The second test concentrates on the evolution of the budget and the business cycle in the course of
reforms (from period BR to period SA).
3.2 Features of pre-reform periods
In order to detect typical features of pre-reform periods the following equation is estimated51:
(1) Y = X β + γ Reform + ε
Here, Y is the scrutinized fiscal or economic variable and X is a complete matrix of country and
year dummies. Reform is a dummy equal to 1 for the two years before a reform event occurs
unless these periods are at the same time years following a reform event with lag one or two. The
purpose of this procedure is to limit the test to clear cases of pre-reform years that are not at the
same time affected by the consequences of recent reforms.52 In terms of the above suggested
stylized scheme of reform cycles this tests amounts to searching for significant differences
between period BR and S. Note that the time and country dummies neutralize year specific and
country specific factors. The impact, for example, of global business cycle trends is thus
neutralized.
50 European Commission (2005), for example, tests for differences between reform and no-reform periods. This test does not allow whether
resulting differences are related to changes occurring within a reform cycle or to differences between periods of stable institutions and (multi-
annual) periods of reform cycles.
51 This test has been suggested in a completely different context by Caplan (2002b).
52 This refinement is only relevant in the presence of a consecutive sequence of reforms with some reform pauses in between. In these cases, our
algorithm only treats the years prior to the start of the whole sequence as pre-reform years.
143