Using Gold as a Hedge against Sequence Risk – SWR Series Part 34

January 20, 2020 – Happy New Year! It’s time for another installment in the Safe Withdrawal Series! Here’s a topic that I’ve thought about for a while and that was also requested dozens, maybe even hundreds of times from commenters: What about gold? Gold has been a safe haven asset for many decades (Centuries? Millenials???) and it should have the potential to hedge against Sequence of Return Risk. And I recently found this article on Yahoo Finance: “The world’s super-rich are hoarding physical gold“. Maybe it’s just click-bait. Yahoo Finance must have lowered its standards substantially because they even (re-)published one of my articles last year. 🙂

But seriously, in light of the recent runup in gold prices, rising interest by the world’s super-rich, and the many requests by readers, I’ve finally gotten around to studying this subject in the context of Sequence Risk. Let’s take a look at how useful gold would be as a hedge against running out of money in retirement…

Historical asset class performance

Before I even run any safe withdrawal rate simulations, let’s take a step back and get a sense of how the different major asset classes – stocks, bonds, bills/cash, and gold – performed over the last roughly 100 years. Here’s a table with the annualized, real CPI-adjusted since the mid-1920s. So we would cover the last 10 major bear markets:

Returns 1/1925 to 9/2019. Stock (S&P 500 Total Return), Bond (10Y Treasury), U.S. T-Bill (1Y before 1934, then 3M), and Gold (London A.M. fixing). Based on Monthly data, annualized, adjusted by U.S. CPI inflation

Well, Gold had a positive return, even slightly better than very short-term fixed-income assets like T-Bills. But the average return was lower than that of intermediate-term bonds (10-year Treasuries). On top of that, the risk was also higher. So, should we throw out the idea of gold as a hedge? Well, not so fast! The average return is almost as high as that of bonds and the standard deviation/risk is a two-edged sword, of course. Higher risk doesn’t necessarily have to be bad. Actually higher risk is good if gold vastly outperforms during recessions and bear markets! Diversification is worth giving up a little bit of expected return!     (But for the record, I also still love the Warren Buffett quote: “Gold gets dug out of the ground […]. Then we melt it down, dig another hole, bury it again and pay people to stand around guarding it. It has no utility. Anyone watching from Mars would be scratching their head.”)

So, let’s take a look at those 10 most recent bear markets I studied in my post “Who’s Afraid of a Bear Market?” from a few months ago. Specifically, I like to see how the four different asset classes performed during those severe equity market drops, please see the table below.

Real CPI-adjusted asset Performance following the last 10 prominent equity market peaks (bear market starts). Not annualized, this is simply the performance during the bear market, i.e., the row “2000” refers to the 08/2000 to 09/2002 performance. Note that the mean/median/standard deviation stats are the simple, equal-weighted stats. They are not weighted by the bear market length!

Well, suddenly gold looks pretty good, doesn’t it? It has higher average and median returns during those 10 equity market drawdown events than any other asset class. But just to be sure, there were a few bear markets where even gold lost: 1946, 1968, and 1980. So, gold is still not a fool-proof bear market insurance! But it certainly helped during the two worst market environments for retirees: the Great Depression and the 1970s.

Talking about severe vs. not so severe bear markets, we can make the gold performance look even a little bit better than that if we concentrate only on the five bear markets that posed a serious Sequence of Return Risk problem: 1929, 1937, 1968, 1972 and 2000. And yes, you saw this right: 2007 is not included in that list; despite being the second-deepest equity market drawdown after the 1929-1932 crash, the swift recovery and the record-long bull market that followed (almost 11 years and counting) make the 2008/9 bear market a relatively benign market event for retirees. At least in hindsight!

Asset class return during the worst bear markets from a Sequence of Return Risk perspective.


Before I forget I have to get one big caveat on the record: In the U.S., gold ownership was severely restricted between 1933 and 1974. So, we can certainly run the simulations with this reported price for gold but it’s not clear how the average investor would have used gold as a recession hedge during that 4-decade time span. So, what I’m doing here is not really an exercise of “what if someone had bought a thousand ounces of gold back in 1937” because that was impossible at the time. The simulations here have to be run under the assumption that future asset returns hopefully behave like those in the past. If they do, is it worth investing in gold in the future? And just to be sure, this assumption about the statistical distribution of returns and correlations and comovements over the business cycle can be a bit of a stretch because one could argue that without the government intervention gold prices have evolved differently during those four decades!

Also, I will use gold ETFs in the simulations below. Of course, before 2004 when GLD (SPDR’s gold ETF), came out they didn’t even exist (but correct me if I’m wrong!). One could have bought physical gold, which would save you the annual expense ratio of the ETF (normally around 0.2% to 0.4%) but that would also potentially cost you more due to large transaction costs for trading gold bars and coins: large bid/ask spreads and in some states, you even have to pay sales tax when you buy gold coins/bars!

With all those limitations in mind, let’s do some simulations…

How to add gold to the portfolio

So, how would an allocation to gold impact our safe withdrawal rate analysis? Let’s make some more assumptions about the inputs here:

  • A 30-year horizon,
  • A final value target of 25% of the initial portfolio. Not too different from the average, typical retirement portfolio. For example, this could be a traditional retiree, age around 65, with a 30-yer horizon who likes to hedge a life expectancy to age 95 and leave a modest bequest, 25% of the initial assets, to the kids. Or an early retiree, age 40, who wants to bridge 30 years of early retirement but then still have 25% of the assets at age 70 to supplement Social Security and corporate pensions.
  • No other cash flows,
  • The baseline portfolio has 75% stocks (S&P 500 Total Return Index) and 25% bonds (U.S. 10-year benchmark Treasury bonds).

If we like to add 10% gold to this, there are (at least) three different ways to go about this:

  1. Reduce the equity portion to 65% to add 10% gold
  2. Reduce the bond portion to 15% and add 10% gold
  3. Keep the 75/25 allocation and add 10% gold on margin (e.g., through futures contracts). The way to implement that in the SWR Google Sheet (see Part 28 of this series) is to allocate +10% to gold and -10% to cash. The returns for your 10% leveraged gold portfolio would then be exactly the gold minus the short-term T-bill return, which is, lo and behold, roughly the expected return of gold futures.
Implementing Gold into the SWR simulations is really easy. For example, case 3 involves 10% gold and -10% cash allocation (gold on margin). Note how this raises the weighted expense ratio to 0.075%.

I assume that the gold ETF in cases 1+2 has an expense ratio of 0.25% p.a. (just the widely-used iShares ETF IAU), equity and bond ETFs have 0.05% and cash (CDs, MM, Bills) has a zero expense ratio. Actually, getting gold exposure through futures contracts (case 3) will likely be somewhat cheaper than the 0.25% annual expense ratio.

So, let’s look at the results. See the table below. Indeed, allocating to gold improves the failsafe withdrawal rates from 3.58% to 3.96%, 3.64% and 3.75% in the three different gold allocation scenarios. Moving 10% from equities looks like the overall best route; you get the by far best improvement in the failsafe withdrawal rate. Also, notice how the 4% Rule failure rates are still pretty high if we use all simulations since 1871. But gold does improve your results since 1926 when a lot of other SWR simulations start, e.g., the Trinity Study.

Adding 10% gold, three different ways.

How much gold to add to the portfolio

Having decided that the ideal way would be to reduce the equity portion and replace it with a gold ETF, let’s see how much we should replace. Let’s reduce the equity allocation from 75% to 50% in 5% steps and add gold accordingly, going from 0% to 25% (while keeping the 25% bond allocation) and see how that changes our safe withdrawal stats:

  • You get the highest failsafe at 15% gold, i.e., 60/25/15 allocation so S/B/G. Beyond that, you start lowering your failsafe again.
  • With a 15% gold allocation the post-1920 failsafe doesn’t even occur at the 1929 or 1960s/70s market peaks. It’s right before the 1937 equity market peak.
  • Actually, the failure rates of the 4% rule are still really elevated if we consider the entire simulation period starting in 1871. That means there were many periods when the gold allocation didn’t work pre-1920. But we could argue that the 1871-1920 era also had a very different monetary policy (essentially no monetary policy at all) so it may not be comparable to today.
  • Post-1920, the 60/25/15 allocation also had the lowest failure rate of the 4% Rule.
Adding gold and removing equities in 5% steps.

So, in this particular example, shifting 15% from equities to gold looks best.

Other portfolios involving gold allocations

Three other variations of portfolios that tons of readers have inquired about:

1) The Permanent Portfolio:

  • 25% Stocks
  • 25% Bonds
  • 25% Cash (i.e., short-term, 3-month T-bills or CDs or money market, etc.)
  • 25% Gold
  • Further reading:

The rationale here is that it offers something for every market condition. In other words, you’d always have (at least) one asset class that’s appreciated and thus you’d have assets to liquidate without having to dig into deeply depreciated assets. For example, if you find yourself in an equity bull market, you sell equities. If we’re in a demand-side (deflationary) recession and bear market (2001, 2008/9), stocks may be down but bonds will rally. If we’re in a supply-side and thus inflationary recession (e.g., in the 1970s, early 80s), you sell gold because both stocks and bonds got hammered.

And when all else fails and all risky asset classes (equities, bonds, and gold) are down you got your T-bills, CDs, and money market accounts. Sounds intuitive, right?!

2) The Ray Dalio/Bridgewater “All-Weather Portfolio”

  • 30% Stocks
  • 55% Bonds
  • 15% commodities, of which half are in gold (e.g. iShares’ IAU) the other half in an overall commodity index ETF (e.g. GSG). I don’t have a long enough return history for the overall commodity index – the GSCI index exists only since 1970 – so I can only simulate this assuming a 15% allocation to gold, instead of 7.5% gold + 7.5% all commodities.
  • Further reading:
  • Just as an aside, this portfolio is not the actual allocation of the famous Bridgewater hedge Fund. You have to do something a lot more sophisticated and profitable to justify charging clients 2 and 20 (2% management fees and 20% of the gains)! So, don’t get fooled by some hacks telling you that they can show you the “super-secret” allocation of Bridgewater! Ray Dalio will keep the intellectual property that generates hundreds of millions of dollars in fees every year to himself!!!

The appeal of this portfolio comes from the so-called risk-parity allocation. The portfolio shares are (very approximately) inversely proportional to the asset class volatilities. (side note, the actual calculation is a bit more complicated because you set the weights of the portfolio so as to equalize each asset’s contribution to the overall portfolio variance, which also needs to factor in the covariances!). So, you’ll avoid the issue of one single asset class taking up the overwhelming part of your portfolio risk (such as equities causing 90+% of volatility in a 60/40 portfolio) and it gives you some of the same flavors as the Permanent Portfolio: something for every market and macro condition.

I’m running two versions of this. First, with a 55% allocation to intermediate (10-year) Treasuries. But notice that not all bonds are created equal. Frequently, the recommended 55% bond allocation is comprised of 15% shorter-term bonds (e.g. 5-year Treasuries) and 40% long-term bonds (20+ years maturity). But I only have long-term monthly returns for the 10-year U.S. Treasury benchmark bond, so I have to improvise here.  With a 105% bond and -50% cash allocation (i.e., 55% bonds plus another 50% bonds on margin) we would approximately match the weighted average maturity and duration of the Bridgewater portfolio.

3) The Golden Butterfly:

  • 40% U.S. Stocks, of which half is held in a broad market index (e.g. total market or at least S&P 500) and the other half is in small-cap value stocks
  • 40% Bonds
  • 20% Gold
  • Further reading:

I run the simulations of the GB portfolio once without the small-cap value bias and once with it because I like to see how much of the change in results is due to the 40%/40%/20%  S/B/G allocation and how much comes from the Small-Cap-Value tilt.


Let’s see how these portfolios do from a Sequence Risk perspective, see the table below. I also include the baseline portfolio (75% stocks, 25% bonds 0% gold) and the 60/25/15 portfolio for comparison. Overall a pretty disappointing picture:

  1. The Permanent Portfolio had the lowest failsafe rate (2.61%). What’s really sneaky about this portfolio is that you certainly improve the performance during the well-known stress periods (1929 crash, 1960/70s) but you create a serious deficiency elsewhere (1930s, more specifically the 1937 bear market) with a sub-3% safe withdrawal rate. It’s like squeezing a balloon! You solve the problem of the 1929 crash but create an equal-size problem or worse elsewhere. Not much of a surprise here; I’ve always had my doubts about the permanent portfolio because you have only 25% of your portfolio in truly high-return assets (stocks) and the rest is exclusively invested in assets that had very little real inflation-adjusted return (cash, gold). And bonds certainly had a decent real return historically, but currently yields on 10-year Treasury bonds only just about reach the inflation rate. 30-year bonds have a yield less than 1 percentage point above most inflation forecasts. Not enough to sustain a 30-year retirement!
  2. The All-Weather Portfolio suffers from the same problem. You do remarkably well in the 1929 crash. But then you open the door for really bad performance in the 1930s and 1940s, i.e., during times when the baseline portfolio didn’t have any problems sustaining a 4% withdrawal rate. Robbing Peter to pay Paul! Also, the all-weather portfolio with an elevated bond duration will have a lower fail-safe in the 1960s than even the baseline portfolio without any gold (3.57% vs. 3.58%). That’s a disappointment, but not a huge surprise: The bigger the bond exposure, the worse the pain during the inflationary 1970s! And both AW portfolios and especially the long-duration one have some of the worst failure rates for the 4% Rule I’ve ever seen. Not a pretty picture!!!
  3. The Golden Butterfly Portfolio (with small-cap-value) will give you a decent improvement in the SWR stats relative to the 75/25/0 portfolio. But not noticeably better than the 60/25/15 portfolio. But notice that all of the improvement of the GB portfolio comes from the stellar historical performance of Small-Cap-Value stocks and not so much from the gold allocation. Without the boost from small-cap-value, the SWR stats are actually worse(!) than under the baseline portfolio without gold! Well, if you remember my post from last year “My thoughts on Small-Cap and Value Stocks“, my personal working assumption is that SCV will likely not give you the outsized excess returns it might have given you historically. Blame the fact that this style is now so well-known and there are a lot of ETFs and lot of retail and institutional investor money chasing this premium. It stopped working about 15 years ago, right around the time when it became popular!
Some of the widely-cited exotic portfolios. Curb your expectations!

So, the widely-cited exotic portfolios don’t exactly deliver any notable improvement in the safe withdrawal stats. I’d stay away from them! If you want to use gold to hedge against sequence risk, shift some of the equity portion into gold. But stay away from the “sexy” portfolio allocations recommended by the internet gurus and motivational speakers!


Just for full disclosure, except for a quarter-ounce Gold American Eagle coin I got from my late Uncle Karl I own no gold. So, with a little bit of confirmation bias, I set out to prove that gold has no place in a retirement portfolio. The average return over time is simply too low. But gold shines (pardon the pun) when all the other asset classes are hurting. And that’s a huge benefit! Not so much in the pre-1920s era but at least during the last 100 years and during some of the well-known bear markets and recessions.

I particularly like the fact that gold seems to work well both during inflationary recessions (1970s) but also during the bad demand shocks that conjured up fears of a deflationary scenario (2008/9).

So, am I going to increase my gold holdings? Probably not. At least not right now. Here are two reasons:

  1. Procrastination: Yeah, I admit it, I’m a bit of a procrastinator. But there is even a good rationale to stick with equities; let’s just ride the equity momentum a little bit longer. I don’t see any imminent risk for a recession around the corner and I hope that the music doesn’t stop playing in this financial version of musical chairs anytime soon. Update: oops, that didn’t age well! Just two months later, we had the 2020 recession. And gold would have worked quite well as a diversifier again!
  2. Inertia: Making a major change is hard! For me to go from 0% gold to 10-15% gold allocation would be a major shift and I’d need to see evidence almost “beyond reasonable doubt” to make that move. But with the lingering doubts about whether gold will perform as well as in the past (see the disclaimer and the caveat about the gold ownership restrictions and the government fixing prices above), I’m still on the fence about putting a six-figure sum into some useless metal just sitting around and not generating any dividends. But likewise, if you currently do own 10-15% you probably don’t see any clear and convincing evidence to move out of gold either.

So much for today! Maybe readers can convince me to shift from equities into gold now?!

Thanks for stopping by today! Please leave your comments and suggestions below! Also, make sure you check out the other parts of the series, see here for a guide to the different parts so far!

Title Picture Source:

153 thoughts on “Using Gold as a Hedge against Sequence Risk – SWR Series Part 34

  1. Thr Permanent Portfolio and All Weather usualt have a better Sharpe than 75:25, but the volatility is much lower.

    I wonder if the lower risk is the driver of the lower SWR?

    Did you try comparing All Weather with 1.5x leverage to 75:25?

    Or with the PP you could at least swap the Tbills for more stocks, bonds and gold and get 1.33x the risk without using margin.

    1. If I scale up the weights from the AW portfolio by 1.5 (and also set eh cash weight to -50%) then I get the following SWRs by decade:

      1920s 5.25%
      1930s 3.95%
      1940s 3.85%
      1950s 4.13%
      1960s 4.05%
      1970s 4.82%

      It’s an improvement relative to the unleveraged AWP. The leverage helps getting the expected return up.

          1. I tried 70/15/15 Stocks / Cash / Gold and it had roughly 4% SWR All Time in your toolbox. I think I’m missing something but having basically no bonds and having a Cash allocation instead of Cash Bucket appeals to me. Any thoughts?

              1. “You can have your emergency fund and eat it too”. It’s especially neat since coping with 1929-deflation is supposed to be the main rationale for including long dated treasuries. Potentially being able to leave them out while still having 70% stocks looks to be kind of sweet spot for both decent SWR and long term CAGR.

                1. Yeah, it’s a nice historical performance. Not sure if/when this asset allocation enters the mainstream though. I haven’t seen any sizable gold allocation in target date funds post-retirement. 🙂

  2. Britain and the US both left the gold standard in the early 1930’s. Prior to that gold and silver were money, Sovereign Pound coins for instance in the UK. People may have been inclined to deposit that gold/money, in return for interest. Pre 1934 set gold = cash interest for ‘gold’. When the gold standard ended Americans may have been inclined to ask for their gold (money) back, maybe holding/storing physical gold in Canada/Mexico as it was outlawed in the US. Set Custom1 spreadsheet data to that combination of pre 1934=cash, since 1934 = gold, and 50/50 stock/custom1 supported a 4% SWR in nearly all cases since 1871, just marginal failures when that wasn’t the case. Next factor in the the opportunity costs of holding 50% in gold, the capacity to deploy that into stocks at appropriate times such as after sizeable declines in stocks (at which times gold had often spiked in price), and the average case final portfolio value after 30 years of 4% SWR is inclined to be uplifted, considerably in many cases. Robert Lichello’s AIM for instance might be used as a guide for that, where each time it buys you up stock weighting by 10%, so 50% initial, 60% after the first AIM buy signal, 70% on the second signal …etc. In many cases you’d end 30 years with 100% stock where half of that stock was averaged into at decent discounts (typically 25%).

  3. Hey Karsten,

    Really love the all the work you’ve done. As a actuary, your comment about convincing evidence got me thinking about how the challenges of estimating tail risk might effect the our ability to make confident decisions. Particularly in the case for gold, but more broadly for all asset classes, the SWR is highly influenced by a small number of events / non overlapping periods.

    Have you done any significance testing around the SWR’s and what might be a rough threshold to be confident in saying one strategy outperforms another?


    1. I’ve put some thought into the overlapping window issue.
      An example: you do any kind of regression analysis, say, regress the realized SWR on the CAPE yield at the beginning of retirement. You can run a regression and the OLS estimates are “BLUE” (best linear unbiased estimator) but the t-stats are wrong because of heteroscedasticity (i.e., the estimation errors between two neighboring retirement months are highly correlated). But even the Newey-West t-stats are highly significant. That’s because the overlap does not cause the effective number of observations to drop by a factor of 360 if you’re doing 360-month retirements. The effective number of observations drops maybe only by a factor of 60 (=5y). So, with 150 years of data, you still have 30 effective uncorrelated observations. That’s because of Sequence Risk, where really the success or failure of retirement hinges on the first 5 years of data. So, yes, even with overlapping windows we can make statistically significant and meaningful statements.

      1. Good point about effective sample size. I tried running blocked bootstrapping experiments at different windows and can confirm in simulation the sensitivity did not change much as the degree of overlap varies (at least until the window gets small enough to lose the autocorrelation we’ve seen historically).

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.