One of the most interesting things in the electoral field is try to quantify the quota of voters who change party preference from one election to another. In order to do that, it is possible to carry out a survey (by asking which party the interviewed voted in the present and in the former elections) or to employ ecological data. Both methods have advantages and disadvantages: ecological data, besides being less expensive (sometimes free), consent to avoid the problems related to the survey methods (memory problems, sample coverage etc); survey methods allow to avoid problems related to ecological fallacy (restrictive rules for applying ecological regressions). In order to use the electoral ecological data, several methods have been built. The most employed ones are the Goodman Model (1953, 1959), the King model (1997) and the Brown and Payne model (1986). The first one is more suitable for multiparty systems, and the other ones for two-party systems. This paper aims to apply the three ecological inference models to the same data, and compare the estimates obtained by the three models with the estimates obtained by survey data. For this purpose, the data (both the survey and the ecological ones) concerning the 2006 and 2008 Italian Parliamentary elections will be employed. For the ecological analysis data of the 2006 and 2008 elections at the polling station level (in Italy there are about 60.000 polling stations) will be used. It is important to note that it is the first time these models will be tested on the wholeness of votes. The survey data are provided by ITANES (Italian National Election Studies) and consist in a large representative sample (3000 interviews). Comparisons of the results obtained will be presented: the focus will be on the differences found, discussing the weakness and strength of each method used. The ecological data to be employed are not a sample, but all the votes of both the considered election (2006 and 2008) at the polling station level (in Italy there are about 60.000 polling stations). For our knowledge, it is the very first time that these models will be tested on the wholeness of votes. In order to test the King and the Brown and Paine models, even if Italy is traditionally a multiparty system, we will spare the parties in two main coalitions.
