4. Empirical Strategy
In this section, we present our identification strategy. One approach to studying the effects of legislative
malapportionment on tax sharing is to rely on a cross-section analysis, i.e., run a regression between an
index of the legislative representation of a given district and an outcome variable related to the transfers received by that district. There are, however, two main issues with this approach. First and foremost, the
existence of unobservable variables that may have an impact on the distribution of tax revenues results
in a potential omitted-variable bias. These variables could be, for example, qualitative characteristics of
the relationship between some provinces and the national government that could make some districts
more or less likely to be overrepresented in the legislature and to receive larger or smaller federal
transfers. Party affiliation –a regressor usually included in the cross-section regressions- is just one of the
possible qualitative dimensions involved. Other aspects of political affinity may not be quantifiable at all.
In addition, even if it were possible to include proxy regressors for all the relevant unobservable
variables, the existence of a finite number of observations –in a cross-section analysis of Argentina at
province level there cannot be more than 24- clearly limits the statistical feasibility of such an analysis.
A possible solution for the problems posed by cross-section studies would be to extend the analysis over
time by creating a panel database in which each province is followed over several years. An example of
this approach can be found in the work done by Pitlik, Schneider and Strotman (2006) on Germany’s
intergovernmental transfer system. To the extent that the omitted variables are time-invariant, a fixedeffect
analysis using panel data solves the endogeneity problem. However, this is not always the case:
whenever an unobservable variable changes over time, the endogeneity problem persists. For instance, if
the relationship between political elites in a province and the national government –a variable which is
very difficult to quantify- changes over time, this could independently affect both apportionment in
Congress and the share of resources channeled to that province. To control for this, an instrumental
variable approach needs to be adopted. The potentially endogenous regressor –in our case, the index of
legislative representation- has to be instrumented by an exogenous variable.
We use the last approach and exploit exogenous variations in our index of legislative representation.
First, during the period under study, there were several changes in the minimum number of seats per
district, which produced exogenous variations in the legislative representation of the provinces. Second,
new provinces were established, were given seats in the Chamber of Deputies and were incorporated into
the tax sharing system, and this induced exogenous variations in the legislative representation of the
original provinces