HomerJ wrote:
Nowhere near that exact.

I have never looked at valuations or expected returns or tried to calculate a "required savings rate". We just saved a lot.

I see. So this statement wasn't really accurate then:

HomerJ wrote:<snip> and planned on getting around 4% real long-term on the stock side to figure out how much I need to save.

Nevertheless, your approach of "saving a lot" probably is a good idea for most people, since most people don't save nearly enough for a comfortable retirement (there are lots of statistics on this, and I see it in interacting with young people I come in contact with through my blog, friend's children and their friends, etc.).

Coincidentally, I've been writing a blog series on calculating required savings rates, and have been using 4% real for stock expected returns in most of the examples. Of course I came up with this figure using the types of models we've been discussing here rather than using worst-case historical returns, and I've emphasized that this is not a point estimate, but the mean of a wide dispersion of probably outcomes.

HomerJ wrote:I basically just use 4% now to see how we're doing when playing around with financial calculators. As in, "If we get 4% from now on, here's how much we'll have at 55. Sweet!" I did choose a conservative number so that's there a lot of upside potential. If we get less, I'll work longer or cut back on our retirement lifestyle. So it's not a worst-case scenario. If we get less, I'll adjust.

Gotcha. I'd just comment that I wouldn't consider 4% real a conservative number for "how much we'll have at 55", assuming you are age 35 or older, since we've seen historical real US stock returns as low as about 1% for 20-year time periods (1962-1981).

HomerJ wrote:But I'm going to adjust on ACTUAL returns, not on what you guys guess is going to happen.

Larry also talks a lot about adjusting to actual returns, since he acknowledges that actual return could be at the low end of the dispersion around expected return, so I think there's no disagreement about that.

The thing is that if someone determines their savings rate (or amount of risk in their portfolio, or whatever) based on artificially high estimates of future returns, or simply doesn't save enough because they just don't think about it, then the closer they get to the terminal point (say retirement, if that's the goal), the less ability they have to adjust their savings rate (or amount of risk, or whatever) to compensate. Clearly this hasn't been a problem for you, but it is a problem for many other people. So it seems to me that most people should give some thought to expected return estimates in their planning, and then we can just debate which approaches to determine these estimates make the most sense.

HomerJ wrote:Does it bother any of you that many of the "expected return" predictions in the past (recent past even) have turned out to be wrong? Anyone who used them to carefully plan out an investment strategy or required savings rate has not seen the numbers that were predicted.

Continuing to say it this way misrepresents the meaning of "expected return", which is the mean of a dispersion of probable returns. It is not a prediction. Since it is not a prediction, it can't really be wrong, and if we look at it correctly with approriate error bands, the 20-year expected return forecasts for US stocks using E10/P have been remarkably accurate. All you have to do is look at some of the charts that I've shared to see that essentially all 20-year returns for US stocks since 1926 have been within +/- 4% (percentage points) of the initial E10/P value, and have been dispersed quite symetrically around this value.

I think Larry is very honest and fair about this in his article:

Larry Swedroe wrote:Thus, we must be very humble about how we think about such forecasts. By that I mean we shouldn't treat them as single-point estimates. With U.S. stocks, based on the historical evidence, to include all actual subsequent outcomes, you would have to both add and subtract about 8% from the expected real return.

In other words, while the mean expected real return for the S&P 500 Index going forward is 4.1%, potential outcomes range from a real loss of about 4% to a real gain of about 12%. That's a very wide dispersion of potential outcomes. It also shows how difficult it is to forecast returns.

I'd say that Larry is even much more conservative in terms of the +/-8% error band than the historical evidence suggests, at least for 20-year time periods for US stocks since 1926, which we have seen is about +/-4%.

Of course we can't really know that the future will resemble the past, but to the extent it does, it's highly unlikely that you'll earn more than 8% or less than 0% real over the next 20 years with a current E10/P of about 4%. It's certainly less unlikely than that you will earn no less than 4% real over the next 20 years.

HomerJ wrote:Look, my opinion is you pick a conservative number for planning and adjust to actual returns.

Picking a conservative number for planning seems prudent, in which case we're back to discussing different methods for determining what a conservative number is. As already mentioned, I think we're all in agreement about adjusting to actual returns (after all, what other choice do we have?).

HomerJ wrote:Surely, you wouldn't suggest to someone to cut back on their savings rate if valuations were low, and expected return was high? (The proper answer to that is "Don't call me Shirley")

OK, I won't call your Shirley.

I tend to agree with you that one should simply save as much as possible. Unfortunately most people don't save enough, so probably the best advice for most people is to save as much as possible--actually, probably to save more than they think is possible by cutting back on what they perceive to be necessary expenses that are in reality discretionay expenses (the amount my kids spend on cell phones and entertainment comes to mind).

But economists do discuss the concept of consumption smoothing, the idea being not to deprive yourself excessively during your accumulation years, but also not to consume so much and save so little that your consumption must be severely reduced in your retirement years. To the extent you think this makes sense, it can indeed make sense to adjust your savings rate (and/or the amount of risk you take in your portfolio) based on your esimate of future returns.

HomerJ wrote:Because what if the forecast was wrong, just like it's been wrong many times before?.

Again, when viewed properly, with approrpriate error bands as Larry is advocating, expected return forecasts using either earnings yield or some form of the Gordon model have been remarkably accurate.

HomerJ wrote:You can't really plan out a "required savings rate" on a forecast that has a plus or minus 8% range.

This isn't really so different than planning on what you consider to be a conservative estimate of future returns, which although you say you haven't actually done, you have said more than once that you think this makes sense. If you want to be conservative, pick something toward the bottom of the range, if you want to place more emphasis on consumption smoothing, perhaps pick something closer to the middle of the range.

Interestingly, both Larry's expected-return estimate of about 4% real and your worst-case estimate of about 4% real are the same value! Larry's approach is even more conservative, since he puts the worst-case real return at -4%, while I'd say it's probably closer to 0% to the extent the earnings yield model has predictive power (out to about 20 years), and closer to -1% if you rely just on historical 20-year returns.

Of course the lower bound is higher, at about 4%, if you look at 30-year US stock returns since 1926, but we only have three independent 30-year periods since 1926, which isn't nearly enough to make any statistically valid forecasts. Using a lower bound of 4% real is placing even more confidence that the future will resemble the past than using a model that has been more accurate historically.

Also, it doesn't really make sense to use worst-case historical stock returns as a worst-case estimate for a balanced portfolio of stocks and bonds, which would only be about 60% stocks at age 40 with an age-in-bonds portfolio construction model.

HomerJ wrote:I guess I don't care about valuations, because I plan around a number that is low enough to be nearly off the bottom of the chart of the range of expected returns anyway.

Again, it's not at all off the bottom of the chart if you look at different time periods (e.g., 20 years instead of 30 years), and use slightly more sophisticated models that have been historically much more accurate than simply looking at the lowest 30-year rolling return. But now we're just back to discussing what forecasting model makes the most sense.

I've focused a lot on savings rate because we seem to have more common ground here than with respect to using an expected return forecast to adjust your asset allocation, even though we disagree on which forecasting model makes the most sense (it seems clear to me that planning around a worst-case real return of 4% is based on a simplistic forecasting model, whether it's acknowledged or not). Larry just extends the use of an expected return forecast to portfolio construction, so let's bring the conversation back to that.

You have a model for portfolio construction, which you have said is "age in bonds". Aside from John Bogle's recommendation for using this as a starting point, Bill Bernstein and others have taught us that a rational economic justification for doing something along these lines involves the concept of human capital and financial capital. But this is really just a subset of Larry's broader paradigm of basing portfolio construction on ones ability, willingness and need to take risk. You generally have more ability to take risk when young, since then your human/labor capital is enormously larger than your financial/investment capital, so a risky portfolio is not really very risky when viewed in the context of your total capital (human + financial). Or to put it another way, using a concept that you and Larry both use, you have more time to adjust your plan when young.

Larry adds the concept of need to take risk to the model, and uses expected return as one component of determining need to take risk.

So you have a model, and Larry has a model, so it seems that what we're really discussing is which portfolio construction model makes more sense, and what factors you should incorporate into the model. Does age in bonds make more sense than rationally evaluating your ability, willingness and need to take risk, using whatever factors you think make sense based on your assessment of the evidence? Is an expected return estimate based on earnings yield a reasonable factor to include in your assessment of need to take risk?

I personally agree with John Bogle that age in bonds is a good starting point (and probably better than nothing for someone who will never take the time to study the topic), but I also agree with Larry that evaluating one's ability, need, and willingess to take risk is a more rational approach than simply following age-in-bonds without further evaulation. I'm also becoming quite convinced that smoothed earnings yield is a decent way to forecast expected 20-year returns for stocks (again, keeping in mind that this is only the mean of a wide dispersion of probable outcomes), and I do think that it's rational to include expected returns estimates in determining one's need to take risk.

I also respect that others may come to completely different conclusions based on their own research and study. My perception is that Larry has studied these topics more deeply and rationally than most of us, so I pay a lot of attention to what he says, but I don't always follow his advice (I reserve the right to be somewhat irrational in my investing decisions).

Kevin