The effect of openness on income inequality sometimes arouses emotion and blood
pressure as well as academic curiosity. There exists an active empirical literature on economic
growth and income inequality, and a related and equally active literature on openness and
economic growth. Most of the papers in these two literatures employ cross-country regressions.
Prominent examples include Forbes (2000), Dollar and Kraay (2001a and 2001b), Edwards
(1992, 1998), Sachs and Warner (1995), Frankel and Romer (1999), Rodriguez and Rodrik
(2000) and other papers cited therein. Useful insights have been gained from this literature.
However, analyses based on cross-country regressions have been criticized on two
grounds. The first type of problems has to do with data comparability across countries. This
issue is particularly acute for data on income inequality: the definitions and data collection
methods can be different across countries. As an illustration, for OECD countries, Atkinson and
Brandolini (2001) noted the pitfalls in making cross-country comparisons based on pooled data.
For a few countries where multiple measures of income distribution are available (households
versus individuals, income versus consumption, etc.), the different measures can give different,
sometimes contradictory, patterns even for the same time periods. Since the data that crosscountry
regressions have to rely on come from potentially different methodologies, they can
produce misleading results when pooled together. Atkinson and Brandolini noted further that ?in
cross-country analysis, use of a dummy variable adjustment for data differences is not
appropriate.? Atkinson and Brandolini?s specific criticism is on pooling data across OECD
countries. It is reasonable to assume that the data quality for developing countries is generally
inferior to the OECD countries. Therefore, running cross-country regressions involving data
from developing countries can only make the quality of inference worse.
Aside from the Atkinson-Brandolini criticism just mentioned, we should note another
potential source of data incomparability. The validity of comparing living standards across
countries depends on the validity of the so-called purchasing-power-parity adjustment, which in turn depends on the assumption that a common ?representative consumption basket? can be
meaningfully constructed for all countries. The last assumption cannot always be taken for
granted.
The second difficulty with cross-country studies has to do with the fact that differences in
cultures, legal systems, or other institutions other than openness may also be relevant for the
outcome variable under study (e.g., economic growth or income inequality). These factors are
difficult to be quantified and therefore to be controlled for in cross-country regressions.
Inclusion of fixed effects in panel regressions helps. However, the myriad of country-specific
institutions may also interact with the key regressor under investigation (e.g., openness) to affect
the outcome variable (e.g., income inequality). For example, in response to a terms-of-trade
shock, some countries would let the poor to fend for themselves, while others would have a
social safety net to moderate the negative impact on the poor but the exact size of the income
transfer may not be proportional to the size of the shock depending both on the nature of the
social safety net and on the size of the shock. In this case, the usual fixed effects are not
sufficient to control for the influence of the country-specific institutions.
In an influential paper, T.N. Srinivasan and Jagdish Bhagwati (1999) asserted that crosscountry
regressions are deficient and cannot be relied upon to understand the impact of
globalization on economic growth and presumably nor on income inequality (see the quote at the
beginning of the paper). One may not agree completely with the strong assertion by Srinivasan
and Bhagwati (1999). Nonetheless, given these criticisms, a careful study of cross-regional
experience within a single country can, at a minimum, provide a useful complement to the
literature based on cross-country regressions. Within a given country and over a relatively short
time period, culture, legal system or other institutions can more plausibly be held constant. So
the researchers? ability to isolate the effect of openness is enhanced. Furthermore, the
comparability of data definition and collection method is, in principle, also higher within a single country than across multiple countries.
This paper presents a case study of the impact of globalization on income disparity by
pooling two unique data sets on Chinese regions. There are five reasons that make China a good
case study. Some of them have to do with the fact that China is an important country per se. But, perhaps more importantly, other factors (or peculiar features of China) provide a methodological
advantage relative to typical cross-country studies.
