real output when productivity rises or when their unit costs decline. We consider two alternative
measures of cost changes. The first one is the variation of total long run unit costs, Δln(C∕PQQ).
It is however suspect of endogeneity, except in the limiting case of constant returns to scale. The
second measure is much less prone to these problems, as it uses only the exogenous
component of short run unit cost growth, labor and intermediate inputs,
SIΔln(PI /Pq) + SL Δln(PL /Pq). Regarding Ut-1 , we expect that firms already operating at
relatively higher capacity levels in 1993 faced tighter constraints when trying to increase output.
We also allow for the possibility that export oriented firms may respond differently to the
new economic environment. We do this in two different ways. First, we control for the export
status of the firm in the previous year by including an export dummy variable Second, we assess
whether exporting firms respond more flexibly to changes in relative prices, by interacting the
export dummy and the cost variables.
A key variable on our analysis is productivity. Recall however that our productivity
measure also embodies scale effects. It is bound therefore to be endogenous. To control for
such effect, we also estimate a simple productivity equation:
(6) Δ lnA = α + βFt _1 + σ[∆ l∏(Q )] + εi
where as a dependent variable (ΔlnA), we use again the measure of scale∕productivity
growth computed above. The first two explanatory variables are lagged and represent
characteristics of individual firms. F is a vector of characteristics of the firm at the beginning of
the period considered that includes the export dummy, the share of imported inputs, the share of
foreign capital, and the employment share of expatriates. The inclusion of real output growth,
Δln(Q), is meant to capture the scale component of our productivity measure. Finally, e is an
error term with standard properties.
14