Uncorrelated wrote: ↑
Sat Jun 27, 2020 7:01 am
Beehave wrote: ↑
Fri Jun 26, 2020 10:16 pm
The articles cited do not impress.
They provide allegedly precise predictive models. Then, when they attempt to use empirical data to serve as evidence that the real world behaves in a way that resembles their model, they list all the simplifications they need to make. The simplifications are telling.
1. They model stocks and cash. No bonds. Then they say you need lots of stock. The simplification is self-serving.
They do this for simplicity. In other papers by Irlam, he investigates cash/stock/bond/annuity mixtures. You can read the rest of his papers here: https://www.aacalc.com/about
2. They model annual rebalancing. They purport that frequent rebalancing performs better. And that's what typical life strategy funds do, they rebalance daily. The typical holder of a stock index and bond index and money market fund rebalances far less frequently. Again, self-serving.
Irlam takes an empirical analysis. Many other papers take a mathematical analysis with continuous rebalancing, but find the same basic conclusions. Irlam actually claims annual rebalancing performs better than frequent rebalancing in real-world conditions, although the conclusions are obviously dependent on how likely you think it is that historical mean-reverting behavior persists in the future: https://www.aacalc.com/docs/when_to_rebalance
3. They model simple, discrete lottery choices, not nested sequences (Von Neumann-Morgensstern). But the receipt of a windfall triggers intensely complex nested decisions about how much of the windfall to allocate to each of many competing needs and wants. Childrens' education, vs. retirement savings, vs. pay down mortgage, vs. home improvement, vs. new car vs. pay down student loan, vs. needed vacation vs. kids' summer camp vs. vs. vs. And there are complex interrationships and time dependencies. The DCA choice is intimately affected by these nested factors in many ways, not the least of which are (a) to compromise between acting and keeping options open.
This does not affect the basic conclusion that the optimal allocation is a lump sum investment. For example, look at the graph here: https://www.aacalc.com/docs/variable_wi ... tock_heavy
. Imagine that you are at age 80 with $500k in capital which gives you a balanced asset allocation. Suppose that you receive a windfall of $400k. Now the optimal allocation is to switch immediately to the optimal allocation according to that chart.
If you disagree that this chart is relevant for you, you should adjust model parameters until they fit your specific scenario. However, the basic conclusion that there is one optimal allocation for each unique combination of age and net worth does not change. This basic conclusion indicates that DCA is never optimal.
4. Claim they are modeling Social Security as an income stream only, and then persist in attributing a decreasing net present value to it as we age as a basis for calculating comparative stock allocations (Irlan)
That is not an explicit assumption, that is the result of the optimizer.
5. They start the modeling of target funds as if age 50 as a begin date to compare with index funds makes sense. A fair comparison would start at age 35 (or earlier) or 40, to show a reasonable glide path. The one modeled by Forsyth is way truncated and self-serving. It looks like the real world data plugged in for an earlier target date fund start would show his analytics to be totally unsupported by real-world behavior. He barely has a case with age 50 (way too high for a start date), annual rebalancing (way too long for a target fund) and cash and stock with no bonds.
I recommend reading https://papers.ssrn.com/sol3/papers.cfm ... id=1149340
if you're interested in optimal asset allocation for the savings phase. I only refer to the Forsyth paper because it contains a direct proof that DCA is suboptimal.
The oversimplifications are not only largely very self-serving, they also point out the complexity of formally or informally modeling complex financial decisions - - including whether to DCA or lump sum, and whether or not new money is the same as an equal amount of old money. The complexity itself can explain many DCA choices - - I want to start but I want to take my time because there are lots of moving parts and lots of unknowns and many complex interrationships. So move, but take it slow.
tl;dr life is complex, so use this asset allocation that has been proven to be suboptimal at least a dozen of times in this thread.... No, just no.
Suboptimal in one context can be optimal in another. Racing slicks are great, unless you're concerned about rain. Contexts matter. That's the concern with simplifications that enable complete analyses of otherwise intractably complex systems. These logically complete analyses are of interest. They contribute to rational understanding. But, because of the simplifications that make their internal logic complete, they become less complete in larger contexts. Life requires all-weather tires, not racing slicks for most people.
The Von Neumann-Morgenstern system is logically complete at the cost of ruling out consideration of nested probabilities. But life is filled with nested probability issues. The investment I make in my child's education affects my retirement in many, interrelated ways, and I have more than one child, and they have different interests and capabilities. If I receive a windfall, I need to prioritize. Nested probabilities make this difficult. So it can be the rational choice is to compromise between optimization and flexibility by things like DCA, even if one believes LS is more effective.
Regarding Forsyth, I stand by my previous observations. He has stacked his modeling cases to "prove" his point, but more realistic model parameters would disprove it.
Regarding the Ales and Nalebuff article intended to correct Forsyth, it seems to agree that target date and fixed allocation up to 70/30 are pretty much equivalent over time, and that the really best policy would be to diversify over your lifespan's investment and intellectual capital assets by leveraging stock holdings early in career and tamping down on that as your wages and investments increase but your intellectual value grows shorter in duration and approaches a period of declining value.
So for Ales and Nalebuff, a 100% stock allocation is not good enough, they suggest amplification by taking out loans, buying options, or taking on margin early in one's career. They say:
"Of course, borrowing on margin creates a risk that the savings will be entirely lost.
That risk is related to the extent of leverage. If portfolios were leveraged 20 to 1, as we
do with real estate, this risk would be significant. We propose a maximum leverage of
2:1. It is worth emphasizing that we are only proposing this amount of leverage at an
early stage of life. Thus, investors only face the risk of wiping out their current
investments when they are still young and will have a chance to rebuild. Present savings
might be extinguished, but the present value of future savings will never be. Our
simulations account for this possibility and even so, we find that the minimum return
under the strategies with initially leveraged positions would be substantially higher
compared to the minimum under traditional investment strategies."
(From: LIFE-CYCLE INVESTING AND LEVERAGE:
BUYING STOCK ON MARGIN CAN REDUCE RETIREMENT RISK)
I'll simply say - they're probably right about what is likely to maximize investment results over time. But if it's me, I do not want to run the risk that at the very time the market plummets, my employment gets interrupted and it ends up that my optimization strategy has in a larger context put me at risk of being in a deep, deep hole of nested disoptimization into which I've fallen.