Portfolio Diversification: REITs and International/Emerging Bonds

Posted on September 11, 2017 by Siamond — 1 Comment

Vanguard and others have put a lot of emphasis on bonds diversification using international bonds in recent years, while the Bogleheads community mostly shrugged. This article studies the effect of such diversification through backtesting techniques, looking at both regular International bonds and Emerging Market bonds. We’ll take a close look by studying monthly returns to better analyze the volatility and correlation properties of various portfolios. Then we’ll perform a similar study about diversification of equities with domestic, global or international real estate funds.

Actual Funds

This study assumes an investor in the US having access to the following funds (in addition to more classic funds for Total Stock Market, Total Bond Market and Total International):

Vanguard Total International Bond Index Fund Investor Shares (VTIBX) – ER 0.15% (Admiral 0.12%)
Vanguard Emerging Markets Government Bond Index Fund Investor Shares (VGOVX) – ER 0.49% (Admiral: 0.32%)
Vanguard REIT Index Fund Investor Shares (VGSIX) – ER 0.26% (Admiral 0.12%)
Vanguard Global ex-U.S. Real Estate Index Fund Investor Shares (VGXRX) 0.35% (Admiral 0.15%)
iShares Global REIT ETF (REET) 0.14%

Indices and Data Source

Since most of the funds involved are fairly new, this study will make use of more extensive data series based on the indices tracked by the various funds. Both Vanguard and BlackRock (iShares) have an impressive record of reliably and tightly tracking their index benchmarks, therefore using such past index data to get more historical data seems reliable enough. Monthly data (total return) was extracted from Morningstar. Here are the index data series of relevance.

Total International Bond (currency-hedged):

Bloomberg Barclays Global Aggregate Ex USD TR Hdg USD (1990+)

Emerging Markets Government Bond:

J.P. Morgan Emerging Market Bond Index (EMBI) Global TR USD (1994+)

US Real Estate Investment Trusts (REIT):

FTSE NAREIT All Equity REITs TR USD (1972+)

Global ex-U.S. Real Estate (referred later as “International Real Estate”):

S&P Global Ex US Property TR USD (1990+)
Note that this index is quite different from the FTSE NAREIT index. This S&P “Property” index is about all types of real estate related activities, hence a broader scope than REITs (which form roughly half of the holdings). It is rather strange that Vanguard didn’t choose to track the S&P Global Ex US REIT TR USD index instead.

Global Real Estate Investment Trusts (REIT):

S&P Global REIT TR USD (1990+)
FTSE EPRA/NAREIT Global REITs TR USD (2006+)
Note that iShares REET tracks the FTSE NAREIT index, which was pretty much equivalent to the S&P REIT index in overlapping years, so it should work fine to extend the data series further in time with the S&P index.

Individual Data Series – Analyzing Monthly Returns

Analyzing annual returns keeps things simpler in a spreadsheet, but misses intra-years events which can significantly alter volatility and drawdown metrics (both directly linked to behavioral challenges for investors). In order to take maximum advantages of the limited historical data available (~20 years in some cases), the author assembled monthly data series for the indices corresponding to the various funds of interest. For simplicity and consistency, since most funds of interest are very new anyway, only index returns have been used. The same logic has been used for broader and more ‘classic’ asset classes, e.g. Total-US-Market (TSM), Small-Cap-Value (SCV), etc.

Note: to keep things as consistent and realistic as possible, the total returns from index data series have been adjusted with the expense ratio (ER) of the corresponding Admiral or ETF fund.

In addition to total return compound growth (CAGR), standard deviation, maximum drawdowns, and correlation have been assessed with the longest data series available for each asset class. Some statistics were also assembled for the 1994-2016 period to more directly compare asset classes during shared years of existence. Here is the outcome (click on the image for a bigger display):

A few observations:

International Bonds (currency-hedged) looked fairly similar to US bonds for 1994+, actually slightly better (CAGR, standard deviation/volatility).
International Bonds (hedged) correlation to any type of equities is close to zero, quite a remarkable result (while US Bonds are a bit higher, albeit still very low).
- Interestingly enough, Int’l Bonds displayed a correlation with stocks of various types nearly as low as Gold. The currency hedging actually helped a lot in creating such effect.
- A side note about Gold: when correlations are computed on an annual basis, Gold displays solid negative correlation outcomes. But that’s very coarse math. When computing on a monthly basis (like we did here), the negative correlations disappear in favor of near zero correlations.
EM Bonds displayed relatively contained volatility and drawdowns compared to its excellent returns. And a fairly weak correlation with pretty much everything else, including broader types of bonds (only slightly higher with EM stocks).
US REITs displayed a similar pattern, very solid returns, fairly weak correlation with broader markets, a higher volatility though.
Global REITs behaved quite similarly to US REITs, displaying solid potential as a diversifier.
International Real Estate looked quite poor, low returns, high volatility, high correlation with the broader international market, nothing much to like.

This being said, looking at a specific and limited period of history (here barely more than 20 years in some cases) can be quite misleading. The Total International numbers look nearly as poor as International Real Estate for 1994-2016, and yet we know from a broader historical perspective that they have been basically on par with US returns since the end of WW-II. When looking at such limited period of time, volatility and drawdowns are probably more significant results than absolute returns.

Individual Data Series – In Time of Crisis

Correlations are interesting per se, but what happens in time of crisis is probably more relevant to what happens over the overall time period. Here is a correlation table restricted to the Internet and Financial crises.

How things change… International Bonds (hedged) went straight in negative correlation territory (good), and so did US Bonds, but only for the Internet crisis. REITs displayed a much lower correlation during the Internet crisis, but not during the Financial crisis. A similar thing happened to EM Bonds, with a lower correlation during the Internet crisis than the Financial crisis.

Building a portfolio

Contrary to common wisdom, regular bonds were NOT particularly safe on their own, notably in periods of high inflation which can be especially devastating to bond investors in terms of purchasing power. Where bonds did shine was their consistently low correlation with stocks over time. This allows to construct portfolios which should be more resistant to deep crisis, while retaining solid returns. When acknowledging such facts, one can wonder what effect would have further diversification with other low-correlation assets, in this case REITs and EM/International bonds, and maybe fret a little less about their higher volatility.

Continuing to work with monthly returns series, we’ll compare a 60/40 Base Portfolio (Total US Market: 30%, Total International: 30%, US Bonds: 40%) with various portfolios using some level of additional diversification with either REITs or EM/Int’l Bonds. We’ll compare over the 1994-2016 time period, and focus on the trajectory of the portfolio balance (with the usual initial investment of $10,000) as well as the portfolio drawdowns.

Diversifying with International Bonds

Our base portfolio has 40% US Bonds. Let’s first split the bonds 20% US and 20% International – see the chart below for the (nominal) trajectory of $10,000 invested on Jan 1st, 1994. Well, if you see a difference, you have really good eyes. There is not much point providing more statistics or charts about this one! This being said, International Bonds certainly didn’t hurt. But it’s just not worth the additional complexity of using an extra fund in one’s portfolio. Stepping back from empirical data, central banks (from the largest developed countries) have been working closely together in the past decades and interest rates moved in relative sync, so this outcome isn’t that surprising, notably in presence of 60% stocks which provided the bulk of the action in this time period.

This outcome is somewhat surprising though from the perspective of the negative correlations displayed by Int’l Bonds in times of crisis. This shows that correlations are one thing, but the dynamics of absolute returns at the portfolio level are another. We often forget that correlation is about directionality, but doesn’t capture the amplitude of the change.

The author ran a quick test using 40% Global Bonds (currency hedged) instead of 20% US Bonds and 20% Int’l Bonds, and the results were unsurprisingly very similar. Now if such a Global Bonds fund were to exist with a low ER, it seems that there would little reason to stick to a US bond fund, as maximizing diversification should always be one’s starting point.

Now let’s play with Emerging Markets (EM) Bonds. Let’s split the 20% Int’l Bonds, and allocate 10% of it to EM Bonds (hence 20% US, 10% Int’l, 10% EM). Here the difference is starting to be noticeable, with a higher portfolio balance most of the time. The respective CAGRs were 6.8% and 7.2%, a non-negligible difference.

Here is the drawdown chart (note that a monthly drawdowns chart tends to exaggerate things a tad, as it picks the highest and lowest point on a monthly basis, hence a very temporary situation – hopefully our investors minds are not THAT bad with anchoring). There is a slight difference between the two trajectories, but not terribly meaningful. The diversification didn’t quite help in time of crisis, in other words.

Note that EM Bonds dipped a good deal (~20%) for a couple of months during the financial crisis, while their returns stayed positive during the entire Internet crisis. There is no magic bullet, and no crisis is the same as previous ones.

Note: allocating 20% of the bonds to EM bonds would have resulted in a more significant difference for the portfolio balance (and a slightly worse drawdown in 2008), bringing the portfolio CAGR to 7.5% for this time period. It would seem doubtful that a 60/40 investor would want to go that far in diversifying their bonds though.

Just out of curiosity, let’s replace EM Bonds by Gold (hence 20% US Bonds, 10% Int’l Bonds and 10% Gold). Gold has quite a good reputation as a diversifier, let’s check if those deep drawdowns would have been better mitigated.

Er, that is NOT convincing, to say the least. It didn’t hurt, but didn’t help any further. As to the portfolio balance, it would have displayed a somewhat different trajectory, but ended up pretty much at the same point as the base portfolio. Regular bonds proved as effective as a diversifier as Gold during those recent crisis, interestingly enough. The author’s opinion is that Gold is quite overrated, mostly because it acted so well during the oil crisis of the mid-70s, but this was really a one-time event, directly following the ending of the convertibility of the US dollar to gold in 1971.

Diversifying with REITs

Our base portfolio includes 30% US Stocks and 30% Int’l Stocks. Let’s allocate 10% of that to REITs, starting by US REITs. The resulting Asset Allocation for the diversified portfolio is 25% US Stocks, 25% Int’l Stocks, 10% US REITs, 40% US Bonds. Here is the trajectory of the portfolio balance, and then the drawdowns.

Here, the diversification paid off, with better returns for most of the trajectory (overall CAGR 7.2% vs. 6.8%), and a smoother ride before and during the Internet crisis. On the other hand, this didn’t help whatsoever during the financial crisis, which is hardly surprising as the primary root cause of said crisis was subprime mortgages. Overall, it seems to make sense that real estate would go through cycles which are often independent from the overall stock market, but not always.

Now, if we had used Global REITs in lieu of US REITs, the effects would have been very similar, albeit a tad more subdued (overall CAGR 7.1%). The trajectory is so similar that we’ll skip providing the corresponding graph.

But replacing those REITs by 10% International Real Estate (e.g. the index followed by Vanguard VGXRX) would NOT have been beneficial. Not only was the performance sub-par pretty much all the time (overall CAGR down to 6.6%), but the drawdowns during the two major crisis would have been as painful as with the base portfolio. Int’l Real Estate was just not a good diversifier (at least in this time period).

Note: the author ran a quick test with an S&P index truly capturing the ex-US REIT market, and things would have been significantly better — IMHO, Vanguard may want to revisit VGRLX/VNQI…

Another type of diversification

Coming back to the correlation table displayed at the beginning of the article, one can observe that REITs (US or Global) were more correlated with Mid-Cap-Value (MCV) and Small-Cap-Value (SCV) than with the US market as a whole. This begs the question of using MCV or SCV as a diversifier in lieu de REITs. Here is the trajectory when using 5% of each (hence 25% US Stocks, 5% MCV, 5% SCV, 25% Int’l Stocks, 40% US Bonds).

The trajectory would have been very close (slightly better, actually) as the 10% US REIT scenario (and the drawdown chart is very similar as well). In the past, the ‘value effect’ has been highly dependent on the time period, so maybe this is just a coincidence, or maybe not, as many real estate companies are not that large and often a tad troubled (hence likely to fall in the MCV/SCV category). If this hypothesis is true, then one should probably choose between a (small/medium) value tilt and a REIT tilt, but not do both.

Combining the diversification ideas

Coming back to the observation that EM Bonds displayed weak correlation with pretty much everything else, we can try to combine the stocks and bonds diversifications. An intuitively appealing idea would be to diversify stocks with 10% Global REITs, and diversify bonds with 10% Int’l Bonds and 10% EM Bonds. Here is the outcome for the time period being studied, displaying an overall CAGR of 7.5% for the diversified portfolio.

Overall, the trajectory of the diversified portfolio definitely improved, with significantly higher returns and riding the Internet crisis better than the base portfolio. The financial crisis was just the perfect storm though, and would have hurt equally bad (while recovering in pretty much the same way).

Practicalities and fees

In practice, some of those investment vehicles tend to come with higher expense ratios than regular index funds, but also with purchase and/or redemption fees. Case in point (at the time of writing):

Vanguard Emerging Markets Government Bond Index Fund (VGOVX/VGAVX) comes with a hefty 0.75% purchase fee (but no redemption fee).
Vanguard Global ex-U.S. Real Estate Index Fund (VGXRX/VGRLX) comes with a 0.25% purchase fee and a 0.25% redemption fee.

The corresponding ETFs do not have anything similar, but part of the extra cost probably goes in the bid/ask spread when buying or selling such funds. Note that it appears that Vanguard rounds up such purchase/redemption fees to the nearest quarter of a point, while returning the extra income (if not spent in transaction costs/etc) to the fund itself.

Buy and hold investors should not be overly impacted by such fees (after all, they occur only once), but rebalancers might cringe a bit. This is a maturity issue, once (and if) those funds become large enough and once buys & sells transactions become more balanced, Vanguard should logically eliminate such pesky fees. In the mean time, it might make sense to use the ETF funds, while paying close attention to the bid/ask spreads (ETF.com shows interesting statistics in this respect).

Conclusion

Diversification is a wonderful thing, and one should not hesitate to explore the option of using asset classes with a low/weak correlation with the overall market. The outcome may not always be worth the additional complexity though:

International Bonds (at least in a currency hedged flavor) did not seem to add much (nor hurt in any way). If Vanguard or BlackRock would offer a low cost Global Bond fund, it would probably be reasonable to go for it, but splitting US Bonds to create room for International Bonds doesn’t seem terribly useful.
Emerging Bonds are more intriguing. It might be perceived as a risky asset class (and it is when analyzed on its own), but when used as part of a portfolio to replace some of the bonds portion, its historical high returns combined with fairly weak correlation with pretty much everything else would have added a little zing to one’s portfolio without significantly impacting volatility or drawdowns.
US REITs (or Global REITs), probably thanks to the decoupled cycles in real estate, would have helped quite nicely to navigate the Internet crisis, and displayed better returns overall. International RE would have proved to be quite the bust though.
Finally, there was clearly a certain redundancy between the (small/medium) value effect and REITs, and both would have helped in pretty much the same way.
The Financial crisis was just the perfect storm, and nothing we explored would have helped more than good old regular bonds.

Again, two decades of historical records is a short time period, there are multiple examples of historical cycles in finance which lasted longer than that, so we should not jump to conclusions here. Also every crisis is different from the previous ones. It is therefore important to associate such numbers analysis to some intuitive perspective. REITs (and Value companies) acting on their own does seem to make intuitive sense (although they may sync up, e.g. financial crisis). Maximizing bonds diversification would seem desirable out of principle. As to Emerging Bonds, their very nature should lead to distinct dynamics from anything else. Time will tell, but those seem reasonable hypotheses, at least to the author.

PS. feel free to provide feedback and comments on the following Bogleheads forum thread: https://www.bogleheads.org/forum/viewtopic.php?f=10&t=227434