Assessing the long term risks of equity

[StacheEngineer]

Hey everyone,

I want to discuss the risks of long term equity investments, specifically that risk of equity decreases over time. Here's a previous discussion that didn't fully satisfy my curiosity as it ended without a strong consensus: http://forum.mrmoneymustache.com/investor-alley/asset-allocation-100-stocks-for-how-long/

I started thinking about this when I saw this chart in Investments, 9th Ed. by Bodie, Kane and Marcus.

[1] InvestmentReturns.png

Returns are nominal. It shows that more than 5% of the time in a Monte Carlo simulation of 25 years of equity returns, equities underperformed risk free bonds around 6% of the time. You can also lose your entire investment!

Then, I read this page: http://www.norstad.org/finance/risk-and-time.html It states the arguments for why time diversification for risky assets is a fallacy.

Here's a chart of expected total nominal return showing a range of outcomes based on the appendix from the above link. It shows how returns can be significantly lower than the median return.

[2] EquityReturnSpreads.png

For those that think this is only possible in historical simulations, here's a chart of Japanese equities in the past 30 years.

[3] JapaneseReturns.png

This doesn't include dividends so you should compare the return against 0 returns which should be roughly the risk free bond rate. The nominal capital return is only 1.7% and doesn't include inflation! Thus, in a real-world time series we can have a very small risk premium over nominal bonds.

Finally, to address the use of U.S. historical returns, I reference this paper: http://merage.uci.edu/~jorion/papers/risk.pdf and page 19 and Table VI especially. The U.S. equity market since 1926 has experienced the most growth and lowest volatility of international stock markets. Thus to point to the last 90 years of equity performance in the U.S. as broadly representative of equity returns over the next thirty year period is highly optimistic.

I think you can still make a case for U.S. outperformance continuing into the future (most open, liquid financial markets, large internal market, no true external threats, continuing population growth, etc.) but it shouldn't depend on strict historical returns or that equity always outperforms bonds if you wait long enough.

To conclude, I think there are two parts of maintaining a 100% equity allocation. First, not having the nerve to stay the course during short term crashes (2008 - 2009 for example) and buying high and selling low. Second, the very real possibility that equities underperform bonds or even lose value over decades. No amount of time will guarantee that equities will outperform bonds.

[Interest Compound]

Here's the inflation-adjusted, dividend included chart for Japan:



Me personally, I'm modeling my 3 fund portfolio off the 80/20 Vanguard LifeStrategy fund:

48% VTSAX (US Stocks)
32% VTIAX (International Stocks)
20% VBTLX (US Bonds)

CFiresim's historical model shows 90/10 has the best withdrawal rate for long retirement periods, but CFiresim's Monte Carlo data shows usually between 70/30 and 85/15 as being the best (with scary stuff for 100% stocks). So I think 80/20 is a good middle ground.

[brooklynguy]

I want to discuss the risks of long term equity investments, specifically that risk of equity decreases over time. Here's a previous discussion that didn't fully satisfy my curiosity as it ended without a strong consensus: http://forum.mrmoneymustache.com/investor-alley/asset-allocation-100-stocks-for-how-long/

There have been many more threads that continued the discussion from the thread you linked to that can be found with some searching, but Go Curry Cracker subsequently created a post that succinctly sums up the key arguments in favor of a 100% equity allocation:

http://www.gocurrycracker.com/path-100-equities/

In addition, the reason to believe that equity returns will outpace bond returns in the long run is not because they have historically done so, but because of the underlying characteristics of equities and bonds that have caused them to historically do so and can be expected to continue to cause them to do so in the future.  Equity represents the residual interest in a company's capital structure and, unlike bonds, is not a contractual agreement entitling the holder to any specified return.  Thus, they carry higher risk of loss of capital than bonds do, but that risk is compensated for in the form of expected higher returns.

The Norstad article you linked to is essentially simply making the point that we should never forget that stocks carry a higher risk of capital loss than bonds do, even over long time periods.  He argues that, on an appropriately risk-adjusted basis, time diversification is a fallacy.

He gives the example of a fictional hypothetical investment that costs $5000 and has two possible outcomes:  (1) in the good case, which has a probability of 99%, you make $500, and (2) in the bad case, which has a probability of 1%, you lose your entire $5000.  He (quite correctly) points out that even though this investment has a positive expected rate of return of 8.9%, in order to determine the appropriateness of the investment, you need to consider both the probabilities of possible outcomes and their magnitudes.

A 100% equity portfolio most certainly has greater potential for loss of capital than a 100% bond (or cash) portfolio (again, that is the reason it has higher expected returns--to compensate for the increased risk).  To use completely made up numbers in line with Norstad's example, let's say the total equity portfolio has a 99% of outperforming the bond/cash portfolio, but a 1% chance of portfolio depletion down to zero.  And let's say the bond/cash portfolio has a 1% chance of outperforming the equity portfolio, but a 0.0001% of total portfolio depletion.  Consistent with Norstad's point, you can't look at the probabilities of the outcomes alone, you also need to consider the magnitudes.  But even on a risk-adjusted basis (taking into account both the probabilities and the magnitudes of the outcomes), I would take the total equity approach, because there are other risks (besides capital loss) to consider -- for example, the risk of portfolio underperformance such that the returns are inadequate to fund my living expenses (I would have to work longer and defer retirement in order to amass a bond-heavy portfolio that can be expected to fund the same level of expenses as a stock-heavy portfolio).  Another example is inflation risk.  A total equity portfolio for a given extended time period (say, 30 years) provides much better protection against inflation than, say, a fixed-rate "risk free" US treasury bond covering the same duration -- although the treasury bond provides better protection against the risk of nominal capital loss (because that risk is as close to zero as you can get in the real world), it provides worse protection against the risk of inflation eating away at the investment's value.

[StacheEngineer]

Here are my comments on the Go Curry Cracker post:

I believe Cfiresim's underlying model of return distributions is wrong (see description here: http://www.prospercuity.com/modified-distributions.htm) as I think it is overfitting to the returns in the U.S. over the past 100 years and doesn't reflect the global experience. The paper I linked to above shows that only the U.S. market shows a mean-reversion behavior (see page 8). You can assume U.S. equity won't act like other global markets but then you are assuming more strongly that the past will be like the future.

For the chart of terminal values, I believe that the historical record in the U.S. is exceptional and might not persist. The ending value medians increase with more equities (expected) but the ending value lowest and the cycles that dip >60% below initial value do NOT increase with equities which is not expected. Mathematically, you should not expect to take on more risk and not face more downside. You can assume that U.S. will continue to provide exceptional risk-adjusted returns but with so few truly independent even 30 year periods in history, I cannot make that assumption.

Tolerance to risk doesn't just include the ability to avoid selling in major market downturns. Risk must necessarily include the ability to weather Japan style 30 year long periods of flat growth. We haven't seen those periods in U.S. history but they could happen in the future. I do think that early retirees do have different risks compared to later retirees as we must worry more about return risk.

Thus, [equities] carry higher risk of loss of capital than bonds do, but that risk is compensated for in the form of expected higher returns.

But that's just what they are "expected". You can't depend on them. If you could, then you would ell bonds to buy equities. They always go up so you would always come out ahead? There must be significant chances of underperforming bonds (say 2000 - 2009) or else everyone would always just buy equities. Also, the Norstad link shows that the risk doesn't decline over time. Stocks won't necessarily outpace bonds over the long run. They probably will but not always.

such that the returns are inadequate to fund my living expenses (I would have to work longer and defer retirement in order to amass a bond-heavy portfolio that can be expected to fund the same level of expenses as a stock-heavy portfolio)

What do you think is the probability that your 100% equities portfolio will not generate enough return to fund your retirement? This probability must necessarily be less than 1 as equity is risky. If a 2% withdrawal rate from bonds works 100% of the time, a 4% withdrawal rate from equities and bonds must not work some of the time as in some cases equities don't grow enough or go to 0 (note that diversification isn't enough to bring returns up 200 basis points). I think we differ especially on what are the chances that equity underperforms bonds or especially does worse.

Again my assumption is that the past 100 years of U.S. equity returns is exceptional and isn't likely to repeat it. I consider the cases of Japanese equities over the past 30 years and say the European returns where you can lose 60+% of your portfolio over 10 years as possible in the future of U.S. equity returns. These kinds of events haven't been seen in the U.S. historical record and so don't show up in simulations using historical U.S. equity returns.

Thank you for your thoughts.

[milesdividendmd]

Stachengineer. Thank you for this thoughtful post.

I think you make several excellent points here,

1.  Our model for equity performance may be overly optimistic based on the random country in which we were born.

2.  Our statistical models for predicting the future  may be flawed (garbage in/garbage out).

To me the take home is obvious:

1.  Keep costs down.
2.  Diversify
3.  Avoid home country bias.
4.  Don't skate to where the puck was by blindly and dramatically overweighting that which has historically done well to the exclusion of all other assets classes.

That being said. Even in Japan stocks beat bonds in total returns long term in an near zero interest rate environment.

Also while our bonds can get more expensive from here, we are rapidly running out of yield to trim.

And finally, although our equity valuations are relatively high domestically right now, suggesting poor returns going forward, these pale in comparison to Japanese stock valuations pre bubble.

So the smart money is that stocks will perform better than bonds long term going forward.

[innerscorecard]

I think the logical conclusion, given the uncertainty (and unknowability, even) of the inputs involved, is to have a backup plan for what happens if your ball of financial assets can't provide for you in perpetuity. MMM has one. Arebelspy does too. Then, knowing you have one, don't worry so much.

[brooklynguy]

I believe Cfiresim's underlying model of return distributions is wrong (see description here: http://www.prospercuity.com/modified-distributions.htm) as I think it is overfitting to the returns in the U.S. over the past 100 years and doesn't reflect the global experience.

Jus to be clear, your point is not really that cFIREsim is wrong, but that our (and Go Curry Cracker's) general (over)reliance on its results in planning for the future is misplaced.  Despite the fact that it is used (and intended to be used) as a planning tool (and despite its label as a "simulator"), cFIREsim is not a prognosticator about the future and does not attempt to "model" future returns; it is merely a tool that allows you to see how a given set of inputs would have fared historically.  So, unless there's a problem with cFIREsim's underlying historical data, cFIREsim is not "wrong."

EDIT:  Also, it would definitely be nice if it were possible to include international data in cFIREsim's historical simulations, but I believe bo_knows (the creator of cFIREsim) has mentioned that the reason it doesn't is because of a lack of reliable data for international markets going back far enough.

[milesdividendmd]

I think the logical conclusion, given the uncertainty (and unknowability, even) of the inputs involved, is to have a backup plan for what happens if your ball of financial assets can't provide for you in perpetuity. MMM has one. Arebelspy does too. Then, knowing you have one, don't worry so much.

Always good advice. But isn't the question posed here more one of asset allocation rather than how to personally deal with shortcomings in the 4% rule?

[immocardo]

I think the logical conclusion, given the uncertainty (and unknowability, even) of the inputs involved, is to have a backup plan for what happens if your ball of financial assets can't provide for you in perpetuity. MMM has one. Arebelspy does too. Then, knowing you have one, don't worry so much.

Always good advice. But isn't the question posed here more one of asset allocation rather than how to personally deal with shortcomings in the 4% rule?

To me that's like saying "isn't the argument more about what toppings you like to put on the pizza before you bake it rather than what toppings you like eating on a pizza?"

The whole point of asset allocation is to end up with more money.  No matter how you slice it (oh god pizza pun I'm sorry).  You may have different plans, you may care more about safety of investments than final balance, but when you boil it all down it's ultimately the same thing. 

Ultimately you need to consider what the money is ultimately for, and that will always have an impact on your asset allocation. 

[milesdividendmd]

I think the logical conclusion, given the uncertainty (and unknowability, even) of the inputs involved, is to have a backup plan for what happens if your ball of financial assets can't provide for you in perpetuity. MMM has one. Arebelspy does too. Then, knowing you have one, don't worry so much.

Always good advice. But isn't the question posed here more one of asset allocation rather than how to personally deal with shortcomings in the 4% rule?

To me that's like saying "isn't the argument more about what toppings you like to put on the pizza before you bake it rather than what toppings you like eating on a pizza?"

The whole point of asset allocation is to end up with more money.  No matter how you slice it (oh god pizza pun I'm sorry).  You may have different plans, you may care more about safety of investments than final balance, but when you boil it all down it's ultimately the same thing. 

Ultimately you need to consider what the money is ultimately for, and that will always have an impact on your asset allocation.

Not seeing your point here.

The OP raised a specific question about the risks of equity exposure in a  portfolio.

To use your analogy it's as if OP asked

"How many slices of pepperoni are optimal on a pizza slice?"

To which the answer was given:

"A balanced diet is important."