Passing the burden: corporate tax incidence in open economies

country-level omitted variables that may be correlated with the residual. All of my estimations
include random effects and cluster the standard errors at the country level. Using random effects
allows me to look at the effects between countries in addition to over time. Several countries in
my data include only one year of data making fixed effects an unattractive alternative to random
effects.

3.1 Luxembourg Income Study

The Luxembourg Income Study (LIS) is a panel database covering 30 countries over 5
waves of data from 1979 - 2002.⁷ The LIS compiles data from country specific household
income surveys. For example, individual data for the United States is taken from the Current
Population Survey. The dataset includes individual and household data on demographics,
expenditure, income, labor market outcomes and tax variables. This dataset is a rich source of
data that has been underutilized in economic literature.

The LIS database does, however, have some limitations. Not all of the variables are
available in all of the countries and years. Second, and more problematic, many variables are not
defined consistently across countries. For example, the United States’ education variable records
years of education, while Canada’s education variable records levels (i.e., graduated high school,
university degree ...) of education. The LIS does provide a recoding of the education variable for
many countries that helps circumvent this problem. Third, income variables are generally
available as net or gross income but not both.

The LIS education recode program groups each individual’s education into one of three
categories: low, middle or high. High education includes those with a college degree and above.
Middle education includes those with a high school degree but not a college degree. Low

11

<<   9   10   11   12   13   14   15   >>

More intriguing information