# Lifecycle theory: the academic backbone of the plan This book is full of practical rules. A target allocation, a glidepath, an annuity in old age, a withdrawal that responds to wealth. Each one is defended in its own article with numbers and backtests. But none of them was invented by bloggers. Behind almost all of them stands half a century of academic research, three Nobel prizes, and a single framework, **lifecycle theory**. This article lays out that backbone. It shows where the recommendations the rest of the book applies come from, by proof rather than by habit. And it surfaces something just as valuable as the theory itself: the list of places where practice departs from it, and the reasons, good or less good, for each departure. The thread is chronological and logical at once. First the founding idea, smoothing consumption over a whole life. Then the central result of Samuelson and Merton: an optimal risky share exists, and horizon alone does not change it. Then human capital, which derives glidepaths instead of decreeing them. Then Yaari's theorem, which makes the annuity the theory's starting point rather than an old person's oddity. Finally what all of this says about withdrawal rules, and the honest ledger of the gaps between the provably optimal and the livable. ::: cle Three grades of evidence This book tries to keep apart three things the financial literature happily blends. The **theorem**, true by proof under its assumptions (you contest it only by attacking the assumptions, never the logic). The **empirical result**, measured on data, whose strength is judged by replication and sample size. And the **judgment call**, a choice made by the author or the practitioner, defensible but debatable. This article tags every claim with its grade. It is the most academic page of the book, and the referee for the others: when a practical chapter contradicts a theorem, it means one of the theorem's assumptions fails in practice, and that chapter owes you the name of the assumption. ::: ## Smoothing consumption: the founding hypothesis The starting point dates from the 1950s. Franco Modigliani and Richard Brumberg (1954) state the **lifecycle hypothesis**: a household's utility comes from its consumption, spread over its whole life, not from any one year's income. Milton Friedman (1957) reaches the same view through **permanent income**. The consequence is immediate. A rational agent saves when earning above cruising speed, dissaves afterward, and wealth is never a goal, only the reservoir that carries purchasing power from one period of life to another. This work earned Friedman the Nobel prize (1976), then Modigliani (1985), and the framework remains the one all of modern household economics runs on. This is the grade of scientific consensus. FIRE is nothing but this hypothesis taken seriously all the way ([[what-is-fire]]). Compressing the earning phase into fifteen or twenty years, then smoothing consumption over the following fifty, is exactly Modigliani's program, with one unusual parameter. All the plumbing in this book, from the target capital ([[how-much-you-need]]) to the withdrawal rules, is engineering in the service of that smoothing. A word of empirical honesty, because the hypothesis describes the rational agent, not the observed human. Real households smooth badly. They overconsume current income, underconsume wealth once retired, and sort their money into watertight mental boxes. Hersh Shefrin and Richard Thaler (1988) documented and modeled these gaps as the **behavioral lifecycle hypothesis**. That is a robust empirical result, and it returns below, because several of this book's tools, the buffer first among them, are concessions to Thaler rather than to Modigliani. ## The Merton share: a target allocation exists, and horizon does not move it The next question is the portfolio. What fraction of wealth belongs in risky assets when you consume out of it for a lifetime? Paul Samuelson (1969) solves it in discrete time, Robert Merton (1969, 1971) in continuous time, and the answer has a closed form. For an agent with constant relative risk aversion facing returns independent from one period to the next, the optimal risky share is **w∗ = (μ − r) / (γ σ²)** that is, the expected risk premium, divided by the variance of the risky asset and by the risk aversion γ. This is the **Merton share**. Three properties follow from the formula, all three proved rather than observed. The optimal share is constant: it depends neither on wealth, nor on the path so far, nor, this is the counterintuitive one, on horizon. It is finite: even a mildly risk-averse agent does not go all-in on stocks once γ exceeds the premium over the variance. And it justifies the very existence of a target allocation to rebalance, the most ordinary gesture in this book ([[stock-bond-allocation]]). The orders of magnitude speak. With a 4-point risk premium and 17% volatility (the conservative forward-looking assumptions of [[expected-returns]]), the formula gives 69% stocks for γ = 2, 46% for γ = 3, 35% for γ = 4. Measured γ in individuals spreads precisely between 2 and 5. The book's 50-80% plateau is therefore not a practitioners' custom: it is the band the theorem draws for the range of plausible human risk aversions. ::: science The time-diversification fallacy Samuelson's most useful result is a negative one. "Stocks always win in the long run, so a long horizon justifies more stocks" does not survive analysis. Samuelson (1963) named it the **fallacy of large numbers**: with time, the probability of ending up a loser falls, but the size of the losing scenarios grows exactly as fast, and under his assumptions the two effects cancel to the cent. Zvi Bodie (1995) restated it through the markets themselves: the price of insurance (a put) guaranteeing that a stock portfolio beats the risk-free asset **rises** with horizon, it does not fall. If time erased risk, that insurance would become free. This is a theorem, to be handled as one. It does not forbid holding a lot of stocks; it forbids justifying them **by the length of the horizon alone**. The real justifications go through the assumptions: human capital (next section), or a mean reversion in returns that would violate the independence of periods. The latter does exist empirically, valuations predict part of ten-year returns (Campbell and Shiller 1988, [[valuations-and-cape]]), but Ľuboš Pástor and Robert Stambaugh (2012) showed that uncertainty about the predictors inflates long-horizon variance back by a good part of what mean reversion had removed. Empirical result against empirical result: the debate is open, and it is the weakest grade of evidence on this page. The 100%-stocks case of [[anarkulova-cederburg]] should be read with this grid: a serious empirical argument, confronting a theorem by attacking its assumptions, exactly as it should be done. ::: Merton's assumptions deserve their trial, because that is where practice forks. The formula asks for μ, yet Merton himself (1980) showed that a century of data is not enough to estimate a mean return with useful precision ([[expected-returns]]). It asks for γ, which is not observable to better than a factor of two. A factor-of-two uncertainty on γ moves w∗ by a factor of two. The lesson is not to throw the formula away, but to give up on knife-edge optima and settle in the middle of the plateaus, the line of conduct of [[deciding-under-uncertainty]]. Victor DeMiguel, Lorenzo Garlappi and Raman Uppal (2009) gave the now-classic measurement: out of sample, the naive equal-weight portfolio beats most sophisticated optimizations, precisely because estimation error eats the gain from optimality. An empirical result, widely replicated, and one of the discipline's most humbling. ## Human capital: the bond that you are A household's wealth does not reduce to its brokerage account. The present value of future labor income, **human capital**, is often the dominant share before fifty. Zvi Bodie, Robert Merton and William Samuelson (1992) brought that capital into the portfolio problem, and the consequences structure everything this book says about trajectories. For a stable employee, human capital looks like a large bond: regular cash flows, weakly correlated with markets. The young saver therefore already holds, without knowing it, a massively bond-like fortune, and the Merton share applies to **total** wealth. To reach it, the financial portfolio, tiny by comparison, must be very predominantly in stocks, and the stock share must then decline as human capital is consumed. Accumulation glidepaths are this theorem, not a tradition ([[glidepaths]]). Ian Ayres and Barry Nalebuff (2010) pushed the logic to its end, all the way to leverage for the young; the deduction is valid, the implementation perilous ([[leverage-and-margin]]), and the book files it under advanced uses. Calibration, though, remains an open research topic: the human capital of an entrepreneur or a tech employee is correlated with stocks, and the recommendation partially inverts for them (John Campbell and Luis Viceira 2002, who gave the subject its reference treatment). For the retiree, the framework stays illuminating through three balances. Their human capital is not quite zero: the **option to go back to work** remains, and Bodie, Merton and Samuelson showed precisely that this flexibility licenses more portfolio risk. The option is real and wide at 45, nearly extinguished at 70 ([[going-back-to-work]]); an early FIRE plan can lean on it, a late one cannot, and that is an asymmetry the book exploits without always naming it. Second balance, the pension to come is an implicit bond, often worth several hundred thousand dollars in present value ([[us-healthcare-and-social-security]], [[pensions-and-other-income]]): counting it as such mechanically raises the justifiable stock share of the financial portfolio, the reasoning behind the pension bridge of [[vpw]] and the ladders ([[bond-ladders]]). Third balance, human capital is domestic by construction, which makes international diversification of the financial portfolio a corrective, not a luxury ([[international-diversification]]). ## Yaari: the annuity as the starting point In 1965, Menahem Yaari proves the most unsettling result in the corpus. A rational agent with no bequest motive, facing actuarially fair life annuities, should convert their **entire** wealth into annuities. The intuition is limpid: the annuity pools the one risk the market does not pay for, the length of your life, and pays survivors the **mortality credits** of the deceased, a return uncorrelated with any market ([[annuities-and-safety-first]]). Thomas Davidoff, Jeffrey Brown and Peter Diamond (2005) showed the conclusion withstands weakening almost every assumption: even with incomplete markets and loaded annuities, substantial annuitization remains optimal. Against this theorem, the observation: almost nobody does it. This is the **annuity puzzle**, and its dissection is an entire strand of the literature (Shlomo Benartzi, Alessandro Previtero and Richard Thaler 2011 give the best synthesis). Part of the gap has rational explanations, the loadings, the uncovered inflation, the bequest motive, the irreversibility in the face of health expenses. Part of it only has behavioral ones, the annuity is perceived as a bet on one's own death rather than as the insurance it is. The modern consensus that emerges, **partial** annuitization, of the floor, late, is exactly the safety-first doctrine the book applies. The theory's role here is to flip the burden of proof. The economist's default is the annuity, and it is the refusal to annuitize that must argue its case, not the reverse. Whoever waves annuities away is contesting a theorem without knowing it. ## What theory says about withdrawal rules What does this corpus recommend for the withdrawal itself? Something simple and unsettling: no optimization produces a **fixed** inflation-indexed amount. The optimal consumption of a lifecycle agent is recomputed every period, by amortizing current wealth over the remaining horizon, weighted by survival. It is a withdrawal that responds to wealth. The family of rules that embodies this result exists, it is amortization, ABW and VPW ([[amortization-based-withdrawal]], [[vpw]]), and that is why the book holds them to be the theoretically clean form of withdrawal. The fixed 4% rule, for its part, is a practitioner's invention ([[fixed-inflation-adjusted-withdrawal]]), and theory has priced it. Jason Scott, William Sharpe and John Watson (2009), in a paper with a transparent title ("The 4% Rule: At What Price?"), show that a constant income financed by a risky portfolio is an incoherent object: the lucky paths pile up surpluses that are never consumed, which you nonetheless pay for in ruin risk on the dark paths, and the same income floor could be bought more cheaply by backing it explicitly. The order of magnitude of the waste, in their calibrations, runs to tens of points of initial capital. Robert Merton (2014) pressed the same charge on the savings side: the target of a plan is an **income**, not a pot of wealth, and steering the pot while losing sight of the income is the original framing error. Grade of evidence: a theoretical and quantified result, little contested on the merits. Why, then, does the book not recommend pure amortization everywhere? Because two empirical results stand in the way, and this is where the judgment call begins and must say so. First, the utility of a stable income is real: human preferences carry **habit formation** (George Constantinides 1990, an acquired standard of living becomes the reference point, and giving it up costs more than its mirror image is worth), and the loss aversion of Daniel Kahneman and Amos Tversky (1979) hits cuts in lifestyle far harder than raises delight. A withdrawal that faithfully tracks wealth is optimal for Merton's agent and unlivable for Kahneman's. Second, governance: a simple rule, written down, followed in a panic, beats an optimal rule abandoned in March 2009 ([[the-psychology-of-spending]]). Hence the book's position, owned as a judgment call informed by theory: amortization as the skeleton, dampened by corridors and floors ([[floor-and-ceiling]], [[choosing-your-strategy]]), and never a blind fixed withdrawal. ## The ledger of departures All of the above makes the full ledger possible, each line with its grade of evidence. This is the book's epistemic identity card. **Departures grounded in theorems or strong results.** The target allocation and rebalancing (Merton share, theorem). The refusal to justify stocks by horizon alone (Samuelson 1963, theorem). Glidepaths derived from human capital (Bodie-Merton-Samuelson 1992, a theorem under debatable calibration). Partial annuitization of the floor in old age (Yaari 1965, Davidoff-Brown-Diamond 2005). The withdrawal that responds to wealth (the whole corpus, priced by Scott-Sharpe-Watson 2009). **Departures grounded in evidence against a theorem's assumption.** The plateau rather than the knife-edge optimum (Merton 1980 and DeMiguel-Garlappi-Uppal 2009, estimation error dominates refinement). The valuation anchor on the initial withdrawal (Campbell-Shiller 1988, real but noisy predictability, [[cape-based-rules]]). The income corridor (Constantinides 1990, Kahneman-Tversky 1979, real preferences are not CRRA). **Judgment calls, defended but debatable.** The cash buffer: no place in theory, which sees in it only a mental account in Thaler's sense, kept for its governance services and billed honestly, about half a point ([[cash-buffer]], [[the-bucket-strategy]]). The dose of gold and trend ([[gold-in-retirement]], [[managed-futures]]): insurance against regimes that Merton's single-regime framework does not see ([[market-regimes]]), at a weight standard theory would call excessive. The preference for executable simplicity over fragile optimality, everywhere. The reader is the judge, and giving the reader the exhibits is the point of this page. ::: exemple Claire, or the theory reread over a real plan Claire, 46, $1.3M in financial assets, $42,000 of spending, a pension expected at 67. Let us reread her plan as a theorist. Her pension is worth about $320,000 of discounted implicit bond: her total wealth is on the order of $1.6M, already 20% bond-like before the first dollar of fixed income, which legitimizes the 65% stocks of her financial portfolio (Merton share on total wealth, γ ≈ 3). Her option to go back to work, still real for ten years, is the residual human capital that licenses that level rather than 55%. Her withdrawal follows amortization with a corridor: Merton's skeleton, dampened for the Kahneman agent she also is. She keeps 30 months of buffer, a concession to Thaler paid 0.4 point of expected return, which she knows. And her written plan schedules studying an annuity around 75, when mortality credits will have grown large: Yaari, at the moment his theorem bites. Every layer of her plan has a name and a date. That is what an anchored plan is. ::: ## The essentials - FIRE is the lifecycle hypothesis (Modigliani 1954, Friedman 1957) taken seriously: wealth is only a reservoir carrying consumption from one period of life to another. Scientific consensus. - An optimal risky share exists, (μ − r)/(γσ²), constant and independent of horizon (Samuelson 1969, Merton 1969). For plausible human risk aversions it draws the book's 50-80% plateau, and it forbids justifying stocks by the length of the horizon alone (the fallacy of large numbers, Samuelson 1963). - Human capital is an implicit bond that derives the glidepaths (Bodie-Merton-Samuelson 1992); for the retiree, what remains of it is the option to go back to work and the pension to come, two assets the plan must count. - Theory makes the annuity the default (Yaari 1965) and amortization-based withdrawal the clean form (Scott-Sharpe-Watson 2009 for the price of the fixed rule); it is the refusal of these objects that owes an argument, not their adoption. - The book's departures from theory are listed, dated and tagged: evidence against an assumption where it exists (plateaus, corridors, the CAPE anchor), an owned judgment call otherwise (the buffer, the regime-insurance doses). Demanding that tag from everything you read about money may be the best habit this page can leave you. --- ## Going further - Franco Modigliani and Richard Brumberg, "Utility Analysis and the Consumption Function" (1954); Milton Friedman, "A Theory of the Consumption Function" (1957): the founding texts of smoothing. - Paul Samuelson, "Lifetime Portfolio Selection by Dynamic Stochastic Programming" (1969) and "Risk and Uncertainty: A Fallacy of Large Numbers" (1963); Robert Merton, "Lifetime Portfolio Selection under Uncertainty" (1969) and "On Estimating the Expected Return on the Market" (1980). - Zvi Bodie, Robert Merton and William Samuelson, "Labor Supply Flexibility and Portfolio Choice in a Life Cycle Model" (1992); John Campbell and Luis Viceira, "Strategic Asset Allocation" (2002): human capital treated in depth; Moshe Milevsky, "Are You a Stock or a Bond?" (2008) for the general-audience version. - Menahem Yaari, "Uncertain Lifetime, Life Insurance, and the Theory of the Consumer" (1965); Davidoff, Brown and Diamond, "Annuities and Individual Welfare" (2005); Benartzi, Previtero and Thaler, "Annuitization Puzzles" (2011). - Jason Scott, William Sharpe and John Watson, "The 4% Rule: At What Price?" (2009); Robert Merton, "The Crisis in Retirement Planning", Harvard Business Review (2014). - On the behavioral departures: Shefrin and Thaler, "The Behavioral Life-Cycle Hypothesis" (1988); Kahneman and Tversky, "Prospect Theory" (1979); Constantinides, "Habit Formation" (1990); DeMiguel, Garlappi and Uppal, "Optimal Versus Naive Diversification" (2009). - In this book: [[deciding-under-uncertainty]] (deciding when the parameters are unknown), [[amortization-based-withdrawal]] and [[annuities-and-safety-first]] (the two objects theory prefers), [[glidepaths]] (human capital as a trajectory), [[anarkulova-cederburg]] (evidence contesting an assumption, done by the rules of the art).