Misusing the Most Powerful Computer - Part 2

Those who do not understand research  methods have been empowered to make mistakes.  Computers permit backtesting and curve-fitting.  The system developer who takes ALL of the past data develops a perfect model -- perfect for "post-diction."

Background

Using data to develop predictive models is much more challenging.  The process involves -- minimally -- holding out-of-sample data for testing.  Better yet is extensive testing on different stocks and time periods.  Anyone who does the systematic  development of models understands this or quickly fails as a trader.

Misuse of the Most Powerful Computer

The most powerful computer is the human mind.  It can be easily abused.  The mistaken process is a simple one.

Start with the conclusion one seeks.  Then find the pattern.

A few months ago we described how many pundits were using this method by finding charts.  If one starts with the conclusion, it is easy to find charts that support it.

An objective, scientific approach would be to find all chart patterns describing a particular "setup" and then look at what happens.  The best analysts do this.  One often finds that there are not enough cases for reliable inference.

Making Comparisons with Words

We now see many analysts who use words to accomplish the same objective as the chart abusers.  Pundits see the current market as similar to that of 1929, or 1937, or 1974, or 1987, or 2000.  The debate rages among bearish pundits about which disaster scenario is most comparable to the current market.

Here is a useful "sniff test."  Remember back to your college days when you had an essay question that asked you to "compare and contrast."  If the writer makes no effort to do both sides of the analysis, just move on.  That writer is not making a serious effort  to analyze.

The Case of 1987

Helene Meisler, an excellent technical analyst who writes for RealMoney (paid site at the.street.com and worth it for serious investors) was asked to put up a chart comparing 1987 to the current market.  She did so.   The chart  did indeed show a pattern similar to the current market, but her conclusion was that it is not really a similar case.  She cited a number of technical factors.

The Real Story

I posted a comment on RealMoney that provided the "contrast".  This is pretty easy to do in any of the proposed comparisons, but an analyst with an agenda will not provide this for the reader.  Here is the contrasting 1987 analysis, as I wrote it:

Meanwhile, I am also getting questions about the 1987 comparison. It has been written about enough to worry some of my clients. I see (at least) four important differences:

1) Any valuation method that uses both earnings and interest rates shows stocks very overvalued in 1987 and very undervalued now.

2) Individual investors were "all in" on mutual funds and then they headed for the exits all weekend before the Monday crash. The process of handling redemptions was clumsy. The individual investor of today does not have the same equity exposure.

3) The system of 'portfolio insurance' gave institutions an unfounded sense of complacency. Managers believed that they could sell futures contracts to hedge positions after a big decline started, so they were also "all-in." This is akin to buying insurance after your house is on fire. The post-crash discussions (and there are many fine ones) all emphasize this dependence on prospective selling of futures. There were also no circuit breakers. Today's managers have learned this lesson, and put premiums are low.

4) The complacency was so great that individual investors regularly sold naked puts as a means of generating monthly income. You can see the ads for such "systems" in the issue of Barron's that came out the day of the crash. Aggressive put-selling, especially by naive investors, is a good measure of complacency, and something you do not see now.

Conclusion

This is not 1987, but it is easy to be misled by facile comparisons and charts.  If the writer does not seem to be providing a balanced analysis,  just say "no" and move on.