March 21, 2022
A quick recap, the appeal of using leverage in retirement is that we would borrow against the portfolio instead of liquidating assets. Nice! That might help with Sequence Risk if we avoid liquidating assets at temporarily depressed prices. There could also be a tax advantage in that we keep deferring the realization of taxable capital gains, potentially until we bequeath our assets to our daughter who can then use the “step-up basis” for complete forgiveness of all of our accumulated capital gains. That’s the famous “buy, borrow, die” approach popular with high-net-worth folks.
The gist of the post last year: Not so fast! Leverage could potentially even exacerbate Sequence Risk if you are unlucky and retire right before a bad market event that’s deep enough (like the Great Depression) or long enough (like the 1965-1982 stagflation episode) to compromise the portfolio so badly that the margin loan becomes unsustainable relative to the underwater portfolio.
One solution proposed by several readers: instead of always borrowing against the portfolio, maybe we should carefully time when we use leverage. For example, borrow only when the stock market is down “far enough” and use withdrawals from the portfolio otherwise. And if the market is doing well again, potentially pay back the loan again! Sounds like a reasonable and intuitive plan. But I want to put that to the test with some real simulations. Let’s take a look at the details…
A quick recap of the 1929 and 1965 cohorts
In the post last year, I modeled the margin loan with a fixed real interest rate. I realize that a slightly different assumption might be more realistic, namely using a fixed spread over a short-term government reference rate. That’s because if you want to borrow on margin, your broker will likely not quote you a rate as “x% over CPI inflation” but, more likely “x% over the Federal Funds Rate (FFR)” or some other short-term rate like the 3-month T-Bill rate or LIBOR. What would be a reasonable assumption for the margin loan spread?
- A Home Equity Line of Credit (HELOC) often has an interest rate tied to the Prime Rate, which is itself about 3 percentage points above the (overnight) Federal Funds Rate (FFR). I remember back in the good-ol’ days you could get HELOCs for Prime minus 0.75%, even 1.00%. But I think today’s rates are closer to Prime +/-0% or maybe minus 0.25% if you have good credit and shop around a bit. So, if you get a HELOC with a rate equal to prime minus 0.25%, you’ll pay around 2.75% above the FFR.
- M1 Finance offers rates between 2% and 3.5%. That’s a wide range. It looks like the upper edge is actually inferior to the average HELOC. I couldn’t ascertain what benchmark rate they use, but I’d suspect, the range of interest rates will likely move up in line and 1-for-1 with the Federal Reserve policy rate.
- Interactive Brokers will lend to you at a rate of 1.50% above the FFR for smaller amounts, and 1.00% for medium-sized loans ($100,000-$1,000,000). Loans up to $50,000,000 go for 0.75% above the policy rate. And 0.50% for loans $50m+, if you’re loaded enough. This is all for the IBKR PRO account. The IBKR LITE account charges significantly more.
- Borrowing through a box spread trade as I discussed in my post in December 2021, you might get rates as low as 0.3-0.5% above the T-Bill rates. That’s likely the lowest rate you will encounter anywhere. (side note: most likely you’d choose a longer-term loan, maybe 1-2 years. Potentially as long as 5 years. The 0.3-0.5% spread refers to the spread above the T-Bill or Treasury bond of the same maturity, not necessarily the spread above the realized Fed Funds Rate. But over shorter horizons, the T-Bill rate is a pretty accurate prediction for the average FFR over the same time span)
It turns out that the choice of the loan rate will make quite a difference in the margin loan calculations. For example, last time I used real rates of 0%, 1.5%, and 3% with the middle value of 1.5% above CPI as my “preferred” value. I like to recalculate the scenario for the November 1965 retirement cohort with a $1,000,000 initial portfolio withdrawing $30,000 annually and supplementing the withdrawals with a margin loan worth $10,000 annually. For the CPI+x% loan rate, I use the middle value of 1.5% and contrast that with three different FFR+x% loan scenarios: a 0.50% spread (best possible case: box spread trade), 1.25% spread (margin loan with Interactive Brokers), and 2.75% (HELOC, and other not so attractive brokers). The portfolio value net of the 3% withdrawals is the same in all four scenarios but the loan balances are quite different, see the chart below. It turns out that the FFR+x% loan interest looks far worse in the 1965-1995 time span than I had previously assumed. That’s bad news because it means that the margin issues that hamper our “buy, borrow, die” efforts would have been even more constraining than I previously assumed. Inflation-adjusted short-term rates were much higher during the crazy 70s and early-80s than I previously assumed!
That said, what hurt us in 1965 would have helped the September 1929 retirement cohort! If I redo the exercise with the different margin rates we get the following picture, with far lower loan-to-portfolio ratios when using the FFR+x% rule. That’s obviously because the realized short-term real rates were quite low during the Great Depression.
Simulations for the 1965 cohort
Let’s get to work and run some simulations. I want to start with the November 1965 cohort because, as the results above show, that seems to be the most constrained cohort and potentially the one with the most to gain from timing the margin loan draws.
Let’s start with the base case simulation, where we use a 4% annualized withdrawal rate, and a quarter of the monthly budget is financed through the margin loan, every month. I display the simulation results, for this and all the subsequent scenarios, in this simple chart plotting the real CPI-adjusted time series of the portfolio, loan, and net worth balances (left axis) and the monthly margin loan draws (right axis). The monthly budget is $3,333.33 (4% annualized), $2,500 of which comes from withdrawals and the remaining $833.33 from the loan. There is no timing (yet). We confirm again that this would have been at the very least a very unpleasant retirement experience, with the net worth depleted to $133,000 about 17 years into your retirement. The maximum loan-to-portfolio balance would have been 72.3%. That means for every $100 of portfolio value, only $27.70 was your net worth and $72.30 came from the loan. That’s almost a 4x leverage, which would have likely busted your portfolio and caused a forced liquidation by the brokerage, considering that I simulated this only at a monthly frequency, and daily and intra-day fluctuations would have easily pushed you beyond your margin constraints.
Lower WR, less leverage
Since a 4% withdrawal rate and 25% of consumption financed through the loan wouldn’t have worked in 1965, let’s see how much we have to reduce our budget to make this work. Let’s assume that this retiree wants to keep a $250,000 safety cushion after 30 years. Without any leverage and using a 75/25 portfolio, I compute the baseline safe withdrawal rate as 3.58%.
How about with leverage? I use the built-in Excel Solver function to maximize the retirement budget subject to the $250,000 final net worth target and the 50% upper limit on the loan/portfolio ratio (=2x leverage), by changing the withdrawal rate and the “Borrow%” value, i.e., the share of retirement budget funded by the margin loan. This is still without timing the margin loan, i.e., we withdraw the same (real CPI-adjusted) amount every month. The results are pretty disappointing. Even with an extremely inexpensive margin loan rate, we can only finance about 10.76% of the retirement budget (a little over $300/month). And lift the safe withdrawal rate to 3.78. Better than 3.58%, but still no panacea for sequence risk either!
Timing the margin loan
Let’s look at the following timing mechanism: we fund 100% of our retirement from withdrawals unless the S&P 500 drops to a certain percentage below its most recent all-time high. Let’s start with a 20% drawdown target. We can indeed raise the withdrawal rate and margin loan percentage to 3.84% and 26.48%, respectively. Still a bit shy of the 4% Rule but not bad. Notice how the margin loan draws, about $850/month, occur briefly during the 1970 recession and bear market, the 1973-1982 stagflation era, and again briefly during 1987/88 stock market hiccups. Out of the 360 months of retirement, we’d use the margin loan “only” 125 times. Also notice that we easily spot the time when the margin loan constraint binds: When the blue line touches the green line we’d have loan+net worth=portfolio, and thus loan=0.5*portfolio, i.e., exactly 2x leverage.
Pay back the loan when we’re back at a fresh all-time high!
If we look at the time series chart of the previous simulation, we notice that the 50% margin ratio hits you relatively late in retirement (month 349), during the 1994/95 Peso Crisis and the resulting stock market weakness. During the ugly 1982 stock market bottom, we had a loan to portfolio ratio of “only” about 32%. One way we might be able to increase both the margin loan use and the safe withdrawal rate would be to pay back the loan during the 80s/90s stock market rally to relax the 50% margin constraint in 1994.
So, let’s assume that if we reach a fresh all-time high in the S&P 500 Total Return Index, we’ll start paying back the margin loan. I assume that we simply double the withdrawals from the stock/bond portfolio and use the excess to pay down the margin loan, i.e. set the margin loan draw to -100% of the monthly retirement budget. This method allows us to raise the WR to 3.91% and the margin loan portion of the withdrawals to 41.08%. We still draw the margin loan during 125 months but we also pay back the loan for a few months in 1972, and then again starting in 1985, for a total of 47 months. Notice that the maximum margin constraint is now binding in 1982 again at the bottom of that recession when the margin loan touches the net worth line.
Complete Simulation Results
I also played around with more restrictive drawdown constraints, i.e., use the margin loan only when the stock market is down by 25%, 30%, and 35%, respectively. I don’t prepare separate charts, but report the complete simulation results in the table below. Indeed, we can push the safe withdrawal rates even a little bit higher if we move to a 30% cutoff, but at 35% there is a huge deterioration again. Having to wait until the market drops by 35% appears to be too restrictive, at least during the 1970s. But it seems that anything between 20 and 30% seems like a neat option. The no-leverage withdrawal rate would have been only 3.58% and we can improve that by about 33-36bps, or about 9-10%. Not bad at all!
Also, to explain again how the numbers were created: with the exception of the first column, where I just calibrate the WR and the Borrow% at 4% and 25% (and we clearly get way too much leverage), I use the Excel Solver function to maximize the retirement budget subject to the $250,000 final net worth and the 50% upper limit on the loan/portfolio ratio (=2x leverage), by changing the withdrawal rate and the “Borrow%” value, i.e., the share of retirement budget funded by the margin loan.
I also included a reader suggestion in the last column on the right, where the loan draw vs. payback is timed not by the S&P 500 drawdown, but by what I call a “Portfolio on Track” indicator. I check if the projected portfolio value net of withdrawals, assuming a 4% real return rate. And then subtract the current margin loan balances with interest. If you’re below the final bequest target of $250,000, draw down the loan, if you’re on track then withdraw the retirement from the portfolio and – if applicable – pay back the loan as well. I haven’t played around very much with this rule, and maybe it can be optimized some more, but the equity index drawdown rule appears to be superior for the 1965 cohort.
A few caveats
As always, there are a few warnings and caveats to keep in mind here:
- There is no way an actual retiree in 1965 would have been able to borrow at a rate of FFR+0.50%. We’d have to view my calculations here as a “thought experiment” of whether with today’s financial innovations – index funds, low expense ratios, low margin rates – we can use leverage and still survive a repeat of the return patterns of a historical worst-case scenario!
- If your margin interest is higher, say FFR+1.25% (IB margin loan) or FFR+2.75% (HELOC), some of the appeal of the margin strategy evaporates. The withdrawal rates for the 25% drawdown timing mechanism go from 3.92% to 3.87% and 3.75% when raising the loan spread rate from 0.5% to 1.25% and then to 2.75%, respectively.
- If you indeed plan to use the box spread trades, keep in mind that those loans tend to be “lumpy”. I initiated three box spread loans so far, one with a 200-point spread and two with a 1,000-point spread, worth $20,000 and $100,000, respectively. Rather than doing a box trade every month, most people would likely to draw down the IB margin loan at a slightly higher interest rate until they reach a large enough loan balance and then initiate a new box spread loan, in the low-to-mid 5-figures.
- Also, keep in mind that most actual retirees face an even tighter margin constraint than what I model here because only traditional brokerage accounts allow margin loans. You can’t do this in a retirement account!
The 1929 Cohort
Just for completeness, let’s redo the same exercise for the September 1929 cohort, right at the stock market peak before the Great Depression. The unleveraged safe withdrawal rate was 3.61% in this cohort. Quite intriguingly, the margin loan would have provided a tremendous improvement in the safe withdrawal rates. The base case with the 4% WR and 25% borrow share would have stayed way below the 50% margin constraint. So, even without timing the margin, you could have gone to a 4.39% withdrawal rate and financed 31.86% of your expenses with the loan. And it gets better when you time the loan draw and the repayment. With a 35% drawdown criterion, a 4.93% withdrawal rate would have been feasible.
There is no panacea against Sequence Risk. I didn’t expect this to be one either. But I was positively surprised that a small dose of leverage can indeed smooth out some of the headaches of even the historical worst-case retirement scenarios.
A negative surprise: even if you use the margin loan very occasionally, i.e., only after a stock market drop of 25% or more, it would still be too risky to fund all of your retirement expenses with a margin loan. I would use the loan only very sparingly, to replace only about 50% of the retirement budget, even after a 25% drop! Applying this rule to the more recent retirement cohorts, you wouldn’t have utilized the margin loan even during the 2020 bear market because the month-end drawdown in March 2020 would have been just under 20%. Even though the February 19 to March 23, 2020 drawdown was almost 34%! Under the leverage timing rules studied here, you would use leverage only during the 2002-2003 or 2007-2009-style market meltdowns, not the garden-variety volatility we’ve seen post-2009.
I am also surprised about the margin loan working so well in the 1929 case study. I would have expected the reverse, i.e., I thought that the margin loan would have worked better during the 1970s/80s stagflation than during the deflationary 1930s. But of course, the inflation rate alone doesn’t matter so much. Real interest rates were painfully high during the 1970s but relatively benign in the 1930s.
But make no mistake, leverage doesn’t work as well as some hand-waving “experts” on the interweb want it to appear. We can’t “just use the margin loan” and all worries about sequence risk go away. The historical worst-case scenario retirement starting in 1965 needed to tread much more carefully. Don’t even think about using the margin loan to fund your entire retirement. That’s only for the ultra-rich with a nine-figure+ net worth and a 1% withdrawal rate. But, again: using leverage sparingly, say, 40-50% of the retirement budget is funded through a loan and only when the stock market is down significantly, we can make a noticeable dent in the effect of Sequence Risk.
Thanks for stopping by today! Please leave your comments and suggestions below! Also, make sure you check out the other parts of the series, see here for a guide to the different parts so far!
PS: For the math wonks, I posted the Excel Sheet here. This is not a sheet that would work well with Google Sheets due to Google’s inferior solver function. So, it’s an “*.xlsx” file. Since I never intended to share it, it might take a bit of a learning curve to figure things out. It doesn’t have enough documentation to serve as an easy plug-and-play toolbox.
Title picture credit: pixabay.com