Subordination levels are of critical importance in the classic senior-subordinated structure for securitized financing (such as collateralized debt obligations and commercial mortgage-backed securities). Subordination levels determine the amount of credit support that the senior bonds (or tranches) require from the subordinated bonds (or tranches) and are provided by the rating agencies. Thus, ratings agencies play an important role in the pricing and risk management of structured finance products. The finance literature has numerous studies examining whether securities with higher risk (as predicted by asset pricing models, such as the CAPM) earn higher ex-post average returns. In a similar vein, it is of interest to examine whether securities (or tranches) with greater levels of subordination experience higher ex-post levels of delinquencies and default. In this paper, we examine whether bonds (or tranches) with greater levels of subordination do, in fact, experience higher ex-post levels of delinquencies and default. Recent studies have found that rating agencies follow a "learning by doing" approach in subordination structuring (Riddiough and Chiang, 2004). As expected, the rating agencies were conservative in the early stages with regard to subordination levels given the paucity of information about delinquencies, defaults and prepayments on loans. As time progresses and more information is available regarding loan performance, subordination levels adjusted to new levels. This paper focuses on cross sectional differences in subordination levels. We examine if this relationship between subordination and ex-post delinquencies and defaults is conforming to rational expectation. We perform both a deal level and a loan level analysis using commercial mortgage-backed securities (CMBS). Our results show that the expected loss for CMBS pools are a statistically significant factor in explaining both AAA and BBB bond subordinations; however, expected loss accounts for less than 30 percent of the variation. Even considering the rating agencies' practice of incorporating differences in loan terms, borrower quality, deal structural and information quality into their subordination structure, the empirical fit is still too low. These findings indicate the difficulty in determining subordination levels apriori.