Asset return correlation and default probability are key factors for determining the regulatory capital requirement in credit risk model of New Basel Accord. In this paper, we apply the nonparametric estimate smoother to estimate default probability using distance to default. The asset correlation is calculated using the estimated default probability first and then extracting the implied asset return correlation from the joint default probability. Using data of American listed companies between 2000 and 2003, we empirically examine the relationship among asset return correlation, default probability, and firm asset size. Furthermore, we also compare the implied asset return correlation in this study with the correlation calculated from the definition in New Basel Accord. The empirical results indicate that the relationship between asset return correlation and default probability is negative, and asset return correlation is an increasing function of firm size. This finding is consistent with the result of quantitative impact study in New Basel Accord. However, the implied asset return correlation from this paper are much smaller than those calculated according to New Basel Accord. It shows that New Basel Accord tends to overestimate asset return correlation. It also suggests that the variance of default probability should be considered as a factor for calculating asset return correlation.