Welcome to a new installment of the Safe Withdrawal Series! The last post on the Yield Illusion (Part 29) was definitely a discussion starter! 140 comments and counting! Just as a quick recap, fellow bloggers at Millenial Revolution claim that the solution to Sequence Risk is to simply invest in a portfolio with a high dividend yield. Use the dividend income to pay for your retirement budget, sit back and relax until the market recovers (it always does, right?!) and, boo-yah, we’ve solved the whole Sequence Risk issue! Right? Wrong! As I showed in my last post, it’s not that simple. The Yield Shield would have been an unmitigated failure if applied during and after the 2008/9 Great Recession. So, not only did the Yield Shield not solve Sequence Risk. The Yield Shield made it worse! And, as promised, here’s a followup post to deal with some of the open issues, including:
- A more detailed look at the reasons for the Yield Shield Failure over the past 10 years (attribution analysis).
- Past performance is no guarantee for future returns. How confident am I that the Yield Shield will fail again in the future?
- Dividend Yield vs. Value
- Are non-US investors doomed? Probably not!
So, let’s look at the details:
An autopsy: Attribution analysis of the Yield Shield failure post-2008/9
Going from a simple and clean 60/40 portfolio as in the Big ERN world to the convoluted mess of the Yield Shield portfolio took a lot of work. To understand how much of each of the individual pieces of the Yield Shield contributed to its failure, let’s look at the following sequence of portfolios, changing one portion of the portfolio at a time:
- Portfolio 1: 60/40 as defined here in the SWR Series: 60% S&P 500 index and 40% 10-year Treasury Bonds, closely approximated with the IVV and IEF ETFs, respectively.
- Portfolio 2: Replace 10% of the IEF with AGG (U.S. aggregate bond index, i.e., including investment-grade corporate bonds). Note: AGG is an almost identical substitute to the BND.
- Portfolio 3: Replace another 10% of the IEF with LQD (exclusively corporate investment-grade bonds)
- Portfolio 4: Replace the remaining 20% of IEF with PFF (Preferred Shares)
- Portfolio 5: Replace 30% of IVV with VEU (non-US stocks)
- Portfolio 6: Replace 15% of IVV with VTI (US total stock market)
- Portfolio 7: Replace 10% of the IVV with USRT (US RETIs)
- Portfolio 8: Replace 5% of the IVV with VYM (US high-dividend yield)
Let’s see what was the marginal impact of each of the changes. Let’s start with just the plain portfolio stats, not factoring in withdrawals (yet), but simply looking at return, risk, Sharpe Ratio (=risk-adjusted return), worst drawdown, and a simple “factor” model to estimate the exposures to stock and bond betas.
- Returns get worse, the risk gets worse, the Sharpe Ratio gest worse and the drawdowns get worse as you move along the steps to the Yield Shield portfolio. There is one tiny improvement to the mean return in step 6 (replacing the IVV with the extremely highly correlated % VTI) but that’s very minor. Also, the Sharpe was the same and the drawdown a little bit worse.
- Someone in the comments section conjectured that most of the underperformance is due to the overweight in international stocks. And that’s certainly true, notice the big step down in the average returns and Sharpe Ratio came about in step 5 when adding international stocks. But notice that Preferred Shares (added in Portfolio 4) take your portfolio down pretty dramatically as well; not so much in the average return but the drop in the Sharpe Ratio is awful. So is the Worst Drawdown.
- Is the underperformance just a statistical fluke? Of course not! As you go from left to right you see that you keep increasing your equity exposure, as estimated by a simple style analysis (factor model). To the untrained eye, you may think you keep a 60% equity, 40% bond exposure. But in portfolio 4, when you supposedly only reshuffled the fixed-income portion you’ve gone to an effective 75/25 portfolio. No wonder it had steeper drawdowns and less diversification. Shifting from government bonds to corporate bonds and preferreds introduces equity beta through the backdoor!
- Messing with the equity portion (portfolios 5-8) takes the equity beta to above 83%. Again, not by increasing the equity share but by shifting into higher risk and higher beta instruments. The Gov. Bond beta is also above 40% again (thanks in large part to REITs), so the whole Yield Shield portfolio was essentially a leveraged bet on equities plus bonds but minus a pretty bad portfolio alpha.
What about the portfolio performance when we take into account the withdrawals? In the chart below, I plot the time series of the portfolio values when we withdraw 3,333.33 each start of the month (i.e., 4% annualized). This is with double inflation adjustment: 1) withdrawals are adjusted by CPI and 2) the portfolio value is CPI-adjusted. The biggest step down is when you introduce international stocks. But notice how even the fixed-income portfolio adjustments already cost you almost $90,000 by 2019.
Let’s also dig a little bit deeper into the Yield Shield underperformance. In the table below, I display the final portfolio value, i.e., the endpoint of the 8 lines in the chart above:
- We can dispell the myth that this is all due to bad luck of picking international stocks. Or REITs. 28% of the underperformance ($81,205 out of $288,342) is due to the fixed income portion. True, that’s less than the 40% share in the portfolio but keep in mind that the fixed income portion also has lower risk, hence the slightly lower potential for messing up your portfolio.
- It is indeed true that the largest chunk of the Yield Shield underperformance came from the shift to international stocks. Out of $288,342 underperformance, $190,705 or about two-thirds came from shifting into VEU. But all of the other moves also look pretty bad.
- Quite intriguingly, the one single step that – at the margin – increased the final portfolio value was replacing 15% of IVV (S&P 500) with the VTI (U.S. Total Market) and that was the one step that decreased the overall portfolio dividend yield. In other words, the one step that (very marginally) helped you with Sequence Risk was going against the Yield Shield plan of increasing yield! So, if you calculate how much final portfolio value you gained per basis point (bps=0.01%) of additional dividend yield, all seven measures are negative! The inventors of the Yield Shield are either the most unlucky or the most incompetent financial planners: they are 0 for 7! It’s so bad, you almost can’t make it up!
In the bottom portion of the table, I also like to dissect the Sequence Risk a little bit more:
- Start with the nominal return without withdrawals of each of the portfolios. Example: 6.36% for portfolio 4.
- Deduct CPI inflation (~1.73% annualized over this time span). Example: 4.56% for portfolio 4.
- Deduct your 4% real withdrawal mount. Example: 0.56% for portfolio 4. In the absence of sequence risk, Portfolio 4 should have grown by (approximately) this much after the withdrawals.
- But the actual CAGR net of withdrawals can be very different. Example: portfolio 4 actually shrank by 0.36% p.a.,…
- … and that difference is the impact from Sequence Risk! In the case of Portfolio 4, it’s -0.91%!
- We can also assign the marginal impact on this Sequence Risk impact as we move from the 60/40 portfolio over to the yield Shield.
The reason I like this type of attribution is that we’re able to separate the underperformance due to simply the average return vs. “pure Sequence Risk.” Remember: This isn’t so much about the average return, it’s about the order (sequence) of returns, hence the name. The Yield Shield worsened the SoRR effect from -0.51% to -1.20%. Preferred Shares were actually the worst offender. More than half of the deterioration came from the three fixed income moves, almost half from Preferred Shares alone. And again: International stocks drag down the final portfolio value not just because of their inferior average return but also the bad sequence of returns!
So, this was my pretty thorough autopsy of the Yield Shield method. The conclusion is that this horse is really, really dead. No need to beat it any further!
We are all familiar with the “Past performance is no guarantee of future results” disclaimer. And this disclaimer applies here as well. Just because the Yield Shield would have failed the last time around doesn’t automatically mean that it will fail again. But we should also recognize the great irony here: Normally, when people propose “better” asset allocation recipes it’s the critics who invoke the “Past performance is no guarantee…” disclaimer. And the proponents bury that pesky little disclaimer deep in the fine print. Here it’s the other way around! The inventors of the Yield Shield would have to invoke it, hoping that their proposed solution to shielding investors from Sequence Risk doesn’t fail quite so miserably next time. Actually, “not failing” is not enough, right? Let’s not shift the goalpost here because you should do better than the passive 60/40!
So, how confident can we be about the Yield Shield failing again in the future? It depends. Different asset classes will have a different likelihood of repeating the past experience:
We can say with near-certainty that the fixed-income portion of your portfolio will behave very similarly to the 2008/9 episode once the next big bad bear market comes along. In other words, in the next big recession, higher-yielding corporate bonds will underperform the safe and low(er)-yielding government bonds. But, you ask, wouldn’t corporate bonds benefit from a likely rate cut by the Federal Reserve? True, but the recession will also raise credit spreads, i.e., the difference between corporate bonds and safe government bonds. And that means that in financial stress scenarios, low-yielding government bonds will handily outperform high-yielding corporate bonds. With great regularity; it has happened in every recession in the past (see chart below) and it will surely repeat in the future.
Also a word about Preferred Shares: people have pointed out that most of them are issued by financial institutions and that clearly hurt them during the financial sector meltdown during the Great Recession. 100% true! But again, let’s not shift the goalposts; preferreds don’t have to perform as poorly as in 2008/9. They merely have to perform worse than safe government bonds. And preferred shares will not only be subject to that same corporate credit spread issue above, but they will also suffer even more due to their subordinated debt structure. Unfortunately, I don’t have a long enough history of either ETF or a Preferred index performance at my fingertips to test my hypothesis prior to 2007. If anyone has a longer data series of Preferred returns please let me know! I’m sure they will look similar to the High-Yield bonds in the chart above!
U.S. vs. international Stocks:
I am reasonably certain that during the next U.S. recession and the deep bear market that goes along with it, there will be very little diversification from international stocks. As I wrote in the post from a while ago, if we see double-digit losses in the U.S., your non-U.S. portfolio will get hammered, too. It has happened in every single recession in recent history, see the chart below from an old post of mine:
That said, I have no idea how international stocks will perform during the subsequent bull market. If you look at that funnel-shaped cluster of dots above you’ll notice that the widest dispersion of dots occurs to the right. There have been instances where the U.S. return was +20% year-over-year and the non-U.S. return ranged between -10% and +70%! So, in all fairness, there could be an opportunity for international stocks: maybe they will outperform U.S. stocks again during the next Bull Market (whenever that may be). It has happened before, see the summary table from that old post of mine below!
You see, from a Sequence Risk perspective, it matters not just how deep the bear market is and how long it lasts but also how long it takes for the subsequent Bull Market to recover. In other words, your Sequence Risk worries aren’t over until the Bull Market takes your portfolio back to “normal” to recover 1) the loss during the Bear market, 2) Inflation and 3) your withdrawals. There have been instances in the past where the equity market recovery was significantly stronger abroad (e.g. 1982-1987 and 2002-2007). As I wrote in that post in 2017, it is definitely noteworthy that non-US stocks significantly outperformed US stocks following the 2000-2003 bear market (see chart below). But let’s be clear: Nobody in their right mind would claim that this is because of the higher dividend yield! It had to do with (at least) two effects 1) FX movements: the USD depreciated against other currencies, and 2) macroeconomic divergence.
And by the way, the underperformance of non-US stocks since 2009 is also largely due to those two effects simply working in the opposite direction: 1) U.S. macroeconomic performance, though weak in absolute terms, was strong relative to other countries, and 2) the USD appreciated again and that makes the equity performance abroad weaker from a USD-based investor. It had nothing to do with dividend yields.
So, what can we predict now? Nothing. Absolutely nothing! That’s because I can predict neither macroeconomic trends nor FX movements that far into the future. I don’t want to make the case that VEU will underperform again and others shouldn’t make the case that VEU will necessarily outperform the U.S. index. Certainly not because it has higher yields! As a working assumption, I’d simply use the same expected returns here in the U.S. as in the ex-U.S. ETFs. So, in other words, my conclusion on the U.S. vs. International stock market is significantly weaker than for the fixed income bucket. For the fixed income assets, higher yield, almost by definition, means higher macro risk and worse performance next time we have an adverse Sequence Risk event. For equities, the higher-yield approach doesn’t necessarily make Sequence Risk worse but there’s also no guarantee it would necessarily alleviate Sequence Risk. Looking forward, it’s a wash! That’s still a far cry from the absolutely asinine promise of the Yield Shield “protecting” or “shielding” (their words, not mine) you during the next recession.
I’m no big fan of REITs. I think there are superior investment vehicles to get access to real estate: either manage your own properties – check Coach Carson’s blog for an excellent intro (I especially like the post on house hacking) – or take the more passive route through one of the real estate syndication platforms. Or even through traditional private equity funds as we do here in the ERN portfolio; very hands-off and passive! For investors who are absolutely opposed to those routes, sure, go ahead and invest in REITs. They will likely not get hammered as badly as in the last recession. But I see no reason why REITs would deliver reliably superior total returns during the next Sequence Risk scare. It’s the same conclusion as in the U.S. vs. International discussion above. I can provide theories for why REITs will outperform next time – if there’s an inflation shock REITs are a good inflation hedge – but also why they may get hammered again in the next bear market: They’re over-priced already plus too much leverage could bite you if there’s a spike in interest rates. Ex-ante, it’s a wash!
Wow, I only got about halfway through what I wanted to accomplish today and we’re already close to 3,000 words! I think I will cut this post into two and publish the remainder later this week. There, I will summarize some more of my thoughts on the “Looking forward” section, elaborate more on my views on the U.S.-high-dividend tilt and some other issues. Have a great start of the week, everyone!