5 ms·

If I remember correctly, that was suggested half-jokingly by the author to demonstrate that any metric where the numerator is proportional to the stock market w

• YjSe2GMQ 8y ago
  Yup, I should have included the link you posted for a complete picture. The value of the stock allocation metric is in that it seems to be the strongest mean that the stock reverts to, in the sense that it has strongest predictive power. It is not that surprising because it uses to the raw supply/demand dynamics.
  • DavidSJ 8y ago
    But the author’s point is it’s only predictive in hindsight, due to reversion to the mean of the past data (since investment allocation is basically just measuring price over the short run), not to some objective mean that also applies to the future.
    • YjSe2GMQ 8y ago
      Well that's the hypothesis - that since reversion to the mean stock allocation has always happened it will continue to happen. There's things that can invalidate the assumption, like the widespread common wisdom of HODLing index ETFs (till retirement, no matter what). If the hypothesis is false we could see stock market allocation go all the way up to 80%, or even beyond. Am I wrong?
      • DavidSJ 8y ago
        The author was not presenting it as a serious hypothesis. They were showing how compelling it looks at first glance, and how deceptive that naive analysis is, because of the unfair advantage the chart gets (in the sense that it inherently looks better than its predictive value). They did this to knock down many other valuation-based measures, not to advocate for this one. Reversion to what mean, is the author’s point. The mean for the twenty years between 1900-1920 is different from the mean for 1900-1940, is different from the mean for 1900-1960, etc. The mean is continually shifting. All data sets revert to their own mean, on average, by definition. The mere fact that a data set does so is tautological and thus provides no evidence — no confirmation of any scientific hypothesis. Edit: see my other comment for the quote of what I consider the key passage in the article.
• DavidSJ 8y ago
  The key passage: Charts like these (including my own) that attempt to correlate valuation metrics with future returns start off at a significant unfair advantage relative to other types of correlation efforts. Note that a valuation metric is just the current price divided by some variable (earnings, book value, sales, etc.). Neglecting dividends, the long-term future return is just the difference between the current price and some price far out in the future. Notice that “current price” shows up in both of these terms. Is it such a surprise, then, that valuation metrics and future returns seem to correlate well? Roughly: (1) Valuation Metric = Current Price / Variable (2) Future Return = Future Price – Current Price If future prices are inclined to rise at some rate over the long-term, then any time current price falls (and the same fall isn’t exactly mimicked way out in the future), (1) will go down, and (2) will go up. The valuation metric will fall, and the return–the distance between the future price and the current price–will rise. Hence the (inverse) correlation. Now, if you choose a denominator for the valuation metric that is highly noisy, its noise may get in the way. But if you choose a denominator that is smooth over time, the pattern will hold. Notably, the plot of the valuation metric versus future return will end up producing a series of coinciding squiggles and jumps that create the visual illusion of non-trivial correlative strength, when there is none.