“There are three kinds of lies: lies, damn lies and statistics.”
– Mark Twain
One of my primary job functions is analyzing investment manager performance. Thus I spend a lot of time looking at statistical summaries of investment performance. These are usually included in a fund manager’s marketing material. Every so often I catch people playing tricks with data and formatting — not fraud, mind you, just little cognitive tricks designed to focus you on what the manager wants you to see perhaps at the expense of what is actually decision-useful.
Linear Scale Graphs of Compound Growth: Charts are pretty. Data tables are not. The ubiquitous pretty chart investment managers like to show their clients is the growth of $1/$10,000/$1,000,000 over time. The trick is that everyone is inclined to look at the pretty colors in the chart without thinking about the scale. So a successful manager will often show you a growth graph plotted on a linear scale. It will look something like this:
Whoa! It’s parabolic! However, you get a somewhat more nuanced view if you flip the scale to a logarithmic one:
Now, this is the exact same data. The difference is that that the log scale better reflects compounding wealth as an exponential function. It makes it easier to compare the slopes of the lines (which represent performance). In this case I think the manager actually benefits from the comparison because you can see on the log scale the 2015-2016 drawdown was much smaller in percentage terms than in dollar terms. But not all managers come off looking so good. And even in this case the log scale makes it clearer that performance was not quite as parabolic as from 2012-2015 as the linear scale might lead you to believe.
Cumulative Data Not Annualized: There is one reason and one reason only that an investment manager will present you with a cumulative performance number and not an annualized number. That reason is that annualized performance is not competitive. It is hard to contextualize a number like “a cumulative 40% return since inception.” It is easier to contextualize “an annualized 7% return.”
Anchoring: Data mining is easy. There is always some data point you can pick to highlight that will make you look good. All fund marketing material is designed to anchor you on the positives while minimizing the negatives. Sometimes this means bringing more obscure data to the fore while letting more useful data recede into the background. My favorite obscure statistic is “cumulative outperformance when the benchmark is down,” or, in plain language: “how much less bad we have done when the market is down.” In this situation the best response it, “okay, great, but how do you do when the market is up?”
Volume vs. Usefulness: This is the time-honored tactic of disclosing large volumes of useless information versus smaller volumes of useful information. It is most effective if the disclosure uses intimidating technical terminology or jargon. Recall my point about cumulative versus annualized returns above. Not too long ago I was reviewing some material from a manager who provided an enormous volume of statistical information: delta-adjusted net exposures; r-squared values versus a range of indices; betas and alphas; a tremendous number of granular data points. The one data point that was conspicuously absent? Annualized net of fee returns. You can bet that the manager did not look good on that measure.
Lies of Omission: Sometimes marketers will go to great lengths to avoid using certain words. “Leverage” is one of these. Leverage is not inherently evil but it does amplify risk. I recall a recent pitch for a strategy claiming to yield 20% annually (!!) Bear in mind that the average yield spread of the lowest-rated corporate debt to Treasuries is something like 4% these days. Of course it turned out the fund was levering its investments 4-5x on average. I think the word “leverage” appeared one time in the entire pitch book, in a footnote I’m sure they assumed no one would read. More generally, pitching anything based on “yield” versus “total return” is a kind of lie of omission. A 20% yield is not worth losing 90% of your principal.