There has been a lot of talk in the last few days about an article on MarketWatch referring to trend regression. The article argues that we could be near a bottom because we are below the long-term trend line of 7%.

Taking a purely statistical approach I find it hard to understand how this data is at all relevant to today’s  (and more importantly tomorrow’s) market.  A trend-line is nothing more than the average past performance of a market.   An investor in Japan could have easily made the same conclusion in 1990 after the Nikkei had declined over 50% from its high.  From 1920 to 1990, the Nikkei had grown roughly 9% per year.   Using a long-term trend-line you might have concluded in 1995 that the Nikkei was  undervalued because it is expected to grow 9% per year.  But the Nikkei fell another 50% over the coming decade.

The problem with using technical analysis and trend-lines in this manner is that the data can be skewed however you prefer.  Drawing a trend-line on a chart is ambiguous.   For instance, running the S&P 500 numbers since 1900 comes up with a 6% average trend-line.   Skewing the data a few years or a few 10ths of a % point changes the trend line entirely.  Who is to say that Siegel’s 7% growth figure is accurate?  The Nikkei had a compound annual growth rate of 9% per year as of 1990.  Was that wrong?  What on earth convinces Siegel that the U.S. stock market will grow at 7% for the rest of time?  Just the fact that it did so over the prior 137 years?

This sort of data mining is useless.  The data set for the stock market is so small (137 years) that you have to take any study like this with a grain of salt.  There have been just  21 recessions since 1900.   And the stock market of today is a very different place than it was back in 1900 or 1950.  You could make the argument that none of those recessions or data are applicable today because today’s market is so different.

Doug Short comes to the same conclusion although he reaches his answer a bit differently. Technical analysis has its place in investing, however, I find this sort of data mining counterproductive.   Depending on a future rate of return solely based on past performance is a great way to get your portfolio into trouble.