Rim's Table 9 shows real estate hedging the S&P 500 and four of five US sectors better during the Global Financial Crisis than afterwards. For the S&P 500, diagonal BEKK puts the variance reduction at 59.34% during the crisis against 55.07% after. The exception is US Energy: 24.95% in the GFC against 31.44% afterwards. "After" covers 2010 to 2019 in the text and 2010 to 2021 in the Table 9 notes, a discrepancy discussed below. Yet the abstract says real estate's hedging effectiveness "significantly eroded during the GFC". The text supports that claim with the collapse in the long-only weight. The paper's own HE figures run in the opposite direction.
The study behind the claim
Low correlation with equities and a claimed inflation hedge gave real estate its place in multi-asset books. REITs also bring mandatory dividend payouts and lease income, which the paper says can cushion income declines in downturns. Rim asks whether this defensive role survives the financialisation of property, looking beneath the broad indices at sectors in the US and UK markets.
The sample contains 1,149 observations per series: weekly Bloomberg returns from January 2000 to December 2021. The FTSE Nareit Equity and Mortgage REIT indices represent US real estate; FTSE EPRA Nareit UK represents UK real estate. Rim pairs each with its broad index (S&P 500, FTSE 100) and five sectors. The US sectors are Financials, Industrials, Consumer, Energy and IT. The UK sectors are Banking, Industrials, Consumer, Energy and Retail.
For each pair, Rim fits a VAR(1) mean equation and three bivariate GARCH variance models: CCC, DCC and diagonal BEKK. The paper interprets lagged cross terms in the variance equations as volatility spillover. In this reading, a lagged shock or variance in one series increases the other's conditional variance. The fitted covariance matrices supply three measures. The Kroner-Ng weight gives real estate's share of a long-only minimum-variance two-asset portfolio, capped to [0,1]. The hedge ratio sizes the minimum-variance short. Hedging effectiveness (HE) equals one minus the hedged position's variance divided by the unhedged equity leg's variance.
Two findings carry the argument. Real estate receives higher weights against sectors than against the index: US 0.26 against the S&P 500, compared with 0.43 to 0.66 against sectors. The paper also describes weaker spillovers at sector level. Its main pitch comes from splitting the crises: the property-led GFC damaged real estate's portfolio role, whereas COVID-19, an outside shock, improved it.
Did the US stock hedge fail in 2007 to 2009?
For US REIT/S&P 500, HE rises from 20.40% pre-GFC to 59.34% during the GFC. Post-GFC it is 55.07%, followed by 73.55% in COVID. Industrials follow the same pattern, with 64.87% in the GFC against 50.68% afterwards. The collapse occurs in the portfolio weight: 0.49 pre-GFC, 0.05 during the crisis and 0.23 after.
Rim acknowledges the tension: "Hedge ratios rose, yet actual hedging effectiveness was unclear due to the minimal weight of real estate in portfolios". The sentence continues by arguing that, when real estate becomes part of a broader financial shock, "its role as a diversifier is eroded". This shifts the defence to the diversifier role and hence back to the weight. The paper's formula defeats that defence. HE uses r2 minus beta times r1; the Kroner-Ng weight does not enter the calculation. And for the broad pair, the hedge ratio stayed at 0.46 in the GFC, unchanged from pre-GFC and below the post-GFC 0.48. The paper's "rose" does describe some pairs: US Financial moves from 0.53 pre-GFC to 0.88 in the GFC, and UK Banking from 0.61 to 0.75.
A 5% long-only allocation and a 59% variance reduction belong to different trades.
An asset whose own variance has exploded gets little room in a minimum-variance portfolio. A short hedge can favour that same asset when its correlation with equities is high, because the correlation allows the short to absorb equity risk. Both interpretations fit the GFC. Only one supports "eroded".
Which leg does the hedge ratio size?
Equation 17 defines beta as h12/h22. In the paper's description, a long dollar of asset 1, real estate, is hedged with a short of beta dollars in asset 2, equities. The HE portfolio reverses those legs: it holds equities long and shorts beta of real estate. The Table 5 notes describe that same trade, which requires h12/h11. Either the label or the formula is wrong.
The Table 6 notes attribute the S&P 500's 46.09% full-sample DCC result to Table 5's average ratio of 0.55. The text leaves open whether that ratio means h12/h22 or h12/h11. A desk needs the answer before sizing a REIT short.
Three of twelve pairs did worse in COVID than in the GFC
A systematic improvement in hedging during COVID should leave few exceptions. HE increased from the GFC to COVID in nine of twelve pairs. The three exceptions are large. UK Banking drops from 54.43% in the GFC to 38.56% in COVID, and US Consumer from 67.31% to 51.26%. UK Energy changes little, at 21.82% against 21.32%. Section 5.4 qualifies its claim, saying hedging effectiveness "often exceeded pre pandemic levels". The abstract's "remarkably improved" leaves that qualification behind. Real estate's weight against UK Banking also stayed at 0.64 through the GFC, an odd finding given the paper's emphasis on property-to-bank transmission in Britain.
The weakest full-sample DCC pairings are Energy and IT: UK Energy at 19.03%, US IT at 20.30% and US Energy at 23.65%. US Financials lead at 48.88%, just above the S&P 500's 46.09%.
The coefficient estimates also resist the prose. Rim identifies Industrials and Consumer as the largest transmitters into US REITs. Their DCC cross-shock terms, however, are 0.0356 and 0.0370, neither significant. Their cross-persistence terms are negative (-0.0645 and -0.0608). Energy, described as a weak-spillover sector, has a significant cross-shock term of 0.0385.
Rim's DBEKK check places the S&P 500 covariance shock term above Industrials, at 0.1749 against 0.1125. For persistence, the ordering reverses: 0.7749 against 0.8396. Rim reports this himself. Sector comovement starts smaller and persists longer. Section 5.5 restricts "weaker" to the initial impact; the abstract and conclusion omit that restriction.
Short estimation windows add another limit to the crisis comparison: three years of weekly data for the GFC and two for COVID. We did not find a formal test of whether the two sets of parameters differ. For the US Industrial GFC fit, the joint residual Q-test has a p-value of 0.0320. The Table 9 notes also end the post-GFC period in December 2021, whereas the text ends it in December 2019. That column may therefore overlap COVID.
The recommendation outruns the test
Rim recommends "dynamic, sector-tailored allocations and hedge ratios". The HE calculations do not test that advice. According to the table notes, each calculation takes the period's average hedge ratio and applies it to realised returns from that same period. We did not find a rolling or out-of-sample version. Borrow and transaction costs receive no mention in the paper. The 73.55% COVID result is in-sample, excludes costs and covers two years of data.
We could not test it ourselves. Our data start in 2010, leaving the GFC, the half of the comparison that matters, outside our coverage. We also lack UK REIT and UK sector indices at the required depth.
A rolling, out-of-sample REIT short could change our view if it used a consistently defined beta and showed lower GFC effectiveness than in the post-GFC period.