Modeling the price of a stock based upon its dividend history

Question

The value of a stock is the present value of all future dividends. This is sometimes called the Gordon Growth model. This model assumes that dividends increase at a constant rate. In the real world that may not be right. In addition, there is no way to know what the long term rate of the dividend growth will be.

One way, is to look at the current dividend rate and the dividend rate a while back. Say 10 years. You can them compute the annualized dividend rate.

It seems to me a better approach, would be to look at the annual dividend rate, say for 10 years and compute a exponential function for the dividend history. This function would be computed using a least squares approach possible giving more weight to recent dividends.

Would it be better to use the second approach over the first approach in computing the rate of growth in the dividend?

Please comment.

Answer 1

Your question borders on opinion-based, but I'll try.

Corporations' earnings and profits are volatile and unpredictable.

Publicly traded corporations' common share prices are even more volatile, driven by a lot of factors other than expectations of dividends and boards believe that investors prefer dividend rates to be less volatile and unpredictable.

In order to attenuate the inherent volatility, boards often announce plans for dividend rate and its growth. Although the plans are non-binding, corporations sometimes borrow money in order to pay dividends or otherwise use money that arguably might be spent better long term. In other words, people sometimes try hard to wipe out the information contained in share price and earnings time series.

So, what can you do, nevertheless, with a dividend rate time series? You could, for starters, see how well your model price, i.e. the present value of the indicated future dividends, explains the observed share price. Many others have looked at this, so you should have little trouble comparing your findings with published papers.

You could look at further inputs, such as industry, history of earnings, and history of debts, to characterize corporations that historically were better at making their dividends predictable, e.g. regulated utilities. You could try to predict which ones are likely to fail or succeed at this. You could even try to investigate how surprise changes in divident rates affect stock prices.

Answer 2

The Gordon Growth Model (GGM) is just a simple model. I don't imagine any serious user of the GGM believes in its assumptions or that calculating the true value of a stock is as simple as plugging in some numbers.

Your approach might indeed lead to better estimates as it uses more data than the GGM but past dividends alone are not enough to accurately value a stock. Disadvantages of your approach are that it's more complex and requires more data.