“How much income can I count on from my retirement savings?” With traditional pensions, defined benefit plans, becoming as rare as a good plumber, most of us are stuck figuring this out on our own.
The math here is brutal. You don’t know how long you’ll live. You don’t know what markets will do. And the thing that would solve this problem, an inflation-protected annuity, basically doesn’t exist. You can search for one, but good luck finding something that makes sense.
(Believe it or not, in Italy it still exists because the State pension is indeed an inflation-protected annuity. But it has been phased out, if you want to see it this way, and I doubt many will still enjoy it if they are going to retire after 2045)
The retirement spending literature is massive, but most of it obsesses over one thing: not running out of money. That’s obviously important. But there’s another problem that doesn’t get enough attention: spending so little that you die with millions in the bank and a lifetime of vacations you never took. Both extremes are failures, just different kinds.
I found a paper that tackles both problems. Here’s what it gives you: a simple portfolio structure (equity index fund plus a ladder of inflation-protected bonds held to maturity) and a spending strategy that helps you actually use your money without the constant fear of going broke.
The paper includes this beautiful table that recaps past retirement studies and what they cover:
Here’s the brutal choice: spend a fixed amount (adjusted for inflation) every year and (most likely) die rich, or spend what your portfolio can actually support and risk some painful budget cuts when markets tank. (Well, the problem is not dying rich per se, but not enjoying the fruit of your work and the sacrifices made along the way).
Neither option is great. The fixed withdrawal rate sounds stable, but you’re basically hoarding money for a catastrophe that will never come. The variable approach lets you spend more when times are good, but “fringe cases” is a polite way of saying “sometimes you’ll need to slash your budget by 30% or more.”
If you look closer, the fixed withdrawal method can be partially fixed in practice. If you recalculate your safe withdrawal rate every year, you can (slightly at first) increase your spending when markets cooperate or floor it at last year’s level when they don’t. This keeps your portfolio from that fixed-horizon assumption, your budget won’t spike dramatically, and you’ll still die with money left over…just not as much as the traditional backtests suggest.
But even this “improved” version misses the point for most retirees over 60. Your spending doesn’t follow a straight line: it’s U-shaped. You want to blow money right after retirement while you can still hike Machu Picchu without your knees staging a revolt. Then spending drops during the “Netflix ’cause I cannot even bother to drive to the movies” years. Finally, it spikes again when you need round-the-clock care.
That last phase? It has a tail fatter than the belly of some of your friends from high school now in their 50s. Which creates an impossible question: would you rather toast that Japan trip while you’re mobile, knowing you might spend your final years in diapers you can’t afford proper care for? Or save obsessively for medical bills and dementia care that your brain won’t even be present enough to appreciate?
The Funded Ratio
After spending years managing other people‘s defined benefit plans, I’ve become obsessed with funded ratios. Here’s how it works for retirement: split your spending into two buckets: “needs” (the stuff you can’t cut) and “wants” (everything else). Then calculate a funded ratio for each.
Your “wants” aren’t some aspirational Pinterest board because…sky is the limit, if you think in those terms. They’re just a projection of what you can actually afford after covering needs: your total savings (including the present value of any pension) minus the cost of your needs.
The denominator in each ratio is the present value of your planned spending, adjusted for survival probability. I love using survival probability because it kills the “how long will I live?” guessing game. This comes from spending too much time reading mortality assumptions for pension funds, where “improvement”, I kid you not, is corporate-speak for “beneficiaries are dying faster than we budgeted for, which is great for our balance sheet.”
The numerator for your needs ratio is simple: it is the part of the current portfolio value plus the present value of any future income (Social Security, pension, part-time work), adjusted for survival probability, that covers for the related liabilities. You want this ratio to be 1 all the time, neither more nor less.
The numerator for wants is…what’s left in your portfolio after covering for the needs.
Then comes the spending rule: your annual withdrawal for each goal increases or decreases based on how far your funded ratio sits above or below 1.0.
This paper sort of follows the same logic, allocating a ladder of TIPS to cover the needs and a risky portfolio to cover the wants.
The Perfect Withdrawal Rate
The Perfect Withdrawal Rate is the exact amount you could withdraw each year to hit zero dollars on the day you die 30 years later, given the specific path your portfolio actually took.
Of course, nobody has that kind of foresight. But the graph is still useful, mainly because it provides ammunition for literally any argument you want to make.
The risk-averse camp can point to that spike around 1930 and declare that 100% stocks in retirement is financial suicide. Another risk-averse person can look at the same data and argue the 4% rule itself is too aggressive because we’re cherry-picking one of the best-performing markets in history (the U.S.). Meanwhile, the aggressive investor sees proof that you’d need spectacularly bad luck to actually run out of money following the 4% rule, so just VWCE and chill your way through retirement.
Then the conservative side fires back: we just lived through a period of extreme stock valuations, so the future will probably be bleaker than the past.
“Unless one improbably spends at exactly the PWR, it is impossible for there to be any withdrawal scheme or set of “guardrail” rules for a volatile portfolio that reliably delivers income that is free of (possibly large) reductions at unpredictable times, without also excessively preserving funds that one could otherwise spend in one’s lifetime.“
This perfectly captures the retirement spending dilemma. You’re managing a volatile portfolio, so pick your poison: spend conservatively and leave your kids a massive inheritance, or live with a budget that swings wildly from year to year.
If you’re thinking “whatever, the money goes to my kids, no big deal”, reframe that. By the time they inherit, they’re probably in their 50s or 60s. That money would have been transformative at 25 when they were starting their careers. Now it’s just nice. And if it eventually trickles down to your grandkids? Sure, it’s a generous gift, but split between multiple grandchildren, it’s not buying anyone’s first home. Though maybe demographics will work in their favor, who knows.
This is why this blog obsesses over maximizing return per unit of volatility. In retirement, or even before when you’re funding college, apartments, cars, or gap years, you’re literally eating your Sharpe ratio. A higher Sharpe ratio means a more stable budget, which means you can actually plan your spending instead of gambling on what the market might offer you at that specific time.
Most financial planners do this backwards: they plan your spending first, then build a portfolio around it. This approach has major problems, starting with the fact that human brains are poor at understanding the power of compounding. It forces savers into medium-term, unambitious thinking because there’s no real framework for measuring risk-return tradeoffs between volatile assets and changing priorities/goals. You get a false choice: bonds with false-certainty (hello, inflation erosion) or the wild swings of stocks.
And by “ambitious goals,” I don’t mean a 911 instead of a Prius. Check the price difference between a top-tier university and a third-tier one. Or whatever your plan is for your kids, like getting them on the property ladder before they’re 40.
Annually Recalculated Virtual Annuity
“The more one wishes to withdraw from a volatile portfolio, the more volatile the income one must be prepared to accept.“
“A suitably constructed and managed ARVA portfolio can be adapted to the individual retiree’s possibly time-varying preferences for the trade-offs between lower but stable income vs. higher expected but variable income. It can also adapt to changes in the retiree’s consumption needs and beliefs about longevity.“
I’m not sure “framework” is the right word for the ARVA methodology when it comes to designing a retirement spending strategy. Still, the description above does a solid job highlighting its strengths. If you want all the technical details, go read the paper, it’s worth your time.
Here’s what stood out to me after digging into the research:
One of the most interesting takeaways (and yes, it fits nicely with my own confirmation bias) is this: the best “risky portfolio” designed for the “wants” turns out to have the highest compound annual growth rate (CAGR). Over time, this portfolio delivers the most consistently high median income. And most of the variability in withdrawals shows up in the upper percentiles, not in the lower.
This is not some blanket endorsement for going 100% into stocks. The reason this strategy works, at least in the context of the paper, is because they only tested two asset classes and didn’t allow for leverage. So, while the results are compelling, they’re not universal. Context matters.
Let’s be clear: CAGR isn’t the same as Sharpe ratio. Yes, volatility impacts CAGR, but not nearly as much as it does the Sharpe ratio.
If returns were truly independent and identically distributed (i.i.d.), you’d expect the portfolio with the highest Sharpe ratio to also deliver the highest withdrawals. But, as Cederburg’s paper shows, reality is different. Historical returns for a 100% stock portfolio have actually performed better than what you’d predict under the i.i.d. assumption.
Now, here’s where I get a bit skeptical about the whole “stocks always mean revert” narrative, the reason why stocks perform better in real life than in the lab.
Stocks mean revert, eventually. They did in the past. But the data set isn’t exactly massive (as a reminder, you need 10-year non-overlapping periods to test this), and there are two things that keep nagging at me: First, over the period analyzed, valuations didn’t just bounce around, they climbed higher and higher with each cycle. Did that shorten the mean reversion period? What happens if we hit a plateau and the cycle stretches out?
Second, it’s easy to trust in mean reversion when you’re running a backtest. Real life is a different story, especially when a downturn drags on for years. It takes real conviction to believe that “this time, stocks will bounce back,” particularly if you lived through a decade-long slog. Maybe it’s the boomer in me, but I still remember 2009–2011. I was optimistic, but I kept thinking valuations were headed for single digits, that another leg down was coming. It didn’t. But that uncertainty? It’s always lurking.
Also note that this strategy works only if employed for the “wants”. It cannot be used for the whole portfolio because the volatility of the withdrawals would be massive. It works because we have the TIPS ladder to cover for the “needs”.
The following chart really drives home what it’s like to invest in stocks…or any volatile asset, for that matter. If you focus on the average outcome, the math looks simple: you’re better off going all-in with a 100% stock portfolio. The long-term numbers just work.
But if you’re the type who loses sleep over worst-case scenarios? Then stocks might not be for you because your results will be worse.
“A behavioral approach is to distinguish between priority vs discretionary spending. Set the target for priority income to be relatively low, and set the fraction of TIPS to secure the priority income with a high level of confidence.“
To be honest, before reading the paper I was expecting the strategy would deliver a higher lower bound. I am looking at fringe cases, sure, but those scenarios deliver something lower than the 4% rule.
The 50% stocks / 50% TIPS portfolio provides a 8.0% withdrawal rate in the best 20% of case and 5% in the worst 20%. That’s a big swing in budgeting.
Maybe my perception is warped by how much I currently spend in rent but basically even with this strategy I should keep my fixed cost at…30%?…of my standard budget. I guess reading these papers is good because the results are not intuitive, even for someone like me.
Or again, I am fooling myself because of my current circumstances. I pay that rent (also) because it takes me 15 minutes to walk to my office. In 10 months “my second mortgage”, my son nursery, will be finally over. That’s (almost) 2k/month salary increase! My brain cannot really process it after 6 years in the foxhole.
While writing this post, I read this and my feeling about retirement changed again. Maybe the solution is to take the leap earlier rather than later. Maybe the solution is stop fucking listening to my pal Mr RIP and his fretting about his future (YOU WILL BE FINE!).
Anyway, back to the paper.
“The variation [in spending] can be located in the mental account of discretionary income. Instead of setting a high income target where reductions will be painful, designate the discretionary income as a bonus. Frame the plan as having a modest income target, with a smaller likelihood of reductions and the pleasure of frequent bonuses. As Jonathan Guyton points out, retirees have typically spent their working decades learning to adapt to income variation due to life events, set-backs and bonuses (Benz and Ptak 2020)“
This part is quite funny because it equally makes a lot of sense for Americans and not a lot for Europeans. The typical Europoor almost never experiences income variations, neither on the negative or positive sides.
I am not even sure that this framing would allow retirees to spend more of that ‘discretionary income’ in the first part of their retirement, if they know how low they might end up. Most likely, they will stash it away.
And now…just a few final considerations on the paper.
The beauty of ARVA math is that it lets retirees see, in real time, how taking a bigger withdrawal today impacts their future income (or the distribution of). It’s not just about running the numbers but helping people make informed decisions on whether to go beyond their budget.
And if your budget is bigger than what you actually need right now? You’ve got options. You can leave the surplus in risky assets and let it grow, or you can use it to beef up your TIPS ladder, locking in more risk-free income for the future. The ARVA shows clearly the trade-off of the choice.
ARVA also helps tackle longevity risk. Each year, you can update your planning horizon based on how likely you are to live to a certain age. Stick with the same percentile target, and you’ll find your planning horizon gets longer and longer as you age (if this sounds counterintuitive, ask ChatGPT why ;)), meaning your expected withdrawals will shrink. But you can offset that by adjusting your withdrawal targets upward over time, keeping your income more stable as you go.
It’s a flexible, transparent way to manage retirement spending giving you control on all the relevant variables (asset allocation, mortality, probability of failure, consumption profiles), no matter how the future unfolds.
The biggest issue with all the data presented so far is that the range of 10-year real interest rates in the available historical data is between −1.19% and 7.66%. It is a very wide range, making the strategy viable only a limited number of times, because the TIPS yield affects the location of the distribution (the minimum withdrawals).
When I say “limited,” I mean it. The recent past has been dominated by stretches where yields have barely cracked 1%. Sure, maybe real rates will rise in the future, but I wouldn’t stake my retirement on that bet. If anything, it feels more likely that governments could stop issuing inflation-linked bonds altogether, or even start tweaking inflation indexes to keep reported numbers artificially low.
But that’s a whole other conversation, and probably a topic for a future post.
What I am reading now:
Follow me on Bluesky @nprotasoni.bsky.social
0 Comments