William Davidson Institute Working Paper 487
inflation in accession countries. However, research has shown that this effect should not
be overestimated"22.
Practically there are some preliminary steps that we will take for the estimation of
the BS effect. The methodology we use is very similar to the one approved by Halpern
and Wyplosz, 2001.
Step 1. The economic branches are grouped into two sectors: tradable and non-
tradable. Productivity indices are formed for the tradable and non-tradable sectors in
Bulgaria23. There is space for a third sector of semi-tradable goods. For instance De
Broek and Slok (2001) outlined agriculture as an independent sector since it is known to
be a “mixture of tradable and non-tradable activities” in the accession countries. Despite
the great number of methodological studies (De Gregorio and al., 1994), there are no
clear quality criteria for differentiating between tradable and non-tradable sectors so far.
Step 2. Productivity dynamics in both sectors is examined; we check which one is
higher and whether they tend to get close to each other. According to the BS effect,
productivity in the non-tradable sector is expected to be lower and/or to increase
considerably slower.
Step 3. The process of wage convergence in both sectors is studied as a
prerequisite for the presence of the BS effect.
Step 4. Price indices for the sectors are found and their movements are observed24.
Step 5. After analyzing the prerequisites of the BS effect existence, we proceed
with the new direct estimation based on the following regression (1) (see for details De
Gregorio and al., 1994).
22 Speech by Dr. Willem F.Duisenberg, President of the European Central Bank, at the Frankfurt
European Banking congress, at the Alte Oper in Frankurt am Main, 23 November 2001.
23 It is possible to estimate the BS effect in a two - country model (two zones). In the case of Bulgaria
and in the context of convergence it is more appropriate to make a comparison with the EU (or Germany).
See. ECB (1999).
24 Productivity dynamics of prices and wages in both sectors can be estimated by different econometric
techniques (simple regression, cointegration, VAR model) or can be described by common statistical
features of the series (correlation between levels and rates).
14