Abstract:     In this paper we model earnings as a function of observed and unobserved skills and their prices. Observed skills are acquired through education and work experience whereas unobserved skills are related to characteristics such as ability. Previous studies have focused on rising skill prices as the main cause for increasing earnings inequality. The shift in the demand for skills and hence higher skill prices could be driven by skill-biased technology changes (Bound and Johnson, 1992; Berman, Bound and Griliches, 1994), changes in industrial structure (Katz and Murphy, 1992; Murphy and Welch, 2001) or globalization (Wood, 1998).     Besides observed skill prices, widening dispersion of observed skills across the earnings distribution will increase earnings inequality. There have been few studies on the cause and the mechanism of the dispersion of skills across the earnings distribution. In this paper we posit that three interacting factors as causes. First, skill-intensive industries expand as the economy restructures towards high value-added production. Second, firm job levels are hierarchical (Baker, Gibbs and Holmstrom, 1994) with higher level jobs demanding more skills. Third, the education system, in particular higher education, expands to increase the supply of skills for the skill-intensive industries. Given sufficient mobility in both the external and internal labor markets, over time there will be an increasing aggregation of skills among workers of higher earnings, causing a widening dispersion of skills between the top and bottom deciles of the earnings distribution.     The objective of this paper is to test the importance of skill dispersion as a  cause of rising earnings inequality by decomposing the increase in log earnings differential between the 90th percentile and the 10th percentile into three components due to rising observed skill prices, increasing dispersion of skills, and a rise in the prices and dispersion of unobserved skills. Using Hong Kong census data, we show that the 90-10 percentile log earnings differential increases by 50.3 percentage points from 1981 to 2001. Over half of this increase can be attributed to an increasing dispersion of the observed dimensions of skills, such as education and experience.  Hence the most important factor in the rapid increase in earnings inequality in Hong Kong is the widening dispersion of observed skills, not changes in the wage structure in observed and unobserved skill prices. This is very different from the experience of the U.S.. Juhn, Murphy and Pierce (1995) estimate that less than 10% of the increase in earnings inequality from 1964 to 1988 in the U.S. is due to an increasing dispersion of observed skills.     During the 1980s and 1990s there was a rapid expansion in higher education which increased accessibility in Hong Kong. The economy was restructuring from labor-intensive manufacturing to high value-added service and the labor market was flexible. These factors lead to a widening dispersion of education skills between the top and the bottom ends of the earnings distribution. Under these circumstances, instead of being the great social equalizer, higher education expansion has actually caused greater earnings inequality.