I finished reading the truly amusing book, Misbehaving, by Richard Thaler, a couple of weeks ago (see here for my review of the book), and beyond being a highly readable book (the book of the summer, in my opinion) with many entertaining stories, there is a passage which I found particularly interesting. It is in Chapter 17, where Thaler mentions a debate that took place in October 1985 at the University of Chicago on the empirical relevance of the Modigliani-Miller (M&M) dividend irrelevance proposition.
The M&M dividend irrelevance proposition says that the value of a company is independent of dividend policy – i.e. it does not matter whether the firm decides to pay more or less dividends. The argument is based on arbitrage: an individual investor will be indifferent between receiving cash-flows as dividends or as capital gains and, moreover, he or she will be able to undo corporate decisions by creating “home-made” dividends – i.e. selling (parts of) the shares. This result was achieved without taking into account the tax structure, because once the tax structure was included it distorted the relative value of dividends and capital gains for the investor. As it happens, in the 1980s dividends and capital gains were very different taxed in the US. It is worth quoting Thaler’s passage in full:
One of the key assumptions in the Miller-Modigliani irrelevance theorem was the absence of taxes. Paying dividends would no longer be irrelevant if dividends were taxed differently than the other ways firms return money to their shareholders. And given the tax code in the United States at that time, firms should not have been paying dividends. The embarrassing fact was that most large firms did pay dividends.
The way taxes come into play is that income, including dividend income, was then taxed at rates as high as 50% or more, whereas capital gains were taxed at a rate of 25%. Furthermore, this latter tax was only paid when the capital gain was realized, that is, when the stock was sold. The effect of these tax rules was that shareholders would much rather get capital gains than dividends, at least if the shareholders were Econs […] So the puzzle was: why did firms punish their tax-paying shareholders by paying dividends? ” (p.165, italics in the original)
An early answer to this “puzzle”, based on behavioural theory, was offered in a paper by H. Shefrin and M. Statman presented at the Chicago conference. They proposed to resolve the dividend puzzle using a mixture of self-control theory (people may wish to consume just dividend income, so as to enforce themselves to preserve capital for retirement), desire to segregate (if the value of stocks goes down, the dividend gain is viewed as a “silver lining”) and regret aversion (it is not the same to consume out of just-received dividends than to consume out of the selling of shares). They made very clear throughout their paper, that they want to emphasise that the dividend-and-capital-gains equivalence may not hold even in the absence of taxes and transactions costs (the best-case scenario for M&M) due to behavioural considerations.
Mr. Tobin does the math
But there is a more powerful reason on why dividends and capital gains cannot be always treated as substitutes (and why firms were paying dividends at the time): the reason is not behavioural, but rather purely numerical. Paraphrasing Shefrin and Statman, we could say that the dividend-and-capital-gains equivalence may not hold even in the absence of taxes, transactions costs and behavioural considerations. To illustrate why it is simply about math, consider the following graph, which shows Tobin’s q historical series in the US since 1951:
Tobin’s q is defined as firms’ asset market value to its replacement cost. Note that when the conference took place in the mid 1980s, Tobin’s q value was around 0.5. So let’s take this figure for the following calculations. Because for the discussion what is needed is the equity q (i.e. the value of equity at market prices to its replacement cost) rather than Tobin’s q, we have to convert the previous Tobin’s q number to an “equity” q number. Simple algebra shows that the formula to lever the original Tobin’s q is:
Equity q = (Tobin’s q – leverage ratio)/(1 – leverage ratio)
Being the leverage ratio the ratio of total debt to total assets. Quarterly data from the Integrated Accounts shows that in 1980q1 the non-financial corporate sector had loans worth $463.3 billion, debt securities of $412.1 billion and total assets of $4,870 billion, which yields a leverage ratio of around 18% (taking any other quarter will not change the essence of the argument). With these numbers, the equity q would be:
Equity q = (0.5 – 0.18)/(1 – 0.18) = 0.39
We have all the ingredients now to do some math. Following the tax rate numbers given by Thaler, the value of a unit of dividends for the shareholder would be:
Value of one unit of dividends = $1 of dividends * (1 – 50% tax rate) = $0.5
And the value of one unit of realised capital gains (for simplification I assume that capital gains are realised in order to have “home-made” dividends):
Value of one unit of realised capital gains =
$1 of retained earnings * equity q of 0.39 * (1 – 25% tax rate) = $0.29
Obviously, the equity q value matters a lot when a firm decides to reinvest its earnings: if, for whatever reason (e.g. poor prospects), the market says that the equity is worth below book value (when q is below 1), then it means that earnings reinvested cannot raise one by one the value of the stock, so the shareholder will have to sell a higher number of shares in order to make a home-made dividend (in other words, the shareholder will be worse off). Of course, this number is an aggregate, so there will be firms that will command higher valuation ratios (e.g. growth stocks) for which the previous argument will apply to a lesser extent. But on average, that was the situation of the corporate sector as a whole in the 1970s and 1980s. I do not deny the importance of the behavioural explanation, but I am just trying to point out (Ockham’s razor) that a more straightforward explanation may have been more relevant in this case for the “dividend puzzle”.
As a final note, one might conclude that, if the previous argument is correct, then the M&M framework was right because firms were doing the right thing (maximising value) for their shareholders. Not quite. The very fact that Tobin’s q has been persistently different from one (and trending upwards and downwards for whole decades) is direct evidence of the lack of relevance of the M&M propositions. But that’s another story.