Less than two years away from early retirement, we wonder how much cash (if any?) we’d like to hold in a money market account. As many of you might have heard, we currently run a very tight ship with our cash management. We have no emergency fund – our entire portfolio is our emergency fund! But that’s easy to do while the paychecks are still rolling in and we maintain a 60% savings rate. Early retirement will be very different. How would we handle the cash withdrawals in retirement? How do we react to market fluctuations?
In the FIRE community, I often read that the solution (maybe even the panacea) for an equity bear market is to keep a certain percentage of the portfolio in cash (money market account) to sustain cash flows through a bear market. And we should point out that we are not the only ones thinking about this, as evidenced by recent popular posts on the PIE blog and on Retirement Manifesto (also check out the really cool infographic) dealing with this subject. Two to three years worth of expenses (presumably 5-10% of the portfolio) seem to be the numbers floating around (examples: 5% cash allocation for the PIE blog, The Retirement Manifesto recommends 2-3 years, ThinkSaveRetire uses 3 years), obviously calibrated to roughly correspond to the length of the average bear market.
How much of a difference does a cash cushion really make?
How big is the opportunity cost of holding cash when there isn’t a bear market?
Let’s take one step back and confirm really quickly the great risk we all face in the FIRE community: Retiring at the wrong time, right before the equity market takes a nose dive.
Let’s look at how a $1,000,000 portfolio would have performed at the following four retirement dates: 12/31/1996, 12/31/1997, 12/31/1998, and 12/31/1999. In each case, we assume a 3.5% withdrawal rate, 100% equities held in an S&P500 index fund (returns include dividends!), and monthly CPI-adjustment of withdrawals. Your real (CPI-adjusted) final portfolio value on June 30, 2016, would have been very different depending on the retirement date. Specifically, over the four retirement dates, only 36 months apart, you experience the entire spectrum of final real portfolio values:
- from significantly growing your portfolio despite the two bear markets (1996): $1.506 million!
- to roughly maintaining your purchasing power (1997): $956k
- to seriously threatening your long-term viability (1998): $593k
- and pretty much dooming your long-term sustainability for sure (1999): $402k
A little side note: I sometimes read in the blogosphere that the 4% rule did just fine during 2000/2001 and 2008/9. Not true! Starting retirement on 12/31/1999 and withdrawing 4% (rather than 3.5%) would leave you only $263k in 2016. You will run out of money for sure, probably before the year 2025. The 4% rule will most definitely fail the 30-year retirement horizon. Forget about a 50-year retirement!
How to implement the Cash Cushion
This cash cushion approach sounds really appealing in theory but implementing it in practice, in my view, opens up a bit of a can of worms. Here’s why: In my opinion, there are two different ways of implementing this cash cushion and both have the potential to create some major headaches:
Method 1: we always keep exactly 2 years worth of expenses in a money market account. That doesn’t help us at all if there is an equity market drawdown. We may take money out of the MM account every month to fund our living expenses, but at that same time, we have to replenish that MM account out of the equity holdings. It’s “robbing Peter to pay Paul.” This method will essentially create a constant drag (lower average returns!) in the portfolio while still forcing us to do what we tried to avoid, namely, liquidating equity holdings during a drawdown. So, this can’t be what folks in the FIRE community mean when they advocate a cash cushion.
Conclusion so far: the cash cushion in retirement works only if we actively time when to draw down the cash cushion and when to replenish it again. The disadvantage is that we would have to do what we absolutely dread: market timing! If you are registered on the Bogleheads forum, you probably want to keep your market timing ambitions to yourself, or they might send the Spanish Inquisition after you. But for the sake of academic curiosity, here’s my proposal for actively timing the cash portion of the portfolio:
Method 2: a varying cash cushion in a money market account. And when I said that this cash cushion timing methodology is a can of worms, I wasn’t kidding. To do this in a sensible, systematic and replicable way takes a lot of assumptions! Common to all of our simulations are the following assumptions and parameters:
- Initial net worth $1,000,000
- The initial withdrawal rate is 3.5% p.a., so we withdraw 3.5%/12 on December 31, to fund the January consumption, then adjust the amount by CPI inflation each month.
- We keep a cash cushion of x months worth of expenses in a money market account yielding the 3-month T-Bill interest plus another 0.25% annualized. Savvy shoppers that we are, we would have found an interest rate slightly better than the short-term T-Bill! The rest is invested in a stock portfolio (S&P500 index fund)
- Every month we use the dividend income (=difference between equity portfolio total return and price return) to fund our expenses. The portion of expenses not covered by dividends is first taken out of the money market account.
- During normal times, we replenish the money market fund up to the cash cushion target by selling a portion of the equity fund. This is identical to method 1 above.
- If we face a drawdown of at least y% from the most recent peak of the S&P500 index we will refrain from selling the equity fund and rather draw down the money market account that month. Of course, if we ever exhaust the money market fund we have no choice but to sell equities to make ends meet.
- One additional twist: Once the cash cushion is below the target and the equity index reaches the normal level again, i.e., comes within y% of the previous peak level, do we replenish the entire cash cushion all at once? To avoid the sale of one big chunk of equity holdings all at once, we set parameter z equal to the maximum number of months worth of consumption we transfer out of equities and into the money market account. Notice that if we set z=x we do replenish the entire money market shortfall all at once. If we wanted to do this in a more gradual fashion we will set this parameter to maybe 3 to 6 months. We see that this parameter can potentially make a huge difference because replenishing the money market account usually falls into the strong rebound phase after the drawdown. It’s potentially more advantageous to drag out this replenishment phase and let the equity returns run for a little bit longer before filling up the MM account again.
- Simulation details:
- Time span: varying start dates: 12/31/1996, 12/31/1997, 12/31/1998, and 12/31/1999.
- End date: June 30, 2016
- We assume that all realized capital gains are long-term and all dividends are qualified. Since we plan to stay in the lower two tax brackets in retirement and pay zero state taxes we assume that all equity income is tax-free.
- Since the interest income will likely stay below the annual standard deduction plus exemptions we assume that even the interest income is tax-free. Since the average cash yield over this period was quite low that wouldn’t make a huge difference, though.
Let’s see how different withdrawal rules would have performed over those 20 years. The different parameter settings we pick are:
- Equity only, no cash cushion: x=0. We are at the mercy of the S&P500 every month and have to withdraw all cash flow needs not currently covered by the dividend yield.
- Method 1 (no market timing): 24 months worth of cash cushion (x=24) but constantly and fully rebalance back to the target: y=1.00, z=24.
- With market timing: 24 months worth of cash cushion, a 5% drawdown cutoff, and full rebalancing when in a “normal” period (x=z=24, y=0.05).
- Same as 3, but replenish only up to 4 months worth of expenses into the MM account: x=24, z=4, y=0.05.
- Same as 4, but set a threshold of y=0.10.
- Same as 4, but keep only 12 months of expenditures in the MM account.
- Same as 4, but keep x=36 months worth of expenditures in the MM account. In light of the larger cash cushion we also increase the replenish rate to 6 months: z=6.
I run each of the seven parameterizations with the four starting dates, for a total of 28 simulations.
Results:
- The different cash cushion models may or may not help compared to the 100% equity portfolio. The only consistent result is that Model 2 (constant cash cushion, no market timing) is the worst performer.
- Model 4 seems to dominate Model 3 (about the same for 1996 but consistently better in 1997-1999). So, replenishing the cash cushion only gradually helps.
- Model 4 pretty consistently beats Model 5 (10% draw-down cutoff). It helps to dig into the cash cushion when the market is down only 5%, rather than 10%.
- Starting retirement in 1996, all cash cushion models (2-7) would have created a bit of a drag on performance. Not the end of the world because all portfolios were up by a lot.
- In the 1997 scenario, Models 1 and 3-7 are shockingly similar; all within a few thousand dollars of each other.
- The cash cushion models 3-7 would have significantly helped compared to the all-equity portfolio when starting retirement in 1998 and especially 1999.
- But, of course, the cash cushion is no panacea. The advantage over the all-equity portfolio will not change the simple facts:
- Starting in 1998 you’re left with only around 59-64% of your initial wealth, which is a big drawdown, almost too much to recover from, considering that equities are quite richly valued today. I don’t see a quick equity surge of 50-60% returns anytime soon to bring the portfolio value back to $1,000,000!
- Starting in 1999 (right before the 2000 bubble) whether with cash cushion or without, the portfolio is doomed. Also, that’s despite the 3.5% withdrawal rate rather than the often quoted 4% rule. If you had started retirement in 1999 with a 100% equity portfolio, with or without the cash cushion, it’s not so much if you will run out of money, but when.
More results
Let’s look at the time series of the cash cushion size (measured in the number of months of consumption). I do that for models 3-7 (because model 1 has a constant 0 and model 2 has a constant 24 months, so not much to see there).
- All cash withdrawal and replenish rules reached their targets again before the 2008 crisis hit.
- 2-3 years worth of expenses in a money market account (7%-10.5% of the portfolio) is a large cash cushion. It took a lot longer than 24 or 36 months to wipe out that cash cushion if we assume that we first consume the dividend yield: around 40 months with a 24-month cushion and 60 months with a 36-month cushion.
Next, let’s look at how Model 4, i.e., the 24-month cushion with all the bells-and-whistles (including the gradual replenishment) performed vis-a-vis the all-equity rule (Model 1) for the four different starting dates, see chart below:
Before December 1999 you would have suffered a bit from opportunity costs (recall that the equity market went up by quite a bit in 1997, 1998, and 1999). But this cash cushion rule would have outperformed the all-equity portfolio for all retirement dates between 12/1997 and 12/1999. That’s a surprise. I always thought that the cash cushion only works if you perfectly time it, but the opportunity cost of having a cash cushion will wipe out all the gain if you keep the cash for too long before the bear market. But that’s not true. Even in 12/1997, with two more years to go before the market meltdown, the cash cushion approach would have eventually made up for the opportunity cost.
Conclusions:
- Don’t get your hopes up too high! A cash cushion is no magic bullet against market fluctuations. I mean, come on, does anybody really believe that an equity portfolio will be down 60% but with a cash cushion of a few % you suddenly keep all your purchasing power? We didn’t and we confirmed our prior belief.
- A constant cash cushion is a bad idea. It just creates opportunity cost. No surprise here!
- Market timing works – to a degree! And it makes sense! By not selling equities when the market tanks you bet on long-term mean reversion. The market will always recover eventually.
- Of course, the cash cushion will likely help if you find yourself right before the market tanks. But I was amazed that even in Dec-1997 and Dec-1998 when the S&P500 still had a very strong subsequent performance in the 1998 and 1999 calendar years, you beat the passive withdrawal strategy of a 100% equity portfolio. I was very surprised about this result! I learned something new and unexpected today!
- Nothing is set in stone yet, but we might entertain keeping a bit of a cash reserve, come 2018! We will still do some more research on how robust these results are, but we may scale back our cash-aversion a bit.