Gianni Marliani, Ph.D. and Laura Grassini, dr. Department of Statistics, University of Florence, Viale Morgagni, 59, Firenze, 50134, Italy
Performance of Italian Firms: an Analysis of 1998-2004 Panel Data Using Hybrid Log-linear Models
Abstract
The Italian economy is going through a critical period. Italian firms seem to have difficulty in keeping up with competitors, mainly because their productivity (particularly labor productivity) is not increasing.
This general economic trend is not equally shared by the different economic sectors that exhibit different performances.
The aim of this paper is to analyze the performance, for the years 1998 and 2004, of a panel of firms belonging to different manufacturing sectors. The panel data are obtained from the surveys carried out by Istat to provide Eurostat with the Structural Business Statistics (SBS).
The study applies the hybrid log-linear models to transition tables. For each economic activity, a transition table was built by classifying single firms according to the values of a performance index referred to the two different time periods. Two performance measures were considered: “value added/employment” and “value added/labor cost”. Hence, the transition table displays the number of firms that: (i) did not change their initial position (main diagonal); (ii) increased their performance (above the main diagonal); (iii) decreased their performance (below the main diagonal).
Hybrid log-linear models are used to identify specific structural patterns (inertia, symmetry, quasi-symmetry, etc.) that are defined by specific hypotheses on the model parameters. Hybrid log-linear models are particularly suitable to study transition tables, since they have special effect-parameters that can impose particular restrictions on the interaction parameters of a saturated log-linear model. This analysis is expected to provide typical profiles of the different economic sectors.
Keywords: Labor Productivity, Panel data, Log-linear hybrid models.
JEL Classification: D21; L60