Welcome back to the newest installment in our Safe Withdrawal Rate Series! If you are new to our site please go back to Part 1 to start from the beginning. And there are quite a few new visitors these days. That’s because our small blog is one of the finalists in the “Blog of the Year” category at the upcoming 2017 Plutus Awards. How awesome is that? Thank you to all of our faithful readers and followers for supporting and nominating Early Retirement Now!

But back to the topic at hand. It’s been on my mind for a long time. It’s relevant to our own situation and it’s come up in discussions on other blogs, in our case study series and in numerous questions and comments here on the ERN blog:

### Should we have a mortgage in Early Retirement?

The case for having a mortgage is pretty simple: You can get a 30-year mortgage for about 4% right now. Probably even slightly below 4% when you shop around. Equities will certainly beat that nominal rate of return over the next 30 years. Open and shut case! End of the discussion, right? Well, not so fast! As we have seen in our posts on Sequence of Return Risk (Part 14 and Part 15), the **average return** is less relevant than the **sequence of returns**. Having a mortgage in retirement will exacerbate your sequence of return risk because you are frontloading your withdrawals early on during retirement to pay for the mortgage; not just interest but also principal payments. In other words, if we are unlucky and experience low returns early during our retirement (the definition of sequence risk) we’d withdraw more shares when equity prices are down. The definition of sequence risk!

How badly will a mortgage mess with sequence risk and safe withdrawal rates? That’s the topic for today’s post…

### Our situation

Our situation is a bit unique in that we will move from our current location in a high cost of living area in an extremely high marginal tax state to a low cost of living area in a low or even no income tax state. Hence, we will certainly pay off our **existing** mortgage when we sell our apartment. We should be able to pay cash for a modest house in our new location and still have money left over to invest. But should we get a mortgage and invest even more money? It’s so tempting! And remember, I’m not foreign to the concept of leverage and I have praised it many times (see, for example, Seven reasons in defense of debt and leverage: Yes, you CAN have too little of a bad thing!). But that was when we were still saving for retirement. This time is different, though:

**Taxes:**In retirement, we expect to use the standard deduction. No more itemized deductions, hence, our mortgage interest is no longer tax-deductible.**The equity glidepath slope reverses in retirement!**As we detailed in the previous two installments of the series (Part 19 and Part 20), a glidepath shifting from a moderate bond allocation at the commencement of retirement to a mostly equity portfolio later in retirement can serve as a hedge against Sequence Risk. But with a mortgage, we’d do the opposite. Having a mortgage is similar (though not identical, I know) to a**short bond position**, and paying off the mortgage means we shift money out of equities and into bonds. The wrong direction! That can only exacerbate Sequence Risk!**Dynamic Withdrawal Rules:**Another tool we recommended to deal with Sequence Risk is to adjust the withdrawal rates according to how the portfolio performs. We like a CAPE-based rule (see Part 18 for more details) but others prefer the Bogleheads VPW (see Part 11 for a comparison of different dynamic rules). But using variable rules becomes harder when faced with the fixed expense of a mortgage. Imagine you follow the VPW rule and you have to cut your $4,000 monthly withdrawal by 50% in response to a 50% market drop. That’s painful. But with a $1,000 mortgage and a $5,000 initial monthly withdrawal, you’d cut your withdrawal to $2,500. After paying for the mortgage you’re left with only $1,500 in discretionary spending compared to $4,000. That’s a 62.5% drop in consumption! Thus, the**mortgage will magnify the consumption impact of market volatility!**

### Simulations and Limitations

The 30-year mortgage as we know it today didn’t even exist before the Great Depression. What’s worse, I don’t have a very long time series of 30-year mortgage rates. Even if I had a time series for mortgage rates I’d have to make assumptions, lots of assumptions, about if and how each of the cohorts since 1871 would have handled changing interest rates and potential prepayments and/or refinancing of mortgages. A can of worms!

So, studying the pros and cons of a mortgage-free early retirement would have to take a few shortcuts and hacking of my simulation engine. And once you start hacking, always keep in mind one of Big ERN’s fundamental rules:

### There is a fine line between **doing** a hack and **being** a hack!

So for full disclosure, today’s simulation results are mostly a thought experiment with the following assumptions:

- I calculate the mortgage payment of a 30-year and 15-year mortgage with today’s market rates and assume that the real, inflation-adjusted mortgage payments decay due to a projected 2% annual inflation rate going forward, see chart below.
- Given the mandatory real mortgage payments, what would be the experience of a retiree today if the real stock/bond returns of all the past retirement cohorts were to repeat themselves?

In other words, I don’t simulate how a retiree in 1929 with a mortgage in 1929 would have experienced the 1929-1989 equity and bond returns. I calculate how a retiree **today** with **today’s mortgage** parameters would fare if we hit him/her with the 1929-1989 real, CPI-adjusted stock/bond returns. And the 1928-1988 returns and the 1927-1987 returns, and the 1930-1990 returns, and so on.

You be the judge if this crosses the line. But remember, before yelling at me, please keep in mind another one of Big ERN’s fundamental rules:

### It takes a model to beat a model!

In other words, unless you have a better way of evaluating the mortgage vs. no mortgage tradeoff please don’t call me a hack. 🙂

### Simulation Assumptions:

- 60-year horizon, capital depletion target.
- We run monthly simulations through our Safe Withdrawal Rate Google Sheet. See Part 7 of this series for more information.
- We assume a 2% annualized inflation rate, so the mortgage payments, in real inflation-adjusted terms will decline over time.

We look at eight different models/parameter assumptions. They all have the same initial net worth but different assumptions about the mortgage, mortgage term and stock/bond allocation:

**Model 1 (baseline):**$1,000,000 portfolio, no mortgage, 80% Stocks, 20% Bonds.**Model 2:**$1,000,000 portfolio, no mortgage, 100% Stocks.**Model 3:**$1,2000,000 portfolio, $200,000 mortgage, 30-year term, 3.875% interest. 100% Stocks.**Model 4:**$1,2000,000 portfolio, $200,000 mortgage, 15-year term, 3.250% interest. 100% Stocks.**Model 5:**$1,2000,000 portfolio, $200,000 mortgage, 30-year term, 3.875% interest. 80% Stocks, 20% Bonds.**Model 6:**$1,2000,000 portfolio, $200,000 mortgage, 15-year term, 3.250% interest. 80% Stocks, 20% Bonds.**Model 7:**$1,2000,000 portfolio, $200,000 mortgage, 30-year term, 3.875% interest. 67% Stocks, 33% Bonds. Why 2/3 Stocks and 1/3 bonds? That’s because $800,000 in stocks and $400,000 in bonds netted with a $200,000 mortgage brings us exactly back to the 80/20portfolio allocation in the baseline!**Model 8:**$1,2000,000 portfolio, $200,000 mortgage, 15-year term, 3.250% interest. 67% Stocks, 33% Bonds.

### Results

To warm up, let’s start with a simple time series plot of the safe withdrawal rates of the 8 different models. All numbers are percentages of the initial Net Worth to account for the fact the models with a mortgage obviously have a larger portfolio value! I plot this just for the record because you can’t really make out the relative performance of the different rules. But notice how all 8 models drop below the ostensibly safe 4% mark for quite a few unlucky retirement cohorts.

Let’s see when the models do better than the baseline. In the plot below I display the difference between models 2-4 (Model 2 = 100% Stocks, no mortgage, Model 3 = 100% Stocks, 30Y mortgage, Model 4 = 100% Stocks, 15Y mortgage) and the baseline SWR. Notice how most of the time, models 2-4 do better than the baseline (lines above the zero line). Substantially better, in fact, by about a full percentage point on average. So, on average you can withdraw $10,000 more when you leverage your 100% equity portfolio with a mortgage. Nice! Unfortunately, the outperformance comes at a high price. You underperform when it hurts the most, namely, when the Safe Withdrawal Rate is low: During the Great Depression, the late 1960s and around the dot-com bubble.

Below is the same chart for the other four Models. Models 5 and 6 (80/20 portfolio with a 30Y and 15Y mortgage, respectively) have qualitatively the same features as models 3 and 4: outperform substantially most of the time thanks to leverage, but do worse when it hurts the most, i.e., when the SWRs are low during the major market events. Models 7 and 8 have a completely different pattern. They would have helped slightly during the Great Depression and dot-com bubble when bonds offered great diversification benefits. But that’s not thanks to the mortgage. It’s entirely due to the higher bond share. For example, a 67/33 Stock/Bond portfolio without a mortgage would have handily outperformed Models 7 an 8. Also notice that in the 1960s, Models 7 and 8 would have underperformed the baseline because bond returns suffered so badly during the 1970s, due to the inflation surge.

And finally, here’s a summary table of the safe withdrawal rates in the 8 different models. Same story again: Model 1 offers the best failsafe withdrawal rate but the mean/median/max look very mediocre compared to the others. Likewise, the models with the highest mean and median (Models 3,4) have atrocious failsafe withdrawal rates in the low-2%!

### Results conditional on an elevated CAPE Ratio

The unconditional distribution of safe withdrawal rates is interesting but it obviously ignores the fact that in light of today’s elevated CAPE Ratio equity returns could be slightly below average going forward. Let’s look at the safe withdrawal amounts when targeting different failure probabilities conditional on the CAPE between 20 and 30, see table below. The way to read this table: For example Model 1 (80/20 portfolio, no mortgage), an initial withdrawal amount of $33,262 (subsequently adjusted for CPI) would have been the failsafe. With $34,791 you would have run out of money 5% of the time, with $44,658 you would have run out of money half the time, etc.

According to this table, no mortgage and an 80/20 portfolio would have done the best if targeting a failsafe withdrawal amount and all other failure probabilities up to 10%. If you want to maximize the withdrawal amount to target 25% and 50% failure rates (which seems way too risky for my taste), then the 100% equity portfolio with the mortgage become the most attractive. But Models 3 and 4 with the most leverage also have the worst failsafe withdrawal amounts.

The lesson from this exercise: If you are risk-averse and like to hedge out the tail risk it’s best to have no mortgage and a moderate bond allocation. If you are a risk-taker (degenerate gambler?) then you might as well go all-in: Have a mortgage and 100% equities in the portfolio as well. Having both a mortgage and a bond portfolio doesn’t make any sense. And there is a reason for it: the bond return is likely inferior to the mortgage yield. This is beautifully consistent with an old vintage post from last year. If you want to use the mortgage as leverage to juice up your equity returns, that’s fine. It’s a matter of risk tolerance. But make sure you don’t use the mortgage to buy low-yielding bonds; leverage only works when your asset returns more than what you pay for your liabilities!!!

### Exceptions to the no-mortgage recommendation

Our case is special and probably not applicable to everybody in the FIRE crowd. I can think of at least a few scenarios where folks might want to keep their mortgage. For example, you might have a hard time paying down the mortgage all at once because you’d realize too much in capital gains. This could put you into the 15% tax bracket for capital gains and that additional tax burden would negate the mortgage paydown idea.

Another scenario: some early retirees simply don’t have enough in after-tax savings. This is a frequent issue in some of the case studies I have seen where people have saved a huge pile in tax-deferred accounts but have only very little in after-tax accounts. If your small taxable accounts barely last until age 59.5 (to avoid tapping tax-deferred accounts before the cutoff age) you probably can’t afford to pay off the mortgage.

Finally, I can see how at some point down the road interest rates could be much higher than today. If you retire in 5 years and still have 20 years left on your 3.25% fixed rate mortgage but bond interest rates are now 3.5 or 4%, then by all means, hold on to that mortgage. Now the mortgage vs. bond leverage works beautifully!

### Conclusion

The decision whether or not to keep a mortgage in retirement is not trivial. The comparison “expected equity return > mortgage rate” is just too simplistic. The median/average retiree will clearly benefit from the leverage but also remember that the median retiree never runs out of money either. For us, not having a mortgage might hurt us in the long-run but only in the scenarios where we’d become fabulously rich anyway. Who cares if we end up with $6 million instead of $7 million when we’re in our 80s? We are willing to pay that cost for the hedge against Sequence of Return Risk, i.e., the very unpleasant tail risk of running out of money after 30 or 40 years due to poor portfolio returns in the first few years after retirement.

### We hope you enjoyed today’s post. What’s your mortgage strategy for early retirement? Please leave your comments and suggestions below and make sure you check out the other parts of this series:

- Part 1:
**Introduction** - Part 2: Some more research on
**capital preservation vs. capital depletion** - Part 3: Safe withdrawal rates in different
**equity valuation**regimes - Part 4: The impact of
**Social Security benefits** - Part 5: Changing the
**Cost-of-Living Adjustment**(COLA) assumptions - Part 6: A case study: 2000-2016
- Part 7: A
**DIY withdrawal rate toolbox**(via Google Sheets) - Part 8: A
**Technical Appendix** - Part 9:
**Dynamic**withdrawal rates (Guyton-Klinger) - Part 10: Debunking Guyton-Klinger some more
- Part 11: Six criteria to grade
**dynamic withdrawal rules** - Part 12: Six reasons to be suspicious about the “
**Cash Cushion**“ - Part 13: Dynamic Stock-Bond Allocation through
**Prime Harvesting** - Part 14:
**Sequence of Return Risk** - Part 15: More Thoughts on
**Sequence of Return Risk** - Part 16: Early Retirement in a
**low return environment**(The Bogle scenario!) - Part 17: Why we should call the 4% Rule the
**“4% Rule of Thumb”** - Part 18:
**Flexibility**and the Mechanics of**CAPE-Based Rules** - Part 19:
**Equity Glidepaths**in Retirement - Part 20: More thoughts on
**Equity Glidepaths** - Part 21:
**Mortgages**and Early Retirement don’t mix! - Part 22: Can the
**“Simple Math”**make retirement more difficult? - Part 23:
**Flexibility**and**Side Hustles!** - Part 24:
**Flexibility Myths**vs. Reality - Part 25: More
**Flexibility Myths** - Part 26: Ten things the “Makers” of the 4% Rule don’t want you to know
- Part 27: Why is
**Retirement Harder**than Saving for Retirement? - Part 28: An
**updated Google Sheet**DIY Withdrawal Rate Toolbox - Part 29: The
**Yield Illusion:**How Can a High-Dividend Portfolio Exacerbate Sequence Risk? - The Yield Illusion Follow-Up (SWR Series Part 30)
- The Yield Illusion (or Delusion?): Another Follow-Up! (SWR Series Part 31)

In my case i have pension income that more than covers the mortgage so I am not inclined to pay off a low rate mortgage early

That certainly helps. But not having a mortgage would also help with Sequence Risk, so some people would pay off the mortgage even in that situation.

This article is fascinating to me and I wish I could hire you to analyze our financial situation. The previous article on the reverse equity glidepath has been a guiding light for us, but this article threw all of that into disarray. I’m curious if you have comments on a very high-level summary and/or can recommend tools to do a better analysis?

Current investments: ~$2.1M, goal $2.3M

AA: 80/20, goal to be 60/40 at FIRE

Mortgage: ~$800k

We have enough in taxable investments to wipe out the mortgage but that would incur some capital gains taxes as well as leave us cash poor outside our retirement accounts. We intend to be FI in the next year or two, or at a minimum drop to part-time work for a little while to pad the stash.

I realize there are a ton of factors at play (MAGI for ACA subsidies, tax basis, AA, Roth vs trad vs taxable breakdown) so what advice do you have on how to analyze this whole complex situation?

Thanks!

I don’t offer for-pay advice (yet), but here’s my 2-cents:

It’s painful to see the large mortgage plus a 40% bond portion, which implies about a bond portfolio the same size: 0.4x$2.3m=$920k.

If there’s a way (say, through a large chunk of high cost basis stocks and/or bonds in taxable portfolio) I’d pay off the mortgage. How high is your interest rate? It would be painful to see a 4%+ mortgage rate when bond yields are so low.

Also, with a mortgage this large, how big is the house? Do you have enough resources to pay for property taxes, maintenance, etc. while in retirement?

Mortgage rate is 3.625%. We live in a HCOL area, so at 1100ft^2 for a family of 5, we are already on the inexpensive side for our area, as crazy as that may sound. We do have budgeted in for property taxes, insurance, and other expenses in retirement.

If it makes a difference, bonds are all in tax-advantaged accounts and the tax basis of the taxable account is fairly high (~10% is gains).

Well, 3.625% safe return… One could make the point that it’s wise to pay down the mortgage then, if you’re risk averse. Your (esteimated) $4,000 mandatory motnhly mortgage payment increases Sequence Risk .

But if you’re not that risk averse, probably keep everything as is and concentrate on Roth Conversions to fill up the lower tax brackets. first. Maybe in later years fill up the 0% capital gains tax bracket sell high-cost-basis equities in the taxbale account to pay off the mortgage…

If we paid off the mortgage then I see it would make sense to follow this article’s advice on asset allocation for a reverse equity glide path. If we keep the mortgage, what asset allocation would you follow in our shoes?

If we are going purely by gut/risk tolerance, I’d hate to wipe out such a big chunk of my taxable account. Then again, I may feel differently about that mortgage note once no longer working.

A third, intermediate option is to pay down some of the mortgage and have it recast. This is available to us if we pay in chunks of at least $50k.

With the mortgage you’d need some bonds to hedge the Sequence Risk.

I like the idea of paying down pieces of the mortgage and recasting the mortgage (i.e., keeping the same term but lowering the payments).

But I als see that it’s problematic to wipe out too much of the taxable account. It’s always best to keep at least a few years (5+) in the taxable account, certainly until you got a Roth Ladder going.

I finally found some time to do some more analysis using cFIREsim. I modeled everything from keeping the full mortgage to recasting the mortgage by paying off chunks in $200k increments, all the way to paying the entire thing off tomorrow. I used both a reverse equity glide path asset allocation and a fixed 75/25 allocation.

I found that my success rate is highest keeping the full mortgage amount using reverse equity glide path, and decreases somewhat linearly the lower the mortgage is. The same phenomenon is seen with a fixed asset allocation, though overall success rates are lower than with reverse equity glide path.

So what I am trying to understand is why I am getting the opposite conclusions than you did in your analysis? Is it the limitations of cFIREsim (US-centric data)? Is there something else I am missing here? I’m a little surprised at these results after reading everything you have done here.

I can only vouch for my own Google Sheet, so I can’t tell you for sure where cFIREsim differs. I can imagine that if you enter a nominal mortgage payment in cFIREsim at today’s rate of 4% then, sure, during the 1970s you would have done better with the mortgage due to the high inflation. But in today’s environment, your mortgage will be deflated only at 1.5-2% p.a.

That’s why I prefer the mortgage modelling as described in Part 21 of my series. Work exclusively with real returns and assume the mortgage payments shrink by 2% p.a. to account for today’s inflation. Again: if you apply historically high inflation rates (10%+ in the 1970s) it’s a no-brainer to keep the mortgage! This may explain the difference!

Disagree that the correct metric to measure the mortgage interest rate arbitrage against is only the bond component of the portfolio. Instead, you use the return of the total portfolio, not just the bond portion. So if your 50/50 equities/bonds portfolio averages a 5% return and your mortgage interest rate is 4%, you’re earning a plus 1% (over time, on average, of course–but then the mortgage is either 15 or 30 years, so that’s the correct time period to measure over). The mortgage doesn’t “know” that it’s only supposed to be used as leverage against the bond component of the portfolio. Nor does the portfolio “know” that only the bond portion is being leveraged. In fact, this error in ERN’s post is very common, but shows a fundamental misunderstanding of the basic theories of portfolio construction–i.e. the capital markets line. You construct your asset allocation stocks/bonds in accordance with your preferred risk level (Sharpe Ratio), ideally, the “tangency portfolio.” You then lever up or down (i.e. borrow to invest more in the portfolio at the preferred AA, or alternatively, hold cash). No one can borrow at the risk free rate, but if you can get a low enough interest rate, it can make a lot of sense. Remember: When you take out a mortgage (or do any other form of borrowing), you are levering the entire portfolio, not just the bond portion.

Disagree. Believe me, I do undestand portfolio construction quite a bit better than you. Actually, you are the one misunderstanding the issues here.

You seem to be under the false pretense that there’s a contractual constaint that you HAVE to keep the asset allocation at 50/50. But I don’t. I can pay off the mortgage and use the bond portfolio for that and I get a positive carry from that. Plain and simple.

Further, if you maintain your net asset allocation at the same level, and assuming you have a reasonably low mortgage interest rate, having a mortgage doesn’t increase Sequence of Returns Risk. ERN’s post is conflating tax issues (having to withdraw from an IRA to pay the mortgage for example) as well as neglecting to account for the fact that leveraging the portfolio with a mortgage, but insisting on buying more equities rather than bonds with the additional funds (due to the above misunderstanding about portfolio construction theory and the capital market line), means you’ve increased your equity risk level. It is the increased level of equity risk that creates the additional risk, which is not Sequence of Returns Risk. So let’s say you have a $1,000,000 portfolio with a 50/50 stock/bond allocation. You also have a $500,000 house with a $250,000 mortgage on it. Treating the mortgage as a negative bond (in actuality it’s more of a negative mortgage-backed security, which isn’t quite the same thing), your net AA is 500 stocks & 250 bonds, or 500/750 = 66.66666…%, rather than the 50% without the mortgage. Going into retirement, if you pay off the mortgage from your 500k in bonds, you end up with 500 stocks & 250 bonds, and no mortgage. But your AA is exactly the same: 66.66666…%. So you haven’t reduced your Sequence of Returns Risk at all. Your 66.6666% portfolio is subject to the exact same SRR as any other similarly risky portfolio. Now let’s say instead of paying off the mortgage–let’s say you have a 4% mortgage and your average expected return across your entire investment portfolio is 6%–not a very long reach–and you want to take advantage of that arbitrage, so you keep the mortgage into retirement. You don’t have to withdraw any additional funds from your IRA to make the monthly mortgage payments since you have an “extra” $250 k in fixed income (because you didn’t use it to pay off the mortgage early). If you didn’t have it in after tax accounts in the first place, you would have had to have withdrawn it from IRAs to pay off the mortgage early, anyway. The “extra” $250k is what you use to make the mortgage payments in retirement. The mortgage isn’t just leveraging that “extra” $250k–it’s leveraging your entire portfolio. That’s the conceptual error (or one of several) that ERN and many others seem to repeatedly make in this kind of analysis. Another mistake often made seems to be that when people have the choice between either making extra payments on the mortgage or buying more investments, they don’t seem to realize (as ERN doesn’t seem to realize) that you can’t simply use those additional funds to buy 100% equities without increasing your portfolio’s risk level. THAT’s what causes the increase in Sequence of Returns Risk. When making the choice between paying the mortgage down quicker with extra funds, or investing more in the portfolio, you have to use the extra funds to invest in your portfolio in stocks and bonds in the same proportions in order to maintain the same risk level. If you buy all stocks with the extra money you will increase your risk level. What is not “good” for most people to go into retirement with, is a 100% (or higher, if they retain a mortgage and have an all-equities portfolio) equities asset allocation. THAT’s what creates excessive SRR. If you have a relatively conservative portfolio going into retirement, let’s say 60/40 net of your mortgage, it’s perfectly fine to have a mortgage in retirement, and can be quite beneficial, if you have a good interest rate.

Look, you obviously misunderstand both my post and Sequence Risk in general.

I show with the simulations that Model 7+8 (again 80/20 portfolio if netted) DOES INCREASE Sequence Risk.

“Even a 40% stock allocation according to Vanguard returns an average of 7.8%/year so why would anyone not want to keep a 4% mortgage going into retirement in this scenario? ”

Because you still don’t understand Sequence Risk. The MEAN return may be higher, but Sequence Risk doesn’t care about the mean return. By definition.

Thanks for the insightful post. It’s a good reminder that in (early) retirement the goal isn’t so much maximization of returns but protecting against sequence of return risks. Running out of money is a much larger issue then ending up with too much money.

The one place where I think the model falls short is in the 20’s / 30’s. In other times, the 2% inflation assumption coupled with todays mortgage rate works well enough (since both the inflation and mortgage rate move together). But the deflation of the 20’s / 30’s would have made the situation a lot worse. Increasing your mortgage payment with 10% a year in real dollars would not be pretty.

Yes! Good point. Though I hope that we have a central bank today with enough resolve to not let double-digit deflation happen ever again. If push comes to shove they’ll drop money from helicopters (!) to create inflation. 🙂

There might be a mistake in #3 under section “Situation”. Shouldn’t that be $2000 (not $4000) vs. $1500? Still proves your point…just by a smaller margin.

Before:

Total spending: $5,000

Mortgage: $1,000

Discretionary: $4,000

After:

Total spending: $2,500 (-50%)

Mortgage: $1,000 (same)

Discretionary: $1,500 (-62.5%)