On Complexity, Diminishing Returns, and Personal Finance (Part 3)

The Final Frontier?

In the previous two sections (part 1 and part 2), we considered the relative impact of my financial decision over the past decade on my net worth. Most of the advice was pretty standard and the results were only mildly surprising. Today I’m going to wrap up this series with a few things that I’ve discovered in the past year or two that are a bit further afield from the advice I typically hear.

Will there be any more surprises? Let’s find out…

Getting Out There: Riskier Optimizations

Over my career, most of my success comes from the suggestions in parts 1 and 2, which match the approach recommended by most finance professors and many financial advisors. I believe that you can do quite well if you stop right here.

What follows are a few optimizations that I’m pursuing right now that are riskier and may not be for everyone. I follow them because I believe in the analysis behind each approach and my experience navigating a significant financial crash in 2008 leads me to believe that I can handle riskier approaches even under adversity.

Everything I Know Is Wrong (about Lifecycle Investing)

Ayres and Nalebuff’s book Lifecycle Investing is probably the most controversial resource that I recommend in this entire series. That said, it totally changed my view on how to mitigate risk and I believe that it will result in significantly better portfolio performance over my lifetime.

At a high level, the approach tries to mitigate a risk that’s existed implicitly in all the approaches I’ve shared so far. I even mentioned it directly once already, if you’re sharp-eyed.

The problem is this: if prices are high when you start investing and low when you retire, your results will be poor. Conversely, if they are low when you start and high when you retire, you’ll have amazing results. Said another way, young people should pray for a market crash (so they can buy at low prices) and older people should pray for a market boom (so they can sell at high prices).

Unfortunately, while market timing in general is extremely difficult, timing when you are born is impossible. The problem is actually worse than that… since the stock market tends to go up over time, the value of your investments gets more and more sensitive to timing risk as you get older.

The solution is conceptually simple, if harder to implement in practice:

Start investing whenever you can since market timing is extremely difficult and the stock market tends to go up over time; if you happen to start investing right after a market crash, that’s a bonus
Invest all the money you will ever invest (in your whole life) upfront
Ride it out until you have enough money to live on for the rest of your life, then start taking money (and risk) off the table, ideally whenever the stock market is high

(1) is what it is; you’re either lucky or your aren’t. (3) is challenging, but fairly doable and if you take the money off the table gradually, you reduce your market timing risk.

(2) is where it gets interesting. It seems impossible as stated (how do you invest everything upfront?), but it turns out that you can approximate it using leverage; that is, borrowing money to invest.

Ayres and Nalebuff suggest doing the following:

When you start investing, put 200% in risky assets (stocks) at 2× leverage, which means you invest 100% of your money in stocks, then borrow another 100% to also invest in stocks
Using a formula that incorporates your stock portfolio and your earning power from your job (which is kind of like a bond), figure out your lifetime risky investment amount
Keep putting money into stocks at 200% leverage until you hit your lifetime risky investment amount
Keep adding money to stay at your lifetime risky investment amount but with less and less leverage
Once you hit your lifetime risky assets amount at 1× leverage, start buying bonds
The formula from (2) also provides you with a target percentage of risky assets based on your relative risk aversion (RRA); once you hit this percentage, you can just keep adding your money to both your stocks and bonds to maintain the target percentage

This approach effectively allows you to invest more of your money upfront and smooth out your risk over time. As your portfolio gets bigger and bigger, your volatility goes down, which makes you less vulnerable to market declines later in life.

If you do this, you end up far less exposed to timing risk, but it comes with a few extra headaches:

Leverage is scary: if you think you’ll bail because your early stocks are twice as volatile at 2× leverage, then this approach is not for you. This is a big reason why Ayres, Nalebuff, and I don’t recommend that you go past 2× leverage even if it is mathematically optimal to do so
Leverage can be expensive/annoying to obtain: leverage in taxable accounts isn’t so hard to get (Interactive Brokers will give margin loans at ~2.5%, which is pretty cheap, and you can often deduct it from your taxes), and you can get leverage in IRAs using options or leveraged ETFs (just watch out for fees); however, for most 401(k)s you’re out of luck
Leverage can force you to sell (a “margin call”) at bad times: this is not a problem for leveraged ETFs, and if you use dynamic margin calculations and stay under 2× leverage at Interactive Brokers, the changes of a margin call are extremely low
RRA is subjective and lifetime earning power is unknown, so take the lifetime risky investment amount and percentage calculations with a grain of salt; they’re fairly stable though, so you can use them and adjust as necessary

I’ve been following this approach for around 2 years, and with the recent stock market rally, it’s made a huge difference in my returns. So much so that I’m almost to the step (3) where I reduce my stock exposure below 200%.

To implement, I sold all my bonds, began using leverage up to 2× as necessary in my Interactive Brokers account, and added 2× leverage monthly reset ETFs to my IRA (such as SPLX) for asset classes where there exist decent choices. Then, I multiply my value averaging target for each asset class by the computed leverage to understand how much I need to buy. 2× leveraged ETFs are double weighted.

So far so good with this approach, but I have yet to be tested by a significant stock selloff. Since, I’m confident from past experience that I won’t buckle and since my risks will go down over time, I feel fairly secure using this approach. I only wish I had started earlier.

Let’s see how lifecycle investing with up to 2× leverage affects returns over time:

	75% stocks, 25% bonds	Leveraged
Total saved	$142,993	$142,993
Total value	$728,854	$3,653,710

Here I’m comparing a standard 75% of stocks in the S&P 500 and 25% in short term bonds vs. the ramping-down leverage approach, which in this case starts at 200% stock exposure and ramps down to 64% by the end. The results are staggering. In our starting at 1975 back test, the leveraged approach ends up at just over $3.6M, which is nearly 5× the standard approach! Now, the time period that we’re looking at was especially good for stock investment, so maybe this result is just lucky?

Ayres and Nalebuff anticipate this argument in their book and run back tests over the last 138 years (and forward looking Monte Carlo simulations). In all cases, the leveraged approach wins. Sometimes the results are close and sometimes as in my backtest, the leveraged approach wins by a lot. It also performs in foreign markets with different return patterns (like Japan, where the stock market has performed poorly since the early 90s).

None of this should be surprising. The theory of diversification, applied to time, allows this approach to reduce risk or enhance returns for the same risk. It’s counter-intuitive, but I’m willing to bet that it works with my own money.

This is the one optimization I wish I knew about earlier!

Time Series Momentum During Recessions: The Final Optimization (for now)

My final optimization comes from a super-interesting blog called Philosophical Economics rather than a book. I don’t have much background on the author, Jesse Livermore, but some of the ideas are fascinating. This optimization comes from a series of posts about time series momentum and recession timing.

Time series momentum is a sub factor of momentum. If you recall, stocks that have done well recently outperform stocks that have done poorly recently. If you compare a given stock to other stocks and buy the high performers over a timeframe, that’s called relative momentum. Some of the ETFs suggested above capture that. Time series momentum on the other hand compares a stock’s performance now to its own past performance. Said simply, we buy stocks that have gone up in the recent past and sell stocks that have gone down.

Research shows that if you pick the right definition of “recent past” this strategy reduces risk (especially maximum drawdown, which is a measure of how big a crash a given stock experiences in some time period) and probably also enhances returns. Big drawdowns are really tough psychologically and can cause you to abandon your strategy, so avoiding them is key.

Unfortunately, time series momentum is also just a mechanical way to time the market, which I already said was hard. If you trade in and out too frequently there are so-called whipsaw losses that eat up your earnings. Also, if you’re out of the market for too long you miss out on the upside that comes from stocks going up in general.

The author argues then, that a good time series momentum approach:

Is biased towards being in the market
Doesn’t trade that often
Protects you from “significant crashes”

They suggest that most significant crashes that don’t recover quickly happen during recessions and spend some time coming up with a reasonable recession indicator. It turns out that by using the moving average of the unemployment rate, one can predict recessions with fair accuracy and precision.

By combining the recession timer and time series momentum, we can achieve the goals above by being 100% in the market when a recession isn’t occurring and by relying on time series momentum when a recession is occuring. This approach is:

Biased towards being in the market because recessions are infrequent
Doesn’t trade that often because it only trades during a recession
Protects you from a crashes that happens during a recession

So… for the cost of complexity and missing some crashes, we get most of the value. Or so the theory goes. The backtest is compelling, increasing annualized performance by over the past 87 years by 1.5% annualized and reducing max drawdown from 79% to 51%, but this approach is not guaranteed to work. The fact that historically it has only kept the investor out of the market 15% of the time is a big point in its favor, however.

I have incorporated this into my portfolio by creating time series momentum and recession indicators inspired by the blog and some academic articles, then combining them with value averaging and tax considerations as follows:

Every month do the following, check if the indicator suggests a recession, then for each asset in my portfolio check its time series momentum and whether value averaging tells me to buy or sell:

If no recession indicated
- In a taxable account
  - Time series momentum says “buy”
    - Value averaging says “buy”: buy up to the VA level
    - Value averaging says “sell”: hold
  - Time series momentum says “sell”
    - Value averaging says “buy”: tax loss harvest, maintain value
    - Value averaging says “sell”: tax loss harvest down to VA level
- In a tax free account
  - Time series momentum says “buy”
    - Value averaging says “buy”: buy up to the VA level
    - Value averaging says “sell”: hold
  - Time series momentum says “sell”
    - Value averaging says “buy”: hold
    - Value averaging says “sell”: sell to the VA level
If recession indicated
- In a taxable account
  - Time series momentum says “buy”
    - Value averaging says “buy”: buy up to the VA level
    - Value averaging says “sell”: hold
  - Time series momentum says “sell”
    - Value averaging says “buy”: hold
    - Value averaging says “sell”: tax loss harvest to VA level
- In a tax free account
  - Time series momentum says “buy”
    - Value averaging says “buy”: buy up to the VA level
    - Value averaging says “sell”: hold
  - Time series momentum says “sell”
    - Value averaging says “buy”: sell all
    - Value averaging says “sell”: sell all

All in all, it’s kind of complicated, but it works for me. It’s good because:

It reduced my trading costs in comparison to what I was doing before; the hold state is the most common, which results in less trading
It is tax efficient since I never sell in my taxable accounts except to tax loss harvest; I may revisit this decision later
It should cut off downside in recessions, but it hasn’t been tested since we haven’t had a recession in nearly a decade

It may not be easy, but it gives me something to do once a month (or any time there is a major downward market move) that could help and is unlikely to be harmful.

Unfortunately, I’m not able to make an apples-to-apples simulation comparison with the rest of my simulations because this approach requires a more frequent check-in period than the yearly period I used in the other simulations.

Instead of running a full month-by-month simulation, I simply took a look at what the effect would be on our investor if their maximum drawdown since 1975 was reduced by 35%, matching the drawdown improvement from the articles.

	75% stocks, 25% bonds	Capped drawdown
Total saved	$142,993	$142,993
Total value	$728,854	$921,521

This is somewhat of a contrived scenario, but it shows that if this approach does have the drawdown protection it has demonstrated in the past, the rewards are substantial, taking our total portfolio value up to $922k from $729k with less risk. It’s no leveraged lifecycle investing, but the downside protection and enhanced returns seem worth the risk of being out of the marked around 15% of the time.

Conclusion: Was It Easily Done?

It definitely was not easily researched, but perhaps my years of research can make things easier for you. Ignoring all the setup costs, I spend approximately 2 hours per month maintaining my system and doing all the trading it entails. Most of that time is the result of maintaining my value averaging and time series momentum spreadsheets.

Let’s review how much benefit I’ve estimated I’m getting from each optimization.

Starting at 10% (rather than 5%): 50%
Ramping up from 10%: 19%
Lifecycle leverage: 17%
Recession momentum: 14%
Basic diversification: 0% to maybe a bit more, but reduces risk
Fancy diversification: 0% to maybe a bit more, but reduces risk
Value averaging: 0% or even negative, but can reduce risk and help you plan

Now, let’s consider time costs per year, based on my experience doing these:

Starting at 10% – 0 hours
Ramping up from 10% – 1 hour
Basic diversification – 0 hours
Value averaging – 12 hours
Fancy diversification – 6 hours
Lifecycle leverage – 1 hour
Recession momentum – 12 hours

Finally, let’s get a rough ROI ranking over 40 years with our top performing portfolio, assuming it takes 2 hours to set up your initial investments:

Starting at 10% – $913k / hour
Ramping up from 10% – $17k / hour
Lifecycle leverage – $16k / hour
Recession momentum – $1k / hour
Basic diversification – $0 / hour, but reduces risk for free since it takes no time
Fancy diversification – $0 / hour in return, only worth doing as a hobby
Value averaging – $0 / hour, probably not worth doing at all

Seems worth doing, at least a few of them, eh? There are vanishingly few investments of time that net you 5 figures per hour and I defy you to find even one that will net you 6!

And that’s how my personal finances were (mumble, mumble) easily done.

If you have any questions or try out any of these techniques, I want to hear from you. Let’s get better together.