Using Technology To Tell Pensions What They Want to Hear: Alternatives
Suppose the aviation industry had a computer model that allowed planes to fly safely about 95% of the time. Do you think anyone would get on an airplane knowing that 5% of the time the model resulted in crashes? Sadly, pension plans rely on a model named “mean-variance optimizer.” The model is used to determine their asset allocation but is prone to fatal error. Actually, the problem is less with the model than the assumptions that go into it. The assumptions preordain the result, and then the model understates the risk. This a powerfully bad combination. Nonetheless, these models are the intellectual backbone used to justify the foray into alternative investments.
You can’t understand why pension officials are gaga over alternatives unless you spend a little time unpacking one of these models. I am going to spend this post doing just that. I discovered a set of capital market assumptions North Carolina public pensions in a presentation by the Investment Management Division to the Treasurer’s Investment Advisory Committee. The table below isn’t complete because it doesn’t show the assumed relationships, known as correlations, between the various categories. However, it does tell us what the staff estimates will be the returns and risk of each category. These return and risk assumptions are two key ingredients that are pumped into an optimizer.
Just run your fingers down the list. Conventional stocks and bonds are on the bad side of the ledger. They appear to have lower returns and higher standard deviations than the collection of alternatives (see hedge funds, direct real estate, and private equity at the bottom of chart). When the numbers are poured into the optimizer, alternatives scream, “Own me.”
2012 Capital Market Assumptions
North Carolina Department of State Treasurer
November 14, 2012
So what’s wrong with this picture? In no particular order, here’s the critique. As always, I’ve committed every one of these sins.
First, there’s the problem of false precision. Notice that these estimates are shown out to two decimal places. These numbers are, at best, educated guesses, and should probably be rounded to the nearest percentage. The greater the figures placed after the decimal point, the more the numbers look like they are precise answers. This leads to a bad case of overconfidence by the decision makers. Their confidence is further misplaced because all the figures are probably gross returns, and the actual results will be 1% to 6% lower after accounting for fees and expenses.
Second, there’s the lack of consistency in the categories. The stock and bond categories have fairly precise meanings, while hedge funds can be just about anything. There are dozens of highly disparate hedge fund strategies that run the entire gamut of potential combinations of stocks, bonds, derivatives, leverage, and short positions. As a result this figure is meaningless. There are several other groupings, such as commodities, private equity, and real estate where there’s way more art than science to producing these estimates.
Third, there’s a high degree of mispriced risk. The hedge fund and the direct real estate are great examples. Even if you don’t understand the first thing about risk (standard deviation), wouldn’t you want to own as much as possible of something called hedge funds that are estimated to return 9.2% with a risk of 6.2%? Doesn’t direct real estate look attractive, with an estimated return of 9.1% and a risk of 6.6%? Why would you ever own large cap equities? With a return of 8.9% and risk of 17.1%, they look like a relatively horrible investment. It appears that you’re going to get a smaller return at nearly twice the risk. If I believed these figures, I wouldn’t need an optimizer to tell me to invest my capital in alternatives.
While the risk for the various types of stocks and bonds looks relatively realistic based on historical trends, most of the other risk estimates don’t begin to capture the true risks of those strategies. The comparison between direct real estate and REITs is a perfect example (red box). Direct real estate means that the pension plan owns properties in a fund that doesn’t trade on a market. REIT means that the pension owns real estate in a fund that trades on a market. The REIT is estimated to only return 6.3% with a risk of 19.1% compared to direct real estate’s estimated return of 9.1% with a risk of 6.6%. If these figures were correct, no one would own an REIT. In reality, owning direct real estate has a risk much closer to the number cited for REITs – the risks are merely hidden because the portfolio doesn’t trade.
Fourth, there’s the fundamental misuse of statistics. The return and risk statistic should only be used to help construct an investment portfolio if the returns for each category are normally distributed. Over time they would have to produce a shape that approximates the Liberty Bell. While stocks and bonds come fairly close to this ideal, almost all the other categories have a very different shapes. Moreover, the potential distribution is further distorted and asymmetrical because of performance fees in the alternative categories. Nonetheless, all these distorted estimates are put into an optimizer in order to get the most efficient portfolio possible. The actual result is actually quite dangerous.
The model dramatically understates the risk of extremely bad outcomes. As I mentioned at the outset, if an airplane experiences tolerable turbulence 95% of the time, but gets smashed apart about 5% of the time that’s unacceptable, and it shouldn’t be permitted for pension plans. Even worse, at those extremely bad moments when you’re hoping that your hedges will keep your financial Dreamliner aloft, they fail completely. Nonetheless, we cram assumptions into the old optimizer because it produces the result pension professionals desire: a shift toward alternatives.
I was very fortunate when I ran the NC pension plan. Our permissible commitment to alternatives (remember the law limited everything to 5%) was tiny. So when I ran an optimization, I was essentially mixing stocks, bonds and a bit of real estate. The results were a tad more reliable than the full-blown optimizations undertaken today. By the way, I turned away dozens of software marketers who wanted to sell me a professional optimizer. They said it would produce more precise results. I knew that our assumptions were so prone to error that the model I constructed in Excel would provide a good estimate without infecting the Treasurer or me with overconfidence.
 https://www.nctreasurer.com/inv/IAC%20Resources/IACRiskAnalysis-111412.pdf at page 59. These assumptions aren't unique to North Carolina. Every pension plan that migrates into alternatives use similar figures.
 In addition to the huge guesswork that goes into estimating private equity and real estate returns, the actual gross returns to the pension plan will almost certainly be lower than these estimates because the IRR calculation that underpins these assumptions is flawed. IRR assumes that all cash flows are reinvested at the same rate of return as the PE and RE investments. This is seldom the case. There is often a considerable and unavoidable cash drag.
 These estimates imply that NC expects hedge funds on average to return 9.2%, but there’s about a two in three change that those returns will be somewhere between 3% and 15.4% (one standard deviation)
 The estimates for stock are actually a tad optimistic and just a bit unusual. An 8.9% return for large cap stocks over the next ten years is about 1% too high based on the long-term record and a 10.6% return over 30 years makes no sense. There’s no logic to having a 30-year horizon that has significantly higher returns than the 10-year forecast for large cap stocks in a mature economy.
 In moments of extreme distress, the correlations all tend to align. In fact, the portfolio appears to have only two assets: risky assets and safe assets (cash and US Treasuries). So when you are counting on diversification to protect the portfolio it breaks down. One need only look at the experience of the endowments in the midst of the credit crisis to see this phenomena.