GMO’s New Asset Management Platform

More volatility, better chances

The result is a higher chance of achieving those spending and wealth goals than the conventional approach will typically provide. In the conventional approach, too low a level of volatility is often selected. The Nebo platform does allow the user to constrain the volatility, but in doing so, it enables the user and the client to see and understand the tradeoff between volatility and the probability of achieving the goals. Often, lower volatility leads to lower probability of achieving goals. In products from other vendors, this tradeoff is less obvious, and needs to be explained with historical charts. Nebo makes it explicit in its software. As a result, Nebo typically recommends a glidepath with a higher average commitment to equities than other products, but also a greater chance of achieving spending and wealth goals.

In most applications of mathematics in the portfolio management field, the mathematics is little more than a smokescreen. It may impress the client, and is intended to do so, but usually does little if anything of practical value. In the Nebo product, this is different. The math derives the optimal glidepath.

In this case, the mathematics does what it claims to do; it minimizes the risk of not having what you need when you need it. To put it the other way around, it maximizes the chance that you will have what you need when you need it.

As with all such applications of mathematics in the investment field, it must be stated as a caveat that the result depends on the assumptions that are input to the model. But here the result is a consequence of a sensible, straightforward, well-defined statement of the problem, together with a mathematical solution that is appropriate to the task. Furthermore, and most importantly, the result is not a single allocation in time, as it is with mean-variance optimization. The latter optimization leaves the user with no advice about what to do next. The default assumption was that whatever was done in the first place – whatever asset mix – must be what also needs to be done in the future; hence, “rebalancing.” There is no reason to believe this. It should have been obvious from the beginning that a dynamic approach is much more appropriate; that is to say, changing the allocation over time – or allowing it to change – not only to implement the initially prescribed glidepath, but to alter it in response to feedback from the current value of the portfolio and the market, and from changes in plans for future spending and desired terminal wealth. The Nebo product allows for these changes. It prescribes a glidepath of asset allocations over time, but it allows this path to change as circumstances, planned outlays, and market assumptions change.

Separating the substance from the hype

When I first encountered this product, I had an intensely skeptical reaction. Although I agreed with the basic premise, my reaction was due to claims made for the product in two Nebo white papers. One of these claims is based on conventional wisdoms that I believe to be wrong, and one that I believe is not as important as the papers say it is. Because of my reaction, however, I did not initially read Tarlie’s mathematical papers carefully enough and thus did not realize the fundamental coherence of the approach and its mathematical underpinnings.

Nevertheless, while these quibbles could be dismissed in light of the overall quality of the product, let me take up three claims made by the two white papers mentioned above. I will agree with the first claim; express skepticism about the second; and disagree strongly with the third.

1. The Nebo approach takes care of sequence of returns risk.
2. It attacks random walk, claiming its superior results are due to “Next-gen” Monte Carlo simulations.
3. It implies it creates a buy-low-sell-high benefit by going against investor irrationality.

Does Nebo take care of sequence of return risk?

Here the answer is yes. Nebo correctly identifies sequence of return risk as financing risk, akin to the risk of illiquidity versus insolvency. This risk can only occur if funds are being withdrawn from the portfolio. If returns have been temporarily much lower than average when funds need to be withdrawn, it can mean running out of money.

But this is taken care of by Nebo’s approach, the approach of treating risk as “not having what you need, when you need it.” Sequence of return risk is merely a corollary to this risk. If the risk of not having what you need, when you need it, is minimized, it also minimizes sequence of return risk. Sequence of return risk is not a distinct and separate risk.

The idea that sequence of return risk was a new and unrecognized form of risk was a consequence of the conventional approach to investment planning, based around mean-variance optimization. Since this method of analysis has no over-time component, sequence of return risk appeared to be a separate source of risk – separate from what, in that conventional approach, is fashioned to be the main source of risk, namely volatility. Jettisoning that approach to risk also dispenses with the need to identify a separate category of risk called “sequence of return risk.”

Is Nebo better because it uses “Next-gen” Monte Carlo simulation?

In one of the white papers, Nebo launches a vicious attack against random walk, calling it “a world view that passed as conventional wisdom in the 1970s.” What Nebo is really attacking is only one specific assumption of one implementation of random walk in the modeling of investment returns – that the expected return is constant over time.

Nebo’s different assumption – as expressed in one of Tarlie’s papers – is that expected equity return is not constant but varies randomly. But if it only varied randomly then the result would be equivalent to simply assuming a higher volatility. This would make equities’ returns more volatile without increasing their expected return. Thus, it would be detrimental to their portfolio allocation. However, expected equity return is not assumed in Nebo’s model to vary randomly without bound; it is assumed to mean-revert to a long-term mean expected equity return. That is, while it can vary randomly, it gets pulled back as if on an elastic band to the long-term average expected return. The latter feature thus counters the increased volatility resulting from assuming expected equity return varies randomly.

What net effect will this have on the resulting allocations and glidepath? I find it difficult to intuit the answer to this question. It depends on the assumptions about the variance of the expected return and the speed of mean reversion. But I don’t believe that the impact of this tweak to an ordinary random walk on Nebo’s results is explained well enough in Nebo’s white papers – a tweak, incidentally, that is often made, and that I have even made at times in my own implementations of Monte Carlo simulations.

This tweak is in part an attempt to link Nebo to a signature GMO insight, namely that market P/Es revert to a mean over long periods of time, such as seven years, and also to link Nebo with a recent increased emphasis in academic literature on discount rate variability. GMO is reputed to have been successful at dodging market bubble collapses, presumably due to this insight. But the feature above of Nebo’s random walk assumption is not necessarily closely related to the long-term mean reversion of GMO’s forecasts.

Nebo allows the user to either include or not include this mean-reversion feature in its use of Monte Carlo simulation. Even without this modification to a simple random walk, Nebo prescribes a higher commitment to equities than most other products; the tweak increases the equity commitment even more, but it is not clear to me why.

Does Nebo create a buy-low-sell-high benefit by going against investor irrationality?

This is where my skepticism with Nebo rests – or at least with the hype for it in its white paper.

Nebo’s second, most recent, white paper invokes the recently faddish hokum about how investors, due to their irrationalities as revealed by experiments such as those by Kahneman and Twersky and others, invest as a herd and therefore all you need to do is to do the opposite of what the herd is doing and you’ll beat the market. Somehow – apparently because of the mean-reversion assumption – Nebo is supposed to do this.

This is nonsense. This idea was sparked by studies that have either been dead wrong in their methodology (as is the case with DALBAR) or have reached weak conclusions based on better but still questionable methodologies – all of which conclude that “investors underperform their investments.” But even if it were true, it doesn’t mean that there is some automatic way of going against the herd – a herd that is somehow almost always wrong.

What I would like to see more of

Nebo’s method of deriving an optimal glidepath is so crucial that it needs more explanation and illustrations. The objective of the optimization is defined clearly and well – to minimize the chance of not having what you need, when you need it. This is the essential first step. Then, in principle, a computer could generate all possible glide paths to see which one achieves both the minimum probability of achieving this goal and the minimum shortfall. Tarlie’s utilization of the calculus of variations is a short cut that makes this computer-intensive method unnecessary. But it would help if it were better explained. Perhaps moving videos could show glidepaths being varied and the probability and shortfall results of each. This would be much better than trying to tout Nebo’s mean-reversion algorithm as if that were its secret sauce, which it isn’t.

Could this put an end once and for all to MPT?

In an ideal world, this would put an end to articles in finance journals about MPT, mean-variance optimization, “rebalancing,” “rebalancing-diversification return,” and even “sequence-of-return risk” – and on and on. There could be plenty of articles that can be written about optimizing glidepaths to minimize the risk of “not having what you need, when you need it.” Nebo is far from the last word on that. But at least it points in the right direction. And it serves its purpose much better than anything that has gone before.

Economist and mathematician Michael Edesess is adjunct associate professor and visiting faculty at the Hong Kong University of Science and Technology, managing partner and special advisor at M1K LLC. In 2007, he authored a book about the investment services industry titled The Big Investment Lie, published by Berrett-Koehler. His new book, The Three Simple Rules of Investing, co-authored with Kwok L. Tsui, Carol Fabbri and George Peacock, was published by Berrett-Koehler in June 2014.

Read more articles by Michael Edesess