The question
Real estate is unusually exposed to uncertainty. Buildings produce cash flows over decades, they are costly to adjust or sell, and their income depends on tenants, occupancy and refinancing conditions that can all change when the outlook darkens. Real Estate Investment Trusts (REITs) make this exposure visible: they are listed and traded every day, but their income comes from property portfolios.
A growing literature shows that uncertainty helps forecast the stock market, including our own earlier work on financial uncertainty. It is not obvious that the same holds for real estate. Property has its own cash-flow dynamics and financing frictions, and it has often moved differently from equities: REITs did well while the dot-com crash hit stocks, and lagged badly after COVID-19 while equities rebounded. Most existing evidence on uncertainty and REITs is also in-sample, and in-sample predictability often disappears out of sample. We ask whether uncertainty genuinely forecasts real estate returns, whether investors could have used it, and why it works.
What we do
Data. We study monthly excess returns from January 1987 to December 2022 on six series: all U.S. REITs, equity REITs and mortgage REITs (FTSE Nareit), the Case–Shiller house price index, the overall stock market, and the real-estate industry portfolio from Kenneth French’s data library.
Five measures of uncertainty. We compare the financial uncertainty index of Ludvigson, Ma and Ng (2021); economic policy uncertainty (Baker, Bloom and Davis, 2016); news-based equity market volatility and its real-estate version (Baker et al., 2021); and realised stock market volatility. We use both their levels and their monthly changes, and benchmark them against the fourteen standard Welch–Goyal predictors.
Tests. In sample, we use the IVX-Wald approach of Kostakis, Magdalinos and Stamatogiannis, which stays valid when predictors are persistent. Out of sample, we forecast recursively against the historical average using Clark–West tests, and add forecast combinations and principal-component methods. Because the uncertainty index is published only twice a year, we also run a release-consistent version that updates the signal in June and December only. We then measure the gains for a mean–variance investor, decompose returns into cash-flow and discount-rate news, and look across lags, horizons, market regimes, REIT characteristics and property segments.
What we find
Finding 1. The change in financial uncertainty is a strong, robust predictor.
Several uncertainty measures look useful in sample but fail out of sample. One does not: the monthly change in financial uncertainty. A one-standard-deviation rise predicts next-month REIT returns roughly 1.5 percentage points lower. Out of sample it beats the historical average for all six return series, with R² of 8.9% for all REITs, 8.2% for equity REITs, 6.0% for mortgage REITs and 10.5% for the real-estate equity portfolio, all statistically significant. Those are large numbers for monthly returns. Even when the signal is updated only when the index is actually published, the gains fall but stay positive and significant, at 2.5–3.2% for the three REIT series.
Notes: Campbell–Thompson out-of-sample R² relative to the historical mean, expanding window with a 200-month initial estimation period, 1987–2022. All values are significant at the 5% level or better (Clark–West). Source: Table 3.
Data table
| Out-of-sample R² | |
|---|---|
| Real-estate equity portfolio | 10.51 |
| Stock market | 8.94 |
| All REITs | 8.87 |
| Equity REITs | 8.17 |
| Mortgage REITs | 5.98 |
| Case–Shiller house prices | 5.50 |
Pooling predictors helps too. Scaled principal components (Huang et al., 2022) deliver the best combined forecasts, with out-of-sample R² of 7.6% for all REITs and over 18% for house prices.
Finding 2. The signal is valuable to investors.
For a mean–variance investor choosing between REITs and Treasury bills, timing with the change in uncertainty raises the certainty-equivalent return by 10–15% a year, far more than any other predictor. Most of that survives conservative trading costs of 50 basis points.
| Return series | No costs | 50 bp costs | Release-consistent, no costs |
|---|---|---|---|
| All REITs | 10.25 | 8.34 | 7.27 |
| Equity REITs | 10.66 | 8.82 | 7.22 |
| Mortgage REITs | 14.58 | 12.59 | 12.90 |
| Stock market | 7.26 | 5.11 | 4.21 |
| Real-estate equity portfolio | 8.13 | 5.96 | 11.23 |
Certainty-equivalent gains over the historical-average strategy for an investor with risk aversion of five and portfolio weights between 0% and 150%. Source: Table 5 of the paper.
Simple trading rules tell the same story. A long-only REIT strategy that holds REITs only when the forecast is positive roughly doubles the Sharpe ratio of buy-and-hold (0.26 against 0.12) and earns a Fama–French five-factor alpha of about 0.8% a month. A long/short version earns about 1.6% a month.
Finding 3. Uncertainty works mainly through expected cash flows.
Using the Campbell–Shiller decomposition, with each asset’s own dividend–price ratio, we split unexpected returns into news about future cash flows and news about future discount rates. Rising uncertainty is associated mainly with negative cash-flow news: investors revise down the income they expect from property, rather than simply demanding higher returns.
When and where it works
The response is slow. Returns keep falling for up to five months after a rise in uncertainty, and the predictive effect fades gradually over ten months with no reversal. Predictability is also much stronger when conditions are strained: the out-of-sample R² is around 10.6% in periods of high financial uncertainty, against 2.7% when uncertainty is low, and a similar split appears between tight and loose credit conditions.
- High financial uncertainty
- Low financial uncertainty
Notes: High uncertainty: months when the financial uncertainty index is above its historical mean. Forecasts use the change in financial uncertainty as the sole predictor. Source: Table 9.
Data table
| High financial uncertainty | Low financial uncertainty | |
|---|---|---|
| All REITs | 10.56 | 2.69 |
| Equity REITs | 9.64 | 2.22 |
| Mortgage REITs | 6.52 | 4.85 |
| Case–Shiller | 5.79 | 2.81 |
| Stock market | 12.70 | 2.14 |
| Real-estate equity | 11.16 | 6.51 |
Across 620 individual REITs, the effect is significantly stronger for smaller REITs and for those followed by fewer analysts. Media visibility, institutional ownership and lease length make no significant difference. Every property segment responds, with lodging and resorts the most sensitive, followed by diversified, retail and industrial/office REITs.
Notes: Slope coefficients from one-month-ahead predictive regressions of value-weighted CRSP/Ziman segment returns on the standardised change in financial uncertainty. All are significant at the 1% level (IVX-Wald). Source: Table 12.
Data table
| Return response | |
|---|---|
| Lodging/resorts | −2.30 |
| Diversified | −1.75 |
| Retail | −1.68 |
| Industrial/office | −1.60 |
| Residential | −1.22 |
| Self storage | −1.13 |
| Health care | −1.07 |
Why it matters
For investors, the results show that a freely available, publicly released index carries information about the next few months of REIT returns, and that this information is most valuable when markets are under stress. Because the signal works mainly through expected cash flows, it also says something about how uncertainty reaches property income: through tenant demand, occupancy, rent growth and the cost of refinancing.
For research, the paper shows that the link between uncertainty and returns documented for the stock market carries over to real estate, and is stronger there. It also makes two practical points for anyone forecasting with uncertainty indexes. Changes in uncertainty matter far more than its level, and the index’s publication schedule should be respected when judging how useful a signal would have been in real time.
Takeaways
- Financial uncertainty forecasts real estate returns. Its monthly change predicts lower REIT returns, in and out of sample, across REIT types and property segments.
- The gains are economic, not just statistical. Utility gains of 10–15% a year, and higher Sharpe ratios, survive trading costs and the index’s publication schedule.
- The channel is cash flows. Uncertainty mainly signals lower future property income rather than higher required returns.
- The adjustment is gradual. The slow response, and its strength among smaller, less-followed REITs, fits gradual information diffusion (Hong and Stein, 1999). Slow-moving leases, occupancy and refinancing could produce the same pattern, so we do not claim a single mechanism.
Key references
- Baker, S. R., Bloom, N., and Davis, S. J. (2016). Measuring economic policy uncertainty. Quarterly Journal of Economics, 131(4), 1593–1636.
- Campbell, J. Y., and Shiller, R. J. (1988). The dividend-price ratio and expectations of future dividends and discount factors. Review of Financial Studies, 1(3), 195–228.
- Clark, T. E., and West, K. D. (2007). Approximately normal tests for equal predictive accuracy in nested models. Journal of Econometrics, 138(1), 291–311.
- Henry, Ó., Kerestecioglu, S., and Pybis, S. (2024). Can financial uncertainty forecast aggregate stock market returns? Financial Markets, Institutions & Instruments, 33(2), 91–111.
- Hong, H., and Stein, J. C. (1999). A unified theory of underreaction, momentum trading, and overreaction in asset markets. Journal of Finance, 54(6), 2143–2184.
- Kostakis, A., Magdalinos, T., and Stamatogiannis, M. P. (2015). Robust econometric inference for stock return predictability. Review of Financial Studies, 28(5), 1506–1553.
- Ludvigson, S. C., Ma, S., and Ng, S. (2021). Uncertainty and business cycles: Exogenous impulse or endogenous response? American Economic Journal: Macroeconomics, 13(4), 369–410.
- Welch, I., and Goyal, A. (2008). A comprehensive look at the empirical performance of equity premium prediction. Review of Financial Studies, 21(4), 1455–1508.