45th Congress of the European Regional Science Association: "Land Use and Water Management in a Sustainable Network Society", 23-27 August 2005, Amsterdam, The Netherlands
Shift-Share analysis is a well-known methodology frequently used to obtain insights into the determinants of regional growth processes. It can address many issues, such as output growth, employment growth and productivity growth. After the initial equation proposed by Dunn (1960), several extensions have been suggested in order to overcome some conceptual problems. One of the most important undesirable properties that have been mentioned is the so-called "non-uniqueness” of the results. That is, numerous decomposition forms are equivalent to the classical shift-share equation from a theoretical point of view, but the results often depend strongly on the choice of a specific one. In this paper, we propose a methodology based on maximum entropy econometrics to incorporate additional information to select the unique shift-share formula that fits this information best. We illustrate the method empirically by investigating the sources of change of employment growth in Spanish regions, 1986-2000.