Demographics and Returns
With longevity rising and fertility leveling-off, unless retirement ages rise commensurately, the ratio of non-workers to workers will go up. This will tend to create more demand for fixed-income assets, driving down yields. Something has to give, and retirement ages are headed higher.
What about equity returns? There is a sense in which the stock market is fundamentally a pyramid scheme. New participants are important, and equities will tend to rise along with population and participation. While the population of developed countries has been stable in the recent past, globalization has picked up the slack in this regard. There is little doubt that globalization is responsible for much of the non-illusory gains of the past twenty years. Now, even if we assume that this trend will continue, will the rate of change slow? Are we past the inflection point of global development? That would not be particularly supportive of equities. Of course it's not just a matter of bodies in the system, but what minds produce and consume.
On the consumption side, there are some disturbing demographic arguments out there, but one of the more initially plausible ones does not appear to hold up to closer examination. The "age wave" theory promoted by H.S. Dent is based on the observation that personal consumption peaks around the time that one is 48 years old. Dent graphs the US stock market alongside immigration-adjusted birth statistics lagged by 48 years. The peaks and declines seem to correspond, suggesting that equity returns are driven by this demographic consumption effect, and that the best days are behind us for some time.
There are a few problems with this argument, one being that such a model does not actually seem to correspond to aggregate consumption. For one thing, US population has drifted upwards in the last 20 years, swamping the consumption-by-age effect. But what if overall population levels-off? Higher retirement ages could again help, as the peak consumption age would presumably also rise.
More importantly though, there are two distortions in the data as Dent presents it. He bases his consumption-by-age curve on the annual BLS Consumer Expenditure survey. One issue is that the survey is based on households, not individual consumers, with the breakdown by age cohort referring to the "reference person" of each household or "consumer unit". This means that if a household has children, the consumption of the reference person as a function of their age will be overstated. If the goal is to model future consumption rates based on consumption-by-age, the model should isolate the latter and not try to predict future fertility rates. When child consumption is mixed with that of the reference person, consumption-by-age is conflated with past fertility rates. Now, when you adjust the BLS data for the average number of children in each household (assuming that children consume some small percentage of what adults consume), the peak consumption age rises to the mid 50s from the late 40s. Not to imply that relative fertility-by-age is constant, but that peak is drifting higher, pushing the conclusion in the same direction.
Furthermore, you will notice on the first graph above that the y-axis has no scale, giving the impression that the consumption of 65 year-olds is roughly that of 25-year olds. When you actually look at the BLS data, adjusted or not, this is simply untrue. The 65-74 cohort consumes less per capita than the 25-34 cohort, but the fall-off is nowhere near as dramatic as Dent presents it. While the granularity of the data does not allow us to compares 65- and 25-year-olds directly, their consumption certainly is not equal. If anything, consumption does not decline back to the 25-year-old level until around 75. When you put all of this together, along with UN projections of US population, aggregate US consumption as a function of demographics only will continue to rise despite a greater percentage of Americans being past their peak spending age. One is much more confident then in dismissing the second graph as an accident of history.
All of this just begins to address some questions that are very pertinent to prospective returns.