Pooling Heuristics and Credit Risk Distribution in Conduit CMBS: Evidence from a Controlled Re-Pooling Experiment

Abstract

Conduit commercial mortgage-backed securities (CMBS) pool loans from multiple borrowers and properties and allocate cash flows and losses across tranched bonds. While prior work emphasizes how collateral characteristics and deal structure affect tranche performance, less attention has been paid to how the pooling rule itself shapes credit risk distribution across pools. This thesis studies that question in a controlled experiment that holds the loan universe fixed and varies only the pooling heuristic. Using SEC ABS-EE Exhibit 102 loan-level disclosures, the thesis builds a dataset of 389 commercial mortgage loans from ten U.S. conduit CMBS transactions issued between 2022 and 2025. The loans are reallocated into four synthetic pools under four pooling rules: Random, DSCR stratification, LTV stratification, and a Defensive rule that disperses jointly weak loans (low DSCR and high LTV). Each pool is evaluated using a stylized refinance-at-maturity loss mapping and a common three-tranche waterfall (10/20/70). Results are reported for a base case and seven stress scenarios. Three findings emerge. First, single-signal stratification reduces cross-pool dispersion relative to random assignment, with DSCR stratification achieving the lowest dispersion in every scenario. In the combined rate-plus-NOI stress, cross-pool standard deviation is 0.98 percentage points under DSCR stratification versus 2.81 under random assignment. Second, dispersion minimization and worst-pool containment are distinct objectives: the Defensive rule can improve worst-pool outcomes in jointstress states even when overall dispersion remains higher. Third, distance-to-impairment margins for mezzanine and senior tranches vary materially across heuristics, indicating that pooling design can affect tranche-level tail risk even when total collateral loss is fixed within a scenario. The results imply a practical tradeoff for issuers between dispersion control and tail containment under joint stress. For investors and credit analysts, they motivate supplementing weighted-average metrics with worst-pool diagnostics. While the framework is intentionally stylized and results are comparative rather than predictive, the findings suggest that pooling warrants explicit attention in deal structuring and credit assessment.