The relative importance of economic policy uncertainty and geopolitical risk on U.S. REITs returns Free

In this article, we analyze the relative impacts of economic political uncertainty, as defined by Baker et al. (2016), and geopolitical risks, as recently introduced by Caldara and Iacoviello (2022), on the performance of U.S. Nareit returns since the “new REITs era” (1993) (Pagliari et al., 2005).

As reported by Caldara and Iacoviello (2022), the Bank of England defines economic and policy uncertainty, with geopolitical risk, as an “uncertainty trinity” with significant consequences on real business cycles dynamics. Since the work of Niederhoffer (1971), quoted by Jorion and Goetzmann (1999) (among many others), an extensive economic literature has been devoted to the analysis of world events and global economic and financial markets. After the recent introduction of economic policy uncertainty indices, EPU (Baker et al., 2016), and geopolitical risk indices, GPR (Caldara and Iacoviello, 2022), a renewed and growing interest for this important economic topic has been observed, with special focuses on asset prices, returns and volatility (energy, commodities, gold, financial assets … and in a lesser extent for real estate) ^[1].

In their “2023–2024 Top Ten Issues Affecting Real Estate” “political unrest and global economic health” is ranked first issue by the Counselors of Real Estate, while it was ranked 2 issue in their 2022–2023 report, as mentioned by Korogolos (2022) (before “obsolescence and devaluation”, ranked 2, “AI”, ranked 4, “population shock”, ranked 6, “interest rates and inflation”, ranked 7, and well before “energy” and “ESG requirements”) ^[2].

Following Baker et al. (2016) and Caldara and Iacoviello (2022), who respectively report a positive exposure to the economic policy uncertainty and to the geopolitical risks for the real estate industry, a recent housing and real estate economic literature has been growing. André et al. (2017) show that economic policy uncertainty significantly affects US real housing returns and volatility. Using a time-varying parameter factor augmented vector autoregression modeling (TVP-FAVAR), Christou et al. (2019) demonstrate that uncertainty shocks negatively impact the US housing prices. Yuni et al. (2024) analyze in a quantile regression framework the effects of GPR indices on a small sample of five daily global and regional house prices indices during the COVID-19 pandemic period. Ahiadu and Abidoye (2024) review existing literature on the impact of economic uncertainty on property performance. They show that economic uncertainty negatively influences investment volumes, returns and performances. Very recently, Coën and Desfleurs (2024) use an extended conditional version of Fama and French (1993) asset pricing model to analyze the influence of GPR indices on the dynamics of REITs returns. For U.S. REITs, grouped into portfolios, they report that the geopolitical risks are priced.

In this article, we extend Coën and Desfleurs (2024) study, focusing on U.S. Nareit property sectors and subsectors. Our contribution is twofold. First, using an extension of the standard Fama and French (1993) model (as suggested by Hsieh and Peterson (2000), Bond et al. (2003), Lee et al. (2008), Van Nieuwerburgh (2019) among many others in the literature), we consider the EPU indices and the GPR indices, separately and simultaneously, as potential risk factors to analyze the decomposition of U.S. Nareit indices returns since the “new REITs era” (1993). With this approach, we are able to compare the relative importance of economic policy uncertainty and geopolitical risks on U.S. REITs returns. Second, we shed light on the different sensitivities of Nareit sectors to these two distinct exogenous sources of risks. These differences may have indeed important consequences and implications for investors, insurers, bankers and other stakeholders in the real estate industry. Therefore, we give weight to the concerns and “top issues” mentioned by the real estate industry, as described above by the Counselors of Real Estate, and suggest a partial solution. To obtain robust economic results, we correct for the problems of errors-in-variables (biased estimates: see Durbin (1954), Davidson and MacKinnon (2004)) in the linear asset pricing models, using the instruments suggested by Dagenais and Dagenais (1997) ^[3]. Our results report that geopolitical risk, GPR, and economic policy uncertainty, EPU, are priced for U.S. REITs. Based on the critical value introduced by Harvey et al. (2016), the GPR index may be reasonably considered as a relevant risk factor in the pricing of REITs returns.

The article is organized as follows. In Section 2, we introduce the econometric method and the asset pricing model. In Section 3, we present the data. In Section 4, we report empirical results and provide robustness checks. In Section 5, we draw our conclusions by summarizing our main findings.

Following the financial literature and the seminal work of Fama and French (1993), we use a linear multifactor asset pricing model to analyze REITs returns.

where R, stands for the excess return, $X̃k,t$ is the true (unobserved) risk factor k realization in period t, β_i,k is the factor k loading, i = 1, …10 is the US Nareit index, t is the time index and ϵ_i,t is the error term. α_i is a constant term defined as the security’s abnormal return in this unconditional asset pricing framework.

Our aim here is to report robust and consistent estimates for our analysis of linear asset pricing models (APM). Therefore, following Coën and Racicot (2007) and Carmichael and Coën (2008), we advocate the use of the Dagenais and Dagenais (1997) higher moment estimators as instruments in a Generalized Method of Moments (GMM) procedure as defined by Hansen (1982) (more precisely, we use iterated GMM: iGMM). As it is well acknowledged in economic literature, the presence of errors-in-variables (EIV) leads to biased and inconsistent OLS parameter estimates (Durbin, 1954; Hansen, 1982; Pal, 1980; Erickson and Whited, 2000, among many others).

To deal with this inference problem, a common solution is the use of instrumental variables. This approach has drawbacks as highlighted by Pal (1980): the main problem is the choice of instruments. We follow Coën and Racicot (2007), Carmichael and Coën (2008), Erickson and Whited (2012), and Coën and Lecomte (2019) (among others), who suggest the use of Dagenais and Dagenais (1997)’ higher moment estimators (DDHME) as instrumental variables to deal with the problem of EIV. As put forward by Dagenais and Dagenais (1997) and detailed thereafter by Carmichael and Coën (2008), the relevant instruments are: z₁ = x∗x, $z2=x∗x∗x−3x[E(x′x/N)∗IK]$ and a constant. x_ij are the elements of the matrix x and x = AX where $A=IN−ii′/N$⁠. The matrix x is the T × K matrix X calculated in mean deviation, standing for the matrix of K factor loadings where T is here the number of observations. The symbol * is the Hadamard element-by-element matrix multiplication operator. As suggested by Davidson and MacKinnon (2004), we run Durbin-Wu-Hausman (DWH hereafter) type test (see Hausman, 1978) to detect the presence of EIV ^[4].

We consider the “new REITs area” (Pagliari et al., 2005; Ambrose et al., 2007), including 360 monthly returns from January 1994 to December 2023, to analyze the impacts of economic policy uncertainty and geopolitical risk.

We use the FTSE Nareit U.S. Real Estate Index Series to track the performance of U.S. REIT industry at the sector level to shed light on the differences among the different sectors during the full period. Our sample includes the following monthly property index returns: (1) Apartments; (2) Diversified; (3) Health Care; (4) Industrial; (5) Lodging/Resorts; (6) Office; (7) Regional Malls; (8) Residential; (9) Retail; and (10) Shopping Centers ^[5]. The detailed descriptive statistics are reported in Tables 1 and 2.

As we use the standard three factor Fama and French (1993)’s asset pricing model, the market excess return MKT = r_m − r_f, and the factors addressing the anomalies for size (SMB) and value (HML), are taken from Kenneth French’s data library ^[6]. The variations of the geopolitical risk (GPR) index, with its two broad components, the GPA index (geopolitical acts), the GPT index (geopolitical threats), and the domestic U.S. GPR index, GPRUS, are extracted from Caldara and Iacoviello (2022) data base ^[7]. The variations of the U.S. Policy Uncertainty indices, as defined by Baker et al. (2016), are from Policy Uncertainty website (https://www.policyuncertainty.com/ ^[8]). As suggested by Baker et al. (2016), we consider both the “Three Components Index”, EPU1, and the “News Based Policy Uncertainty Index”, EPU2.

Here, we extend the Fama and French (1993) model (the benchmark model in the financial literature), first with the geopolitical risk factors (GPR, GPRUS, then with GPT and GPA), second with the U.S. economic policy uncertainty indices (EPU1, EPU2) to test their explanatory power on monthly returns between January 1994 and December 2023. Tables 3 and 4 report the results of the extended Fama-French model with GPR, the global geopolitical risk factor, and GPRUS, the domestic U.S. geopolitical risk factor. In the  Appendix, we also report the results for the geopolitical risk threats, GPT, and the geopolitical acts, GPA respectively (Table A1). First, as expected, we note that MKT is highly statistically significant for all Nareit indices, followed by the book-to-market factor, HML, and by the size factor, SMB. The robust coefficient estimates for MKT rank from 0.740 for Residential to 1.361 for Regional Malls. Very interestingly and as suggested by Coën and Desfleurs (2024), GPR is highly statistically significant at 1% for all Nareit indices. All robust t-statistics are indeed higher than 3, the critical value reported by Harvey et al. (2016) to valid a potential risk factor in linear asset pricing models. It suggests that GPR is a relevant factor to price REITs. The estimates, although highly statistically significant, are low compared to Fama-French factors. They rank from 0.021 for Diversified (t-stat: 3.86) and Health Care (t-stat: 4.77), to 0.042 for Regional Malls (t-stat: 6.42) and 0.041 for Shopping Centers (t-stat: 6.93), followed by Retail (0.038, t-stat: 6.83) and Office (0.038, t-stat: 5.53), with an average coefficient of 0.0312. As a robustness check of these results in a GMM framework, we can highlight that the baseline model’s J_T statistics (test of overidentifying restrictions) exhibit all p-values well within the usual acceptance region. This leads to the interpretation that the inclusion of GPR significantly improves the Fama-French model. It contributes to the pricing of U.S. Nareit indices, giving weight to Korogolos (2022) and to the Counselors of Real Estate (CRE) position on the importance of geopolitical risk. This result is confirmed by the average pricing errors, (α): all coefficients are not statistically significant (except for Lodging). The null hypothesis of no average pricing errors cannot indeed be rejected for this asset pricing modeling.

The pricing of geopolitical risk is also confirmed in Table 4, where we report the relative contribution of the U.S. geopolitical risk, GPRUS. As previously observed for the global geopolitical index, the U.S. index is statistically significant at 1% for all U.S. Nareit indices. The estimates rank from 0.022 for Diversified (t-stat: 3.56) and Health Care (t-stat: 3.38) to 0.047 for Regional Malls (t-stat: 5.58), followed by Shopping Centers (0.044, t-stat: 6.67) and Retail (0.041, t-stat: 6.19) with an average of 0.0327. Interestingly, we also report in the  Appendix that the geopolitical acts, GPA (average: 0.0236), are overall highly statistically significant for U.S. Nareit indices, while geopolitical threats, GPT (average: 0.0205), are less significant during the last 3 decades. As observed by Coën and Desfleurs (2024), we confirm that the Nareit indices are more sensitive to geopolitical acts than to geopolitical threats. Based on these robust results, we may conclude that global (GPR) and U.S. (GPRUS) geopolitical risks have a significant impact on the dynamics of U.S. Nareit indices from January 1994 to December 2023. We also notice that Retail Reits, including Regional Malls Reits and Shopping Centers Reits are more sensitive to geopolitical risks. This observation is also valid for Office Reits. As defined by Nareit, ”Retail Reits include Reits that focus on large regional malls, outlet centers, grocery-anchored shopping centers and power centers that feature big box retailers”. As highlighted by Caldara and Iacoviello (2022), a rise in the GPR index predicts lower expected GDP growth with possible impacts on households’ expectations, sentiments and consumption (as suggested by the Survey of Consumers, University of Michigan ^[9]). On the other hand, Healthcare Reits that include senior living facilities, hospitals, medical office buildings and skilled nursing facilities, are less sensitive to geopolitical risks. This point supports Caldara and Iacoviello (2022)’s observation. They report indeed in their  Appendix that Healthcare industry is indeed less sensitive to GPR indices than Real estate industry (as a whole). These results are confirmed with Carhart (1997)’s model (See Table A2 in the  Appendix).

In this section, we follow Korogolos (2022) suggestion, and analyze the impact of economic policy uncertainty, EPU, on the dynamics of U.S. Nareit returns. As reported by Baker et al. (2016), “policy uncertainty is associated with greater stock price volatility and reduced investment and employment in positive-sensitive sectors”. As introduced by Baker et al. (2016), we consider two metrics for the EPU index in the U.S.A.: the “Three component index”, EPU1, and the “News based policy uncertainty index”, EPU2. The detailed results are reported in Table 5. First, we focus on EPU1, in the top part of the table. For all Nareit indices, the EPU factor is statistically significant at 10% (for 6 indices at 5%: Apartments, Health Care, Office, Residential, Retail and Shopping Centers), except Lodging. The coefficients rank (when significant) from 0.040 for Diversified (t-stat: 1.72) to 0.106 for Health Care (t-stat: 3.33), followed by Shopping Centers (0.067, t-stat: 2.43), and Retail (0.059, t-stat: 2.32), with an average coefficient of 0.058. This result is consistent with Baker et al. (2016) who highlight the impact of policy uncertainty on policy-sensitive sectors like defense, health care, finance, and infrastructure. Second, we consider, as robustness check, EPU2 that is statistically significant at 10% for eight Nareit indices. As previously observed, the coefficient is not statistically significant at 10% for Lodging, and now Diversified. The estimates rank from 0.026 for Office (t-stat: 1.65), followed by Regional Malls (0.031, t-stat: 1.65) to 0.075 for Health Care (t-stat: 3.59), followed by Shopping Centers (0.053, t-stat: 2.89).Very interestingly, these results shed light on the policy-sensitivity differences among U.S. REITs sectors. Health Care Reits are indeed the most sensitive Reits to economic policy uncertainty, confirming Baker et al. (2016)’s results, who report that the volatility for firms in the health care sector significantly increases with the EPU index. Besides, Lodging REITs including different classes of hotels based on features such as the hotels’ level of service and amenities, are not sensitive to EPU indices. This point could be seriously considered by investors, banking and insurance sectors, policy makers and other stakeholders in the real estate industry [¹⁰].

As reported by Caldara and Iacoviello (2022), economic policy uncertainty and geopolitical risks are two different phenomena and induce different shocks. Focusing on Table 2, we note indeed that the correlation coefficients between EPU indices and GPR index are practically zero (0.008 and −0.002). In this case, it would be relevant to consider simultaneously in a parsimonious asset pricing model their relative contribution on the analysis of U.S. Nareit indices returns since the “new REITs era”. We report the results of this valuation in Table 6 for the GPR index and for the EPU indices. As expected, the geopolitical risk factor is still highly significant at 1% for all Nareit indices. All t-stats are higher than 3, the critical value introduced by Harvey et al. (2016). The coefficient estimates rank from 0.021 for Lodging (t-stat: 3.32) to 0.048 for Shopping Centers (t-stat: 6.48) with an average of 0.0356. We observe the same trend in the second panel. EPU indices are statistically significant at 10% for 9 U.S. Nareit indices (except Lodging in the first panel). Thus, we note that Nareit indices are sensitive to both geopolitical risk and economic policy uncertainty. Nevertheless, they are more sensitive to geopolitical risks. These results should shed new light for investment, financing and insurance in the real estate sector.