Commercial Real Estate Pricing with Regime-Switching Macro Dynamics

Overview

Commercial real estate (CRE) prices behave like long-duration leveraged claims on income streams that are sensitive to macroeconomic regimes — output gap, inflation, monetary policy stance — and have notoriously slow-moving cap rates that are hard to price under single-regime models. This topic covers the line of work that connects an NK / DSGE-style macro model (output gap, inflation, short rate) to CRE prices via no-arbitrage. The distinguishing feature relative to a pure term-structure literature is that CRE cap rates under regime-switching macro dynamics generate exponential-quadratic (rather than exponential-affine) pricing factors, which require Riccati recursions over an unbounded horizon and are sensitive to the no-bubble condition (NBC) on the discount-factor process. The deep theoretical foundation for the long-run / asymptotic behavior of these pricing recursions comes from the operator / semigroup approach of Hansen–Scheinkman (2009).

Timeline

• 2009 — Hansen & Scheinkman (Econometrica): operator approach to long-term risk provides the eigenvalue / eigenfunction structure that governs convergence of pricing recursions for assets with persistent cash flows.
• 2013– — Ghysels et al. handbook chapter on forecasting real estate prices: surveys econometric and macro-driven models, motivates the need for regime-aware specifications.
• 2020s — Leather & Sagi: Markov-switching rational expectations CRE asset pricing model — NK macro framework + 4 compound regimes (monetary policy × wage rigidity) + exponential-quadratic Riccati pricing for CRE cap rates. Uses Cho–Moreno forward solution for the macro RE block.

Seminal works

• hansen-scheinkman-2009-long-term-risk-operator-approach
• ghysels-forecasting-real-estate-prices
• leather-sagi-markov-switching-cre-asset-pricing
• riccati-equations-leather-sagi

SOTA tracker

• Pricing kernel: affine and exponential-quadratic forms with regime switching are available; closed-form for the affine case, Riccati-recursive for the quadratic case.
• Filtering: continuous-state Kalman recursions inside particle filters over regime paths (RBPF) are the only known way to evaluate the likelihood without exponential blow-up.
• Estimation: simulated MAP / MLE on simulated DGP works locally; global recovery from cold starts is currently bottlenecked by local-optimizer depth, not basin discovery.

Open problems

• Long-run convergence of quadratic pricing recursions depends on whether the spectral radius of the regime-switching second-moment operator is below 1 (Hansen–Scheinkman gives the right framework but the application to multivariate exponential-quadratic factors is not standard).
• No-bubble check for the macro RE solution: the absorbing-NBC variant is the current production gate (≈99% accuracy), but its theoretical justification under the Cho–Moreno forward solution is informal.
• Income process specification: apartment / industrial / office cash flows are modeled as separate VAR shocks without regime-switching coefficients — the most natural extension but expensive to estimate.
• Curvature / identification: the likelihood at the incumbent is extremely sloppy (condition number ≈10²⁵–10²⁶ on the Hamilton surrogate), and the canonical FD diagonal stencil is gradient-contaminated at non-critical points — see the project’s gotchas.

My position

This is the project’s primary topic. The Leather–Sagi paper and the Riccati-equations note are the central artifacts; everything else in the wiki supports them either as macro / TSM infrastructure (markov-switching-term-structure-models), as filtering machinery (switching-state-estimation), or as control-theoretic background (markov-jump-linear-systems-control-filtering).

Research gaps

• Estimation pipeline: moving from local recovery (PASS) to global recovery (FAIL on cold Sobol starts) requires a formal global optimization stack — currently the planned next milestone for the project.
• Out-of-sample validation on real-data CRE indices (NCREIF, Green Street, RCA) under the regime-switching model is unfinished; the simulated-data validation arc is closed but the real-data MAP is currently the published headline.
• Comparative pricing of CRE vs Treasuries under the same regime-switching kernel has no published joint estimation.

Key people

• john-duca — Federal Reserve Bank of Dallas. CRE risk premia and regulatory capital arbitrage decomposition (Duca-Ling 2015).
• david-ling — University of Florida. CRE cap rate error-correction models and institutional investor behavior.
• alberto-plazzi — USI / Swiss Finance Institute. CRE expected returns and cap rate predictability (Plazzi-Torous-Valkanov 2010).
• walter-torous — MIT Sloan / UCLA Anderson. CRE return predictability and present-value models.
• rossen-valkanov — UCSD Rady School. CRE predictability and financial econometrics.