The 1999 reform of the Korean National Pension Program, a funded and defined benefit plan, extended compulsory coverage. As the result of this reform, about nine and a half million people were newly covered; yet some did not participate in the program. By exploiting this situation, I evaluates the effect of social security on private savings. The difference-in-difference estimations results in Section 1 show that the effect of the National Pension program on private savings is negative but statistically insignificant. Another result is that the lower the household income is and the more a household is educated, the greater the effect of the program is. Then, in Section 2, I estimate a stochastic dynamic model in which households facing income and survival uncertainty choose optimal levels of consumption, asset holdings, and retirement. The parameters in the model are estimated by a simulated minimum distance estimator. Estimation results show that social security pensions reduce private assets by less than 10 percent. Bequest and precautionary savings motives are the main reasons of this partial offset, and the inducement effect through retirement decision is very small. In spite of the progressive structure of the pension program, low income households are affected less than high income households because of the very low expectation of pension benefits. The introduction of the pension program improves welfare. The Section 3 extends the two-stage estimation method of Hotz and Miller (1993) for discrete choice dynamic models to include unobserved individual heterogeneity. Under the assumption of finite mixture, I suggest an modified Expectation-Maximization (EM) algorithm that involves nonparametric first stage estimation.