Vous êtes sur la page 1sur 1

Trade, Foreign Investment, and Industrial Policy for Developing Countries 4085

calculated by Dollar and actual tariffs or nontariff barriers. In fact, tariffs and nontariff
barriers enter with the wrong sign if this measure of openness is regressed on tariffs
and nontariff barriers. They then show that this measure is primarily capturing
exchange rate movements.
Both Harrison and Hanson (1999) and Rodriguez and Rodrik (1999) critique a
heavily used measure of openness created by Sachs and Warner (1995). The Sachs and
Warner measure has been updated by Wacziarg and Welch (2003), but the updated mea-
sure may suffer from the same shortcomings as the Sachs and Warner measure. Harrison
and Hanson show that the Sachs and Warner (1995) measure of OPENNESS does not
pick up differences in trade policy but instead reflects differences across countries in
exchange rate policies and political regimes. One way to reinterpret the evidence pre-
sented in Dollar (1992) and Sachs and Warner is that real exchange rate overvaluation
is bad for growth, a theme recently emphasized by Rodrik (2007). One obvious impli-
cation for researchers is that any study which measures the impact of real price distortions
on growth due to protection should also control for exchange rate movements.
(2) Endogeneity problems.
Endogeneity problems could arise for many reasons. Policy makers may prefer not
to open up to trade until firms are capable of competing on world markets, suggesting
that the causality runs from income to openness. Even statutory measures of trade pol-
icy (tariffs, quotas) are endogenously determined. The pattern of protection is likely to
be skewed toward protecting weak sectors, promising infant industries, special interests,
or vocal minorities.
While empirical work in the 1970s and 1980s largely ignored endogeneity pro-
blems, newer studies give much greater weight to constructing plausible identification
strategies. This progress is evident in Appendix Table 1, which lists prominent studies
on the linkages between openness and growth from the 1980s onward. Most of the
early studies had no identification strategy at all, as indicated in column (4). More
recent work addresses this omission, using one of two general approaches.
The first approach is to use granger-causality tests that exploit lags in studies that use
time-series datasets. As indicated in Appendix Table 1, these studies often find that cau-
sality runs in the reverse direction, from Y to OPENNESS: more successful economies
(or sectors) are more likely to open up to global competition. Related to this approach
is the use of lags as instruments, which depends on some strong assumptions about the
lack of correlation between the instruments and the error term.
The second general approach to identification has been to seek additional instru-
ments for OPENNESS. One path-breaking study along these lines is Frankel and
Romer (1999). Frankel and Romer use the insights from gravity models to derive an
instrument based on geographic proximity. Gravity models predict that countries closer
to each other trade more with each other. This means that distance can be used as an