Yes, Mr. Market is really manic-depressive

We all know that stock prices fluctuate far more than fundamental values. It’s glaringly obvious, isn’t it? But what hard facts do we have to back it up?

As a first step, we have Robert Shiller. Substantiating that insight was part of the reason he was awarded the Nobel memorial prize in economics in 2013. Shiller did this by estimating the net present value of future dividends and comparing the development of these present values with the actual stock price movements. He then concluded that stock prices vary way too much to be explained by changes in their fundamental value.

I’m not too keen on using dividends to value companies, as this places too much emphasis on a distant terminal guesstimate. Furthermore, we keep saying that we view shares with an owner’s perspective, not that of a purely financial investor, meaning that we are concerned with the progress of each company, not its stock price. This approach aligns better with using earnings rather than dividends as an expression of progress, although we often prefer more complex figures like free cash flow or cash conversion.

As Shiller and John Campbell wrote in a later article: ‘A long moving average of earnings is a very natural variable to use to represent fundamental value, and (…) there are not many competitors for this role.’ Logically, then, present values of long earnings series should capture some of the same information.

Admittedly, earnings represent a form of double counting, as retained earnings give rise to higher earnings going forward. So I’ve produced a time series of earnings adjusted for the effect of previously retained earnings – based on adjustments I’ve made for a number of years when calculating my own version of the Cyclically Adjusted Price Earnings multiple (CAPE). This produces a slightly higher income stream than dividends, but because income streams far into the future are heavily discounted, the difference is not that big.

There is another problem inherent in his original article: What do you do when future income streams are limited to three years … two years … one year? We can’t discount unknown future dividends – or earnings, for that matter. As I understand it, Shiller concatenated data as it neared the final observation date. But that was almost 45 years ago. Today, tools like Excel provide an easier way out, by simulating adjusted earnings 50 years into the future, based on historical trends and volatility.

Do note that I’m not comparing present S&P 500 values to the calculated fundamental values in my spreadsheet – I’m not saying that the market is over- or undervalued. I’m just concerned with volatility. I have to admit, though, that using a real discount rate of 3% – which I believe is pretty much in line with Shiller – I arrive at a lower value than the present S&P 500 level. You might also argue that, with data stretching back to 1871, I should have used a variable discount rate, but I’m trying to keep it simple (it’s miles simpler than some of the academic articles on this topic).

A first observation is that earnings are a lot more volatile than dividends. However, when adjusted for retained earnings, volatility falls by about one-third, indicating that dividend changes make earnings more volatile. The present value of these adjusted earnings is remarkably stable, with volatility below 1%, as opposed to 18% for the S&P 500 (real values and log differences, for those of you who care). It is also less volatile than the present value of discounted dividends, despite the underlying time series being more volatile. Shortening the number of discounted years from 50 increases volatility, but it’s still nowhere near actual stock price movements.

What we see here is mean reversion in practice. Over the course of many years, deviations from the trend tend to cancel each other out, especially with accounting accruals that really just change the timing of costs and revenues. For a truly diversified portfolio like the S&P 500 (at least before the dominance of the ‘Magnificent Seven’), fundamental value deviates very little from its long-term trend.

I haven’t made similar calculations for individual companies, but logic dictates that the same holds for them – fundamental values are strikingly more stable than their prices. As Shiller and Cambell reference other researchers (like Eugene Fama and Kenneth French): ‘Stock returns are more highly predictable when measured over several years than when measured over one year.’

So it’s not just something we say.