The dividend discount model (DDM), a variation on the idea of present value, is one of the older methods for evaluating a stock. It is sometimes called the Gordon model, after Myron Gordon, an academic who wrote about it in 1959.

The two main ideas behind the DDM are:

1. a stock is a funny kind of bond and so can be valued more or less the same way one would a bond (an idea that hearkens back to the day when the stock market was the exclusive province of coupon-clipping descendants of industrial tycoons); and

2. by far the most important thing to an investor is the dividends he receives, so they, not the company’s profits, should be what is evaluated.

**Background: a. Time Value of Money**

The general idea of the *time value of money* is that the worth of a dollar I have in my hands today is different from the value of a dollar I will receive only at some future time. Figuring out today’s value of a dollar that I will receive in, say, a year starts with determining an interest rate to use in converting the future payment into its equivalent in the here and now. The rate used is typically either what it would cost me to borrow that dollar for a year, or what I think I could earn during a year with a dollar I had now.

Suppose the interest rate I choose is 5%. Then the value to me today of a dollar paid in one year is the amount, *x,* that I would need to invest at 5% to have a dollar a year from now. In other words, x times 1.05 = $1. This means x = $1/(1.05), which is $.9524. Similarly, the value today of a dollar I would get in two years is $1/(1.05 x 1.05), or $.907. Again, that’s because the money I would have from investing $.907 for two years at 5% is: $.907 x 1.05 x1.05 = $1.

** b. Bonds**

Now, think of a 10-year government bond with a coupon yield of 3%. I get twice a year interest payments of $15 each during the term of the bond and $1000 back at the end of ten years. What is this stream of income worth today? –to figure this out, add together today’s value of each of the 20 semi-annual interest payments I will receive, plus today’s value of the return of principal in ten years.

This calculation is really tedious. But give me an Excel spreadsheet and a little time, and I can set up a template that will allow me to calculate a precise value for this bond, even if I assume a different interest rate for each time period.

**The Gordon model**

Finally, to get to the Gordon model, imagine the same process, only–

1. the periodic payments of dividends I receive have a *rising trend*,

2. potentially *go on forever,* and

3. there is no return of principal.

**Its Main Stock Market Usefulness**

I worked for a while for a firm that was heavily influenced by academic theory and which had employed the dividend discount model extensively over a number of years. The firm’s conclusion? The model was good at identifying sectors or industries where stocks were significantly overvalued or undervalued. But these valuation differences could remain there for long periods of time. Although the model would confirm that a one- or two-year investment strategy was centered on the cheapest stocks, it was of little value by itself as a practical money management tool.

Also, as we’ll see below, the model must be extensively modified to be useful at all.

**What’s wrong with this picture?**

Let’s take care of a simpler issue first. During times of great stress in the markets, like late 1974 or the final months of last year (during both times, companies began to trade below the value of their net cash), investors may distinguish sharply between cash in their hands and in the hands of “those idiots” running the company, but that’s not usually the case.

Normally, investors consider profits that the firm is keeping for reinvestment–and not distributing to shareholders– to be just as much their property as dividends. So in valuing companies, investors typically use profits instead of dividends (this is very similar to the more detailed procedure merger and acquisition specialists use to value a target company). People still sometimes call this a dividend discount model calculation, though. (*Another note* (sorry): you have to be *very careful* in using profits or cash flow in a model like this. Some industries, like utilities, require reinvestment of large amounts of the cash the business generates just to keep the following years’ profits from falling. In such industries, using profits or cash flow instead of dividends will give a result that’s much too high.)

A second, kind of silly-sounding, problem: what about companies that don’t pay a dividend? As a practical matter, someone who wants to use this model will make up a story about how the company in question will begin to pay a dividend in x years, and calculate from that point. But this is really a fudge that highlights the model’s difficulty in dealing with growth stocks.

Take Oracle as an example. It has been around since the late Seventies and a part of the S&P 500 since 1989. But it has only started to pay a dividend in 2009. Admittedly, ORCL is not a growth stock any more, but how would you have dealt with it twenty years ago? How could you have used the DDM if you knew that the stock wouldn’t be paying a dividend until this year?

**The Practical Issues**

1. If you project growth for almost anything until the end of time, you get a very big number. The fact that the absolute numbers look crazy may be ok if you’re trying to establish relative valuation within a universe of stocks, but if you’re trying to determine whether stocks *in general and in absolute terms* are cheap or expensive, it’s not. People usually deal with this problem by terminating the calculation after a set number of years.

2. If you project *fast* growth for a long time, that is, any growth significantly above the interest rate you use to bring the future earnings back into today’s dollars, the values get really huge. Not only that, but the difference between fast growers and moderate growers gets progressively bigger, the farther out you go.

Usually, this problem is fixed by dividing the projection period into three (or more) parts:

–allowing a period of, say, five years, of high growth, where professional analysts’ projections are used;

–followed by a period of “fade,” that is, gradual decrease of the growth rate back to the growth of nominal GDP;

–then followed by a a period of average growth to the endpoint of the calculation period.

3. In practice, professional securities analysts have a hard time projecting earnings accurately* **one *year in advance. So even allowing a short window for stock-specific projections can introduce big distortions. You can’t rely on using brokerage house analysts’ projections, either, because they are notorious for being way too high. So you probably have to have a large, well-trained, in-house staff to do this properly.

4. The model isn’t built to deal with true growth stocks, like *Wal-Mart, Microsoft, *or *Cisco*. On the one hand, if you allow yourself to project high growth for a long time, then everything looks like a gigantic home run. On the other hand, you can force the model to work better by not allowing anything to be more than a metaphorical single or a double. In doing so, however, you define away the chance to find a company in the early stages of a ten- or twenty-year growth run. This loss is probably a reasonable price to pay for a value investor, who is probably rarely going to be involved in the high-growth area of the market anyway. But it’s a real disaster for a growth investor, whose performance depends on finding at least a couple of long-term winners.

Pingback: the stock market during the pandemic « PRACTICAL STOCK INVESTING