Leesbro63 wrote: Sat Sep 27, 2025 6:40 am
gips wrote: Sat Sep 27, 2025 6:33 amIt seems to me nick and Ben got it wrong. they forgot or didn’t realize the authors of the trinity study didn’t take into account social security income. I claim next year at 70, my wife will claim a spousal benefit and it moves our spend from around 4% to 2.2%
Best,
Social Security or pension income has nothing at all with how long a pool of money will last given various inputs. Nick and Ben got it right. Your point, I think, is actually that for many, once you have Soc Sec income, you don’t need 4%. This is person dependent. If you have a $300,000 lifestyle, Soc Sec is a minor factor. If you and your spouse each earned a lot but live fairly frugally, Soc Sec might provide even more than you’re used to spending.
I agree the data set of past returns are what they are, and 4% for a 60/40 stock/bond portfolio can easily be restated using the same return set for a portfolio consisting of the present value of SS for X amount plus a 60/40 stock/bond portfolio of Y amount, basically equivalent to an ‘investment portfolio’ of (X+Y) of <60% stock and >40% bond, how much different depending on size of X v Y. And we also know that in that data set the SWR wasn’t highly sensitive to asset allocation over a pretty broad range. So no, leaving out Social Security isn’t a problem with 4% result.
There are serious problems though, including:
1. it’s actually a pretty small data set. There’s no statistical validity to using overlapping periods unless we’re confident the larger context, not just the noisy movement of financial markets over short periods, is staying the same. If the macro story can change, 1950-1980 and 1951-1981 are basically a rerun of the same thing which adds no real information. 100 yrs is just 3 actually independent trials of 30+ year retirement.
2. The macro story obviously can change. The Great Depression shook but didn’t topple the basic US system…but it could have. The US was on the winning side of WW’s and CW…but didn’t have to be. The opponents might have been stronger. Imagine a main rival for example that produced >4 times as much steel pa as the peak US production in the 1970’s and whose peacetime ship production made the US shipbuilding effort of WWII by tonnage look like a blip on the long term graph… which of course is not imaginary Seriously, the US might do well in rivalry with China, which might also remain peaceful or even a win-win for the two, we don’t know, but it’s important to returns and past rivalries aren’t very relevant.
3. Valuations of risk assets were lower over the great majority of that period than now. Our wish for good outcomes might tempt us to quibble whether valuation and expected return are connected but they are by basic financial fundamentals. See graph of stock earnings yield measure (expected real return) v real 10 yr yield 1900 to 2017. Expected return of a risk asset by definition does not ‘predict’ the subsequent realized return in a given period…hence the name *risk asset*. But the graph shows expected return often hung around at a level similar to the average realized return in that period, something like 6% real. More recently it hasn’t, and stock return generally higher than previous in recent times, as fundamentals would say, is partly a product of higher and higher valuation, which it’s not plausible to assume continuing as midpoint expectation (assuming valuation *deflation* as expected mid point is a different debate, but assuming now’s valuation as future expected valuation still means E[r] lower than at lower starting valuation). Extending that graph to now, stock measure would be slightly lower than 2017 (3.3) but bond real yield 1.8% v 0%, IOW the margin between the lines much narrower than average though not unprecedented. A study averaging everything over time homogenizes everything as if prices available to us in the market now make no difference at all, which doesn’t make sense.
There’s no 2025-2125 data. So it’s reasonable to look at how things went in the past as one possible signpost. But enormously too much weight is often put on such results IMO given the severe limitations relative to the actual question, which is about the future.
