LeatherSagiKnowledgebase

The Other (Commercial) Real Estate Boom and Bust: The Effects of Risk Premia and Regulatory Capital Arbitrage

5 min read

Problem

The 2000s boom and bust in U.S. commercial real estate (CRE) prices was at least as large as the residential boom-bust (103% peak-to-peak vs 75% for housing) and had a larger effect on bank failures, yet it received far less research attention. The authors ask: what drove the dramatic swings in CRE capitalization rates and risk premia across the four major property types (apartments, industrial warehouse, CBD office, retail malls) during 1996-2014?

Key idea

CRE cap rate movements are primarily driven by shifts in required risk premia, not by changes in expected rent growth. The authors decompose cap rates via a credit-augmented Gordon Growth Model:

CapRate_t = r_t - g_t - mu * CapAvail_t

where r_t = RF_t + RiskPrem_t (risk-free rate + risk premium), g_t is expected NOI growth, and CapAvail_t tracks non-price credit terms. The key insight is that during the mid-2000s boom, CRE risk premia fell sharply due to regulatory capital arbitrage (Basel II reducing capital requirements on CMBS), while the post-2008 bust was driven by a spike in general risk premia (Baa-Treasury spread). The post-2010 recovery was driven by low real Treasury yields, not by renewed risk-premium compression.

Method

Error-correction models (Johansen VECM) estimated on RERC survey data for four property types over 1996Q1-2014Q3. Two-stage approach:

  1. Long-run cointegrating relationships for cap rates and risk premia as functions of required returns, expected rent growth, capital availability, and regulatory capital measures (RegCap index or BaselSEC dummy).
  2. Short-run error-correction dynamics capturing adjustment speeds and cyclical factors (VIX changes, LEI growth, government shutdown dummy).

Separate models for risk premia use the Baa-Treasury spread as a proxy for general risk premia and two measures of regulatory stringency.

Results

  • Long-run cap rate models: Required rates of return have positive and significant coefficients; expected rent growth and capital availability have negative and significant coefficients. Unique cointegrating vectors identified for all four property types.
  • Error-correction speed: 40-72% of the gap between actual and equilibrium cap rates closes per quarter, implying near-full adjustment in 2-3 quarters.
  • Risk premium models: CRE risk premia are positively related to Baa-Treasury spreads and regulatory capital stringency. Liberalized capital requirements (mid-2000s) lowered risk premia by 100-200 bp; Dodd-Frank tightening raised them back.
  • Decomposition: The mid-2000s cap rate compression was mainly driven by declining risk premia (linked to regulatory capital arbitrage). The 2008-09 reversal was driven by spiking general risk premia. The post-2010 recovery was driven by declining real Treasury yields. Expected rent growth contributed very little to cap rate swings throughout.
  • Robustness: Results hold with PWC data in place of RERC, with adjusted capital availability indices, and across all four property types.

Limitations

  • Relies on survey data (RERC, PWC) for cap rates, required returns, and expected rent growth. Survey responses may lag or lead true market conditions.
  • The regulatory capital measure (RegCap) is a hand-constructed index of effective capital requirement stringency, not a cleanly identified exogenous variation. The BaselSEC dummy is cruder but more exogenous.
  • The Gordon Growth Model approximation assumes constant expected growth rates, which is restrictive. The authors acknowledge this but argue the ECM framework handles time variation empirically.
  • Sample period (1996-2014) covers essentially one full boom-bust cycle, limiting statistical power for testing crisis-specific dynamics.
  • No direct structural model of the link between CMBS issuance, capital requirements, and risk premia — the connection is through reduced-form cointegration.

Open questions

  • Does the risk-premium / regulatory-capital channel extend to the 2020-2025 period of COVID disruption, remote-work shocks to office CRE, and the 2022-2024 Fed tightening cycle?
  • Can the model’s decomposition be made fully structural by embedding the regulatory capital channel in a no-arbitrage macro-finance framework with regime switching?
  • How does the capital availability effect interact with the regime-switching monetary policy channel identified in leather-sagi-markov-switching-cre-asset-pricing?

My take

This paper is directly relevant to the CRE asset pricing project. The central finding — that CRE cap rate swings are dominated by discount-rate / risk-premium movements rather than cash-flow news — parallels the Campbell-Shiller (1988) result for equities and is the empirical regularity that the Leather-Sagi MS-RE model is designed to capture structurally. The regulatory capital arbitrage channel (CMBS / Basel II) provides an important institutional mechanism for the kind of risk-premium regime shifts that the project models as Markov-switching parameters. The decomposition methodology (RERC survey-based) is complementary to our model-based approach: our model derives risk premia from the equilibrium pricing kernel, while Duca-Ling measure them directly from surveys and corporate bond spreads. The fact that they find expected rent growth contributes very little to cap rate variation is consistent with our model’s focus on the discount-rate side.

Graph View

Backlinks