Last year the Harvard Management Company (HMC) earned 5.8% for the university's endowment. This ranks as the 5th worst year out of the past 16 years (or even the past 24 years if one stretches their historical window that far). From what seems like a childish perspective, we can note that Yale's returns were nearly double. And this all happens at a time when the leading university endowments -with their overrenumerated staff and their consultants (investment middle men)- are only outperforming their investment benchmarks by an untenably narrow edge, all while utilizing "customized" policy that reflects increasingly easier self-hurdles. When thinking about competitive performance, one can revisit shrewd and memorable commentary by former university president and former Treasury Secretary Lawrence Summers regarding grade inflation:
"...the four minute mile was a miracle when Roger Bannister did it and most good college track teams have somebody who can run a four minute mile today. But we still only give one gold medal in the Olympics."
The once fabled HMC had for some time generated awesome returns. These powerful returns paralleled Harvard's preeminent global academic rank and incomparable endowment size. An endowment now $400 million larger not from market gains but rather from hedge fund manager and alumnus John Paulson's recent contribution. How many more of those are coming again? But now, HMC's performance is badly lagging their peers (e.g., Harvard's endowment was the worst among the Ivy Leagues in 2014 and many of them are part of HMC's formal peer group). The mystique is diffusing and HMC president Stephen Blyth (similar to me, a Harvard statistics department alumnus, and he's also a current professor there) now whistles a different tune, insisting that such horse races are "counterproductive". Sure, when the stakes are high and your chips are down big, then keeping score (of who's coming in last place) can be crushing to fragile egos.
HMC's success was such an important business story that I had studied it as a HBS case study there under Nobel Laureate and Professor Robert Merton nearly 15 years ago. It was a glorious age for HMC, during the legendary Jack Meyer era. In 15 years he oversaw an endowment mushrooming by a factor of 5. Simultaneously revolutionizing the business of endowments, and hurling HMC into an enviable position, masterfully dominating any other school endowment on the planet. Providing for the Harvard community and securing it's distinguished rank for generations to come. Let's see the returns though, since that time. On page 6 of the 2015 performance report we see the quartile performance of 5-year trailing returns.
We re-summarize this information below; along with an estimate for 2015 (since not all peer universities have yet reported 2015 results). One of HMC's formal objectives is to be top quartile (the blackened region in the top row of the chart below). And there was a nice 5-year run of that, including under Mohamed El-Erian's (former PIMCO executive's) brief 2-year leadership, on through the climax of the global financial crisis. But there has also since been an equal streak at the bottom quartile (the blackened region in the bottom row of the chart below). If success is a skill that persists with a visibile trend, then one can see appalling clumsiness trending in the chart here.
This is why the more recent HMC president Jane Mendillo -heralded on her way in from Wellesley as one of the most powerful women in finance- was then fired. Venturing into ever more exotic "asset classes" and "sustainable" gambits she didn't understand. Yet seeing the performance slip into the worst quartile is something everyone understands. If these rankings were based on a 10-year smoothing, instead of 5 years, her dreadful performance would leave a more lasting stain. Now we've shown the impossibility of always being in the top quartile (Quartile 1) for anyone, including last year in the New York Times (here, here). Yet we will show with the probability math below, why the HMC saga is more about losing their once-vaunted investment muscle. We'll understand why others have been simply "stealing Harvard".
Let's lay out the investment statistics for both HMC, and their 60/40 stocks/bonds benchmark (only as reference). HMC has an annual standard deviation of about 13%. This comes to a theoretical 5-year annualized standard deviation of just below 6%:
Where σ is the annualized standard deviation, m is for the number of years being averaged and n is for their annualization of the statistic. In reality, we saw that HMC has an annualized standard deviation in their 5-year returns of just over 6%. So very close to our theoretical value.
The 60/40 benchmark on the other hand revealed a standard deviation of 5% on this same statistic. Now if we look instead at the changes in the rolling 5-year annualized returns (from year to year), the standard deviation among them is 5% for HMC and 4% for the 60/40 benchmark. So here it just happens to be the case through the serial correlation conversation further below, that both are a little less than the former values (for non-change or non-difference). The table below summarizes this discussion fo standard deviations on the 5-year rolling returns, as well as the differences in these from year to year.
5yr annualized difference in 5yr annualized
HMC 6% 6%
60/40 5% 4%
Given the 6% standard deviation, the probability of a university endowment transitioning to another quartile depends on two things. First does the university fund truly have skill beyond that of other universities, and second which quartile are they currently are in. For example, see this distribution below for the historic, relative performance over time, typically seen in each of these quartiles:
Quartile 1 >7%
Quartile 2 2%
Quartile 3 -2%
Quartile 4 <-7%
For a university endowment in the outer quartiles therefore, they would need a >3% change in one direction to just barely pull out of the same quartile, and a >7% change in the same direction to jump at least two quartiles. Since we showed earlier that the standard deviation of such a move is about 6%, this amounts to a roughly 30% probability and 17% probability respectively. Note that the 47% (30%+17%) is the probability of simply leaving the current quartile. So weaker than a ½ chance, but higher than some might expect. Also note that we don't need to utilize the 60/40 benchmark statistics for this relative performance analysis, since the benchmarks here the other universities. There is no doubt in strong markets, all boats rise. Here we focus on the probability of return quartiles relative to peers. However we should also take the time here to suggest that the 60/40 perhaps shouldn't be the appropriate benchmark, given violation of at least 1 of the 8 presumptions of a benchmark, which is that the variability in the same should be equal to the variability of one's portfolio (something we see in the table further above to not be true).
Now for universities in the inner quartiles, they would need a few percentage point performance change in the outward direction to just barely pull out of the same quartile. A 28% probability. And due to asymmetry, in the other direction, one would need just a 2% or a 6% relative change to enter those new further away quartiles. Decomposing the same 37% probability would give us 21% and 16%, respectively. We now have all the information needed to complete a delightful probability transition matrix for university's performance, by quartiles. See it in red color below.
A different way to visualize this is to take it in through the contour image below. On the left is the red matrix above, which is for a random performer, and on the right we see the matrix for a highly-skilled investor; both being auto-correlative. We can see that neither are stable matrixes (there are diagonal trends in both) nor a matrix where we see performance terminating in in node -say an outer quartile. We track the changes in quartiles we discussed above with red color arrows. Note first that, barring any rare once-in-multiple-decades luck, HMC is far from succeeding back into their top quartile objective anytime soon (absolutely not this year)! Second the fact that HMC has been stuck in the bottom quartile for many years implies through this illustration that they are likely not performing within the highly-skilled performance matrix on the right.
And third we see that the overall arrow paths follow the theoretical high-density transition tracks, mostly in the left matrix. It is improbable that HMC is a skilled investor and still moving in the enormous whitish space on the right matrix. While we don't expect any university endowment to be a consistent star performer (hence there is a weak basis for active management), HMC certainly is not of late.
Having concentrated our focus on the overall quartiles, let's shift into a deep dive on just serial correlation (here, here) of the sequence of overlapping five year returns. Look at the hypothetical returns (mapped back to the quartile format) below. We show 5 random returns (green), for the same 2000 through 2015 time frame. We also show 5 (red) at the same serial correlation level we have for HMC (ρ>0.8).
In the absence of further calculations, which of the two sets of 5 returns (green, or red) most simulate what we've seen for the (black-color) HMC series in the middle? Of course it is the auto-correlated set on the right. We see streaks of exceptional underperformance and streaks of exceptional overperformance. It's just as we've seen further above that in the case of HMC, this very lengthy streak of recent underperformance is highly unlikely due to simply bad luck. It's due to bad proficiencies and performance, and with a Harvard statistics professor at HMC's helm this should be obvious. Also, if the relative results to Yale continue, Harvard would ultimately concede their top position within a decade.
From the illustration immediately above it is clear to see that the HMC performance has been in the bottom ½ among its peers for 7 years (the report link above notes that 2015 is shaping up to be the same). This is only another vantage into performance, by softening the contrast to simply a binary result: above expectations or below expectations. Per this view a winning fund manager should have no justification to persevering donors and the Harvard community for being in the bottom ½ for 7 straight years, something we'd not see more than once every several hundred years, due to chance. Even Harvard, America's oldest college and predates the founding of the United States, hasn't been around that long.
The overall results we have seen above would be analogous to a fine high school track and field athlete, who just happens to perform in the bottom ½ of his class every semester, and consistently finishes in the bottom ½ in every distance run. Yet the athlete still gets accepted into Harvard. And he or she still receives a laudatory gold medal just to feel good.
Short term update: without surprise, recent market risk articles have been cited and shared widely. "Pending massive rebound?" was discussed in Bloomberg and by the Editor of Barron's. "Market risk; model smash" shared >200 times, including by a top chief of the PBGC (government pension insurer), Ritholtz's Big Picture, a business show, author of one of this year's top business books,
Business Insider, StockTwits and CEO Lindzon, and Zero Hedge. Finally, we see Rick Perry and Scott Walker both quitting the RNC primary race. As predicted more than a month ago, more than nine will inelegantly drop out before we can start getting honest with the GOP polling.
"...the four minute mile was a miracle when Roger Bannister did it and most good college track teams have somebody who can run a four minute mile today. But we still only give one gold medal in the Olympics."
The once fabled HMC had for some time generated awesome returns. These powerful returns paralleled Harvard's preeminent global academic rank and incomparable endowment size. An endowment now $400 million larger not from market gains but rather from hedge fund manager and alumnus John Paulson's recent contribution. How many more of those are coming again? But now, HMC's performance is badly lagging their peers (e.g., Harvard's endowment was the worst among the Ivy Leagues in 2014 and many of them are part of HMC's formal peer group). The mystique is diffusing and HMC president Stephen Blyth (similar to me, a Harvard statistics department alumnus, and he's also a current professor there) now whistles a different tune, insisting that such horse races are "counterproductive". Sure, when the stakes are high and your chips are down big, then keeping score (of who's coming in last place) can be crushing to fragile egos.
HMC's success was such an important business story that I had studied it as a HBS case study there under Nobel Laureate and Professor Robert Merton nearly 15 years ago. It was a glorious age for HMC, during the legendary Jack Meyer era. In 15 years he oversaw an endowment mushrooming by a factor of 5. Simultaneously revolutionizing the business of endowments, and hurling HMC into an enviable position, masterfully dominating any other school endowment on the planet. Providing for the Harvard community and securing it's distinguished rank for generations to come. Let's see the returns though, since that time. On page 6 of the 2015 performance report we see the quartile performance of 5-year trailing returns.
We re-summarize this information below; along with an estimate for 2015 (since not all peer universities have yet reported 2015 results). One of HMC's formal objectives is to be top quartile (the blackened region in the top row of the chart below). And there was a nice 5-year run of that, including under Mohamed El-Erian's (former PIMCO executive's) brief 2-year leadership, on through the climax of the global financial crisis. But there has also since been an equal streak at the bottom quartile (the blackened region in the bottom row of the chart below). If success is a skill that persists with a visibile trend, then one can see appalling clumsiness trending in the chart here.
This is why the more recent HMC president Jane Mendillo -heralded on her way in from Wellesley as one of the most powerful women in finance- was then fired. Venturing into ever more exotic "asset classes" and "sustainable" gambits she didn't understand. Yet seeing the performance slip into the worst quartile is something everyone understands. If these rankings were based on a 10-year smoothing, instead of 5 years, her dreadful performance would leave a more lasting stain. Now we've shown the impossibility of always being in the top quartile (Quartile 1) for anyone, including last year in the New York Times (here, here). Yet we will show with the probability math below, why the HMC saga is more about losing their once-vaunted investment muscle. We'll understand why others have been simply "stealing Harvard".
Let's lay out the investment statistics for both HMC, and their 60/40 stocks/bonds benchmark (only as reference). HMC has an annual standard deviation of about 13%. This comes to a theoretical 5-year annualized standard deviation of just below 6%:
(σ*√m)/n
= (13%*√5)/5 Where σ is the annualized standard deviation, m is for the number of years being averaged and n is for their annualization of the statistic. In reality, we saw that HMC has an annualized standard deviation in their 5-year returns of just over 6%. So very close to our theoretical value.
The 60/40 benchmark on the other hand revealed a standard deviation of 5% on this same statistic. Now if we look instead at the changes in the rolling 5-year annualized returns (from year to year), the standard deviation among them is 5% for HMC and 4% for the 60/40 benchmark. So here it just happens to be the case through the serial correlation conversation further below, that both are a little less than the former values (for non-change or non-difference). The table below summarizes this discussion fo standard deviations on the 5-year rolling returns, as well as the differences in these from year to year.
5yr annualized difference in 5yr annualized
HMC 6% 6%
60/40 5% 4%
Given the 6% standard deviation, the probability of a university endowment transitioning to another quartile depends on two things. First does the university fund truly have skill beyond that of other universities, and second which quartile are they currently are in. For example, see this distribution below for the historic, relative performance over time, typically seen in each of these quartiles:
Quartile 1 >7%
Quartile 2 2%
Quartile 3 -2%
Quartile 4 <-7%
For a university endowment in the outer quartiles therefore, they would need a >3% change in one direction to just barely pull out of the same quartile, and a >7% change in the same direction to jump at least two quartiles. Since we showed earlier that the standard deviation of such a move is about 6%, this amounts to a roughly 30% probability and 17% probability respectively. Note that the 47% (30%+17%) is the probability of simply leaving the current quartile. So weaker than a ½ chance, but higher than some might expect. Also note that we don't need to utilize the 60/40 benchmark statistics for this relative performance analysis, since the benchmarks here the other universities. There is no doubt in strong markets, all boats rise. Here we focus on the probability of return quartiles relative to peers. However we should also take the time here to suggest that the 60/40 perhaps shouldn't be the appropriate benchmark, given violation of at least 1 of the 8 presumptions of a benchmark, which is that the variability in the same should be equal to the variability of one's portfolio (something we see in the table further above to not be true).
Now for universities in the inner quartiles, they would need a few percentage point performance change in the outward direction to just barely pull out of the same quartile. A 28% probability. And due to asymmetry, in the other direction, one would need just a 2% or a 6% relative change to enter those new further away quartiles. Decomposing the same 37% probability would give us 21% and 16%, respectively. We now have all the information needed to complete a delightful probability transition matrix for university's performance, by quartiles. See it in red color below.
And third we see that the overall arrow paths follow the theoretical high-density transition tracks, mostly in the left matrix. It is improbable that HMC is a skilled investor and still moving in the enormous whitish space on the right matrix. While we don't expect any university endowment to be a consistent star performer (hence there is a weak basis for active management), HMC certainly is not of late.
Having concentrated our focus on the overall quartiles, let's shift into a deep dive on just serial correlation (here, here) of the sequence of overlapping five year returns. Look at the hypothetical returns (mapped back to the quartile format) below. We show 5 random returns (green), for the same 2000 through 2015 time frame. We also show 5 (red) at the same serial correlation level we have for HMC (ρ>0.8).
In the absence of further calculations, which of the two sets of 5 returns (green, or red) most simulate what we've seen for the (black-color) HMC series in the middle? Of course it is the auto-correlated set on the right. We see streaks of exceptional underperformance and streaks of exceptional overperformance. It's just as we've seen further above that in the case of HMC, this very lengthy streak of recent underperformance is highly unlikely due to simply bad luck. It's due to bad proficiencies and performance, and with a Harvard statistics professor at HMC's helm this should be obvious. Also, if the relative results to Yale continue, Harvard would ultimately concede their top position within a decade.
From the illustration immediately above it is clear to see that the HMC performance has been in the bottom ½ among its peers for 7 years (the report link above notes that 2015 is shaping up to be the same). This is only another vantage into performance, by softening the contrast to simply a binary result: above expectations or below expectations. Per this view a winning fund manager should have no justification to persevering donors and the Harvard community for being in the bottom ½ for 7 straight years, something we'd not see more than once every several hundred years, due to chance. Even Harvard, America's oldest college and predates the founding of the United States, hasn't been around that long.
The overall results we have seen above would be analogous to a fine high school track and field athlete, who just happens to perform in the bottom ½ of his class every semester, and consistently finishes in the bottom ½ in every distance run. Yet the athlete still gets accepted into Harvard. And he or she still receives a laudatory gold medal just to feel good.
No comments:
Post a Comment