Since the 1970s wage inequality has been growing in the United States, yet another measure of inequality, the difference between women's and men's mean wages, has been declining. Some argue that the gender wage gap would have decreased even more, had overall wage inequality not grown. According to these researchers, the increasing dispersion of wages pushed women's mean wage further away from men's, so women had to swim upstream to reduce the gender wage gap. This reasoning makes intuitive sense: as wage inequality increases, the disadvantage of those who earn below the average wage worsens, and the gain of those who earn above the average increases. Given that the proportion of women who earn below the overall mean wage is greater than that of men, when wages become more dispersed, women's mean wage should fall further behind that of men.

However, the female wage dispersion is different from the male one, and has undergone a different transformation, as men and women operate in different labor markets. Relatively low-skilled men suffered the biggest decline in wages during the 1970s and 1980s, and as their wages fell, wage inequality among men increased. As growing wage inequality among men meant lower male wages, it led to a narrowing of the gender wage gap, so women did not have to swim against a current. Since the 1990s, however, the wages of low-skilled men stagnated, and the highest male wages grew even higher, so the gender wage conversion slowed down, because women's wages had to catch up with a moving target.

My dissertation will make an important contribution by offering an explanation for the slowdown in gender convergence. It also offers an alternative solution to a methodological problem. The statistical method currently used to calculate the effect of inequality on the gender pay gap assumes that there is only one wage structure, and miscalculates the relationship between wage structure and gender pay gap. This dissertation introduces a new method, which takes into account gender differences in wage distribution.