To encourage tax compliance towards the Italian tax contributive system, the Italian Social Security Institute (INPS) develops a number of audits intothe Italian firms. The aim of these inspections is to detect possible evasions and to threaten cheating enterpreneurs with penalties, if necessary. In our case "to cheat" means to hide a part of the labor forces to the authority, underdeclaring their real dimensions and thus to evade a certain amount of social-insurance taxes. In this paper we particularly focus on showing how it is possible to use individual audit data to better understand the relation between inspections and tax compliance, and consequently the relation between the policy of auditing and undeclared work. A new source of data was built for this purpose, merging information about firms with individual audit data. Our analysis is developed as follow: after a brief introduction, we describe the dataset and we give some details on the procedures used by inspectors. Then we show a simple model of auditing in order to enlighten relation between audit policy and work force declaration. The second part of the analisys, mainly empirical, attempts to explain how to estimate undeclared work starting from our new source of data, after that, we assess two microeconometric policy evaluation analysis. Our aim is to nderstand the relation that lies between the policy of auditing and a) the propensity to declare workers; b) the number of black workers implied in the labour market.