Aim of the paper is to empirically test the relevance of the following statement: "high tech cluster are the future of industrial district".
Through an original database (based on census data from USA, UK, France and Italy) on the number of high tech firms and their employment at two geographical levels FLA and SLA (First and Second Level of Analysis) the different models are tested.
The first model identifies the most relevant locational factors for high tech firms (based on a cross section estimation of US States data).
The second model (based on the complete international data set) tests the relative strength of agglomeration versus scale economies in determining the geographical specialisation in high-tech industries and analyses a series of phenomena connected to the interaction between FLAs and SLAs (i.e. locational shadowing).
The third model (built on a cross section estimation of US States data and based on Glaeser, 1992) try to discriminate between three alternative interpretation (which can loosely be attributed to Arrow-Romer, Porter and Jacobs) of the growth of high-tech clusters in order to check similarities and differences with the "classical" definition of industrial districts