April 14, 2023 – Welcome to a new installment of the Safe Withdrawal Rate Series. Please check out the SWR landing page for a summary of and a link to the other posts.
Today’s topic is homeownership. I’ve already made the case that not just rental properties but even homeownership can be a great tool in building assets (“See that house over there? It’s an investment!“). But what if you are already retired? What are some of the benefits of homeownership in the context of (early) retirement? Does homeownership reduce Sequence Risk? Do homeowners enjoy a lower inflation rate in retirement? If so, by how much can homeowners raise their safe withdrawal rate? How do we properly account for homeownership (with and without a mortgage) in the SWR simulation toolkit?
Lots of questions! Let’s take a look…
Do homeowners enjoy a lower inflation rate?
In my safe withdrawal rate simulations, I adjust the withdrawals and the portfolio values with the U.S. Consumer Price Index (CPI). It’s the most widely-used inflation gauge, and we also have decent historical estimates going all the way back to the 1800s, so I can run historical simulations long enough to cover some of the early market meltdowns, like 1907 and 1929.
So, is the overall CPI the correct measure for homeowners? As homeowners, we certainly have a different consumption basket because we don’t pay rent. Or, more precisely, a model-based estimate for owner-occupied equivalent rent indeed enters the CPI, but because homeowners effectively pay rent to themselves, any rental inflation will be a wash because it raises both your implicit expenses and your BLS-imputed income.
How big is the effect on your personal CPI when you’re not paying rent? Multiple people have reached out to me and asked me to comment on a recent phenomenon in the FIRE community that seems to have gotten a lot of buzz and attention, namely, the notion that homeowners enjoy a significantly lower CPI inflation rate and thus have an additional sizable safety cushion in their withdrawal rate.
My buddy Frank Vasquez at the excellent Risk Parity Podcast seems to be the origin of this new insight; see episode 209 (at around the 34:00 mark) and – so I have been told – at a breakout session at the March 2023 EconoMe conference. The story goes as follows: The housing component in the CPI makes up almost one-half (roughly 44%, to be precise), so if the long-term historical inflation rate was 3% for overall CPI and you are a homeowner, then your personal inflation rate is only 1.68% (=3% times 0.56) because you are not facing any rent increases.
Because the 4% Rule uses the overall CPI to adjust for annual spending increases in the historical simulations, we strongly understate the safe withdrawal rate of homeowners. Potentially by around 1.32 percentage points! Suddenly, the 4% Rule becomes the 5.32% Rule. Sweet, that’s a 33% increase in your retirement budget! And even if the decrease in the CPI rate doesn’t increase your SWR 1-for-1, but maybe only by 0.4x (see Part 47 for details), you still generate a major boost in your retirement finances. It sounds like a very intuitive and convincing story. Unfortunately, it’s also totally false. Here’s why…
The BLS is helpful enough to construct a special “CPI-Less-Shelter” series. In the table below is the average, annually-compounded overall CPI and CPI-less-shelter since 1947, which is as far back as the FRED data series goes. CPI was 3.53%, and CPI-Less-Shelter was 3.31% p.a.
So, if you don’t pay rent, your CPI would have been 0.22 percentage points less, translating into a 6% lower annualized inflation rate. Better than zero, but it’s a far cry from the purported 44% reduction.
And even the 0.22 percentage points would be a bit of an overstatement. As a homeowner, I still have to cover other housing-related costs, such as maintenance, repairs, regular updates/renovations, property taxes, property insurance, etc. So, for example, in my current neighborhood, homes would fetch about a 6% rental yield. But as a homeowner, I’d likely have to budget at least one-third of that as my out-of-pocket expense for keeping, maintaining, and insuring our single-family home. If my bills for property taxes, repairs, and insurance roughly go up with shelter inflation, then I lose another one-third of the CPI advantage, and we are down to about a 0.14 percentage points advantage. Maybe a 0.06 percentage point in the SWR. So, in a nutshell, as a homeowner, you may enjoy a marginally lower inflation rate, but it’s not exactly economically significant.
Why is the difference in CPI inflation rates and safe withdrawal rates so small?
The difference in inflation rates computed by Frank & Friends (1%+) vs. the BLS (0.22%) boils down to three misunderstandings.
1: The 44% weight is wrong!
Let’s start with the smaller one, which is due to a quantitative misunderstanding: The shelter component in the current CPI (Feb. 2023) makes up 34.5%, according to the most recent U.S. Bureau of Labor Statistics (BLS) data release. The Housing component that Frank & Friends use has a 44% weight, but that’s because all other housing-related costs, namely energy (electricity & heating), trash collection, etc., are all rolled into that housing cost figure as well. And you can ask any homeowner: you don’t get electricity, natural gas, etc., for free just because you own your home. Sorry for being pedantic, but as a homeowner, you hedge out at most your shelter inflation, not your entire housing inflation. And as mentioned above, not even the entire shelter inflation because you still have inflation exposure through the repair/maintenance/tax/insurance costs.
Also, I’m the first to concede that the most recent 12-month window saw much higher overall inflation (4.99%) compared to only 3.44% ex-shelter inflation. So, the deviation between non-shelter and overall inflation can fluctuate substantially in any given year. The most recent runup in rental inflation is because rental prices initially lagged the overall inflation pressures post-pandemic due to rent price “stickiness” and are now in a “catching-up” mode right when headline inflation is already subsiding. But if we average over the last three years, overall inflation was lower(!) by 0.20 percentage points than CPI-less-shelter (5.35% for CPI vs. 5.55% annualized for CPI-less shelter, between 3/2000 and 3/2023). Over the last five years, we’re again back to CPI-less-shelter below CPI by about 18 basis points. So, don’t focus on any one year. Look at the longer-term averages, all about 20 basis points!
And just as an aside, here’s another fly in the ointment for Frank’s theory: CPI vastly overstates the importance of rental inflation because the headline CPI figures look at urban consumers only. The average nationwide expenditure on rent accounts for only about 15% of consumption expenses, see BEA Table 2.4.5, where Line 50 (Housing) was approximately $2.4t or 15% of overall consumption (Line 1, about $15.9t).
2: CPI math
The more substantial misunderstanding in Frank’s CPI calculations has to do with the way the CPI index and its subcomponents are constructed.
Let’s get wonky and note that the CPI index is constructed as a so-called Laspeyres Price Index. As such, it will generate inflation rates that are simply the weighted average of the underlying component price changes. Please see the derivation from my Intro to Macroeconomics class at UC Berkeley Extension below:
So, if we assume there are only two components, Shelter and CPI-Less-Shelter, the CPI %-price change is simply the weighted average.
And we can also reverse the formula and solve for the CPI-Less-Shelter component; see below:
Notice how this formula is different from what Frank & Friends use:
So, whenever you calculate a CPI index for a subcomponent, like CPI-less-Shelter or CPI-less-Food-and-Energy (=core inflation), etc., it’s not enough to take out the undesired components. We must also reweight/rescale the index, hence the “1-w” in the denominator.
Let’s look at the following numerical example to demonstrate how significant an effect the sloppy CPI calculation can have. Imagine shelter inflation was 3.5%, and all other categories had 3% inflation. CPI-Shelter had a weight of 34%. Then overall inflation was 3.17% (=0.34×3.5%+0.66×3.0%). We can now confirm that CPI-less-Shelter inflation was indeed 3% through our formula:
Frank & Friends would have used 3.17%x0.66=2.09%, almost a full percentage point error.
3: Adjustments to the personal CPI don’t translate into 1-for-1 changes in the SWR (added 4/17/2023)
And a bonus item: even with a 0.15% difference in your personal CPI, you can’t necessarily translate that into a 1-for-1 improvement in your SWR percentage. The direct impact on the SWR is likely much smaller. For example, in Part 47, I looked at how a CPI+x% spending drift would lower your SWR. For x=1%, you lower your SWR by about only 38bps. For x=2%, you reduce your SWR by about 73bps. Of course, the exact figures depend on a lot of parameters, chiefly the retirement horizon and the final bequest target. That’s the exact reason we have to run some simulations to gauge the exact effects.
+ + +
So, enough for beating up on poor Frank Vasquez! I met him at the 2019 FinCon in Washington, D.C., and he’s a really nice guy. So, let’s look at some case studies for how we should adequately account for homeownership in the Safe Withdrawal Rate simulations.
SWR analysis: renter vs. homeowner with a mortgage-free home
Can a homeowner still afford a larger safe withdrawal amount than a renter, independent of the whacky CPI calculations on the internet? The rationale would be that a renter has larger mandatory expenses (rent) than the homeowner, who must cover taxes, insurance, and maintenance. Thus, the renter may be more impacted by Sequence Risk, i.e., withdraw from the portfolio while the stock market might go through a potential bear market early in retirement.
Another way to rationalize the homeowner advantage in retirement is to observe again that if a house has a gross rental yield of 6% and 2% overhead from taxes and maintenance, you have an asset that pays out a net real, CPI-adjusted yield of about 4%. Because safe withdrawal rates for early retirees are often less than 4%, you will likely benefit from having an asset that pays a “safe,” bond-like, and CPI-adjusted dividend of 4%. Compare that to current TIPS yields of only 1-1.5% or the current I Bond fixed rate of 0.40%. And notice that I’ve budgeted pretty conservatively. Some folks in the real estate world plan for an annual 12% gross rental yield and 6% overhead for a net 6% yield on their rental properties. That would certainly help you even more with your safe withdrawal math! But I like to go with the 4% real net yield to be on the safe side.
As always on the ERN blog, there is only so much you can accomplish with bloviating, so we need to run some simulations to know for sure and to quantify the effects of homeownership. As usual, I will do so with my Safe Withdrawal Rate Toolkit; see Part 28 for the link to the Google Sheet.
Let’s start with a simple baseline case for a renter:
- A $3m initial portfolio: 75% stocks and 25% bonds (10-year U.S. Treasury benchmark bonds).
- 0.05% annual expense ratio on your portfolio holdings.
- A 50-year horizon.
- A bequest target (or safety cushion) of $1m at the end of the retirement horizon. This figure is in CPI-adjusted terms! So, I set the final portfolio target in the main parameter tab to one-third.
- I model the rent as a monthly $2,500 outflow, to be adjusted by CPI inflation. Thus, it will show up as a “real” flow in the cash flow tab.
- For simplicity, there are no other supplemental cash flows (like Social Security, pensions, etc.). Or we assume that those cash flows later in life are used to fund higher healthcare costs. Since they would show up in both the renter and homeowner versions of the sheet, those additional flows would really only be a wash when comparing the two.
Also, notice that the way I modeled the cash flows in the SWR toolbox, the safe consumption amount is the non-housing retirement budget only. That’s because the housing cost is already factored in as a negative supplemental flow (i.e., outflow), so we can later compare the safe consumption amounts for the renter vs. homeowner.
Let’s compare the renter with the following homeowner:
- A paid-off home worth $500k.
- A $2.5m initial portfolio with 75% stocks and 25% bonds (10-year U.E. Treasury benchmark bonds). Notice that the renter and homeowner both have the same initial net worth. We want to compare apples to apples and keep the net worth the same. Of course, a $3m portfolio plus a paid-off home is better than a $3m portfolio and no home. I don’t need to run a simulation to know that.
- 0.05% annual expense ratio on your portfolio holdings.
- Assume that the home appreciates in line with CPI inflation.
- The same bequest target: $1m, which is now 40% of the initial financial portfolio. Notice that the house will count toward that bequest target. So, if the home is worth $500k at the end of the retirement horizon, we only supply another $500k from the portfolio!
- I model the cost of homeownership as a monthly $833.33 outflow, to be adjusted by CPI inflation, i.e., a 2% annualized cost of owning and maintaining the house.
- As in the case of the renter, there are no other supplemental flows.
And here are the results; please see the table below. Instead of safe withdrawal rates, I display the (fail-)safe withdrawal amounts for both the renter and the owner. As usual, I slice the data in different ways and also display the fail-safe for different market peaks (1929, the 1960s, 1970s, and 2000). Notice that for the 2000 dot-com market peak, I don’t have 50 years of return data, so I use actual data to 2023 and then conservatively calibrated constant real returns post-2023. Certainly, we should interpret the numbers with a grain of salt, but also note that SWRs are determined overwhelmingly by the first 10-15 years of return data, as established in my earlier research. Also, note that I’m not really after the absolute SWR figures but only the relative performance of renters vs. owners. Any error we make with the calibrated return data will impact both the renter and homeowner and likely be close to a wash. So, the relative performance of the year-2000 is still informative!
In any case, if we calculate the failsafe withdrawal amounts at or close to the historical market peaks, homeownership easily beats renting. We can express the advantage of homeownership in three different ways, 1) in annual spending, dollar terms, 2) in %-gain over renting, and 3) in % of the initial Net Worth (i.e., the percentage point increase in the SWR). Owners can afford a roughly 9% higher non-housing budget than renters. The percentage point change in the safe withdrawal rate is somewhere in the low double-digit percentage range.
If the stock market is not close to its all-time high, the advantage slowly melts away and even reverses. If equities are 35% or more below their most recent all-time high, the renter comes out ahead. It makes sense because equities will be close to the bear market bottom, and in the subsequent recovery, your financial asset portfolio will likely outperform a housing asset with a puny 4% real return. Of course, we might want to use these numbers with a grain of salt. I wouldn’t necessarily recommend selling your house and putting all the money into the stock market. If the stock market is beaten down, then likely, so will the housing market. It’s not something easy to model because I don’t have historical housing prices going back far enough to simulate this property.
Also, a few words about the absolute numbers. The overall failsafe consumption budget of $65,640 and $71,283 translates into respective safe consumption rates of only 2.19% (renter) and 2.38% (owner) of the initial net worth of $3m. The reason these figures are so small is that this is the non-housing consumption basket only. If we were to add the $30,000 in annual housing services consumed by the renter (and implicitly by the owner as well), we’re again in the 3.19-3.38% safe withdrawal rate for overall consumption. That’s as expected for a failsafe withdrawal rate over 50 years with a sizable bequest target of one-third of the initial nest egg.
And finally, a quick word about the 1960s vs. 1972/73 market peak. Notice how the mid-1960s cohorts faced more severe Sequence Risk than the cohort right before the first oil shock. It’s not that the returns from 1968 to 1972/73 were really bad, but they were lackluster, and combined with several years of withdrawals, the 1968 cohort portfolio was already underwater when the 1973-1975 recession hit. So, the 1972/73 cohorts didn’t even come close to two of the other worst-case retirement cohorts in 1929 and 1968!
So far, so good. I also like to look at some more advanced case studies, namely, what happens when housing inflation outpaces non-housing inflation. This brings me to the next section.
What if shelter inflation exceeds overall CPI?
Now assume that housing costs have a 0.50% higher inflation than the overall CPI. Specifically, let’s assume that all housing-related costs – rent and also the owner’s housing-related cost – grow 0.5% above inflation every year. I also assume that the value of the house will appreciate at a real rate of 0.50% over the 50 years. Thus, I assume that your rent goes up by 0.50% more than CPI inflation on every retirement anniversary date. See the screenshot from the SWR toolkit, specifically the cash flow tab:
How would this change impact the safe withdrawal amounts? Please see the table below. As expected, the renter is worse off than before. But the safe consumption amounts don’t change much for the homeowner ($71,069 vs. $71,283). The higher housing cost inflation along the way is balanced against the higher appreciation of the housing asset. Thus because the renter loses significantly and the owner is almost not impacted, the advantage of homeownership is now even larger. At the market peaks, we now see an advantage in the failsafe non-housing consumption of around 13%. We also push the crossover point of market drawdowns at which the renter will overtake the owner all the way to -45%. But also notice that the impact of the differential housing vs. non-housing inflation is tiny. In the base case, the homeowner had an SWR that was higher by 0.19 percentage points. Factoring in reduced non-housing inflation, we can expand the advantage by a “whopping” (sarcasm!) 0.08 percentage points to 0.27%, not exactly the 1+ percentage points as some folks on the internet are touting. But it certainly helps!
Summary so far: Homeownership hedges against Sequence Risk to a small degree. But this effect is mostly independent of any inflation rate differentials between shelter and overall CPI. Specifically, even if all expenditure categories have the same inflation rate, you already get a slight boost in the historical failsafe non-housing consumption budget by about 8.6%. With the difference in shelter vs. overall inflation by about 50 basis points, you gain another 4 percentage points (a total of 12.86%) in your non-housing consumption budget. It’s nice to have, but no panacea for Sequence Risk.
Could we increase the effect with a mortgage? That brings me to the next section…
Accounting for a home with a mortgage
Let’s now compare a homeowner with and without a mortgage. Let’s assume our retiree has a $500,000 home but also a $300,000 mortgage with a remaining 25-year term and a 4% nominal interest rate. As before, I like to keep the initial Net Worth at $3m, so we can compare apples-to-apples. That means the retiree now has a $2.8m portfolio together with $200k in home equity.
The way to model this setup in the SWR toolkit is to enter the mortgage payments as negative cash flows in the “nominal” column; please see the screenshot below. Also notice that I maintain the other assumptions about the housing expenses: $10k annualized (or 833.33 monthly), but rising at 0.5% every year. And at the end of the retirement horizon, month 600, the house with 0.5% annualized gains compounded over 50 years enters the net worth again.
The reason why one might suspect this setup works better is that we indeed replicate some of the flavors in Frank’s story: first, the nominal mortgage payments are being eroded through inflation, and second, once the mortgage is paid off, there is an additional sharp decline in real out-of-pocket expenditures for the homeowner. So, the home with the mortgage displays the desired behavior of spending increasing much slower than the overall CPI.
But alas, it still doesn’t work. In fact, if you have a home with a mortgage, you will have a smaller failsafe than with a mortgage-free home. Specifically, if we compare the owner with the owner+mortgage case, we see that the home+mortgage will offer less hedging against Sequence Risk at or close to the equity market peaks. And as expected, the advantage of owning the home outright melts away the further the equity market is below its all-time high. It’s a point I already made a while ago in Part 21: the larger initial out-of-pocket expenses due to the mortgage will exacerbate Sequence Risk. But of course, more leverage helps you if you expect a sharp recovery in the stock market, for example, if the market is beaten down and the next bull market is right around the corner.
Homeownership helped me to a degree in reaching FIRE; see my old post My best investment ever: Homeownership?! Owning a home in retirement – mortgage-free at that – also gives me a certain peace of mind because I own an inflation-protected asset with a safe real yield much higher than most other safe, diversifying assets. I’ve always defended homeownership against some of the unfair and often economically illiterate attacks from some corners of the FIRE and personal finance community (see How To “Lie” With Personal Finance – Part 2).
So, when I heard about Frank Vasquez’s interesting theory about how homeownership can pretty much solve all of your Sequence Risk headaches, I certainly wanted it to work. But it doesn’t work. I’ve pretty thoroughly debunked the idea of homeownership being the definitive cure for Sequence Risk. Homeownership probably helps you more consistently than some other whacky ideas I’ve come across (flexibility, the “Yield Shield,” a bucket strategy, and many others), but the small positive effect of a paid-off home on your safe withdrawal rate math has mostly to do with the respectable real yield of a home. Very little of the improvement in the SWR comes from the fact that shelter inflation runs a little bit hotter than overall inflation.
In conclusion, the renting vs. homeownership discussion is likely still a lifestyle choice. If you want to put down roots in retirement, you will find that there’s an advantage of owning over renting. But if you want to be a global nomad in retirement, I don’t blame you for renting. Just set your withdrawal rate a little bit lower to hedge against rental inflation.
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!
All the usual disclaimers apply!
Picture Credit: Pixabay.com
Someone in the comments section asked if I could share the Google Sheets. Here are the links; please see below. These are “View Only,” and I will not grant permission for you to edit. You need to save your own copy of the sheet!