The Glidepath Illusion, Estrada (2014)

This is my favorite overall paper on glidepaths prior to retirement. Estrada does a few things that put this at the top for me:

• He uses multiple datasets, not just the usual US-centric S&P 500. Uses the Dimson-Marsh-Stauton database to look at historic performance in the US, Europe, and “the world”. In an appendix he also shows all 19 countries separately and a cross-sectional look across all 19.
• He uses a variety of glidepaths: 15 in total. You could always ask for more variety here; all of his glidepaths are linear. None are exactly what any of the major providers use. But the glidepaths presented are usually pretty reasonable. Some other papers have glidepaths that end up with 100% bonds at age 65, which isn’t very realistic.
• Estrada shows lots of bits of data about each result. The more data shown, the more confident you are in the results. Arnott and Basu, in their respective papers, only showed a couple of deciles of data. Estrada shows: mean, median, max, decile-10, quartile-4, standard deviation, min, decile-1, and quartile-1.
• Finally, Estrada provides more more context and more interesting discussions via reference to other research and findings.

What do I mean by the last point? Estrada has many passages like

• “This is in fact consistent with evidence showing that, when approaching retirement, most target-date fund users prefer downside protection, rather than upside potential.” This was something that I felt Arnott and Basu both failed to discuss in a satisfying way.
• “this controversy perhaps ultimately comes down to how risk is defined” Estrada seems more sympathetic to the idea that others could look at the same data and come to different conclusions due to have different prior assumptions.

Because Estrada does a better job of grappling with the opposing viewpoint, I come away feeling like his conclusion is more solid.

I won’t keep you in suspense: Estrada comes down firmly against declining equity glidepaths. Where Estrada does a better job than Arnott and Basu is his analysis of his results. Sure, you have a higher final portfolio in the average case. But that’s not the best measure of success for most people.

However, as already mentioned, the goal of lifecycle strategies is not to maximize expected terminal wealth, but rather to provide an acceptable balance between risk and return. Hence, it is essential to explore not only differences in the capital accumulated at retirement, but also differences in the risk borne by investors.

He makes a convincing argument that, while they have less volatility, declining equity glidepaths don’t provide you with better downside protection.

If we look just at the absolute worst case (Min) we can see:

• When we look at US data, the 100x20 glidepath does best (103.3) with the 100x30 and 100x40 slightly behind. Other glidepaths are substantially worse (e.g. 54.3).
• When we look at European data, 100x40 glidepath does best (25.1). The gap between the rest of the pack isn’t as staggering as in the US case: other glidepaths are around 18–20.
• When we look at the World data, the 100x40 glidepath does best again (65.8) with other glidepaths in the range of 55–65.

The 100x40 glidepath is actually just staying at 100% equities your entire working life. 100x30 means staying at 100% equities and then, in the last decade, glide down to 50/50. 100x20 means staying at 100% equities for the first 20 years and then in the last 20 years gliding down to 50/50. All three of those approaches seem strictly superior than any of the more gradual glidepaths.

Estrada’s claim is that a heavy equity approach provides similar downside protection as more cautious approaches but substantially better upside. That’s a little bit hard to see because he’s laid out his results in two separate tables, so I’ll reproduce them. Below I’ll be using the World dataset (to remove American success bias), where the 100% equities for your whole life wins out.

Difference from the 100x40 portfolio, Global historical data

The minimum (that is, worst case scenario) is less bad. The average of 1st decile performance is less bad. The average of 1st quartile performance is less bad. The median is higher. And the average of 4th quartile (that is, the lucky scenarios) performance is better.

Hence, the higher variability of equity-driven strategies simply reflects more uncertainty about the upside potential; that is, how much better, not how much worse, investors are expected to fare with these strategies than with lifecycle strategies.

Estrada appears to be in the camp that investors should just suck up the additional volatility of high-equity portfolios. He has two main arguments for ignoring volatility

• Most of the volatility being measured is actually upside volatility. And, realistically, investors don’t consider upside volatility bad. You don’t lose sleep over whether you’ll make $100,000 or $150,000 from equities this year.
• You may experience a large portfolio cut right at retirement but your portfolio is falling from a higher place, leaving you better off. When I wrote that 1921, right before the Great Depression, was the best year to retire in American history you can see some of this effect in action.

Despite how much I like this paper, there are still things I don’t like much about it.

Estrada comes out in favor of high-equity positions (and the title of paper seems to suggest he is anti-glidepath) but it looks like the 100x20 and 100x30 portfolios actually do quite well. Remember, the 100x20 portfolio is 100% equities for the first 20 years (i.e. until you are 45 years old) and then glides down to 50/50 at age 65. That means a shift of 2.5% from equities to bonds every year. That seems awfully close (though not exactly) to what, say, Vanguard does.

Vanguard starts 90% equities, instead of 100%. And they begin gliding down at age 40, instead of 45. But it looks pretty damn similar, if you ask me. So why not use that? It seems to rate better than either a constant 50/50 or a constant 60/40 in Estrada’s analysis.

So where does all of that leave us?

• Estrada seems to find that 100% equities is actually best on every metric except volatility. So if you can ignore that, just stick with 100% equities.
• Glidepaths, contra Arnott and Basu, actually seem to work pretty well…as long you aren’t dogmatic about them. If you want a constant glide — like in Bodie, Merton, and Samuelson’s academic paper or in rules of thumb like 100-minus-age-in-bonds— well, that doesn’t do so well. But so long as your target date fund has a hefty allocation of equities until relatively late in life…they seem to do well.