# The psychology of spending: why spending it is the hard part Here is the best-kept secret of decumulation, the one no simulator shows and the one new retirees discover with a shock. After twenty years of fearing you were not saving enough, the problem flips: **most affluent retirees cannot bring themselves to spend their money**. Every study points the same way. A large share of portfolios **keep growing** through retirement. Actual withdrawals run well below what the plans allow. And the gap between the life you could fund and the life you live, the "retirement consumption gap", runs into the tens of percent. This is not prudence. It is a plan failure as real as running out of money, paid in trips never taken, in generosity postponed until the estate is settled, in working years given up for capital that will die intact ([[one-more-year]]). This chapter takes that side first, then its mirror image: panic in a crash, the other psychological failure that wrecks plans. Then come the biases specific to living off a portfolio, and the toolbox. The good news is that the psychology of spending can be **equipped**, like everything else, with written rules, automatic transfers, permission in numbers. Half the instruments in this book, from the buffer to the guardrails, are psychological prosthetics dressed up as finance. ::: cle The reversal to make Accumulation rewarded an identity: the disciplined saver who never touches the principal. Decumulation demands the exact opposite, methodically spending what took twenty years to make sacred. An identity does not flip on command. The retiree who does not dare spend is not irrational. He is applying, to the letter, the reflexes that made him rich, in a game where they have turned false. Willpower is not the answer. **Rewire** the environment instead: turn withdrawals into a paycheck, permission into a number, fear into a rule. ::: ## The evidence: money that never gets spent Start with the numbers, because the pattern is large and well documented. The research on retirees who hold financial assets (Blanchett, Finke, the Employee Benefit Research Institute, the work on actual withdrawals) finds the same thing from every angle. Most affluent retirees spend **less** than their current income alone (pensions, dividends, interest), which puts their effective withdrawal rate near zero. Households with substantial wealth typically consume 2 to 3% where their plans would support 4 to 5. And more than a third of them watch their wealth **grow** through retirement instead of drawing it down. The simulation had said so from day one. Alongside the failure probability, most simulators produce the distribution of terminal wealth, the bequest, whatever is left at the end of the horizon. For a prudent plan, that distribution piles up far **above** the starting capital ([[reading-a-fan-chart]]). You die rich in the great majority of futures. The simulator knows it; the retiree, in the gut, does not. The cost of this failure is lopsided and quiet. It produces no red statement and no alert, just a string of noes to a life you could afford: the "unreasonable" trip, the help to your children at the moment it would change their lives rather than after you are gone ([[spending-in-retirement]]). And upstream of all that, extra years of work to build capital that will never be spent ([[one-more-year]], the same failure at a different age). ## Why it is so hard: four mechanisms **1. Unbounded fear.** Ruin, as you imagine it, has no shape. It is infinite, undated, irreversible. Against a danger with no bounds, no amount of spending ever feels safe. That is not a figure of speech. In de Berker and colleagues (*Nature Communications*, 2016), the most stressed subjects, by cortisol and by every physiological measure, were not the ones **certain** to receive an electric shock. They were the ones with a fifty-fifty chance of receiving one. What weighs on you is irreducible uncertainty, not the bad outcome itself. This is where the technical machinery of this book is already the therapy: a failure probability **puts bounds on** the fear ([[failure-probability]]), one number, dated failure modes, warning lights that fire ten years ahead ([[when-to-worry]]). You can only negotiate with a risk that has a face. **2. The saver identity.** Twenty years of dopamine from a rising balance, of pride in the savings rate, of never touching the principal. The first **sell** order placed just to live feels like a transgression. Plenty of new retirees put it off for months, living on cash "for now". The bill arrives later, and it has been measured: Kivetz and Keinan (2006) show that regret over the sensible choice, the evening out you skipped, the pleasure deferred, **grows** with time, while regret over indulgence fades. Twenty years on, nobody regrets the trip. The defense is a reframe. A withdrawal does not eat into your capital. It is the **paycheck** the capital was built to pay. Automate it and the reframe becomes literally true (see the toolbox). **3. Mental accounting.** Principal is sacred, income is spendable. That is not a metaphor but a tested model. Shefrin and Thaler's behavioral life-cycle hypothesis (1988) splits wealth into three mental accounts that do not substitute for one another, current income, current assets and future income, with a propensity to spend that collapses as you move down the list. Hence the irrational preference for dividends ([[defensive-assets]]): Hartzmark and Solomon (*Journal of Finance*, 2019) establish that individuals, funds and institutions treat dividends and capital gains as two separate quantities, never registering that the dividend mechanically knocks the price down by its own amount. A dividend gets spent where a sale eats into something. Hence, too, the soothing appeal of buckets ([[the-bucket-strategy]]). You will not root mental accounting out. Put it **to work** for the plan instead: a checking account fed by a monthly transfer is a mental account labeled spendable, and that is exactly what you want. **4. Nobody gives you permission.** An employee has an employer who caps his spending, and it is called a paycheck. Someone living off a portfolio has nobody. Every dollar spent is an open decision, up for renegotiation, faintly guilty. The research on retiree well-being turns up one striking result, and its cleanest measurement belongs to Blanchett and Finke, whose title says it all ("Guaranteed Income: A License to Spend", 2021). At equal wealth, a dollar held as **guaranteed income**, a pension or an annuity, funds roughly **twice** the spending of a dollar of invested savings ([[annuities-and-safety-first]]). Their survey adds the subjective version of the same fact: 60% of respondents would feel more comfortable taking a vacation with an extra $10,000 of lifetime annuity income than with an extra $140,000 of savings. Permission to spend is therefore literally half the value of an annuity, and half the value of any pension that pays for life. Permission can be manufactured. That is what rules are for. ## The mirror failure: panic in a crash The other failure is symmetric. The same retiree who will not spend on a cruise will sell **everything** at the bottom. Capitulating in March 2009 or March 2020, the "let's save what is left" moment, is the most expensive decision of an investing lifetime (20 to 40% of ending wealth, [[cash-buffer]]). It gets made in a few days, driven by the same unbounded fear in its acute form. Someone has counted them. Across 653,455 brokerage accounts, Elkind, Kaminski, Lo, Siah and Wong (2022) isolate the "freakouts", the liquidations of more than 90% of an equity position inside a month, show that they cluster on violent drops, and show that everything then turns on the ability to get back in, which usually fails. One detail of their profile should give this book's reader pause: the people who panic most often are men over 45, married, with dependents, who rate themselves excellent judges of markets. The anti-panic file is already assembled in these pages, so here it is as a list: the **written rule**, which decides in the calm what you will do in the heat ([[choosing-your-strategy]], and a crash is never a reason to change the rule); the visible **buffer**, which guarantees the next 24 to 30 months of living expenses ([[cash-buffer]], the structural sedative); the **thresholds** that separate noise from signal ([[when-to-worry]], and while the warning light is green a decline is a non-event the simulation **already** contained, your fan chart held that path, [[reading-a-fan-chart]]); and information **hygiene**, since the portfolio gets looked at at the annual review and when a threshold trips, not every morning, and the financial press is a machine for manufacturing procyclical decisions. ::: science The retiree's biases, named A useful catalog, because a bias you can name loses half its power. - **Loss aversion** (Kahneman and Tversky: a loss weighs about twice what an equal gain does). In the withdrawal phase every red statement counts double, which is the argument for reading fewer of them. The experiment is unambiguous: Thaler, Tversky, Kahneman and Schwartz (1997) showed that investors who check their results more often take less risk, and so earn less. That is **myopic** loss aversion, and how often you look is the one ingredient you can set. - **Anchoring on the peak.** "I am down 20% from 2021." The all-time high becomes the reference point, and everything below it becomes a loss. But the plan was never calibrated on the high ([[how-much-you-need]]): the legitimate anchor is the plan's target, not the peak. - **Recency bias.** Three bad years "prove" a permanently bad world, and a bubble proves the opposite. The defense is the weight of the evidence, several models at once, which already contains both worlds ([[historical-vs-parametric]]). - **The illusion of control** turns the portfolio into a permanent tinkering project taken as a sedative: you change something in order to be doing something. Every small adjustment outside the rule is a statistical cost. The written plan exists to make inaction comfortable. - **Social survivorship bias.** You measure your plan against the bubble stories you hear at dinner, the neighbor in crypto, the colleague in leveraged real estate. Your plan has a goal those stories do not: lasting 45 years in the worst worlds ([[simulator-traps]]). ::: ## Adapting to a cut: what we know, and on what conditions Every flexible rule rests on a psychological bet nobody spells out: the household will get used to living on less. A 10% cut ([[guyton-klinger]]), a 2.5% glide each year ([[floor-and-ceiling]]), an income that tracks the market ([[fixed-percentage]]): in every case the assumption is that after a few months the new level is simply the level. That bet has been tested, and it mostly wins, subject to three conditions worth knowing. **Yes, and in about four years.** The reference study follows 7,812 Germans from 1984 to 2000 (Di Tella, Haisken-DeNew and MacCulloch, 2010). It cannot reject the hypothesis of **complete** adaptation to income after four years: four years past a change, life satisfaction is back where it started. What does not fade, in the same study, is relative **status**, your position in the social comparison. The lesson is directly usable. A drop in lifestyle that nobody outside the house can see is far easier to absorb than one that moves you down a rung. Skipping a far-flung trip goes down fine. Moving to a less desirable neighborhood does not. **How big a cut are we talking about?** The question has to be asked, because a duration with no magnitude attached means nothing, and the study does not answer it head on. To see why, look at its equation, which has the rare charm of putting happiness on the left-hand side: > Happiness(i, t) = α₀ log y(i, t) + α₁ log y(i, t−1) + α₂ log y(i, t−2) + α₃ log y(i, t−3) + α₄ log y(i, t−4) + controls + individual fixed effect + year fixed effect Life satisfaction as reported by person i in year t, scored 0 to 10, is explained by this year's income and by the four previous years of income, each with its own coefficient. The fixed effects strip out what does not move, the person's temperament on one side and the mood of the year on the other. Testing adaptation then comes down to asking whether α₁ + α₂ + α₃ + α₄ equals zero: if past income no longer explains anything, it has been forgotten. In this panel it explains nothing, and that is the whole result. Two limits are visible in the equation itself. It is linear in log income, so it **imposes** the same speed of adaptation on shocks of every size rather than measuring it: independence from magnitude is an assumption, never a finding. And the four lags are a choice the authors themselves call arbitrary. So "four years" means "four years back, nothing shows up any more", not "it takes four years to digest". The magnitude anchor has to come from somewhere else, from the data itself, and it is vivid. In that panel, household income moves by 14.1% a year on average in absolute value, and **one fifth of all annual changes are smaller than 2.5%**. Hold on to that number. Vanguard's lower bound sits in the bottom quintile of the income noise an ordinary household absorbs every year without anything happening. It is not a cut in the sense your body understands the word, it is background noise. A 10% step, on the other hand, is larger than the typical annual change and cannot go unnoticed. **And a steady glide does not pile up forever.** This is the reassuring consequence of the same structure. If each cut creates a discomfort that fades over a few years, and the effects add up, then a steady descent converges on a **permanent** level instead of getting worse without end. The discomfort settles at a level set by the **annual pace** of the descent, not by how long it lasts. As an order of magnitude, gliding 2.5% a year for fifteen years amounts to carrying, permanently, the sting of a single cut of about 6%, and never more. Not painless, but bounded, which is exactly what a bounded rule promises. **But not symmetrically.** Across 28,723 Germans and 20,570 Britons, Boyce, Wood, Banks, Clark and Brown (2013) establish that a fall in income costs more well-being than an equal rise adds, and they do it on losses people **lived through**, not merely feared. Their conclusion fits in one sentence that doubles as a design instruction: well-being is better served by small, regular raises than by large moves in either direction. That is word for word the case for the asymmetry of Vanguard's bounds (+5% / −2.5%), stated seven years before the rule appeared. The German study puts a number on the same thing in passing, and the order of magnitude is worth knowing. Reaching a given income **by way of a fall** of 4.5% costs, on top of the fall itself, the equivalent of about 3.6% of income. A drop therefore feels roughly like a level gap almost twice its size. The authors call that test exploratory, but it points the same way as everything else. **And not to every kind of change.** Clark, Diener, Georgellis and Lucas (2008) find complete adaptation to marriage, divorce, widowhood, the birth of a child and even to being laid off, but no adaptation at all to long-term unemployment among men. You get used to a change in **level**, not to a change in **role**. Carried over to spending, the implication is sharp. A cut you experience as "I am spending a bit less this year" gets absorbed; the same cut experienced as "my plan has failed" does not. The rule written in advance, which presents the cut as designed behavior rather than as an incident, does half the work here. ::: science The condition everyone forgets: adaptation runs on forgetting The mechanics of adaptation have a shape. Wilson and Gilbert's AREA model (2008) lays out the sequence: you **notice** a change, you **react** to it emotionally, you **explain** it, and because you have explained it you stop paying attention to it, which is what it means to **adapt**. Adaptation is not resignation, it is attention withdrawing. Sheldon and Lyubomirsky proved the converse on positive changes: deliberately keeping your attention on an improvement, through gratitude or variety, **stops** you from getting used to it. Attention is the lever, in both directions. The practical consequence matters, and it cuts against instinct. Anything that reminds you of the cut prevents you from adapting to it. A budget kept by hand, line by line, turns a 2.5% cut into some thirty reminders a month, each one reopening the emotional file that adaptation needed closed. The automatic monthly transfer does the opposite: it concentrates the decision into a single moment, once a year, then makes the cut invisible day to day. That is why it opens the toolbox below, and it is not a comfort feature, it is the mechanical condition for adapting at all. ::: One last mechanism pushes the same way, and it is almost too convenient. Vanguard's lower bound is −2.5% **a year in constant dollars**. With 2 to 3% inflation, it barely moves the number on the bank statement ([[floor-and-ceiling]]). Money illusion, our habit of thinking in nominal terms rather than in purchasing power (Shafir, Diamond and Tversky, 1997), works in the plan's favor for once. The cumulative effect is real all the same: six years at that pace is −14% of purchasing power, and no statement ever announced it. Nothing shows up in the currency, everything shows up in the grocery cart, and that is precisely why the rule goes down easily. One caveat to close, because these results have a domain. They cover income changes above the point where material need begins. Coming down toward your real floor is no longer a question of adaptation but of trade-offs you did not choose ([[how-much-you-need]]), and no literature says you get used to that. This is why the book insists so much on the floor: it marks the line between what you adapt to and what you simply take. ## The toolbox: equipping the psychology **The automatic paycheck.** Tool number one, and its effect is out of all proportion to its simplicity: an automatic monthly transfer from the brokerage account to the checking account, for the amount the rule says ([[fixed-inflation-adjusted-withdrawal]] on paying yourself monthly, [[building-it-with-us-etfs]] on scheduling the sales). Spending becomes what it always was, living on your income. The open decision disappears, and mental accounting starts working for the plan. **Permission in numbers.** Ratchets and ceilings exist to **authorize** spending, not only to cut it ([[fixed-inflation-adjusted-withdrawal]] on the ratchet, [[floor-and-ceiling]] on the upper bound, [[amortization-based-withdrawal]] on the annually repriced payout that **tells you** what you can spend). A plan with no upward mechanism is a plan that organizes its own underspending. The argument to prescribe to a chronic underspender is the distribution of terminal wealth. In the median world, rationing leaves two to three times the starting stake behind. Look at that number, then go read the "unreasonable" budget again. ::: figure mourir-riche The book's real US 60/40, a fixed inflation-adjusted withdrawal never adjusted afterward, held for thirty years: what was left at the end for each of the 43 complete vintages, 1954 to 1996. The household that spends the 4% it is allowed ends above its starting stake in 25 cases out of 43; the one that rations itself to 3% ends above it in 41 cases out of 43, at 2.6 times the stake in the median. ::: **A second person, and a date.** The dated **annual review** is the only moment when decisions get made ([[the-annual-review]]). Between two reviews you execute, nothing more. Keep your **spouse** or a trusted third party in the loop ([[couples-and-family]]); requiring two signatures on any decision outside the rule is the best firewall against panic and against miserliness alike. And where the pattern is entrenched, get professional help without embarrassment. Pathological underspending and chronic financial anxiety are treatable. A good flat-fee advisor then mostly serves as permission in human form. **The mandatory fun budget.** The simplest field technique in the chapter: a named budget line (travel, generosity, projects, 5 to 15% of the comfort tier) where **failing to spend it** counts as a deviation from the plan, exactly as an overrun would. What gets budgeted and tracked gets spent. It is the automatic-savings mechanism, run backwards. ::: exemple Two years learning how to spend A typical field report, a composite of forum accounts and case studies. Hélène, 54, retires with $1.7M and an impeccable 3.4% plan ([[choosing-your-strategy]]). Year 1: she spends $31,000 of the $58,000 she is allowed. She lives "on cash, just to see", turns down a $4,000 trip, and checks the portfolio every morning. The annual review is what flips the switch. The distribution of bequests gives her a 92% chance of finishing above her starting capital, and her husband puts the paradox in one line: "You worked four extra years for a standard of living you refuse to live." Year 2, the equipment: an automatic transfer of $4,800 a month (the paycheck), a $8,000 travel budget she is **required** to spend, statements quarterly and no more often, two signatures on anything outside the rule. Year 3: $49,500 spent, and the portfolio went up anyway. That is the whole lesson. The anxiety followed the spending, not the other way around, and manufactured permission turned into permission she actually felt. Not one parameter of the plan changed. The environment did. ::: ## The essentials - The dominant psychological risk for an affluent retiree is not panic. It is chronic **underspending** (portfolios that never get drawn down, effective withdrawals of 2 to 3% under plans that allow 4 to 5): a plan failure paid in life not lived, as real as running out of money. - Four mechanisms: unbounded fear (put bounds on it, with a failure probability and a set of warning lights), the saver identity (rewire it, a withdrawal is a paycheck), mental accounting (put it to work for the plan), and the absence of permission (manufacture it, with rules that raise spending, with guaranteed income, with the distribution of bequests). - The mirror failure, capitulating in a crash, takes the same arsenal: a written rule, a visible buffer, thresholds that separate noise from signal, information hygiene. A crash is never a reason to change the rule. - Biases can be named (loss aversion, anchoring on the peak, recency, the illusion of control) and disarmed by structure: fewer statements, the plan's target as the anchor, several models side by side, inaction made comfortable. - You adapt to a fall in income, and almost completely within four years, but less well than to a rise of the same size, and not at all if it moves you down a social rung or reads as a failed plan. Magnitude matters as much as duration: at 2.5% a year you stay inside the bottom quintile of ordinary income noise, and the discomfort of a steady glide settles instead of compounding. Adaptation runs on attention withdrawing (the AREA model): a budget kept by hand reminds you of the cut every day and blocks the adjustment, where the automatic paycheck makes it invisible. - The toolbox has four pieces: the automatic paycheck, permission in numbers (ratchet, ceiling, ABW, the distribution of bequests), a second person and a date (the annual review, two signatures), and the mandatory fun budget. Half the finance in this book was equipped psychology all along. --- ## Going further - Kahneman, *Thinking, Fast and Slow*, for the foundations; Thaler, *Misbehaving*: mental accounting told by the man who found it. - On adaptation to income, the chapter's four references: Di Tella, Haisken-DeNew & MacCulloch, "Happiness Adaptation to Income and to Status in an Individual Panel", *Journal of Economic Behavior & Organization*, 2010 (complete adaptation in four years, and the status that resists it); Boyce, Wood, Banks, Clark & Brown, "Money, Well-Being, and Loss Aversion", *Psychological Science*, 2013 (the asymmetry between falls and rises, on losses actually lived); Clark, Diener, Georgellis & Lucas, "Lags and Leads in Life Satisfaction", *The Economic Journal*, 2008 (what you adapt to, and unemployment as the exception); Wilson & Gilbert, "Explaining Away: A Model of Affective Adaptation", *Perspectives on Psychological Science*, 2008 (the AREA model, and the role of attention). - Shafir, Diamond & Tversky, "Money Illusion", *Quarterly Journal of Economics*, 1997: why a real cut that leaves the nominal statement alone goes down so easily. - Blanchett & Finke, the work on the "retirement consumption gap", and above all "Guaranteed Income: A License to Spend" (2021): guaranteed income funds twice the spending that the equivalent savings do. - On the four mechanisms: de Berker et al., "Computations of Uncertainty Mediate Acute Stress Responses in Humans", *Nature Communications*, 2016 (uncertainty costs more than certain bad news); Kivetz & Keinan, "Repenting Hyperopia", *Journal of Consumer Research*, 2006 (regret over the sensible choice grows with time); Shefrin & Thaler, "The Behavioral Life-Cycle Hypothesis", *Economic Inquiry*, 1988 (the three mental accounts); Hartzmark & Solomon, "The Dividend Disconnect", *Journal of Finance*, 2019 (the dividend treated as income of its own). - On panic: Elkind, Kaminski, Lo, Siah & Wong, "When Do Investors Freak Out?", *Journal of Financial Data Science*, 2022; Thaler, Tversky, Kahneman & Schwartz, "The Effect of Myopia and Loss Aversion on Risk Taking", *Quarterly Journal of Economics*, 1997 (looking more often makes you take less risk). - Early Retirement Now, Part 27 ("Why is Retirement Harder than Saving for Retirement?") ([[the-ern-series]]). - Bill Perkins, *Die With Zero* (2020): the case for spending, to be read with the guardrails of [[spending-in-retirement]]. - In this book: [[voices-from-real-retirees]] (the accounts), [[one-more-year]] (the same failure, upstream), [[couples-and-family]] (the second person), [[when-to-worry]] (the warning lights that put bounds on the fear).