The issue of inequality received much attention in the economic literature in the last two decades, motivated by the recognition that inequality is not only an outcome of growth but also a determinant of growth. The increased availability of survey data has led to numerous empirical studies of inequality. A strand of this literature deals with the impact on inequality of the diversification of income sources by farm households. As farmers worldwide are gradually shifting to part-time off-farm work, it is an empirical question whether this raises or lowers income inequality among them. It could lower inequality if off-farm income compensates for low farm income, but it could raise inequality if the more qualified farmers earn more off the farm. For policy purposes, it is not sufficient to know whether off-farm income increases or decreases inequality, because we need to know the impact on inequality of various determinants of income, which are sensitive to policy measures. These include education, landholdings, etc. The purpose of this research is to offer an empirical framework for identifying determinants of income inequality in societies in which significant fractions of households have multiple income sources. This is relevant for many low- and middle-income countries, and in particular for agricultural societies. The usefulness of this framework will be demonstrated using data on farm households in Korea. Korean agriculture has experienced structural transformations similar to many other developed economies. In particular, rural-to-urban migration resulted in a massive loss of rural population, which declined from about a third in 1970 to less than 10% in 2000. Among farm households, about a third are engaged in off-farm labor activities, and the fraction of off-farm income is increasing. Farm size disparities are also increasing over the years. The level of income inequality among farm households is larger now than among all Korean households. In this paper, two major empirical methods are proposed to study the sources and determinants of income inequality among Korean farm households. The first is inequality decomposition by income sources, which provides evidence on the relative importance of various income sources to inequality, and allows the computation of the impact on inequality of a uniform increase in income from each source. It was found that farm business income, which accounts for 42% of total income, is responsible for 58% of total income inequality. On the other hand, off-farm labor income, which accounts for 20% of total income, is responsible for only 12% of inequality. This implies that off-farm labor income is an inequality-decreasing income source. In particular, marginal effect computations show that a one-percent uniform increase in off-farm income is expected to reduce the Gini coefficient by 0.033%. The second method is regression-based inequality decomposition, which allows the quantification of the contributions of various factors, which affect household income, to income inequality. In particular, a linear regression of income on these factors is estimated, and each factor multiplied by its estimated coefficient is treated as an “income source”, and then the method of inequality decomposition by income source can be applied. It was found that education and landholdings combined explain about a third of the explained part of income inequality. Moreover, a uniform one-percent increase in education is expected to reduce the Gini coefficient by 2.5%, while a uniform one-percent increase in landholdings is expected to reduce the Gini by 0.02%. These results provide a solid basis for policy implications. Another set of variables that was found to contribute to inequality is family size and composition. For example, a uniform one-percent increase in family size is expected to increase the Gini coefficient by more than 4%. We extend the above two decomposition techniques by breaking down the inequality contributions of the income determinants by income source. We do this by estimating a separate regression for each income source, corrected for selectivity, and then repeating the above analysis. It was found that family size and land ownership contribute to income inequality mostly through farm income, while education contributes mostly through non-farm labor income. The conclusion is that the process of farmland concentration and farm size polarization is likely to increase income inequality within the farm sector in Korea. However, at least part of this effect can be mitigated by extending education in the rural population in an equitable manner, so that those who lose due to the structural changes in the farm sector could benefit from the off-farm income opportunities.