April 01, 2018
Science Versus Art
Investing is part of economics, which is a social science. Economists and investment professionals like to take a scientific approach to their work, employing computers using advanced mathematics to build forecast models. Much stock trading today is driven by computer algorithms. And with advances in Artificial Intelligence, there is speculation in the media that many investment professionals (possibly even yours truly?) may be replaced by smart computers and robo advisors
Back in graduate school, having studied history, philosophy, and sociology, I wrote a short essay entitled “Can There Be Social Science?”. Natural science is focused on discovering immutable laws of nature that can be verified or disproven by real world experiments. It is hard to conduct such experiments in sociology or economics (although many would argue that politicians do this all the time). The point of my essay was that there are no immutable laws to be discovered in the social sciences because humans keep moving the yardsticks. For example, today slavery is outlawed as an denial of basic human rights yet it was the basis of many economies from the time of the ancient Egyptians to as recently as 150 years ago in the southern U.S.A. Human values change, which changes economics.
Also, people often act on emotion rather than logic. In fact, one of the main arguments for “scientific” investing - indexing, computer trading, and robo advisors - is that this removes human emotions from the investment process because studies in behavioural finance have shown that most people impair their long-term portfolio performance by reacting emotionally to market movements. It would be wonderful to be able to reduce investing to a science, build a computer model, press “enter”, then just focus on your golf game.
However, the scientific approach often creates a false impression of precision. Hundreds of Ph.D.’s employed by central banks build econometric models to forecast GDP growth and inflation but the record shows that their predictions generally do not come close to actual results. Similarly, equity analysts build computer models to forecast corporate earnings and calculate stock valuations. The primary method they employ is the “discounted cash flow” model. It is based on a fundamental concept of classical investment theory that the value of any given stock reflects the future stream of dividends you expect to receive over time. Analysts forecast dividend payments (or cash flows) in future years and then “discount” each amount back to its present value using an assumed discount rate.
That may sound complicated but the process itself is simple. Money has a time value. If you receive a $10,000 bonus today and invest it for, say, two years before you need it to pay for a new car, at the end of the period you will have not only the $10,000 but also an amount of earned income. But if you receive the bonus two years from now, you will only have the $10,000. Looked at another way, if you need $10,000 to buy a car in two years, how much would you have to put aside now if you could earn 5% (net of tax) on the money in the meantime? The answer is $9,070.30. This is the “present value” of $10,000 two years from now at a discount rate of 5%.
While it is easy to calculate present values for amounts to be received in future years and add them up on a computer to generate an estimated present value for a given company’s shares, this discounted dividend methodology is not nearly as precise as it appears. First, it is not easy to forecast future cash flows, earnings or dividends. For many kinds of businesses, the margin of error can be very large. Only specific investments, like apartment buildings, pipelines, mines, toll roads, etc. allow for reliable predictions of future cash flow. For most other businesses, there are far too many variables to allow any analyst, no matter how smart or how sophisticated their computer model, to forecast earnings more than a year or two into the future. Long term forecasts are more art than science.
And although some predict that analysts will be replaced by robots, no one would suggest that a CEO of a major corporation could be replaced by a machine. And one of the first things an equity analyst learns is that management is a critically important factor in valuing a stock. Would a robo analyst be able to interview a CEO and make a determination of management’s ability going forward? Assessing a company’s management and business plan is not something that can be reduced to a computer algorithm.
But a greater problem with the dividend discount model for stock valuation is the choice of an appropriate discount factor. In general terms, an investor would choose a discount rate equal to the rate of return available from competing investment opportunities, adjusted for the perceived level of risk in the choices available. The discount factor is simply the minimum rate of return you would require in order to buy something. But others may apply different discount factors depending upon their particular view of a given security and their personal investment objectives. There is no single correct discount factor provided by “science” because it is everyone’s personal choice to make.
The choice of discount rate can make a huge difference in stock valuation. With current low long-term bond yields, the stock market values may be reflecting discount rates in the order of 5%. But if inflation rises causing interest rates to climb, investors would revalue stocks using higher discount rates. A rate of only 7.5% would cut estimated present values of most companies by about one-third and the market would suffer a sharp decline, even without any reduction in cash flow forecasts.
In short, to quote one industry analyst, Rick Neaton of Riverside Investment Research:
The price of an individual stock represents a present value based upon an expectation of a certain amount of future revenue and/or earnings. The calculation of that value is subjective. Even what constitutes a reasonable PE ratio is a subjective calculation. Certain systems make attempts to award values based upon “objective mathematic criteria”. In the end, they are no less subjective…….
Forecasting earnings and cash flows, and evaluating managements and business plans, requires judgement and a view about how the future will unfold. Computers only run on historical data and experience shows that you cannot use a computer model to extrapolate past trends into the future with any accuracy. Discount rates investors apply to value stocks change when market perceptions change in regard to inflation, interest rates, and economic growth – there is no single “scientific” discount factor to input into a computer model. Determination of the value of a stock - or the market as a whole - is up to the humans who happen to be buying and selling. Plato said it best over 2200 years ago: “Beauty lies in the eyes of the beholder”.
An analyst or advisor needs to assess not only what a given company will do in future years but also how other investors are going to react to future developments and on what basis they will make their determination of share values. Can a robot do that? Computers may be good at science but there is as much art to investing as there is science, and humans are better at art. The moral is: don’t dump your faithful investment advisor for a robot.
This information, including any opinion, is based on various sources believed to be reliable, but its accuracy cannot be guaranteed and is subject to change. CIBC and CIBC World Markets Inc., their affiliates, directors, officers and employees may buy, sell, or hold a position in securities of a company mentioned herein, its affiliates or subsidiaries, and may also perform financial advisory services, investment banking or other services for, or have lending or other credit relationships with the same. CIBC World Markets Inc. and its representatives will receive sales commissions and/or a spread between bid and ask prices if you purchase, sell or hold the securities referred to above. © CIBC World
Markets Inc. 2020