8.1 Goal Planning: Education, Retirement and Withdrawals
Turn life goals into numbers: state a goal in future rupees, find the SIP that funds it, size a retirement corpus with a growing annuity, test withdrawals against sequence-of-returns risk by Monte Carlo, plan an SWP, and set the safety nets — an emergency fund and the right life cover, with term and endowment compared by IRR.
📍 The situation
Vikram (36) and Neha (34) live in Hyderabad with their daughter Ira, who is three. Together they take home about ₹1,80,000 a month, about ₹1,15,000 of it Vikram's. Household spending, including the ₹30,000 home-loan EMI (about ₹30 lakh still owed), is about ₹1,25,000, so they can save roughly ₹55,000 a month. They have three goals and a worry. Ira's higher studies in 15 years: a course that costs ₹20 lakh today. Retirement when Vikram is 60, 24 years from now, on ₹50,000 a month in today's money (the loan will be repaid and Ira independent by then). Safety: what if one of them loses a job, or Vikram dies young? An insurance agent has offered them an endowment policy — "₹1 lakh a year for 20 years, ₹32 lakh at maturity, plus life cover" — and a colleague says "just keep 25 times your annual expenses and withdraw 4% a year".
Each question is a calculation you already know how to do. A goal in future rupees is compound growth with inflation (§1.1, Time Value of Money); the SIP that funds it is an annuity (§1.2, EMIs, SIPs and Annuities); a retirement corpus is the present value of a growing annuity; whether withdrawals survive real markets is a Monte Carlo question (§3.3, Random Walks and Monte Carlo); comparing an endowment with "term insurance plus investing the rest" is an IRR (§1.3, Measuring Returns); and moving money to safer assets as a goal approaches is the glide path of §4.2 (CAPM, Factors, Asset Allocation and Rebalancing). This closing chapter puts the course to work on one family's life.
Every return, inflation rate, premium, bonus and maturity value on this page is illustrative. An "assumed 11% return" is a planning input, not a forecast or promise; market-linked returns vary from year to year and can be negative for long stretches. Insurance premiums and benefits depend on the insurer, the policy terms, age and health; tax treatment of insurance, mutual funds and withdrawals changes — check the policy documents, the insurance regulator (IRDAI), AMFI and the Income Tax Department for the current position. The family, the products and the agent are hypothetical; nothing here is investment, insurance or tax advice.
Goals in Future Rupees and the SIP That Funds Them
The first planning mistake is to aim at today's price. A course, a wedding or a house costs more every year, often faster than general inflation, so a goal must be stated in the rupees of the year you will pay.
A financial goal is a future spending need described by its present cost $C_0$ (what it would cost today), its horizon $T$ years, and its goal inflation $\pi_g$ (the expected yearly rise in its cost, which may differ from general inflation). Its future cost is $G=C_0(1+\pi_g)^T$.
Given an assumed annual return $R$ on the money set aside for the goal and existing savings $S$ already earmarked for it, the required SIP is the monthly amount which, invested at the start of each month for $n=12T$ months, makes the savings plus the SIP corpus equal $G$ at the horizon. We use the monthly rate $i=(1+R)^{1/12}-1$ of §1.2.
With the notation of Definition 8.1, the shortfall to be funded by the SIP is $$G_{\text{net}}=C_0(1+\pi_g)^T-S(1+R)^T.$$ If $G_{\text{net}}\gt0$, the required flat SIP is $$A=\frac{G_{\text{net}}}{\dfrac{(1+i)^n-1}{i}\,(1+i)},$$ and the required starting amount of a step-up SIP (raised by $g$ once a year, $R_1=(1+i)^{12}=1+R$) is $$A_0=\frac{G_{\text{net}}}{s\,\dfrac{R_1^{\,T}-(1+g)^T}{R_1-(1+g)}},\qquad s=\frac{(1+i)^{12}-1}{i}(1+i),$$ with the denominator replaced by $s\,T R_1^{\,T-1}$ when $1+g=R_1$. Both are proportional to $G_{\text{net}}$.
Why. The savings $S$ grow to $S(1+R)^T$, so the SIP must supply the rest. The SIP and step-up SIP corpus formulas of §1.2 (EMIs, SIPs and Annuities) are "amount × factor", so each required amount is the shortfall divided by the factor. ∎ Two consequences are worth saying aloud. The required SIP is linear in the goal: double the goal, double the SIP. And the factor grows faster than linearly in $n$, so starting later costs more than proportionally — this is compounding working against you.
- Write the goal down: what, when, and what it would cost today. Separate goals with different dates.
- Inflate it: $G=C_0(1+\pi_g)^T$ with a goal-specific inflation (education and health costs often rise faster than the general price level).
- Subtract what is already earmarked, grown at the assumed return: $G_{\text{net}}=G-S(1+R)^T$.
- Choose the return assumption to match the horizon and the allocation. Money needed within about three years cannot sit through a large equity fall (§3.1), so it belongs mostly in low-volatility assets with a lower assumed return; long goals can hold more equity (§4.2).
- Solve for the SIP with Theorem 8.1 — flat or step-up.
- Stress-test: recompute with a return 2–3 points lower, and estimate the probability of reaching the goal by simulation (§3.3).
- Follow a glide path: shift the goal's money from equity to debt over the last few years, so a late crash cannot wreck it (§4.2), and review every year, re-solving with the actual balance.
Retirement: How Big a Corpus, and How Much Can You Withdraw?
Retirement is two goals in one. Before it you accumulate a corpus; after it you withdraw from that corpus for perhaps 25–35 years, while prices keep rising and the remaining money stays invested. The size of the corpus is therefore the present value of a stream of withdrawals that grows with inflation — a growing annuity.
Let $E_0$ be the annual spending one wants in today's rupees and $T$ the years until retirement. With general inflation $\pi$, the first-year retirement spending is $E_R=E_0(1+\pi)^T$. An inflation-indexed withdrawal plan takes $E_R$ at the start of the first retirement year, $E_R(1+\pi)$ at the start of the second, and so on, for $N$ years.
The retirement corpus $C$ is the amount that funds this plan exactly at an assumed return $R$ after retirement. The initial withdrawal rate is $w=E_R/C$, and the corpus factor is $F=C/E_R=1/w$: "corpus = first-year spending × factor".
Payments $W,\,Wq,\,Wq^2,\dots,Wq^{N-1}$ made at the start of each of $N$ periods, valued at rate $R$ per period, have present value $$PV=W\,\frac{1-x^N}{1-x},\qquad x=\frac{q}{1+R}\ (x\ne1),$$ and $PV=NW$ if $x=1$. (For payments at the end of each period, divide by $1+R$.) With $q=1+\pi$ and the real return $r=\frac{1+R}{1+\pi}-1$ (the Fisher relation of §1.1), $x=\frac1{1+r}$ and the retirement corpus factor is $$F=\frac{C}{E_R}=(1+r)\,\frac{1-(1+r)^{-N}}{r}\quad(r\ne0),\qquad F=N\ \ (r=0).$$ If $r\gt0$ and the plan has no end date ($N\to\infty$), $F\to\frac{1+r}{r}$: a withdrawal rate of $w=\frac{r}{1+r}$ can be kept up for ever.
Derivation. The payment at the start of period $k+1$ is $Wq^k$ and is discounted by $(1+R)^k$, so $PV=W\sum_{k=0}^{N-1}x^k$, a geometric progression — Theorem 1.6's trick from §1.2 again (multiply by $x$ and subtract). Substituting $x=1/(1+r)$ gives $1-x=r/(1+r)$ and the formula for $F$, which is exactly the annuity-due factor of §1.2 at the real rate. As $N\to\infty$, $x^N\to0$ when $r\gt0$. ∎ Reading it: in a plan indexed to inflation, only the real return matters. A nominal 8% with 6% inflation is a real return of just under 1.9%.
The "4% rule". The widely quoted rule of thumb — withdraw 4% of the corpus in the first year, then raise the rupee amount with inflation — came from studies of historical returns in another country's markets over 30-year retirements. It is not a law. Theorem 8.2 shows that the sustainable rate depends on the real return and the horizon; Example 5 below shows that it also depends on volatility and the order of returns. For a plan with no end date at a 1.89% real return, the deterministic rate is only $r/(1+r)\approx1.85\%$.
Sequence-of-Returns Risk and Testing a Plan by Simulation
Theorem 8.2 assumes the same return every year. Real portfolios rise and fall, and once money is flowing in or out, when the good and bad years arrive matters as much as their average.
Sequence-of-returns risk is the risk that the order of returns, not just their average, determines whether a plan with regular contributions or withdrawals succeeds. For a withdrawal plan, the success probability is the probability, under a stated model of returns, that every planned withdrawal is paid in full for the whole horizon.
Let a balance $B_0$ face yearly growth factors $g_1,\dots,g_N$ (with $g_k=1+R_k\gt0$), with a cash flow $F_k$ added at the start of year $k$ ($F_k\gt0$ a contribution, $F_k\lt0$ a withdrawal), and suppose the balance stays positive. Then the final balance is $$B_N=B_0\prod_{j=1}^{N}g_j+\sum_{k=1}^{N}F_k\prod_{j=k}^{N}g_j .$$
- With no cash flows, $B_N=B_0\prod g_j$ does not depend on the order of the returns.
- Swapping the returns of two consecutive years $k$, $k+1$ with $g_k\lt g_{k+1}$ (so the bad year comes first) changes $B_N$ by $F_{k+1}(g_{k+1}-g_k)\prod_{j\gt k+1}g_j$ compared with putting the good year first. So a bad year early hurts a withdrawal plan ($F_{k+1}\lt0$) and helps a SIP ($F_{k+1}\gt0$).
Proof. The formula follows by induction from $B_k=(B_{k-1}+F_k)g_k$: each flow is multiplied by the growth of every year after it is made. For (1), a product does not depend on the order of its factors. For (2), only years $k$ and $k+1$ change. Entering year $k$ with balance $X$, the balance after year $k+1$ is $(X+F_k)g_kg_{k+1}+F_{k+1}g_{k+1}$ with the original order and $(X+F_k)g_{k+1}g_k+F_{k+1}g_k$ after the swap; the difference is $F_{k+1}(g_{k+1}-g_k)$, and the later years multiply it by the positive factor $\prod_{j\gt k+1}g_j$. ∎ In words: a withdrawal plan is largest at the start, so early returns act on the most money; a SIP is largest at the end, so late returns matter most.
To judge a real plan we need many possible sequences, not two. That is a job for the Monte Carlo method of §3.3 (Random Walks and Monte Carlo): simulate thousands of plausible return paths, run the plan on each, and count.
Model the yearly growth factors as independent lognormal, $g_k=e^{\nu+\sigma Z_k}$ with $Z_k\sim N(0,1)$ — geometric Brownian motion sampled once a year (§3.3), with median yearly growth $e^\nu$ and volatility $\sigma$. Simulate $M$ independent paths; on each, run the plan (withdraw, then grow) and record whether every withdrawal was paid. If $K$ paths succeed, $$\hat p=\frac KM,\qquad \text{SE}(\hat p)=\sqrt{\frac{\hat p(1-\hat p)}{M}},$$ and $\hat p\pm1.96\,\text{SE}$ is an approximate 95% interval for the model's success probability.
Why. Each path is an independent trial with the same success probability $p$, so $K$ is binomial and $\hat p$ has mean $p$ and variance $p(1-p)/M$ — the Monte Carlo accuracy rule of §3.3. ∎ The standard error measures simulation noise only. It cannot tell you whether $\nu$, $\sigma$, inflation or independence are right; that is why a plan is tested under several models, not one.
- Fix the plan: corpus $C$, first withdrawal $W_1=wC$, inflation $\pi$, horizon $N$ years, withdrawals at the start of each year.
- Fix the model: median return $e^\nu-1$ and volatility $\sigma$ of the whole portfolio (which depend on its allocation, §4.1–§4.2), and a seed.
- For each path: set $B=C$, $W=W_1$; for $k=1,\dots,N$: if $B\lt W$ the plan fails in year $k$; else $B\leftarrow(B-W)\,e^{\nu+\sigma z_k}$ and $W\leftarrow W(1+\pi)$.
- Count successes, compute $\hat p$ and its SE, and record the failure year and the final balance in today's rupees ($B/(1+\pi)^N$).
- Vary the inputs one at a time — withdrawal rate, return, volatility, horizon — and report a table, not a single number.
SWP Income, Emergency Funds and Insurance
In practice a retiree's withdrawals usually come monthly, for example through a systematic withdrawal plan (SWP) from a mutual fund (§2.2, Mutual Funds by the Numbers). The arithmetic is an EMI in reverse.
The sustainable SWP for a corpus $C$, horizon $n$ months and assumed monthly return $i$ is the largest withdrawal (flat, or rising by a factor $q$ each month) that the corpus can pay at the end of each month for $n$ months under that return.
A bucket approach holds the next one to three years of withdrawals in low-volatility assets (bucket 1) and the rest in a growth portfolio (bucket 2), refilling bucket 1 from bucket 2 by periodic rebalancing (§4.2). It reduces the need to sell growth assets right after a fall — a practical defence against sequence risk, paid for by a lower expected return on bucket 1.
With end-of-month withdrawals for $n$ months at monthly rate $i\gt0$:
- Flat: $$W=\frac{C\,i}{1-(1+i)^{-n}},$$ the EMI formula of §1.2 with the corpus as the "loan". If $W\le iC$ the corpus never falls.
- Rising by the factor $q$ each month, $x=q/(1+i)\ne1$: the first withdrawal is $$W_1=C\,(1+i)\,\frac{1-x}{1-x^n}.$$
Why. The corpus must equal the present value of the withdrawals. For (1) this is the annuity of §1.2; for (2) it is Theorem 8.2 with end-of-period payments: $C=\frac{W_1}{1+i}\cdot\frac{1-x^n}{1-x}$. The rule "$W\le iC$ never falls" is the SWP-duration result of §2.2. ∎
Safety nets come first. A goal plan assumes the SIPs continue. Two events can stop them: losing an income for some months, and losing an earner for good. The first is met by an emergency fund, the second by life insurance.
An emergency fund is money set aside, in assets that can be withdrawn quickly with little risk of loss (for example a savings account, a sweep FD or a liquid fund), to cover $m$ months of essential spending — food, rent or EMI, school fees, insurance premiums — during a loss of income or an unexpected large expense. Its purpose is not return but to avoid selling long-term investments, or borrowing at high rates, at a bad time.
Suppose that after losing a job, each month independently brings a new job with probability $q$. Then the number of months without income, $D$, is geometric: $$P(D\gt m)=(1-q)^m,\qquad E[D]=\frac1q.$$ A fund of $m$ months of essential spending is enough with probability $1-(1-q)^m$, and the smallest $m$ that is enough with probability at least $1-\alpha$ is $$m=\left\lceil\frac{\log\alpha}{\log(1-q)}\right\rceil.$$
Why. $D\gt m$ means "no job in each of the first $m$ months", probability $(1-q)^m$ by independence. Solve $(1-q)^m\le\alpha$ by taking logarithms; $\log(1-q)\lt0$ reverses the inequality. ∎ The model is deliberately simple: the real chance of finding work changes with the economy and the length of the search, and in a downturn $q$ falls for everyone at once — which is exactly when the fund is needed.
- The human life value (HLV) of an earner is the present value of the income he or she would have contributed to the family (income minus taxes and own consumption) until retirement. A needs-based estimate instead adds liabilities and goals (loans, children's education, the family's living costs) and subtracts existing assets. Either gives a rough target for life cover.
- A term insurance plan pays the sum assured only if the insured dies during the term; nothing is paid on survival, so its premium is the price of cover alone.
- An endowment (or money-back) plan pays the sum assured on death during the term and a maturity benefit (sum assured plus any bonuses) on survival. Its premium buys cover and savings together.
If the net contribution is $I$ in the coming year (paid at the year's end), grows by $g$ a year for $N$ years, and is discounted at $d\ne g$, then $$\text{HLV}=I\,\frac{1-\left(\frac{1+g}{1+d}\right)^N}{d-g}.$$
Why. This is Theorem 8.2 with end-of-year payments: $\frac{I}{1+d}\cdot\frac{1-x^N}{1-x}$ with $x=\frac{1+g}{1+d}$, and $(1+d)(1-x)=d-g$. ∎
Let an endowment with sum assured $SA$ cost a premium $P_e$ at the start of each of $n$ years and pay $M$ at maturity, and let a term plan with the same cover and term cost $P_t$. Holding the endowment is equivalent to buying the term plan and saving $P_e-P_t$ a year in a "savings account" whose survival-case return is the IRR $x$ solving $$(P_e-P_t)\,\frac{(1+x)^n-1}{x}\,(1+x)=M .$$ "Buy term and invest the rest" at a return $y$ gives a larger maturity amount than the endowment if and only if $y\gt x$, and in the death case it pays $SA$ plus the fund built so far, against $SA$ plus the bonus accrued so far.
Why. In both routes the family receives $SA$ on death, so the cover cancels and what is left is a stream of savings $P_e-P_t$ that becomes $M$: its IRR (§1.3, Measuring Returns) is $x$. The annuity-due factor increases with the rate, so the invested route beats $M$ exactly when $y\gt x$. ∎ This is the "fair comparison" that §1.3 promised: a policy's return should be measured after separating out the price of its cover.
- Planning in today's prices. ₹20 lakh of education today is ₹63 lakh in 15 years at 8% (Example 1). Always state goals in future rupees — or do everything consistently in real terms.
- "25 times expenses" with today's expenses. For a retirement 24 years away it gave ₹1.5 crore against ₹5.63 crore needed (Example 3).
- "The 4% rule is safe." It is a rule of thumb from one set of historical data. In our model it succeeds about 58% of the time (Example 5); the answer depends on real returns, volatility, horizon and costs.
- "Only the average return matters." With withdrawals the order matters: the same returns gave ₹83 lakh or ₹92 lakh (Example 4). Average return and CAGR cannot see the sequence.
- "A plan that works at the assumed return is safe." If the assumption is the median, the plan fails on about half the paths (Example 5).
- A flat SWP as a lifetime income. A fixed ₹1 lakh a month loses most of its purchasing power over 25 years (Example 6).
- Keeping the emergency fund in equity. A job loss and a market fall often come together; the fund must be there when needed.
- "Insurance should give returns." Cover and savings are different jobs. Separate them (Theorem 8.8) and judge each on its own — the savings part of Example 8 earned about 4.65%.
- Under-insuring. Cover of ₹20 lakh against an HLV of about ₹2 crore protects the family for a year or two, not until the children are grown.
- Starting late "with a bigger SIP". Three years' delay raised the education SIP by about 51% (Example 2).
Illustrative assumptions throughout: inflation 6% (education 8%), 11% on long-term growth money, 8% after retirement; all numbers are from Examples 1–8 or follow from them.
- Emergency fund first. Six months of essentials, ₹4.5 lakh, in a liquid form (Example 7). If they have only ₹2 lakh, they build the rest from the monthly surplus before starting new SIPs.
- Life cover. Vikram's take-home pay is about ₹13.8 lakh a year; after his own personal spending, about ₹12 lakh of it goes to the family. Expect it to grow 6% a year for 24 years. Discounting at 8% (Theorem 8.7): $$\text{HLV}=12\text{ lakh}\times\frac{1-(1.06/1.08)^{24}}{0.02}\approx\text{₹}\,2.17\text{ crore}.$$ A needs-based check (₹30 lakh home loan + Ira's education + the family's spending gap) gives a figure of the same order. They choose pure term cover for each earner and decline the endowment (Example 8): at 4.65%, its savings part is below their planning return, and its cover is far too small.
- Ira's education. A 10% step-up SIP starting at about ₹6,363 a month, plus the ₹3 lakh already saved (Example 2), moving gradually to debt in the last 3–4 years (glide path, §4.2).
- Retirement. A 10% step-up SIP starting at about ₹14,518 (Example 3) plus the existing ₹10 lakh, in a diversified allocation that is rebalanced yearly (§4.2).
- Budget check. Year 1: $6{,}363+14{,}518\approx$ ₹20,900 a month in SIPs, plus term premiums and the emergency-fund top-up — well within ₹55,000. The step-ups assume income also rises about 10% a year; if it does not, they re-solve with the actual numbers.
- Stress test. Lowering the pre-retirement return to 9% raises the retirement step-up SIP to about ₹19,800; the post-retirement plan has only about a 48% success probability in the model of Example 5. Levers that raise it: a lower withdrawal rate (a larger corpus, or spending 10% less), retiring two years later, keeping costs low (§2.2 and §8.2), a bucket for the first years of withdrawals, and spending flexibly after bad years.
The mathematics, not a recommendation. The plan is a set of linked calculations, each from an earlier chapter, and its value is that every assumption is visible and can be changed. A real plan would add health insurance, taxes, other goals and the family's own preferences, and would be reviewed every year with the actual balances — the XIRR of §1.3 tells them whether they are on track.
Exercises
All rates are illustrative assumptions. SIPs are invested at the start of each month with $i=(1+R)^{1/12}-1$; retirement withdrawals are at the start of each year and rise with inflation unless stated otherwise; SWPs are at the end of each month. Round rupee answers sensibly.
- 1 A wedding that costs ₹12 lakh today is planned in 10 years; its cost rises 7% a year. What is its future cost?
- 2 Find the flat monthly SIP for part 1 at an assumed 10% a year.
- 3 Repeat part 2 if ₹2 lakh is already saved for the goal and earns 10%.
- 4 Find the starting SIP for part 2 if it is raised by 10% every year (no existing savings). Which case of Theorem 8.1 applies?
- 5 A goal costs ₹10 lakh today, is 8 years away, and its cost rises 6% a year. What lump sum invested today at 9% meets it? Show that discounting today's cost at the real rate gives the same answer.
- 6 Find the flat SIP to reach ₹50 lakh in 20 years at 11%, and the SIP if the start is delayed by 5 years (same goal date). By what factor does the SIP rise?
- 7 Explain why the assumed return for a goal three years away should be lower than for a goal twenty years away.
- 1 Spending today is ₹40,000 a month and retirement is 25 years away. Find the first-year retirement spending (annual and monthly).
- 2 Find the corpus factor for a 25-year retirement.
- 3 Find the corpus for parts 1–2.
- 4 What is the plan's initial withdrawal rate?
- 5 At what initial withdrawal rate could the inflation-indexed plan run for ever?
- 6 A retiree with ₹2 crore wants ₹10 lakh in year 1, rising 6% a year. How many full years of withdrawals does the corpus pay?
- 7 Compare part 3 with "25 times today's annual spending". Why is the rule-of-thumb figure so different?
- 1 A corpus of ₹50 lakh pays ₹5 lakh at the start of each year. Find the balance after 3 years if the returns are +15%, −10%, +5% in that order, and in the reverse order.
- 2 With no withdrawals, show that both orders in part 1 give the same balance.
- 3 An investor puts ₹1 lakh in at the start of each year with the same two orders. Which order is better now, and why?
- 4 A simulation of 2,500 retirement paths has 1,900 successes. Estimate the success probability and a 95% interval.
- 5 How many paths are needed for an SE of at most 0.5 percentage points when $p\approx0.8$?
- 6 Using Theorem 8.3, explain why the first five years of retirement are the most dangerous, and how a bucket approach responds.
- 1 A corpus of ₹80 lakh earns an assumed 6% a year. Find the flat monthly SWP that lasts exactly 20 years.
- 2 If instead ₹60,000 is withdrawn every month, how long does the corpus last? What monthly SWP would never reduce it?
- 3 With $q=0.25$ in Theorem 8.6, find the chance that a 6-month emergency fund runs out, and the months needed for 95% confidence.
- 4 Give two reasons why an emergency fund is kept in low-risk, liquid assets even though their return is low.
- 5 An earner's net contribution is ₹8 lakh next year, growing 5% a year for 30 years. Find the HLV at a 7% discount rate.
- 6 An endowment costs ₹50,000 at the start of each year for 15 years and pays ₹10 lakh at maturity. Find its survival-case IRR.
- 7 A term plan with the same cover costs ₹3,000 a year. Find the IRR of the endowment's savings part, and the maturity amount if the ₹47,000 difference is invested at an assumed 7%.
E1. NISM-style A goal costs ₹10 lakh today. If its cost rises at 6% a year, compounded annually, its cost after 12 years is closest to (A) ₹17.2 lakh (B) ₹20.1 lakh (C) ₹17.9 lakh (D) ₹16.0 lakh.
E2. NISM-style For a given premium, the type of life insurance that generally provides the largest sum assured is (A) an endowment plan (B) a money-back plan (C) a pure term plan (D) a whole-life endowment with bonus.
E3. NISM-style Sequence-of-returns risk matters most for (A) a lump sum held for 10 years with no flows (B) a retiree making regular withdrawals in the early years of retirement (C) a T-bill held to maturity (D) a fixed deposit with interest reinvested.
E4. Case study Asha, 45, spends ₹50,000 a month today and plans to retire at 60 for 25 years. Assume 6% inflation, 8% after retirement and 10% before, and ₹40 lakh already saved for retirement. Find (a) her first-year retirement spending, (b) the corpus, (c) the flat monthly SIP needed for 15 years, (d) her initial withdrawal rate.
E5. Case study A retiree has ₹1 crore and withdraws ₹6 lakh at the start of year 1, rising 6% a year. Count the full years of withdrawals if the portfolio returns (a) 8% every year, (b) −25% in year 1 and 8% every year after, (c) 8% every year except −25% in year 15. Explain the result.
E6. Interview "Instead of a fixed rupee amount, I will withdraw 5% of whatever my corpus is worth at the start of each year. Then I can never run out." Is that true? What is the catch? (Use median growth 8% a year and 6% inflation.)
Suggested plan — 2 lectures × 60 min.
- Goals and retirement. Open with the 📍 situation and ask: "How much should they save for Ira?" — most answers use today's ₹20 lakh. Theorem 8.1 and Examples 1–2; project Interactive 1 and move the delay slider. Then Definition 8.2, Theorem 8.2 (derive it as a geometric series in real terms) and Example 3, ending with the "25 times expenses" trap.
- Risk and safety nets. Example 4 on the board (same returns, opposite orders), then Theorem 8.3 with the two-year swap proof. Theorem 8.4 and Example 5 with Interactive 2: start at σ = 0 (the red and green paths coincide), then raise σ; reverse a bad path. Finish with SWP (Example 6), the emergency fund (Example 7), HLV and term vs endowment by IRR (Example 8) and the 📈 Market case.
Class activity. Each student writes one personal goal (a bike, a course, a trip, a down payment), states it in future rupees, and computes a flat and a step-up SIP under two return assumptions; then pairs swap and check with Interactive 1. In the second lecture, groups are given a corpus and must choose a withdrawal rate that reaches 80% success in Interactive 2 under a model the teacher announces.
Misconceptions to watch for. Goals in today's prices; "25× expenses" with today's expenses; average return decides everything; the 4% rule as a law; Monte Carlo SE as model accuracy; insurance as investment; equity for an emergency fund.
Suggested homework. I(1–4, 6), II(1–6), III(1, 3–5), IV(1–3, 5–7); Practice corner E4, E5, E6.
Project. Build a one-sheet plan for one goal and one retirement, then stress-test the retirement with your own Monte Carlo simulation.
Why it matters. Online calculators give one number under one assumption. A sheet you built yourself shows every assumption, lets you change them, and turns "will it work?" into "how often does it work, and what would make it work more often?"
What to do.
- Goal sheet. Inputs: cost today B1, goal inflation B2, years B3, assumed return B4, savings B5. Future cost
=B1*(1+B2)^B3; monthly rate=(1+B4)^(1/12)-1; shortfall=B1*(1+B2)^B3-B5*(1+B4)^B3; flat SIP=PMT(rate,12*B3,0,-shortfall,1)(the final 1 means start-of-month). Check it against Theorem 8.1. - Corpus. Real rate
=(1+R)/(1+pi)-1; corpus=PV(real,N,-E_R,0,1), which is Theorem 8.2's annuity-due factor at the real rate. - Simulation sheet. Row = one path, column = one year. Return cell
=EXP(LN(1+median)+sigma*NORM.S.INV(RAND()))-1. Balance column:=MAX(0,(prev-W)*(1+ret))withW=W1*(1+pi)^(year-1), and a "failed" flag whenprev<W. Fill 1,000 rows; success rate=COUNTIF(flags,0)/1000. - Python version.
rng=np.random.default_rng(8); G=np.exp(nu+sigma*rng.standard_normal((M,N))); loop over years:B=B-W; dead|=B<0; B=np.where(dead,0,B*G[:,k]); W*=1+pi; thenp=1-dead.mean(),se=np.sqrt(p*(1-p)/M). - Sequence test. Pick the path that fails earliest, reverse its returns (
G[j,::-1]) and rerun it. Then make a table of success probability against withdrawal rate (3%–6%) and volatility (5%–20%).
Hand in: the goal SIP (flat and step-up), the corpus and withdrawal rate, the success table with SEs, one failing path and its reversed twin on a chart, and three sentences on which lever (spending, retirement age, allocation, costs) moves the success probability most.
Going further. Add a "guardrail" rule (cut the withdrawal by 10% after any year the portfolio falls, restore it after recovery) and measure both the success probability and the worst real income; or add a bucket of two years' spending held at a fixed low return.
🧪 Self-check quiz
Q1. A goal costs $C_0$ today, its cost rises at $\pi_g$ a year, and it is $T$ years away. The amount to plan for is
Q2. For an inflation-indexed withdrawal plan, the corpus factor $C/E_R$ depends on
Q3. The same set of yearly returns is applied in two different orders. The final values are equal for
Q4. A Monte Carlo test of a retirement plan with 1,600 paths gives a success rate of 80%. The simulation standard error is about
Q5. An endowment's savings part (premium minus the cost of equal term cover) has an IRR of 4.6%. "Buy term and invest the rest" gives more at maturity exactly when the invested money earns