I ran a $10 million international account in 1988 and 1989 using a quantitative methodology -- a black box -- while at
Morgan Stanley Asset Management
. Morgan had recently purchased the
database, and my backtest on the data showed compelling excess returns associated with a country-neutral approach emphasizing low price-to-book and low price-to-cash flow.
The backtest was adopted by Morgan's London-based asset management arm as central to what evolved into a successful marketing strategy, and the findings were written up by
for the weekly
. I was excited, and lost no time investing $10 million in client funds.
Every month's end, I would use the freshly updated MSCI data to screen among about 20 non-U.S. countries, arriving at the most attractive 20% of companies within each country based on a combination of low price-to-cash flow and low price-to-book. After arriving at a buy list, I would generate a series of purchase and sell orders relative to the current portfolio. The list would be faxed to Morgan's international trading desk, and the desk would execute the orders. Confirms would be available the following day.
Over the next year, the portfolio matched the EAFE index. Despite reasonable overall performance, there were some major problems with the strategy that I think offer useful anecdotal lessons.
First of all, the prices I was using for my screens were very different from the prices at which the trades were actually transacted. Some of the prices were off by as much as 15%. Luckily for me, the stocks I paid up for were often the winning positions. It is doubtful that these successes should be credited to the value-oriented quantitative strategy I was trying to implement.
Secondly, the lack of knowledge regarding individual companies resulted in some embarrassing circumstances. I remember one instance in which a low price-to-book company went bankrupt three months after it was purchased for the account. Upon reviewing the company, I learned that the book value published in the MSCI database did not reflect a recent writeoff. The database had maintained an inaccurate book value, reported by the company at the end of the previous fiscal year.
The lessons learned are still applicable to quantitative investment strategies, both domestically and abroad. Data timeliness and accuracy are as important as the specific quantitative algorithm in determining investment results. Purchase candidates should be checked on the fundamentals, and if the numbers do not accurately reflect the current situation, then an individual override should be allowed.
Finally, backtests should be rigorous and detailed; summary results should be treated with skepticism. The bankrupt company mentioned earlier was based in Australia, one of the countries that backtested particularly well on the price-to-book factor. In retrospect, my particular backtest had not included dead companies and was therefore affected by survivorship bias. A more rigorous backtest might have heightened my awareness of the danger of bankruptcy in low price-to-book companies.
Ted Murphy ( email@example.com) operates the MarketPlayer Web site. Prior to MarketPlayer, he was a partner at
Equinox Capital Management.