Economic inequality is of interest not only at some intrinsic level, but also for its close connections to diverse variables, ranging from economic indicators such as growth rates to sociopolitical outcomes such as collective action and conflict. It is only natural, then, to study the evolution of inequality in an economic system. It is fair to say that the dominant view on this topic is that inequality is the outcome of a constant battle between convergence and “luck” (Gary Becker and Nigel Tomes, 1979). Current asset inequalities may echo into the future, but their natural tendency is to die out (owing to a convex investment technology). Disparities are only sustained through ongoing stochastic shocks (see also David Champernowne, 1953; Glenn Loury, 1981). A second approach emphasizes that initial conditions determine final outcomes, owing principally to a nonconvex investment technology (see e.g., Mukul Majumdar and Tapan Mitra, 1982; Abhijit Banerjee and Andrew Newman, 1993; Oded Galor and Joseph Zeira, 1993; Ray and Peter Streufert, 1993). Historical disparities may persist and magnify, if such differences straddle some bifurcation threshold. Inequality is not inevitable in this view: stable steady states with inequality coexist with others that involve perfect equality. In this paper, we examine a third view which emphasizes an intrinsic tendency of the market mechanism to create inequality. In this approach, economic inequality appears as an inevitable outcome, even if all agents are identical to begin with and even if there is no uncertainty or technological nonconvexity. This view has received attention in Ray (1990), Debasis Bandyopadhyay (1993), Lars Ljungqvist (1993), Scott Freeman (1996), Kiminori Matsuyama (2001), and Mookherjee and Ray (2001). The basic argument is as follows. Suppose that an economy is populated by several dynasties, each of which allocates resources to current consumption and bequests to descendants. Assume for the moment that the latter consists entirely of educational expenditures preparing children for their chosen professions. Of course, the returns to such professions are endogenous; they depend on the distribution of individuals across different professions. Now, if several professional categories are necessary, wages must adjust so as to force separation in choices even if all individuals are ex ante identical. To be sure, this “broken symmetry” (Matsuyama, 2002) has no payoff implications for the generation alive today; identical individuals must receive identical payoffs. Starting with the next generation, however, there must be inequality (not just in wages, but in payoffs). There must be individuals in lowpaying professions that involve low training costs, whose parents invested little; and there must be others in high-paying high-trainingcost professions whose parents invested a lot. Once such inequality sets in, it may well magnify. Richer offspring will find it easier to invest in better-paying professions for their children. Two ingredients are used in this argument. First, credit markets must be missing (or imperfect). Otherwise the necessary finances may be borrowed, and wages net of costs must be equalized over all professions. However, the assumption that parents cannot borrow against their children’s future earnings seems innocuous enough. Second, parents cannot make financial bequests to supplement or substitute for educational expenditures. Such bequests could conceivably neutralize earnings inequality. This motivates the question: When might financial bequests compensate for earnings inequality in steady state? Moreover, starting from perfect equality, will equality be preserved thereafter, or will inequality emerge and persist? We discuss an extended example from an ongoing research project. * Mookherjee: Department of Economics, Boston University, 270 Bay State Road, Boston, MA 02215; Ray: Department of Economics, New York University, 269 Mercer Street, New York, NY 10003, and Instituto de Analisis Economico (CSIC), Barcelona, Spain. We thank Kiminori Matsuyama, who encouraged us to write this paper.