In Praise of Auto-Pilot Investing

Winging It This Wednesday, Bloomberg detailed an investment experiment. Sixty-one "quantitatively trained test subjects" were given $25 with instructions that they could wager as often as they wished, as many dollars as they liked, on a metaphorical coin that landed on heads 60% of the time. The maximum payout was $250. Obviously, it made sense to enter the game, as the bet paid off six times out of 10. But how to play?

One approach would be to stake the entire $25 on the first bet. If that wager pays off, the player now possesses $50, and places it all on the second bet. If that pays off, the player has $100, and places it all on the third bet. If that pays off, the player has $200 and can achieve the maximum payout by making and winning a $50 wager.

It's an idea, all right, but it's a terrible idea. On 78.4% of occasions, the player who follows such a strategy would lose one of the first three wagers and would then almost immediately be busted. Going to zero while playing a game rigged in one’s favor is, shall we say, disappointing.

(Note that the evaluation differs greatly for one bet versus a series of bets. The highest expected value for a single wager, $5, comes from staking the entire $25. That is the correct approach for the risk-accepting player who is not worried about losing his entire position, should the virtual coin land on tails. But for a series of bets, putting up anything close to the full amount on a single occasion is foolish.)

Superior strategy yields a much superior result. Rather than immediately losing everything almost 80% of the time, optimal tactics give a 95% chance of reaching the maximum $250 payout. Quite a difference!

Despite their quantitative training, most of the test subjects flopped. Only 20% of them gained the maximum payout, as opposed to the 95% predicted by best practices. Conversely, whereas almost nobody should have gone bankrupt, 17 of the 61 players managed to turn gold into lead. The reason for the collective failure: Few did the math. That left them relying on their judgment and, thus, adrift. Without an analytic “framework to rely upon,” wrote the authors, the “subjects exhibited a menu of widely documented behavioral biases such as illusion of control, anchoring, over-betting, sunk-cost bias, and gambler’s fallacy.”

From Gambling to Investments The merits of auto-piloting extend far beyond this experiment. Whether participating in a gambling test or managing one's life savings, the best results are likely to come from thinking hard when initially forming the plan. After that, one should think much less--or perhaps not at all. As with gamblers, investors are unlikely to improve matters by adjusting on the fly. The better approach is to build the investment adjustments into the initial plan, then automatically implement those changes according to the pre-established rules.

This recent study reinforces a finding from the middle of last decade, when academic researchers found that brain-damaged subjects excelled at investment games. These people were not intellectually shortchanged; rather, they had inhibited emotions due to lesions their brains had developed. They were, in short, fearless--and robotic. While normal participants in Stanford's investment game became "reactive" as results occurred and used their judgment to alter their tactics, the brain-damaged participants just chugged along, following their original plans. And they won. The auto-pilots beat the human pilots.

In Retirement Retirement spending offers one example of how auto-piloting can be used. As I've written elsewhere, retirement-spending models typically are static. The retiree spends a given amount one year; then that same amount the following year, adjusted for inflation; then that same amount the year after that, once again adjusted for inflation; and so forth. That approach is unwise. It is needlessly rigid, thereby forcing the withdrawal percentage to be lower than it otherwise could be, thereby reducing the retiree's standard of living. Better to plan for a flexible spending schedule.

For example, a retiree might decide on a spending rule that consists of withdrawing 4.5% of her year-end asset pool, regardless of the size of that pool. Thus, if the financial markets rise sharply in a year, such that her pool grows by more than her 4.5% withdrawal rate, then the following year she will spend more money--perhaps significantly more. Conversely, if the financial markets were to drop, she would spend less in Year 2. The program is scripted. There is no human intervention.

The same principle describes the optimal strategy for the Bloomberg article's experiment. According to a formula called the Kelly criterion, the best approach for a gambling game with a series of bets that have 60% chance of success is to bet 20% of one's holdings. That is exactly akin to the retiree's flexible-spending plan. In each case, one determines a withdrawal rate (4% for the retiree, 20% for the gambler) that effectively raises the stakes after each success (a winning year for the financial markets, or a successful coin flip) and cuts them after a failure. Once again, there is no room for improvisation.

To be sure, sometimes there are occasions when judgment must intervene. If circumstances change so much that the original assumptions are invalidated, then the original plan cannot stand. That occurs, however, far more in investors' imaginations than in reality. In most cases, for most shareholders, the wisest reaction to an apparently unforeseen event is to pretend that the event was not, in fact, seen.

John Rekenthaler has been researching the fund industry since 1988. He is now a columnist for Morningstar.com and a member of Morningstar's investment research department. John is quick to point out that while Morningstar typically agrees with the views of the Rekenthaler Report, his views are his own.