have poor economic policies in othe

have poor economic policies in other areas. For instance, many ex-communist
economies have high tariffs and many market interventions and inefficiencies as
well as having low GDP. Therefore, the strong correlation in Figure 9.6 might re-
ally reflect the influence of a third variable rather than trade liberalization itself.
To see the problems this causes econometrically consider the case where


GDP per capita


a



c


Financial Sector Reform
Openness
u


b


Capital Stock


Where a, b, and c are positive coefficients and u is a random variable that on av-
erage equals zero. Let us also assume that countries tend to do financial sector
reform at the same time as they reduce tariffs and increase openness so that Fi-
nancial Sector Reform and Openness are correlated:


Financial Sector Reform


d


Openness


If the econometrician mistakenly does not include Financial Sector Reform in
the regression and instead estimates


GDP


b


Capital Stock


f


Openness


then the estimate of f will be c


a


d. In other words, the econometrician will

overestimate the effect of openness as c

a

d rather than the true value, c. If c

were negative, i.e., trade liberalization decreases GDP, then this bias may even
be large enough to think that trade liberalization was beneficial when it was actu-
ally harmful.
This omitted variable bias is an important problem as estimates of the impact
of openness are sensitive to which other variables are included in the regression.

• Simultaneity

It may be the case that countries with high GDP have low tariffs.

If this is the case, then we will once again overestimate c and the impact of trade
liberalization. Why might countries with high GDP have low tariffs?
Poor countries tend to have weak systems for collecting taxes from labor in-
come and so have to rely more on tariffs.
As we saw in Chapter 8, trade leads to some sectors expanding and some de-
clining. Rich countries have more substantial welfare systems and so will
be able to support the losers from free trade more easily and so will lower
tariffs more. As a result, rich countries will have lower tariffs.

Both these problems can be overcome in our econometric estimation if we focus on a
component of openness which is not affected by GDP (although it may in turn influence
GDP) and is unlikely to be correlated with other policy variables. We can think of this
as the exogenous component of openness—the portion of a country’s openness that is
unaffected by policy variables. Because this component is unaffected by GDP and un-
correlated with policy variables, we will be able to correctly estimate how openness af-
fects GDP.
A natural way of estimating this exogenous component of openness is to focus on
the role of geography. Countries that are physically near to other countries with large
markets will have lower transport costs and so will do more trade than isolated coun-