italy. In the time window under scrutiny ? 2002-2006 ? Trentino had in place a subsidy policy for hotels ? 3-digit sector 55.1: hotels and similar accommodation, as defined in the NACE Rev.2 ? within the Provincial Law 6/99. There is no confounding effects coming from other policy measures given that Trentino hotels can only have access to subsidies related to Provincial Law 6/99. We rely on a database built relying on different sources and contains administrative information, structural characteristics of hotels and exhaustive information about the subsidies they received during the period. The sample size is 415 and consists of the Garda lake hotels active from 2002 to 2006.