MRW conclude that human capital must be an omitted variable.
The author finds that both mean and variance heterogeneity are statistically significant issues in MRW's work, while only variance heterogeneity is an issue for Islam.
The existence of mean heterogeneity in MRW's models supports Islam's argument.
Section two lays out the theoretical and econometric models used by MRW and Islam, while section three defines the data sets used to test each author's statistical specifications.
In practice, MRW use income per capita when performing the estimation.
In contrast to MRW, Islam converts income per unit of effective labor into income per capita prior to estimating his equation.
Like MRW, g + [delta] is restricted to equal 0.05 for all countries.
In both MRW and Islam, there are three samples: a cross-section of 22 OECD countries, a cross-section of 75 "intermediate" countries with populations of less than one million in 1960, and a cross-section of all countries that had data available when their papers were written, excluding oil producing countries (non-oil).
The data set used to test MRW's results is a cross-section of averages spanning the years 1960 to 1985 and is taken directly from the appendix of their paper.