Be Careful With Statistics

Mark Hulbert recently published a post,

"This is what the stock-market indicator with the best track record is telling us.",

 on MarketWatch.com, based on Ned Davis research arguing against those who believe that high "sideline cash" could lead to a strong upsurge in equities.

Let me say right off that I don't know where the market is going.  I believe that over the next 10 years stocks will do better than bonds and bonds will do better than cash equivalents. I believe that stocks could do a lot better given that the information age puts all kinds of information at the fingertips of very smart, creative, energetic people around the world. But...I could be wrong. Furthermore. I will be the first to stand up and admit that Hulbert is a lot smarter than me and I enjoy reading his posts.

Still, I think one needs to be careful with time series data on stock investing and this post by Hulbert is a prime example.

I wasn't investing in 1951 but I was investing in the 1960s. In the 1960s the investing world was dominated by defined benefit plans and the defined benefit plans were the responsibility of Trustees. Recency bias dominated the the investment process. Trustees looked over their shoulders and saw the markets of the 1930s. As a result allocations to stocks were low.

As an individual my investment choices were actively managed mutual funds that charged a lot expense wise or individual stocks at a high commission. If I wanted to buy or sell I had to talk to a broker who inevitably tried to up sell me by pressuring me into high commission product. To see how my stocks did the previous day I went to the mailbox and got the newspaper.

Today of course I can buy ETFs with the push of a button that track markets around the world. Today we live in a defined contribution world. Today, according to the Census Bureau, 79% of Americans have access to a 401 (k) at work. Today the small investor can easily and economically  invest in the overall market, a choice that has been shown to outperform most active managers over the longer term.

The point here is that using time series data from the 1950s on up even to the 1990s compared to today is comparing apples and oranges. In effect, arguing for the "reversion to the mean" maybe misleading as  the mean could be changing significantly over the period examined.

So maybe you say "no harm no foul", the data just needs to be taken with a grain of salt. Unfortunately I think the harm is greater in the following sense. Many investors can be persuaded by slick talking advisors using this kind of information to alter their investment approach. Using terms like "mean reversion", R-squared, and performance by "quintiles" gives the report seemingly expert opinion.

Just saying!