Modeling Reduced Short-Term Returns Expectations

I am performing Economics-Based Planning. On the Economic Assumptions for the Nominal Rates of Return for both Regular Assets and Retirement Accounts, I'd like to model the expectation of reduced returns over the next 10 years.

To illustrate, let's say my expectations on nominal returns to stocks are the following:

- over next ten years ~4% nominal
- over a 20+ year horizon ~6% nominal

To model this I have considered three approaches:

1) Be conservative and use 4% for the Nominal Rates of Return on the Economic Assumptions tab
2) Use 4%, but then use the Future Rates of Return to specify 6% starting in 10 years
3) Use a weighted average return assuming 10 years at 4% and 20 years at 6% (4*0.33+6*0.67=5.3)

Approach #1 seems too conservative and yields the lowest Living Standard per Adult. Approach #3, which yields the highest Living Standard may be too optimistic since returns over the first 10 years will likely be lower than 5.3% which would result in carrying too high a savings level into years 11 and beyond. Is Approach #2 the best way to model my assumption?


dan royer's picture

Yes, it's hard to know about the future! I use 4% on mine all the way through age 100, but I guess I want my model to be conservative, and I invest like an 80-year old. I think #2 represents best what you imagine might happen. In number 3 you have all that early compounding that might skew things.

This doesn't address the original question, but, taking into account sequence-of-returns risk (considered in ESPlanner only in the Monte Carlo analysis), and what intuitively seems like an era of increasingly likely black swans, a 4% nominal return doesn't seem exactly conservative. Personally, I use 3% to age 100 (i.e. just matching inflation), and feel a bit nervous even about thinking that aggressively.

stddeviant's picture

Well, there is no doubt that assuming a nominal equity return of 4% (or even 3% as chertz suggests) is more conservative that assuming anything higher. However, in retirement planning with the objective of maximizing lifestyle subject to the constraints of not running out of money or taking on too much volatility, there is a risk of being too conservative on market returns as well as being too aggressive.

In a joint survivorship model, planning for 100-year lifespans is already somewhat conservative (but not that much given that the probability of one partner living to 100 is about 5%). Adding to that the assumption that corporations won't grow earnings or pay dividends is quite strident over a 30+ year planning horizon. Such a model would be less likely to run out of money but would constrain spending. If the couple's spending is such that such a constraint would not be a problem, then great, but many are not in that situation.

With PEs at their current lofty levels, sequence-of-returns risk does seem to be a major factor. However, can't this be at least partially mitigated by holding 2 to 3 years' spending in cash, providing the flexibility not to have to sell at market lows to fund current-year spending?

The more I think about my original question, the more I realize that what I was really thinking about is a strategy for updating ESPlanner each year as I go through retirement. The process I am thinking about is:

1) update ESPlanner with year-end account balances in both taxable and tax-favored accounts (reflecting actual spending and market results for the prior year)
2) for nominal market returns use the lower of current short-term expectations or long-term expectations
3) for inflation use short-term (e.g., <10 year) expectation

The idea is to update the planner each year so the plan's starting point always matches reality. The updated plan can be compared to prior years' plans to see how things are trending.

Is this a reasonable approach?

dan royer's picture

Absolutely you want to update the balances each year. Not me so much, but many earned 15% and 19% for example in 2017 and thus have to adjust up their retirement or even regular asset balances. Of course I hope they don't enter 17% as an average annual return.

I don't really view my conservative assumption of 4% as a risk. Even if I entered 2% and lost money to inflation, if that's my goal and I like that model and I'm happy with the consequent discretionary spending through age 100, then it's not risky since I have a really good chance of meeting my goal. If I invest for 2%--but my goal is to get 7%--then, I agree, it's risky in that sense because I probably won't get my goal.

But I take more of a goal driven approach and build a model that I think will match my goal rather than to have the goal of getting as much as possible from the market and adjust for risk.

But anyway, adjusting each year makes good sense, but I'm not sure about the rates of return to assume. I guess some of this depends on how you view the model you are building. I'm not modeling for the probable--die at 83--but for the possible--die at 100.

Stddeviant makes good points. It occurred to be belatedly that my comments stemmed in part from the fact that I will retire in a year or so, at 75+, and they probably would have been different in I were 20 years younger. Plus I have some immediate annuities and am over 15% in cash/CD’s/conservative bonds, which, as stddeviant points out, mitigates sequence of returns risk. So maybe I am being a bit over-cautious. And you can’t eliminate every risk – thermonuclear holocaust, asteroid strike, smallpox pandemic, etc. – so sometimes you just have to hope life will be reasonable, and enjoy the money you’ve saved, within reason, unless you’re okay with living in a closet on cat food.

I do feel a lot of people don’t grasp the concept of sequence-of-returns risk, so let me just remind anyone reading this that, if you’re in or approaching the retirement phase of your life, even if you know with absolute certainty (because, I don’t know, maybe God told you) the average rate of return of your investments over the next 30 years, it would be dangerous to plug that into ESPlanner, because if your bad years come early, as well they might, you will not achieve the results that a straight-line planner projects. In more general terms, there’s maybe (I can’t really do the math) only a 50% probability of achieving those results. That’s the whole idea supporting the utility of Monte Carlo analyses, though I confess I still find ESPlanner’s implementation of this kind of hard to follow. (My fault, not theirs, I'm sure.)

dan royer's picture

Good points, and for some further perspective, I note the following:

A portfolio of 20% large cap growth (S&P 500) and 80% Total Bond (AGG index) since 1987 yielded a compound annual growth rate of 7.28% since 1987. In this 40-year period (which is the furthest back this tool would go on these asset classes) there were two years when the return was negative: 1994 was -1.55% and in 2008 it was -3.62%. It's best year was 22.16% in 1995.

The standard deviation on this conservative 20/80 portfolio is 4.67%, compared to the inverse, a 80/20 ratio of the same asset classes where the variance is double, or almost three times as high. With the 80/20 the best year is 39% and the worst year is about -30%. With that kind of variance the sequence of returns is really important. I feel like that sequencing is not as worrisome when the worst year is -3.62 on a 40-year history, and the target in the model is not the "expected" 7.28% but rather 4%.