September 5, 2023 – Welcome back to a new blog post in the Safe Withdrawal Rate Series! It’s been a while! So long that some folks were wondering already if I’m all right. Nothing to worry about; we just had a busy travel schedule, spending most of our summer in Europe. First Italy, Switzerland, Austria, and Germany. Then, a cruise through the Baltic Sea from Sweden to Finland, Estonia, Latvia, Poland, Germany again, and Denmark. But I’m back in business now with a fascinating retirement topic dealing with Social Security timing: What are the pros and cons of deferring Social Security? If we set aside the ignorant drivel like “you get an 8% return by delaying benefits for a year” and look for more serious research, we can find a lot of exciting work studying this tradeoff. Earlier this year, in Part 56, I proposed my actuarial tool for measuring the pros and cons of different Social Security strategies, factoring in the NPV/time-value of money consideration and survival probabilities. A fellow blogger, Engineering Your FI, has done exciting work studying this tradeoff using net present value (NPV) calculations. And Open Social Security is a neat toolkit for optimizing joint benefits-claiming strategies.
But those calculations are all outside of a comprehensive Safe Withdrawal Rate analysis. How does Social Security timing interact with Sequence Risk? For example, can it be optimal to claim as early as possible to prevent withdrawing too much from your equity portfolio during a downturn early in retirement? In other words, if you’re interested in maximizing your failsafe withdrawal rate, you may feel tempted to pick a potentially suboptimal strategy from an NPV point of view. Sure, you underperform in an NPV sense on average if you claim early. But hedging against the worst-case scenarios may be worth that sacrifice.
Let’s take a look…
Preliminary calculations
Before we even get started, I like to perform some preliminary calculations. Sometimes, folks in the FIRE community ignore their future Social Security benefits, whether out of laziness or excessive caution, due to potential future policy changes. Depending on your retirement age, you could leave a lot of money on the table! Let’s put that to the test. Assume that we have a retired couple, both 35 years old, with a $2,000,000 portfolio. I assume a 60-year horizon and capital depletion, a 75% stocks and 25% bonds portfolio (10-year Treasury benchmark bonds, and no supplemental benefits in the future. I calculate a safe consumption amount (=retirement budget) of $65,073, i.e., the safe withdrawal rate is about 3.25%.
Assuming a $10,000 supplemental annual income ($833.33/month) at age 67 (as always, in today’s dollars) would raise the safe withdrawal rate to $66,444 or about 3.31% of the initial capital if we assume we reduce the withdrawals later in retirement 1-for-1 when Social Security kicks in. Almost $1,400 a year more; that’s better than nothing! Also, the effect on the sustainable retirement budget is approximately linear (within bounds, of course). So, for example, for a $30k/year retirement income at age 67, we’d bump up the retirement budget by more than $4,000 per annum.
Once you get closer to age 67, supplemental income will have larger benefits on your retirement budget. In the table below, I list the results for different retirement ages and retirement horizons. If you retire in your mid-to-late-40s, like most folks in the FIRE community, don’t ignore future Social Security benefits! First, you will likely have accumulated significant Social Security credits with a long enough work history, and you will bump up your retirement budget by about 30-35% of the future expected income, equivalent to raising your SWR by 0.15 to 0.18 percentage points. Don’t leave that on the table! But maybe apply a small haircut to account for the likely policy uncertainty. But it is certainly true that for extremely young retirees, say, at 30 years old, future Social Security benefits don’t make a huge benefit.
Social Security Timing for Early Retirees
I established that early retirees should certainly incorporate future benefits into their safe withdrawal rate calculations. But should they worry about the exact Social Security timing, too? Sure, at age 45, you would not lock in your retirement age. You can defer that decision until at least age 62. But will a change in the timing of benefits make a meaningful difference in the sustainable retirement budget? Without running explicit simulations, we can already suspect the answer is no. If you retire 20 years before your eligibility age, you don’t get much of a hedge from Sequence Risk because Sequence Risk arises from large portfolio drawdowns during bear markets in your first 10-15 years. But we can certainly study how much of an impact different claiming strategies would have.
So, let’s assume we now have a 45-year-old couple, $2,000,000 in their portfolio, a 50-year horizon, and no bequest target (i.e., capital depletion is OK). They expect combined monthly benefits of $2,500 at age 67, i.e., $30k a year. How much would we alter the retirement math by claiming early ($1,750 at age 62) or late ($3,100 at age 70)?
Before I dig into the details, let me first report the safe consumption amounts for this 45-year-old couple for the different claiming strategies.
- Age 62: $74,862
- Age 67: $74,412
- Age 70: $74,049
It turns out that claiming earlier, at age 62, will give us a slight edge. Not by much; we’re talking about $400 annually in additional retirement budget when claiming at 62 instead of 67. That’s about a 0.5% increase in the budget or a 0.02 percentage point rise in the sustainable withdrawal rate. The range between the best timing (62) and worst timing (70) is about $800 a year or 0.04 percentage points. Not much, but better than nothing.
I want to understand what’s going on here and whether the superiority of claiming as early as possible is generally valid for all cohorts. I plot the advantage of claiming at 70 vs. 62 in a time series chart below. The blue line is the advantage of deferring benefits (plotted using the left axis), and the orange line is the safe withdrawal rate when claiming at age 62, using the axis on the right. There have been historical cohorts where deferring benefits would have been advantageous, namely between 1933 and 1955. Still, all historical worst-case cohorts, like 1929 and the mid-to-late-1960, when the SWR dropped below 4%, were all instances where claiming earlier would have been advantageous.
Side note: Social Security wasn’t even around in 1929. So, why would I simulate that cohort planning for future benefits? It is very simple: think of my simulations as thought experiments where today’s retirees who enjoy today’s government programs want to model safe withdrawal strategies that could withstand a replay of past asset returns. Just because you believe that 1929-1932 could happen again doesn’t necessarily mean that we would also lose Social Security and go back into the 1920s in all other respects.
And again, just for reference, not modeling the Social Security income at all would give you a safe retirement budget of only $66,986. Generally, modeling future benefits is essential for 45-year-olds in early retirement. But it’s likely too early for most of us in the FI community to worry about the precise Social Security timing, at least from a Sequence Risk perspective. I recommend deferring that decision until you get closer to the eligibility age. But eventually, that day will come, which brings me to the next section…
Social Security Timing for a 62-year-old
Now assume that we’ve reached the earliest possible benefits-claiming age, currently 62. Assume that our retired couple has a 33-year horizon up to age 95, a $2,000,000 portfolio, and monthly benefits at age 67 of $2,500. As before, claiming at age 62 would imply $1,750 in monthly benefits (30% reduction), while deferring until age 70 would give you $3,100 or a 24% benefits boost.
Here are the safe consumption amounts when claiming at different ages:
- Age 62: $95,840
- Age 67: $96,198
- Age 70: $96,858
Remember that when claiming at age 62, the $95,840 annual safe consumption level already includes the $21,000 Social Security benefits. That’s one of the reasons why I prefer the term “Safe Consumption Rate.” You are withdrawing only $74,840 from the portfolio, while the rest comes from Social Security!
It’s intriguing that the order is reversed relative to the 45-year-old retriees; you do better deferring benefits. Clearly, there may be a benefit to claiming early and preventing portfolio withdrawals when prices are down. But the worst-case scenarios in past retirement cohorts faced very long portfolio drawdowns, sometimes 15-20 years. Thus, getting those higher benefits at age 70, only eight years into retirement, ensured that your portfolio recovered eventually. Below, I plot the same style plot again as before. Notice how the blue line is consistently above zero around the historical worst-case cohorts (1929, 1964-1968).
But also note that for some cohorts with very high SWRs, say post-1980, when SWRs would have been 10+%, it would have been reversed: You were better off claiming early, at age 62, to avoid drawing down the portfolio during the roaring 1980s bull market. That’s very intuitive.
Even the cohort that retired at the 1937 market peak would have been roughly indifferent between claiming at 62 vs. 70, indicated by the blue line being close to the zero line around that time. That makes sense because the 1937 bear market was much shorter than the market drawdowns starting in 1929 and the 1970s and 80s.
Social Security Timing Case Study: September 1929
I want to drive home the point from the previous section and showcase how deferring benefits until age 70 would have improved the retirement experience. Assume that our 62-year-old couple had retired right before a replay of the stock market crash preceding the Great Depression, i.e., in September 1929. This was indeed the historical worst-case scenario, so a retirement budget of $95,840 (the fail-safe amount quoted above) would precisely deplete the $2,000,000 portfolio after 33 years. What if we had used the same $95,840 retirement budget but deferred Social Security benefits until age 70, then? Let’s plot the portfolio value of the two strategies in the chart below. I plot the two portfolio values as well as the difference (70 vs. 62). Claiming benefits at age 70 would have drawn down the portfolio even more for the first eight years, but eventually, our retirees would have come out ahead by about $126k after 33 years. The crossover point occurs after 317 months, i.e., more than 26 years into retirement.
Summary so far: A replay of the Great Depression market event is long enough for you to find it advantageous to defer your benefits rather than claim Social Security early at age 62. The larger benefits starting in year eight will help during the recovery period. In fact, the higher benefits would have kicked at precisely the right time in 1937 when the stock market took another severe nosedive.
Social Security Timing Case Study: December 1968
Next, let’s do the same exercise for the December 1968 cohort. Again, I set the retirement budget to exactly deplete the portfolio after 33 years, implying a slightly higher budget ($95,983 p.a.) than in 1929. I apply the same retirement budget but defer benefits to age 70 and plot the portfolio time series in the chart below. This time, deferring benefits beats claiming at 62 by about $432k after 33 years, though we’d again draw down the portfolio during the first eight years. The crossover point is now 232 months, about 19 years into retirement.
Again, the higher benefits starting in year eight would have been well-timed when the stock market was still down in the mid-70s and facing another two recessions and accompanying bear markets in the early 80s.
A Social Security “Bridge Strategy”
Notice that if you are 62 years old and would like to defer benefits until age 70, it doesn’t necessarily mean that you must suffer through Sequence Risk for the first eight years of retirement. You could, of course, create your own quasi-Social-Security payments. Simply set aside a portion of your portfolio to build a bond ladder, or even better, a TIPS ladder, to bridge the years until you can claim your maximum benefits. Especially with bond interest rates back to multi-year highs again – around 2% for real-inflation adjusted TIPS – that plan looks more attractive now.
Let’s assume our retired couple, aged 62, defers benefits until age 70 but sets aside enough money to bridge the first eight years with $3,100 of monthly (real, CPI-adjusted) income from a TIPS ladder. I assume the real interest rate is 2% p.a. How much money would we shift from the 75/25 portfolio? We can use the Excel PV function…
=pv(0.02/12,96,-3100,0,1)… which is about $275,000.
(Side note: some folks prefer the compounding formula for the monthly interest, i.e., 1.02^(1/12)-1, but it makes no noticeable difference for small enough interest rates.)
Your $1.725m portfolio plus $3,100 monthly in supplemental income from the bond ladder in years 1-8 and Social Security after that affords us a safe consumption level of, get this, $101,739. That’s significantly higher than the $96,858 in the model without the bond ladder.
But make no mistake! The bond ladder would have been effective only for the cohorts that retired at or close to the historic market peaks, right before your retirement portfolio would have tanked and returned significantly less than the 2% real return in the TIPS ladder. Most of the time, the TIPS ladder would have underperformed your 75/25 portfolio. But of course, that’s the nature of this Sequence Risk hedge: you do better in the worst-case scenarios, but you lose a little bit and leave a slightly less spectacular inheritance to your heirs when the market rallies during your first eight years of retirement. Most retirees are willing to pay this “insurance premium.”
Conclusion
After 7+ years of blogging and 50+ posts in the SWR Series, I’ve finally looked deeper into the Social Security timing question. Contrary to popular belief and my initial intuition, claiming benefits early will not necessarily hedge against Sequence Risk. That’s because some of the historical bear markets were far too long, and getting benefits at age 62 instead of 70 would not have made much of a difference from a Sequence Risk perspective. Quite the opposite, claiming at age 70 proved to work better around the historical worst-case scenarios like 1929 and 1968. Also, to effectively hedge some of the Sequence Risk, it’s best to defer Social Security (maximize the monthly benefits) and rather use a bond ladder to fund retirement between ages 62 and 70.
But of course, timing Social Security is not exactly an urgent issue for me personally. It’s best to defer the exact timing decision until we reach our earliest benefit-claiming age, likely 13 years from now. But as an academic exercise, it’s still helpful to run the math. I hope you enjoyed it, too!
Thanks for stopping by today! Please leave your comments and suggestions below! Also, check out the other parts of the series; see here for a guide to all parts so far!
All the usual disclaimers apply!
Picture Credit: Wikimedia
Thank you. I’m as concerned about longevity risk as I am about Sequence Risk. I’ll delay social security to get a bigger life long, inflation adjusted annuity.
As James states, the value in delaying social security is protection against declining mental faculties. It’s an inflation adjusted annuity, guaranteed by the federal government.
Distilling the decision down to NPV, even accounting for market returns, may be overly reductionist. Though the conclusion of that qualitative factor is the same – delay.
This decision may affect early retirees sooner than the analysis suggests. Specifically – If their parents are taking a conventional retirement. Right as the FIRE’d individual is pulling the plug, their parents are making the social security decision. Smart choices around social security (and Medicare) can substantially impact the familial burden. Shaping the early retiree’s first 1-2 decades.
I’ve encouraged my parents to delay social security. As they could afford it, I also advocated for original Medicare with Part G and Part D. Both options meant less money today, but the hedging of longevity and morbidity risk, provides peace of mind. I know they won’t starve and can always see a doctor.
Good point. If you’re 45, then your parents may be right at that point.
Looking in the other familial direction, those who anticipate the possibility of leaving a legacy may want to consider their heirs tax situation, when exploring claiming options. Someone without heirs, who plans to leave their estate to charity, might claim early, keeping the maximum amount in tax-differed accounts. Alternatively, someone whose heirs are in a higher marginal tax bracket (or may/will be) than their own might delay till 70, either drawing down for expenses or performing ROTH conversions, to minimize their heirs post-demise tax burden. The latter example is especially significant given the SECURE Act’s 10-year withdrawal timeline, which could drive many heirs into a higher marginal bracket. When exploring this issue, all tax impact issues come into play, e.g., ACA subsidies, EITC, CTC, etc. I’m sure other legacy scenarios exist.
Yes, all good points. Estate planning issues should also involve maximizing the after-tax estate, and that should factor in the step-up basis (if that survives long-term) and other issues.
Good point. And that’s the reason deferring is so attractive. Running out of money usually comes toward the end of a long retirement.
As someone who just turned 62 and am semi-retired, I agree with James above about delaying as a security measure for later in life. I planned at 14 to live to be 115 years old. If I even get close to this mark, I’d like the added security that waiting to claim can give me:-)
That’s the spirit. My wife wants me to live until 100, which is a stretch. But I will try!
Speaking of longevity risk, what if most of the “catastrophic” nature of failure was eliminated? What could we do with a SWR then.
So for example, married couple claims SS at 70, at the age of 60 buy a QLAC that starts at 85. (right now, immediateannuities.com has joint life bought at 60 of $100k yields $2200 a month). And then SS plus QLAC is expected to cover all needs of 85 to death.
Could that change the SWR of the remaining nest egg, if the definition of “failure” means we don’t get to travel anymore, rather than cat food
Then you don’t need a SWR analysis. If all of your needs are covered, then just take the VPW rule and take money out of a stock portfolio. Those withdrawals are volatile, but this is the fun money and volatility in this shouldn’t be too painful.
Thanks, that was simple. I wanted to overcomplicate it. And thank you for all your very actionable work.
Thanks! Glad this was useful!
Thanks for digging into this. I’m curious if you also take into account that social security benefits are annually adjusted? Before 62, the base amounts increase annually by the change in AWI and after 62, by the change in CPI-W.
So, if the 35 year old couple have a $10,000 annual supplemental income, this is going to be around $23,000 by age 62 by age 62 assuming 3% AWI growth. And $23,000 is real instead of nominal due to CPI-W cost of living adjustments, if that framing makes sense.
Once you’re receiving benefits, the distinction between CPI-U and CPI-W is negligible.
Before you claim, there is certainly a potential for higher growth in the expected benefits than the CPI if AWI grows at slightly above CPI But certainly not at CPI+3% as you seem to allude. Not sure where you got that estimate.
Either way, It’s best to not rely on AWI=CPI+x when you do your SWR planning because if you’re years and especially decades away you also have to worry about benefit cuts in the future.
Also: When I assume $10,000 in supplemental income in the future (say, for a 45-yo couple today as in the section above), then I mean exactly that. It’s not $23k. The $10k is already the CPI-adjusted future benefit, that might have grown by AWI. You can’t adjust that again. That would be double-counting.
I was interested in understanding the model and how $10,000/yr was used in it. It sounds like you start adding $10,000/yr in benefits starting at age 67, increasing annually from then for CPI.
The reason for the confusion is that this “benefit at age 67” is usually presented in a social security statement as a number in “today’s dollars”. As an example, if someone is 35 and pulls up their social security statement, it might read that at age 67, the benefit will be $2,000/mo. However, for the next 27 years, this $2,000/mo will increase annually even if the person never pays payroll taxes on another penny. These increases are tied to increases in national AWI. I guesstimated that AWI increases by around 3% (nominal) per year, so a $10,000/yr statement benefit will be roughly $23,000/yr by the time the 35 year old reaches 62.
It sounds your $10,000/yr is in “future dollars” starting at age 67, which is totally fine for modelling, but may not be what a reader would guess if they see a $10,000/yr benefit in their social security statement. I wasn’t sure, and wanted to clarify.
I completely understand you picked $10,000/yr as simply a round number. It is probably way too low for a couple with a $2,000,000 portfolio though. Even if their taxed earnings had been only half of that, they’d end up with a “benefit at age 67” of at least $17,000. And by the time they were 67, this $17,000 value would have more than doubled from AWI adjustments described above. The difference is more significant too.
Incorrect. All of my calculations are in today’s dollars, i.e., the $10,000 future benefit is in CPI-adjusted real dollars, not in nominal dollars as you falsely assumed.
As I said (and you wrote it too), $10k is just a round number to illustrate how much of an impact a fixed future cash flow impacts today’s SCR. You can scale up/down the $ numbers, as I stated, because the effect is approximately linear. So, if you expect $17k/y you can also factor in about 30-35% in your current FIRE budget. Just scale up the numbers in that table by 1.7x.
I wonder if this analysis could be extended to my ‘defined benefit’ index linked pension (so works in a similar way to social security). It is officially payable at age 67 but can be taken as early as age 55 with an approx. 4% pa reduction for each year early it is taken. So similar to your 62 vs 67 vs 72 analysis but with a different adjustment rate. Current plan was to use a tips ladder to ‘bridge’ 55 to 67 but perhaps it would make more sense to take the reduced pension at 55?
As James Su mentions above, one reason for delaying social security is to protect against longevity risk which is why I was leaning towards taking it late and using the TIPs ladder.
That’s exactly why I provide the Google Sheet, so people can run their own scenarios.
I suspect that with a reduction of only 4% p.a., it might be a good idea to claim early, because that’s much better than an actuarially fair program would demand. But you’d have to run the numbers to be sure,
Although I modelled the outcomes to see which was most attractive I have chosen to take my Final Salary pension at 55 and my state pension at 67 which is currently the earliest my age group will be allowed to take it ( I have modelled 68 to account for a potential future policy change).
My reasoning is that regardless of the Government Actuarial Dept’s life expectancy calculations individuals do not know where they lie on the distribution. Unlike Mr Moustache I am not confident that exercising a lot or eating healthily is a guarantee as I’ve seen too many friends fall regardless of their lifestyle or BMIs. Again, we can only influence our probabilities and hope for the best. So, I’d hate for my last thought to have been, damn I should have taken advantage of that pension money, I could have left a bigger bequest and had the reassurance of an index linked income.
There is never a guarantee. All we can do is to shift the probability distribution.
Interestingly, for my inputs taking the income sooner produced a slightly higher SWR for a 30 year time frame but for a 40 year the advantage was with deferring. This is similar to my back of a fag packet estimate that I did when I first looked at it. I don’t think I am overly worried about what happens beyond 30 years (especially as I model with a minimum of 50% residual value).
Actually ignore the last post’s comment on SWR being improved on a 40 year timescale. I mixed up my sheets, regardless of timeframe the uplift for deferring my final salary pension was insufficient to compensate for the income lost during deferral no matter the lost psychological comfort blanket. Even after 720 months.
Once again thank you Karsten.
Got it. Makes sense! 🙂
This is the type of situation I alluded to modelling above. Where you modelled as deferring, did you build a linkers bridge or just the standard ‘swr’ for the cashflows in the intervening period? (I think UK index linked govt bonds of real returns somewhere between 0.5% and 1%).
Another factor for me is that once I start taking the DB I can give up a small amount to give 100% spousal benefit rather than 35% and given my wife has very limited provision beyond full state pension this is of value although I could possibly achieve the same with life assurance. My overall plan is that on the first death the survivors income reduces only by the lost state pension (about 20% loss).
My model was simply my portfolio, SWDA.L plus bonds which are the weak spot of my model as I use the 10 year treasuries data as opposed to IGLT.L and IGTM.L, that is gilts plus GBP hedged treasuries), 50% minimum terminal value, 30 years (I tried others too), and comparing a copy of the model with Cash Flows showing the pension at 55 and the pension at 60 with appropriate quotation values, the schemes original retirement age. I didn’t model the uplifts available for deferring beyond 60 as I was unimpressed with their mean spirited calculations. An indicator of this tight fisted approach is the poor transfer value offered, which when gilts were at their lowest remained 20x annual income (i.e. assuming I could buy index linked income at 5% starting yield with dependents lump sum and pension rights on top). In other words, the benefits of the guarantees are worth more to me than I can buy with the transfer value. At least assuming that everyone involved doesn’t die in the first few days of the pension being in payment.
In short, I didn’t do a bond ladder. In the short term they are appealing but in the long term an ETF is likely to work out cheaper given the spreads involved for modest buyers. However for an early sequence of returns period a bond ladder has some virtues.
ERN, it appears you ignored the survival rates in the analyses in these simulations.
What is the chance I will live 19 years to even reach that break-even point?
Certainly much less than 100%. Adjusted for this risk the break-even time point will be much longer.
Remember that survival rates are similar to unconditioned probabilities. For example, if you make it to age 70, the probability that you survive beyond the average lifespan is pretty good.
Good point!
A healthy 62-yo female has a 77% chance of surviving 19 years. A healthy male has a 68% chance of surviving that long.
The joint probability of at least one of them surviving that long is 93%.
What about taking the bridge option a little bit further? Imagine a couple wants to retire at 50 and knows they will get $50k from SS starting at 62. Can they increase their yearly spending by buying 12 years of TIPS paying $50k/year, then switching to SS at 62? If they bridge w/ your TIPS method, that will cost a bit less than $600k. They’ve traded $50k in income for $24k in income (4% of the $600k).
You’re right, one could take this further. With 2% real yields, these kinds of strategies now look more attractive.
I’m curious about whether your bond ladder is just another form of buckets, which is my preferred way to manage sequence of return risk. For the record, I’m waiting for 70, and my wife is waiting for 67 to maximize COLA-indexed benefits for the rest of our lives.
I don’t like the bucket strategy because it doesn’t work concisely, as I pointed out in Part 55. A true bucket strategy would rebalance (occasionally) to the target weights again, while the TIPS ladder will shift out of TIPS over time.
Thus, the TIPS ladder is more akin to a glidepath, which I have shown is a good way to partially hedge against Sequence Risk.
If one has large traditional IRAs, so large that RMDs will put you in a very high tax bracket, then delaying SS and spending down the IRAs (or Roth converting) before age 70 also points to delaying…that is our situation.
Good point. But you’d still pay taxes ion the withdrawals. But I agree, during the years before you claim benefits you should be a low tax bracket and you should do some tax optimization (Roth conversions, etc.) during that time. Another reason to defer!
But what about the fact that this is trading higher RMD taxes for higher S.S. taxes later – which is better? I guess S.S. is taxed at max 85% so that’s a factor, but RMDs would almost certainly be lower, at least for a few years after starting at age 75. We will face this decision in a few years and I haven’t mathed this all out yet, but I have some time.
I still prefer to max out the income that’s only up to 85% taxed.
And again, you deal with the RMDs through higher withdrawals pre-SocSec and Roth conversions. That should take the IRA balance down.
I did a case study on this a while back:
https://earlyretirementnow.com/2019/11/08/safe-withdrawal-rate-case-study-for-becky-and-stephen/
Thank you for this analysis, Karsten. I appreciate your take on the topic.
If I’m understanding your assumptions/method correctly (and I might not be), there are two significant simplifications here:
1) For couples we’re only looking at “both people claim at age ___” strategies, rather than filing decisions where one person files earlier and the other later.
2) By using for example a 33-year horizon for the 62 year old couple, we’re essentially ignoring survivor benefits (i.e., we’re assuming that both people die at the same time, which of course doesn’t usually happen).
Is that correct?
Correct, I assume both survive that long, which is the best possible case from one perspective, but the “worst possible case” from a retirement safety perspective.
One can obviously start with some very particular cases, i.e., male spouse is much older and has much higher benefits. Then it’s likely best for the female to claim earlier and take over the husband’s benefits later in retirement. I certainly recommend for everyone to model this using their idiosyncratic parameters.
But in this blog post I wanted to keep everything as simple and generic as possible.
Thank you. Just wanted to confirm my understanding of the work you are doing.
I wonder if Spousal Benefits should factor into the 62 or 70 argument. You need a lot more saved for your spouse having taken SS at 62 vice 70.
Yes. See my reply to Mike Piper.
I don’t think that you need a lot saved to bridge the 8 years. $275k out of a $2m portfolio should be manageable.
Thanks for taking on this subject, ERN! I have taken away some observations from this analysis that I’m hoping I understand correctly:
1. Since the 45-year-old retiree’s decision on when to take social security is just a mental commitment, it is revokable up until the age of 62. At which point, the decision to take SS or delay becomes practicably irrevocable. My take-away from the 45-year-old retiree analysis was to say that the benefits (at either 62 or 70) are so far away in the distant future (relative to the current 45-year-old) that the benefits don’t make much of a difference from an NPV or SWR perspective. Sequence of returns will be just as present at 45 as it will be at 62, but the decision’s impact will be greater from the 62-year-old’s perspective since SS cash flows fall in the period where Sequence of Returns matters most. In other words, wait until then to make a decision since the 45-year-old will eventually become the 62-year-old retiree albeit with different portfolio and SS benefit numbers depending on how those last 17 years played out in the markets and with the actual inflation experience.
2. At 62, each year SS is delayed comes with the advantage of increased inflation-adjusted benefits but at the cost of not having locked in a smaller amount earlier for an additional year. I took away from this analysis that waiting until 70 to take SS can help hedge against Sequence of Returns in some of the worst-case scenarios where severe drawdowns are long to recover. However, it will come at the cost of better portfolio outcomes in many other market scenarios. Did I understand that correctly?
3. Since SS can help hedge against longevity risk and inflation risk, I can see why it is beneficial to wait, but once you introduce mortality, and the possibility that one spouse might pass early, leaving a much reduced SS benefit for the surviving spouse, things can get complicated quickly. How can one determine the trade-off of higher inflation-adjusted SS benefits that don’t pass to a spouse with those of preserving a portfolio that can pass to the spouse? For one planning an early retirement, I’m assuming one has to conservatively plan based around a portfolio large enough to not only sustain longevity risk and inflation risk, but also strong enough to not have to rely on SS benefits as they can unexpectedly disappear as one spouse passes. As such, should SS be viewed as only supplemental to an early retiree and not use it to try and bolster a life-long SWR? In other words, shouldn’t the portfolio be able to withstand worst-case-scenarios on its own regardless of the timing of SS benefit claims? That way, the SS decision doesn’t have to be a critical one but one where a couple can decide along a spectrum of focusing on marginally protecting the current portfolio vs. marginally protecting higher inflation-adjusted spending for anticipated longer lives.
1: No, I don’t see it that way. As I pointed out, claiming at 62, 67, or 70 doesn’t make much of a difference. But ignoring benefits completely, especially for folks with sizable benefits, easily $30-40k+ p.a., if you paid into the program at the annual phaseout max, you’d leave too much money on the table. Factor in the future benefits to get a better SWR.
2: correct
3: That would be a more specific and involved SWR analysis. For example, I could do the analysis with my personal situation: I claim at 70 and my wife (8.5 years younger) at 62. My wife will eventually get my benefits when I pass, which is likely due to the age difference the lower male life expectancy. The analysis would still hold: After I pass, my wife will get my large SocSec payment and forego her small benefit. And it’s still the best strategy for me to defer from 1) a NPV perspective and 2) SoRR
SWR, as powerful a tool as it is, ignores the primary risks in retirement: dying too soon, or living too long.
Every day my wife (65) and I (61) wake up, our longevity risk increases. Social Security is the best long longevity tool out there, and only gets better every day you delay.
I propose using a generous retirement horizon, not just the life expectancy. 95 for the younger spouse is probably a good start. In that sense, my SWR doesn’t ignore, but very clearly hedges against longevity.
I’d note that in many – but not all – of ERN’s analyses, the fact that the portfolio is not completely depleted *does* effectively cover longevity risk. Certainly when it’s only 50% depletion, but probably even the cases with 75% depletion (25% remainder).
I do agree with Kelly in that, particularly for those who do not defer taking SS, planing for complete depletion at 95 seems to me to be ignoring longevity risk, that said, the impact for a 45 year old retiree is obviously less than for a 62 year old retiree.
Agree with that!
This is probably a side issue as it’s maybe not much use in the US, but I suspect that if you modelled a scenario where a someone retires at 55, and has a choice to claim a significant size defined benefit pension either immediately at a reduced rate, or 10 years later at a higher rate, you would then get different results, finding that taking the pension earlier does indeed hedge significantly against SOR risk.
This is the situation I am in in the UK and after modelling it on a few different tools that do historical modelling I found it made a significant difference.
I other words I guess there may be a sweet spot there which is maybe not relevant to most people.
It all depends on the inputs. But the same result would likely hold if deferring a pension between age 55 and 65. 10 years isn’t long enough to hedge against the worst-case scenarios.
Claiming early works best in the world where you have strong return early in retirement and a short life.
Claiming later works best if you have poor returns for the first 15-20 years and a long life.
ERN,
Could you please confirm what you mean by ‘best’? I think you might mean total payments received and that higher is better. However, is it not also the case that best could also be assessed versus having ‘enough’. That is, is there really any point in delaying SS much beyond the point when it would pay you [combined with your portfolio drawdown] enough? I suspect this is more of a diminishing utility argument. Thanks.
If the monthly budget is the same and it’s enough and you don’t run out of money, then both versions are the same. What I mean by “best” in that case is what leaves the most money at the end (bequest).
Thanks
Karsten,
As always, your keen analysis provides penetrating insight for all to enjoy and make better informed decisions. Bravo!
For the December 1968 cohort analysis, you labeled the graph “Portfolio values for a hypothetical 12/1957 retiree,” — implying a retirement 9 years prior. Typo? Shouldn’t the age 62 decision point be December 1968? Am I misinterpreting this or misunderstand the analysis?
Good catch. I corrected the typo! Thanks!
Not specific to social security, but I recently read Mark Spitznagel’s book “Safe Haven” (https://www.goodreads.com/book/show/35541721-safe-haven), and thought it was interesting and potentially up your alley (historical returns, options, and indirectly SWRs).
The TL;DR is he belives the finance industry is focused too much on “average” returns, instead of focusing on the median of all potential outcomes. He does a simpler version of your historical stock analysis radomized via 120-sided dice aligned to annual historical returns. Through a St Petersburg-paradox point of view, he basically argues that your median expected outcome would be improved by sacrificing a small amount of principle (low single digit percentages) in downside protection though options. Basically because a 50% loss requires a 100% return, major downside protection deserves more consideration. This is apparently what his hedge fund focuses on, with religious protection of 20% drawdowns. This creates a drag on returns most years, except when there’s a market crash and he comes out ahead. The discipline required to stick with the trade, and the fact that it only jumps to the top when there’s a crash, apparently makes it a hard sell and people basically get in and out at exactly the wrong time. And for a “normal”, retail investor he doesn’t really have a paint-by-numbers solution to implement it (i.e., indentifying ‘cost effective’ insurance).
Anyway, not sure if you’ve come across the book, the strategy, etc. The author’s writing style is a little too much, “I need to prove I’m smart”, in my opinion, but content wise it made me think of your blog, this series specifically, and whether there’s a “there” there from a sequence risk perspective that a normal person could implement.
Thanks for continuing to update the series!
Chris
Thanks for the link.
My only “disagreement” is that everyone in the institutional investment world, where I used to work, thinks exactly like that: You can’t afford large drawdown. The methods for hedging against such an event may be different. Buying a put, 20% OTM as Spitznagel proposes is one way. When I used to work at BNY Mellon we had other, more tactical allocation tools to hedge against that. But it’s the same idea.
But for the average retail investor, maybe spending ~1.5% p.a. of the equity portfolio to hedge against that downside risk may be a good start.
Slightly off-topic. How do you feel about the typical “plain vanilla” tactical asset allocation? i.e. setting up typical 60/40 portfolio, preferably with sub allocations (value, growth, large cap, emerging, developed, REIT, etc.) and selling when price dips below 300 EMA and buying back when rising above same? I’ve never quite bitten the bullet on this but thought that it might not be the worst idea to shift 50% of one’s portfolio to this strategy.
It’s on my to=do list: I’d like to check how a momentum strategy performed historically.
Regarding all the other stuff:
Value/growth, Small-cap: likely those styles no longer offfer much excess return
Emerging Markets, REITs: Can’t be backtested long enough. I also don’t think that EM is worth the risk right now.
“ But for the average retail investor, maybe spending ~1.5% p.a. of the equity portfolio to hedge against that downside risk may be a good start.”
Can you explain this, ERN? I’m confused to whom you’re suggesting this as advice. It’s quite opposite your “be in stocks 100% until 2-5 years from your retirement” advice for those not working. And 1.5% is an awfully big annual chunk for a retiree with anything remotely like a 60-40 portfolio; would surely reduce the SWR would it not?
Best I can figure/guess is that you mean this *might* be a good idea for a retiree drawing from their portfolio that is basically 100% invested in the stock market? But even that doesn’t feel right. Can you clarify?
Thx.
If you can hedge out the risk of a 50% drop you may raise the expected equity return by enough to justify a 1.5% annual drag on equity returns.
Thanks very much for the shout-out Karsten!
It’s fantastic to see your analysis of social security timing in the context of safe withdrawal rates. No better person for the job.
I agree 100% with the comment “Contrary to popular belief and my initial intuition, claiming benefits early will not necessarily hedge against Sequence Risk.” These results really surprise me as well. The inverse correlation in the “Social Security Timing for a 62-year-old” plot, especially after age 50, is really surprising.
Looking at “Social Security Timing Case Study: September 1929”, I like how you show the time history of the difference. So for that example you don’t end up ahead by deferring benefits until 317 months after age 62 (26.41666 years). But I believe the current SSA actuarial tables indicate a 62 year old male will live another 19 years and a female will live another 22 years. So in that worst case 1929 scenario, they would have been better off not deferring, right? At least if we assume actuarial lifespans.
Ah, and I see your comment to Marginal Money above. The higher joint probability is interesting.
Looks like for the 1968 example, the crossover point of 19 years exactly matches the (shorter) male life span. So I’m guessing for other even less extreme years examples, the crossover point is even earlier? And thus deferring makes even more sense?
These questions make me think it would be really neat to combine this Safe Withdrawal analysis you’ve done using historical returns with actuarial lifespan values, a la the Open Social Security tool and your work in Part 56.
Thanks again for this fantastic analysis.
Thanks for weighing in, Corwin!
I don’t take the SSA tables too seriously for two reasons. 1) The joint survival probabilities are much higher (as pointed out in the other comment) and 2) healthy retirees will likely live longer than the average US citizen. So, the ERN household will have a much longer than 19 years life expectancy at age 62.
And I agree, a possible extension of this work might be to see how the joint claiming strategy out of the opensocialsecurity.com toolbox would fare in a Sequence Risk world.
“But I believe the current SSA actuarial tables indicate a 62 year old male will live another 19 years and a female will live another 22 years. So in that worst case 1929 scenario, they would have been better off not deferring, right? At least if we assume actuarial lifespans.”
If one assumes actuarial lifespans, then by definition one is ignoring longevity risk.
In real life, the only (non-reckless) people who can afford to do this are people who are spending (i.e. drawing from their nest egg) *far* less than they otherwise could. In that sense and that sense alone, it can be an interesting intellectual exercise to decide which approach will leave a greater legacy amount – on average, at the median, or a certain percentage of the possible outcomes – given SORR.
But for anyone else who *is* worried about longevity risk, then in addition to the benefits for inflation risk, it should be a no-brainer (save perhaps corner cases like Karstan describes with a much older spouse with a much higher benefit) that the correct answer is to defer taking SS as long as possible, precisely because it is at least actuarially fair. Actuarially fair means you lose when you die young, but you gain when you outlive the actuarial “odds”. Which is the essence of worrying about longevity risk.
I always force (I mean encourage) my husband to hiking, biking and eat a healthy diet. Sometimes he asks, why he must eat so much vegetables. And I answer, that I do not like my widow‘s pension tooo early.
Dear Karsten, thank you for you work. Now I have even better arguments.
Beste Grüße aus Österreich
Hanna
Good point. You two got your incentives aligned. Similar to my wife and me!
Viele Gruesse zuruck nach Oesterreich. Wir waren im Sommer kurz im Tannheimer Tal zu Besuch. Sehr nett da!
Karsten,
I’m confused by something. Even given the consistent assumption above throughout that the goal is for the consumption to occur only through age 95, how does it make sense that “the right answer” for an early retiree might be to claim at 62 given that the answer for someone who is 62 is to defer until 70 – and as you do point out no one actually has to make the decision until they turn 62!
I don’t get how there is even a small benefit to “claiming early” when you (for simulation purposes) have to decide 17 years earlier at age 45, given that the results show plainly that when you make the decision at 62, the correct SCR answer is to defer. what’s up with that?
17 years into retirement will put you at 1946 for the 1929 cohort and 1985 for the 1968 cohort. The defer vs. early claim decision is skewed for those cohorts because age 62 falls into some of the strong bull markets. Hence the tendency to claim early.
But once you’re 62 and you look at the 1929 and 1968 cohorts again, the Sequence Risk goes the other way around!
Ah, thx, I get it now. So basically, the take early vs defer decision is correlated with current withdrawal amount being a higher percentage of *current* asset value – which is what happens to those using the calculated SWR after the market tanks and hasn’t come close to fully recovering. Which is also correlated with a higher calculated variable CAPE-based SWR, as you’ve done elsewhere, and is correlated with the % the market is off its inflation-adjusted high then. Which makes sense.
Any chance you will do this calculation in (any of) those terms?
The correlation between the Deferreal decision and the CAPE and the equity drawdown is pretty low. Notice in 1932, at the bottom of the Great Depression bear market, the advantage of a deferral for a 62yo would have been even larger than in 1929.
Thx again. So in my top-of-my-head correlations, I got two of my 3 guesses wrong.
Am I correct – inferring from your lack of comment on low correlation – that the withdrawal percentage of current portfolio *does* correlate with this decision?
I.e. when given a fixed constant (in real dollars) SWR your consumption of your portfolio is a sufficiently high percentage, that’s when it may make sense (at 62, since the optionality can’t be exercised before then even if one wanted to) to claim early, minimizing further portfolio depletion before coming strong returns, rather than defer?
Not sure if I understood the question, but I can see that when the market is now very beaten down and you have 1932-like or 2009-like valuations in stocks you may consider claiming early because from now going forward you’ll likely make more with the stock market than maybe your small real return from deferring SocSec until age 70.
“Most of the time, the TIPS ladder would have underperformed your 75/25 portfolio. But of course, that’s the nature of this Sequence Risk hedge: you do better in the worst-case scenarios, but you lose a little bit and leave a slightly less spectacular inheritance to your heirs when the market rallies during your first eight years of retirement.”
It would be nice to have this strategy (TIPS ladder until SS) evaluated for those non-pessimistic cases as well to see how much it really costs. Does it raise SWR over the usual 70/30 or something strategy?
With 2% real returns you can currently get a SWR of 4.43% assuming depletion over 30y. Not bad. But much lower for longer horizons.
And you get zero upside. So, the worst case scenario for the 70/30 over 30y looks slightly worse. But you have tremendous upside if you like to leave a bequest.
(=PMT(0.02/12,360,-1,0,1)*12)
…plus a whole lot more protection against longevity risk (living beyond the 30 years).
Exactly! Probably you want to use only 80-90% of the portfolio in the TIPS ladder and keep the rest as a reserve in a stock portfolio to hedge against longevity. But then you push the TIPS ladder SWR lower.
oh great, now besides using a 3% SWR, we have to defer SS until we’re dead for 10 yrs+
Brilliant
Oh, you misread that. The SWR is MINUS 2% and you have to defer SocSec until age 90, otherwise your portfolio blows up for sure. Back to work, buddy! Someone has to work and pay SocSec and Medicare taxes while I’m retired and getting ready for another cruise vacation.
This was an awesome analysis and one I’ve been hoping you would do for a long time. Thank you!
Thanks for the feedback! Glad this was useful!