Financial Literacy Bootcamp

Phase: 0 · Orientation & Foundations · Weeks: 1–2 · Estimated hours: ~12 of Phase 0's 74 module-hours.

Where it sits. Third module of the program. The map and the five questions came first, then the arena of exchanges, indices and participants. What follows makes you numerate: it is where percentages, compounding, inflation, and risk stop being words and become numbers you can compute quickly and correctly. Next come your toolkit and knowledge system, and then your first guided annual report. Everything downstream is this arithmetic applied, from ratios in Phase 2 and discounting in Phase 3 to KPI thresholds in Phase 5 and macro arithmetic in Phase 7.

Suggested split of the 12 hours: core teaching ~5h · practice set and timed drills ~3.5h · mini-project ~1.5h · flashcards ~0.5h (then daily) · mastery check ~1h · teach-back and journal ~0.5h.

Gate: timed numeracy exam (Mastery check, Form A or B) at ≥85%. Passing unlocks the toolkit module.

Learning objectives

By the end you can:

1. Convert fluently among percent, percentage points, basis points, decimals, and growth factors, and chain multi-period growth by multiplying factors, never by adding percentages.
2. Compute simple and compound interest future values, explain mechanically why they diverge, and derive the Rule of 72 from the doubling equation, stating where the rule is accurate and where it breaks.
3. Convert any quoted nominal rate (APR) plus compounding frequency into an effective annual rate (EAR) and correctly rank deposits and loans that quote rates differently.
4. Explain inflation mechanically, through basket, weights and index, compute inflation from a price index, and contrast India's inflation character (~4–6% in the modern era) with the US's (~2–3%), knowing where to verify both.
5. Convert nominal to real returns with the exact Fisher relation, use the approximation knowingly, and restate any long-horizon outcome in today's purchasing power.
6. Explain, with a numeric example and no formal statistics, why an asset with more dispersed outcomes must be priced to offer a higher expected return, and why that premium is compensation rather than a guarantee.
7. Sketch the historical character of equities, government bonds, gold, real estate, and cash in both India and the US, meaning rough nominal and real returns and worst-case behavior, and name the primary sources to verify each figure.
8. Compute CAGR with correct year-counting (fence-posts), catch growth-rate-versus-level confusions in one read, and quantify fee drag quoted in basis points, all at drill speed, under a clock.

Prerequisites & connections

Builds on. The five questions. Note that question 2 ("is ROIC durably above the cost of capital?") is a compounding statement and question 5 ("what is the price implying?") is a growth-rate statement, so what follows is the arithmetic those questions run on. You also now know what the Sensex, Nifty, and S&P 500 are; here you learn to measure what they have done. From school math we assume only powers and roots, plus a calculator with an x^y key or any spreadsheet. Logarithms are reintroduced gently where we derive the Rule of 72, and no prior comfort is needed.

Feeds into. Immediately, the toolkit, where you will drill these computations in your spreadsheet, and the guided annual-report read, since annual reports are full of growth rates and margins that you will now read numerically. Later, this is deliberately the informal first pass of three formal treatments: the valuation phase formalizes time value of money (PV, annuities, perpetuities) and then risk, diversification, and the equity risk premium; the macro phase goes deep on inflation, rates, and the yield curve; Phase 9 returns to sequence risk and volatility drag as portfolio problems. When you meet those, the intuition should already be installed. If a computation here feels slow, that is the point. By the end of the drills it will not be.

4.0 Numeracy is the analyst's mother tongue

Is 18% revenue growth good? Is a 350 bp margin decline alarming? Is a 24× P/E demanding? You cannot answer any of them yet, and not one of them is a hard question once the arithmetic is reflex.

An analyst who "knows the concepts" but computes slowly and errs on percentages is like a consultant who knows frameworks but cannot write a sentence. Every judgment you will make in this program rests on a small set of arithmetic moves done fast and without error. There are perhaps ten such moves. What follows teaches all of them, then drills them under a clock, because fluency rather than familiarity is the gate.

One orienting idea before the tools: money has two rulers. The nominal ruler counts rupees and dollars; the real ruler counts what they buy. Most financial mistakes made by intelligent people are ruler confusions: comparing an Indian growth rate to a US one without adjusting for inflation, celebrating an FD return the price level quietly ate, extrapolating a nominal house-price story. By the end of the week, "in real terms?" should be a reflex.

4.1 Percentages: the working language

Percent is a fraction wearing a suit. x% = x/100. So 8% of ₹6,25,000 = 0.08 × 6,25,000 = ₹50,000. Always convert to decimal before computing; most percentage errors are misplaced decimal points.

Percentage change compares a new level to an old one:

% change = (New − Old) / Old

A share moving ₹480 → ₹552 changed by 72/480 = 0.15 = +15%. Note the denominator is the starting value. This creates a famous asymmetry:

Losses and gains are not symmetric. Fall 20% then rise 20%: ₹100 → ₹80 → ₹96. You are down 4%, not flat, because the 20% rise acts on the smaller base. In general, after a loss of L (as a decimal), the gain needed to recover is L / (1 − L):

This one table explains half of risk management. A portfolio that avoids the −50% year has a structurally easier life than one that must find a +100% year. Keep it; Phase 9 builds on it.

Percentage points (pp) vs percent. When a ratio changes, describe the change in points rather than percent, or you will create ambiguity that real analysts get burned by. A company's EBIT margin moving 18.0% → 21.5% rose 3.5 percentage points, which is a 19.4% relative increase (3.5/18.0). "Margin rose 19%" and "margin rose to 21.5%" are different claims; sloppy writers blur them, careful analysts never do.

Basis points (bp). One basis point = one-hundredth of a percentage point: 1 bp = 0.01% = 0.0001. So 100 bp = 1%, and 25 bp = 0.25%. Rates, spreads, fees, and margin moves are quoted in bp precisely to kill the pp/percent ambiguity: when the RBI or the Fed moves its policy rate, the announcement says "25 basis points," never "0.25 percent," and a mutual fund expense ratio of 0.45% is "45 bp." Two conversions to make automatic: bp → decimal: divide by 10,000 (40 bp = 0.0040); decimal → bp: multiply by 10,000 (0.0007 = 7 bp). On ₹2 crore, 40 bp = 0.0040 × 2,00,00,000 = ₹80,000. Small-sounding units, real money.

Growth factors, and the multiplication habit. To grow by rate g, multiply by the growth factor (1 + g). To apply several periods of growth, multiply the factors:

Total growth = (1 + g₁)(1 + g₂)…(1 + gₙ) − 1

Revenue growing +25%, +18%, −7% over three years: 1.25 × 1.18 × 0.93 = 1.37175 → total +37.2%. Never add growth rates across periods (25 + 18 − 7 = 36 is wrong, and the error compounds with size and time). The average annual pace of that three-year run is not (25 + 18 − 7)/3 = 12% either. It is the rate that, compounded three times, gives 1.37175: 1.37175^(1/3) − 1 = 11.11%. That number has a name.

Growth rate versus level: the deceleration trap. "Inflation fell from 6% to 4%" means prices are still rising, just more slowly. "Revenue growth decelerated from 24% to 12%" means revenue hit a new all-time high, at a slower pace. The rate is the slope, the level is the altitude, and a falling slope is not a falling altitude. Newspapers, managements, and beginners conflate these constantly ("India's GDP growth is slowing" ≠ "India's GDP is shrinking"). One subtle second layer: when growth halves from 24% to 12%, this year's absolute addition is 0.12 × 1.24 = 0.149 of the old base, smaller than last year's 0.24, even though the base grew. Check the arithmetic before assuming either direction.

4.2 CAGR: the analyst's speedometer

The compound annual growth rate answers: at what single steady rate would the start value have grown into the end value?

CAGR = (End / Begin)^(1/n) − 1, where n = number of years between the two measurements (End = ending value, Begin = starting value).

The fence-post rule. FY2020 → FY2025 is five growth intervals, even though you are holding six annual numbers. Count the gaps, not the posts. Dividing by the number of data points instead of intervals is the single most common CAGR error in amateur write-ups (it flatters nothing and embarrasses everyone).

Worked example, TCS (India). TCS revenue grew from roughly ₹95,000 crore (FY2015) to roughly ₹2,55,000 crore (FY2025). Illustrative, modeled on actual ~FY15/FY25 scale; pull the annual reports from tcs.com or screener.in for the exact figures. Ten intervals:

CAGR = (2,55,000 / 95,000)^(1/10) − 1 = (2.684)^0.1 − 1 = 10.4%

Worked example, Apple (US). Apple revenue grew from roughly $183bn (FY2014) to roughly $391bn (FY2024). Illustrative, modeled on actual 10-K scale; pull the 10-Ks from SEC EDGAR. Ten intervals:

CAGR = (391 / 183)^(1/10) − 1 = (2.137)^0.1 − 1 = 7.9%

Sanity-check with doubling logic (next section): Apple's revenue roughly doubled in ten years; the Rule of 72 says a double in ~10 years implies ~7.2%; the exact answer 7.9% is in the neighborhood. Cross-checking CAGRs against doubling counts catches keystroke errors instantly, so make it a habit.

So TCS "grew 2.5 points faster" than Apple. Hold that thought. Once the Fisher relation arrives we will re-run this comparison in real terms, and most of the gap will vanish.

CAGR's honest limits. (1) It uses only two endpoints, so a cherry-picked start (a crisis trough) or end (a bubble peak) manufactures any story you like; glance at the full path. (2) It is a geometric average, so it is always at or below the arithmetic average of the yearly rates when those rates vary, for the volatility-drag reason set out later. (3) If the business changed shape between the endpoints (divestitures, mergers, accounting changes), the CAGR mixes growth with perimeter change, and Worked example 2 shows Coca-Cola's revenue CAGR collapsing for exactly this reason.

4.3 Simple vs compound interest: and where the Rule of 72 comes from

Simple interest pays only on the original principal. For principal P, annual rate r (decimal), time t years:

FV_simple = P × (1 + r·t) (FV = future value)

Interest accrues linearly: ₹1,00,000 at 8% earns ₹8,000 every year, forever. You meet simple interest in short-term instruments and quotes (many money-market conventions, penalty interest, coupon amounts between payment dates), and in any arrangement where the interest is paid out and not reinvested.

Compound interest pays interest on interest. The balance itself becomes the base:

FV_compound = P × (1 + r)^t

The reinvestment is the whole trick. Watch ₹1,00,000 at 8%:

At year one they are identical; at year 40 compounding has produced five times the simple-interest outcome. The line is straight; the curve bends away from it forever. Every "the first decade looks unimpressive" story about wealth, and every "how did that company get so big" story about businesses reinvesting profits at high returns, is this table. (A company that earns 20% on its capital and reinvests it is running (1.20)^t on your behalf, which is why Phase 2 will teach you to measure ROIC so carefully.)

Think in doublings. The most useful mental unit for compound growth is not the percent but the double. 1 → 2 → 4 → 8: three doublings is 8×. If you know how long one doubling takes, you can size decades in your head. At 12%, one doubling takes about 6 years, so 18 years ≈ 8×; at 6%, a doubling takes about 12 years, so the same 18 years ≈ 2.8×. Count doublings, not percentages.

Deriving the doubling time, and the Rule of 72. How long does money take to double at rate r? Solve for t:

(1 + r)^t = 2

Take logarithms of both sides. (A logarithm just asks "what exponent produced this number?", and ln is the natural log, the log with base e ≈ 2.71828.)

t × ln(1 + r) = ln 2, so t = ln 2 / ln(1 + r), and ln 2 = 0.6931…

That is exact. Now the approximation that makes it mental math: for small r, ln(1 + r) ≈ r (check: ln(1.08) = 0.0770, close to 0.08). Substitute:

t ≈ 0.693 / r = 69.3 / (r in %)

So the honest rule is the "Rule of 69.3." Why does everyone use 72? Two reasons. First, ln(1+r) is always slightly smaller than r, so dividing by r understates the true doubling time, and using a slightly bigger numerator corrects the bias in the range of rates humans actually meet. At 8%, the exact answer is 9.01 years; 69.3/8 = 8.66, noticeably short; 72/8 = 9.00, nearly perfect. Second, 72 divides beautifully: by 2, 3, 4, 6, 8, 9, 12. The rule is a calibrated instrument, not folklore:

Use it between roughly 4% and 15%; outside that band, compute exactly. Companion rules, same derivation with ln 3 and ln 4: Rule of 114 for tripling (114/r), Rule of 144 for quadrupling (two doublings). And the rule runs on anything growing steadily: at ~7% real GDP growth an economy doubles in ~10 years; at 6% inflation the price level doubles in ~12 years, which means the purchasing power of an unlucky rupee halves every 12 years. The Rule of 72 is also an inflation ruler.

4.4 Compounding frequency: APR, EAR, and the fine print

A quoted rate is incomplete until you know how often it compounds. Banks and lenders quote a nominal annual rate (in loan contexts, the APR, or annual percentage rate) of r, compounded m times per year. Each period credits r/m, and there are m·t periods:

FV = P × (1 + r/m)^(m·t)

The effective annual rate (EAR) is the single once-a-year rate that would produce the same result, and it is the only honest basis for comparing offers:

EAR = (1 + r/m)^m − 1

Twelve percent nominal, compounded at different frequencies:

Two lessons in the table. First, frequency always helps the side receiving interest, since EAR rises with m. Second, the benefit saturates fast: daily to continuous adds almost nothing. (The limiting case, FV = P·e^(r·t), is continuous compounding. You will meet it again in option pricing years from now; for now just recognize it as the ceiling of the table.)

India, in practice. Indian bank fixed deposits typically compound quarterly: an FD quoted at 7% nominal actually yields (1 + 0.07/4)^4 − 1 = 7.19% EAR, which is the "annualised yield" printed beside the rate on the bank's page. Credit cards quote monthly rates of about 3–3.6% per month; 3.5%/month sounds like 42% a year, but the EAR is (1.035)^12 − 1 = 51.1%. Unpaid card debt is a doubling machine running at ln 2 / ln(1.035) = 20.1 months per double. It is also where the Rule of 72 stops being safe, so watch what happens to it. Applied to the EAR, 72/51 = 16.9 months runs three months fast, because 51% sits far outside the 4% to 15% band the rule is calibrated for. Applied to the nominal, 72/42 = 20.6 months is almost exactly right, and only by luck: taking the nominal instead of the EAR pushes the estimate up by roughly what the rule's high-rate undershoot pulls it down. Two errors cancelling is not a method. At these rates, compute exactly.

US, in practice. Deposit products advertise APY (annual percentage yield), which is simply EAR under the Truth in Savings Act, so US savers can compare directly. Loans advertise APR: a 24% APR card compounding daily is (1 + 0.24/365)^365 − 1 = 27.1% EAR.

The analyst's rule: never compare quoted nominal rates across different compounding frequencies; convert both to EAR first. An 8.6% deposit compounded quarterly (EAR 8.88%) beats an 8.8% deposit compounded annually. The fine print is worth 8 bp here; on loan products it is routinely worth hundreds.

4.5 Inflation: the machine that shrinks the ruler

What it is, mechanically. Inflation is a sustained rise in the general price level, not the price of one thing but of the average basket of things people buy, so that each rupee or dollar buys less than it did. Its deep causes (money and credit growth outrunning the economy's capacity to produce; supply shocks; expectations feeding on themselves) are Phase 7's territory. Here you need the measurement machine and the arithmetic consequences.

How CPI is built. A statistics agency (MOSPI in India, the BLS in the US) defines a basket of goods and services with weights reflecting household spending shares, prices the basket every month, and expresses the cost as an index relative to a base period set to 100. Inflation is the percentage change of that index, which means measuring inflation is just computing a growth rate of an index, a skill you already have.

Mini-computation with stylized weights (real CPI baskets have hundreds of items; these three groups make the mechanics transparent):

The index moves 100 → 104.44. Now re-run the identical price moves with US-style weights, where food is ≈ 13% instead of 46%, because the US spends a far smaller share of income on food: 0.13(8) + 0.77(4) + 0.10(−10) = +3.12%. Same world, different basket, different measured inflation. This is why Indian CPI dances with food prices and the monsoon while US inflation is steadier: food and beverages carry roughly a 46% weight in India's CPI against roughly 13–14% in the US index (weights as of the current base series; verify at mospi.gov.in and bls.gov/cpi, and note that India has been preparing a re-based series).

The two countries' inflation character. India, modern era: since formal inflation targeting began (2016), CPI inflation has mostly run ~4–6%, and the RBI's mandate is 4% ± 2%; in the decade before targeting, 8–10% prints were common (as of mid-2026, verify against MOSPI data and the RBI's annual report). The US: roughly 2–3% over recent decades, with the Fed targeting 2% on a related index (PCE, which differs from CPI in weights and scope, with the details left to the macro phase); the 2021–22 episode spiked US CPI to ~9% before settling (verify at bls.gov or FRED).

Two consequences to internalize now. First, by the Rule of 72: at 5% inflation, Indian prices double every ~14 years; at 2.5%, US prices double every ~29. A rupee left idle loses half its purchasing power in a working career's first third. Second, every Indian nominal number carries roughly twice the inflation of its US counterpart, whether it is a growth rate, an interest rate, or a salary hike. Cross-country comparisons that ignore this are systematically wrong, which is why the next section exists.

(One more measure you will hear in India: WPI, wholesale price inflation. It tracks producer-level prices, swings harder with commodities, and is not the RBI's target. When a company's costs track WPI while its prices track CPI, the gap lands in its margins, which is a Phase 5 theme.)

4.6 Real vs nominal: the Fisher relation

The nominal ruler counts currency units; the real ruler counts purchasing power. The conversion is one line, named for the economist Irving Fisher:

(1 + nominal) = (1 + real) × (1 + inflation)

Solve for what you need:

real = (1 + nominal) / (1 + inflation) − 1 (exact) real ≈ nominal − inflation (approximation)

The approximation drops the cross-term real × inflation, so it overstates the real rate; the error is roughly the product of the two rates. At Indian magnitudes it already matters: a 12% return against 5% inflation is not 7% real but 1.12/1.05 − 1 = 6.67%. At high inflation it collapses: 60% nominal against 45% inflation is not 15% real but 1.60/1.45 − 1 = 10.3%. Rule: quote the exact number; use the approximation only as a mental cross-check, and never in high-inflation settings.

Negative real rates, the polite theft. A savings account paying 3.5% against 6% inflation delivers 1.035/1.06 − 1 = −2.36% real. Nothing defaulted, the balance grew, and the saver became poorer. India lived this in 2020–22, the US in 2021–22 (verify the episodes against RBI/FRED series). "Safe" assets are safe on the nominal ruler only.

Taxes bite the nominal ruler too. Indian FD interest is taxed at slab rates on the nominal interest. Take a 7% FD, a 30% slab, 5% inflation: after-tax nominal = 7 × 0.70 = 4.9%; real = 1.049/1.05 − 1 ≈ −0.1%. The entire pre-tax real return of ~1.9% was consumed by a tax levied partly on inflation itself. This single line of arithmetic explains much of Indian household finance, and it previews why analysts always ask whose return, after what frictions. (Mechanical illustration, not tax advice; slabs change, so verify current rates.)

Long horizons: deflate the outcome. ₹10 lakh compounding at 12% for 20 years grows to 10,00,000 × 1.12²⁰ = ₹96.5 lakh nominal. At 5% inflation the price level multiplies by 1.05²⁰ = 2.653, so the real value is 96.46/2.653 = ₹36.4 lakh in today's rupees. Equivalently, the real rate 1.12/1.05 − 1 = 6.67% compounded 20 years gives the same 3.64×. Both routes always agree; use whichever is convenient, but report long-horizon outcomes in today's money or you will systematically overpromise.

The TCS–Apple reveal. TCS ran ≈ 10.4% nominal revenue CAGR (FY15→FY25), Apple ≈ 7.9% (FY14→FY24). Decade-average inflation: India ≈ 5.0%, US ≈ 2.9% (approximate decade averages; verify against MOSPI and BLS data). Real growth:

• TCS: 1.104 / 1.050 − 1 = 5.1%
• Apple: 1.079 / 1.029 − 1 = 4.9%

A 2.5-percentage-point nominal gap shrinks to about a quarter of a point. The two firms grew their real output at almost the same pace, and the rest was the two countries' inflation differential.

This is the single most important habit this bootcamp installs: never compare nominal growth rates across countries or eras. The same correction applies to interest rates (a 7% Indian G-sec and a 4.3% US Treasury may carry similar real yields), to salary offers, and, once you reach Phase 3, to discount rates, where mixing a nominal cash-flow forecast with a real discount rate or the reverse is a classic valuation-destroying error. Currency-and-inflation consistency is one rule, stated once: match the ruler on both sides of every comparison.

4.7 Risk and return: why more doubt must be paid for

Strip away the mathematics and the bond market, and risk-pricing is a lending decision you already know how to make. Your most reliable friend asks to borrow ₹1 lakh for a year; a colleague with a patchy record asks for the same. Same amount, same year. Would you charge the same rate? No. With the colleague there is a real chance you get back less than you lent, so you demand extra return just to break even on average, and something beyond that for carrying the worry. That instinct, priced by millions of participants, is the entire risk-return structure of markets.

Decompose any quoted return (the rate stack). A nominal expected return is compensation for three separable things:

nominal expected return ≈ real riskless rate (waiting) + expected inflation (ruler shrinkage) + risk premium (doubt)

A US Treasury yield is the first two with near-zero doubt premium in nominal terms; an Indian G-sec sits higher mostly because expected rupee inflation is higher; a corporate bond adds a default spread; equity adds the fattest premium because its outcomes are the most dispersed. When Phase 3 builds discount rates (CAPM, WACC) and estimates the equity risk premium, it is formalizing exactly this stack. When you see any return quoted, mentally unstack it: how much is waiting, how much is inflation, how much is being paid for doubt?

Dispersion, without statistics. Compare two one-year investments, each costing ₹100 today:

• A pays ₹108 for certain: return 8%, always.
• B pays ₹80, ₹108, or ₹136 with equal chances: average payoff ₹108, the same 8% "on average."

Same average, different spread of outcomes. Nearly everyone prefers A, because a certain ₹108 is worth more to a person than a lottery averaging ₹108: the pain of landing on ₹80 outweighs the pleasure of ₹136 (Phase 6 explores why that asymmetry is wired into us). So at a price of ₹100, nobody buys B. What happens? B's price falls until buyers appear.

At ₹93, B's average payoff of ₹108 represents an expected return of 108/93 − 1 = 16.1%. The 8-point gap over A is the risk premium, and notice the mechanism. The premium is not a gift attached to risky assets. It emerges from a lower price. Risk premiums are made of discounts.

Three consequences, each load-bearing for the rest of the program:

1. Expected ≠ guaranteed. The premium is compensation for a spread of outcomes that includes bad ones, since otherwise it would not exist. Equity's premium sometimes fails to show up for a decade: US stocks returned slightly less than nothing over 2000–2009; the Sensex in 2003 was still near its 1994 peak; Japan's Nikkei took ~34 years to regain its 1989 high (all approximate; verify against index factsheets). If the premium were reliable, it would be arbitraged to zero. The equity premium is real because those decades happen.
2. Price is the dial. For a given asset, a lower purchase price mechanically raises the expected return, and a higher one lowers it. "What is the price implying?", question 5 of the five, is this sentence turned into a method.
3. Only unavoidable doubt gets paid. If you can cancel a risk cheaply, by holding many uncorrelated assets, the market will not pay you for carrying it anyway. The full argument, covering diversification and systematic against idiosyncratic risk, belongs to the valuation phase. Plant the flag now.

Dispersion compounds badly. Here is the bridge between risk and compounding, and it needs no statistics: −50% followed by +100% is 0% (0.50 × 2.00 = 1.00), even though the arithmetic "average" of those returns is +25%. Spread in returns always drags the compound (geometric) growth rate below the arithmetic average, and the wilder the path, the bigger the drag. Two funds with the same average return but different volatility end at different wealth, and the calmer one ends richer. The long-horizon section quantifies this; Phase 9 weaponizes it.

One honest caveat before moving on. Risk is treated here as dispersion of outcomes because that is the cleanest first window. Practitioners you will study later (Buffett, Marks) argue the deeper definition is permanent loss of purchasing power: a −50% quote you never sell into is pain, not necessarily loss, while a "stable" bond quietly losing to inflation is loss dressed as safety. Both framings earn their keep, and Phase 9 adjudicates. For now: dispersion for pricing intuition, purchasing power for life outcomes.

4.8 The asset-class map: historical character, India and US

You now have the rulers (nominal, real) and the concept (return as payment for waiting, inflation, and doubt). Here is the empirical landscape those tools measure.

Every figure below is an order-of-magnitude historical anchor, not a promise. Long-run averages conceal brutal decades, and the future is under no obligation to resemble the sample. Figures are approximate as of mid-2026, and you should verify against Damodaran's "Historical Returns on Stocks, Bonds and Bills: 1928–current" page (updated each January), the UBS Global Investment Returns Yearbook, RBI's Handbook of Statistics on the Indian Economy, and BSE/NSE index factsheets. Use the TRI, or total-return index, versions, which include dividends; price-only indices understate equity returns by the ~1–1.5% dividend yield.

Reading the table like an analyst:

• Equities compound highest because they are the claim on business itself, and because their path is the most violent, which is why the premium exists. The Sensex went from base 100 (1979) to ~80,000 (2024), ≈ 16% nominal price CAGR, but Indian inflation averaged ~7% over much of that sweep, so the real story is nearer 8%, and it included 2008's −52% and 2009's +81% (verify levels and episodes against BSE factsheets).
• Bonds are the inflation-vulnerable middle. Fixed nominal coupons are exactly what a surprise inflation ruins, and the 2022 US episode was the textbook case.
• Cash is short-horizon safety, long-horizon leak. After the tax arithmetic above, Indian FDs at typical slabs have historically hovered near zero real. Cash's job is optionality and survival, not growth.
• Gold is an anti-currency, not a compounder. It produces nothing, so its return is other people's fear of paper money plus, for Indian holders, rupee depreciation (the INR/USD leg adds ~3–4%/yr over long stretches to gold's USD return; verify). It shines in crises and inflation scares and can then sleep for a generation, spending 28 years under its 1980 peak. Starting point dominates its measured "return": 1971, just after the gold peg ended, flatters it; 1980 damns it.
• Real estate is a business with a bad ledger. Rent yield plus price growth, minus repairs, taxes, vacancy, and enormous transaction costs. Computed honestly (see Worked example 4's cousin in the practice set), the "my uncle's flat quadrupled" story usually resolves to ~7% nominal, ~2% real, plus rent, minus headaches. Its true edges are leverage, and the discipline of illiquidity, since you cannot panic-sell at 2 a.m.

Two disclaimers that are part of the content, not boilerplate. First, these are teaching anchors: this program teaches analysis, not allocation, and nothing here is advice to buy any of these. Second, the ranges above are arithmetic memories of one particular century. Survivorship (the US was history's winning market), regime changes, and starting valuations all matter, and Phase 6 will teach you to distrust extrapolation precisely when it feels safest.

4.9 Long-horizon compounding: the power, and the honest limits

The power, in three exhibits.

Back-loading. ₹1 compounding at 12%: after 10 years ₹3.11, after 20 ₹9.65, after 30 ₹29.96, after 40 ₹93.05. Decade contributions: +2.11, +6.54, +20.31, +63.09. The fourth decade adds thirty times what the first did, at the same rate. Compounding pays almost all its wages at the end, which is why interrupting it (panic-selling, dipping into the corpus, forced liquidation) is so expensive, and why "time in" beats "timing" as a default posture. The canonical human exhibit: the overwhelming majority of Warren Buffett's wealth arrived after his 50th birthday (commonly cited from his disclosed net-worth history; verify the current figure, since the point is the shape rather than the decimal).

Small edges, huge terminal gaps. 12% for 30 years multiplies money 29.96×; 15% does 66.2×, so three extra points more than double the terminal outcome. This cuts both ways:

Costs compound too. A 100 bp annual fee on a 12% gross return (net 11%) turns 30-year wealth from 29.96× into 22.89×, so the "one percent" quietly consumed 24% of the terminal outcome. Every basis point of recurring cost is a permanent passenger on your compounding train. (This is also why Phase 2 obsesses over ROIC durability: a business compounding capital at 20% against 12% for two decades is not "8 points better," it is 1.20²⁰/1.12²⁰ = 3.97× better.)

The SIP arithmetic (India's favorite habit). A monthly investment A for n months at monthly rate r_m grows to:

FV = A × [((1 + r_m)^n − 1) / r_m]

₹10,000 monthly for 20 years at 1% per month (the common calculator convention for "12%"; note that 1% monthly is really 12.68% EAR): FV = 10,000 × [(1.01²⁴⁰ − 1)/0.01] = 10,000 × 989.3 = ₹98.9 lakh, on ₹24 lakh contributed. Now apply the real-versus-nominal discipline: at 5% inflation, that is 98.9/2.653 = ₹37.3 lakh in today's rupees. Still excellent, and comfortably ahead of the ₹24 lakh put in, but a third of the sticker number. Selling the sticker without the deflator is how bad financial marketing works; you now know better.

The honest limits, four of them.

1. The average hides the path (sequence risk, a teaser). For a lump sum with no additions or withdrawals, the order of returns is irrelevant: 0.75 × 1.10 × 1.30 = 1.0725 in any order, because multiplication commutes. The moment money flows in or out, order matters enormously. A retiree withdrawing a fixed sum is hurt worst by bad years early, since each withdrawal locks in the damage; a SIP accumulator is helped by bad years early, since contributions buy cheap. Worked example 3 computes both on the same return set. The full treatment, covering safe withdrawal, cash buffers, and why "average return 12%" can still ruin a retiree, is Phase 9's. The sentence to retain now: compounding math uses one number, life delivers a sequence, and with cash flows the sequence is destiny.
2. Volatility drag, quantified. Steady 6% for two years: 1.06² = 1.1236. "Average 10%" delivered as +40% then −20%: 1.40 × 0.80 = 1.12. The calmer fund wins with a four-point-lower arithmetic average. General habit: judge multi-year records by CAGR (geometric), never by the arithmetic mean of yearly returns; the gap between them is the volatility drag.
3. The engine must survive the whole runway. Projecting 15% for 25 years assumes a business (or an investor) that keeps earning reinvestment opportunities at that spread for 25 years. Almost nothing does. Competition arrives, markets saturate, returns mean-revert. This is why the market pays so much for the rare durable compounder, and why Phase 4 exists, since moats are exactly the thing that defends a reinvestment rate. Damodaran's data and Mauboussin's base-rate work (Phase 3 and 4 anchors) show sustained 20%+ growers are a tiny minority of large firms. When you hear "just compound at 18%," ask: what defends the 18% in year 14?
4. Inflation compounds against you, in parallel. Every nominal projection carries a shadow projection of the price level. At Indian inflation, 25-year nominal numbers roughly triple-count purchasing power (1.05²⁵ ≈ 3.4). Always run the deflator.

One closing connection to the program's spine. The five questions ask whether a business earns ROIC above its cost of capital and can reinvest at that spread. You can now read that sentence numerically: ROIC is the business's compounding rate, the cost of capital is the rate stack the providers of money demand for waiting, inflation and doubt, and reinvestment is what turns a good year into an exponent. A company reinvesting at 20% against a 12% hurdle is 1.20^t racing 1.12^t, and everything in Phases 2–5 is machinery for judging whether that exponent is real and how long it can run.

Common mistakes & how experts think differently

1. Percentage points vs percent. "Margin grew 3%" when it moved 18% → 21.5% (that is 3.5 pp, a 19% relative rise). The expert convention: levels of ratios in %, changes in pp or bp. When reading others' work, always reconstruct which one they meant before trusting the sentence.

2. Averaging yearly returns arithmetically. A fund reporting +40%, −20% did not "average 10%" in any sense that matters to your wealth. It compounded at 5.8%. Experts treat the arithmetic mean of returns as a red flag in marketing material and recompute the CAGR from endpoint NAVs in seconds.

3. Fence-post errors in CAGR. Six annual figures = five intervals. Experts also sanity-check every CAGR against a doubling count (Rule of 72), a habit that catches both fence-post and keystroke errors instantly.

4. Comparing nominal across countries or eras. "Indian companies grow faster" is usually mostly an inflation differential, as the TCS and Apple comparison showed. Experts convert to real terms (or to a common currency, which is related, as you will see in Phase 7) before any cross-border comparison of growth, rates, or returns.

5. Reading "inflation fell" as "prices fell." Disinflation (slower rises) is not deflation (falls). Experts speak in derivatives without confusing them: level, growth, change in growth are three different claims. Same for "revenue growth decelerated", where the company still grew.

6. Ranking rates without converting to EAR. An 8.8% annual deposit vs 8.6% quarterly; a 42%-nominal card vs its 51% EAR. Experts refuse to compare quotes until frequencies match, and read the compounding convention before the rate.

7. Treating expected return as an entitlement. "Equities give 12%" is wrong. Equities have historically averaged something like that in India, nominal, with individual decades ranging from glorious to negative. Experts hold expected returns as the center of a wide distribution and ask what happens to their plan on the left tail. The premium exists because it sometimes fails to arrive.

8. Extrapolating the recent character of an asset class. Gold after a hot decade, real estate after a boom, equities after a bull run: the just-lived sample feels like the asset's nature. Experts anchor on the longest data available (and its limits: survivorship, regime change) and get more skeptical as recent performance improves, not less.

9. Dismissing basis points. "It's only 1%", and yet a 100 bp fee at 12%/30 years costs ~24% of terminal wealth; a 130 bp expense-ratio gap on ₹25 lakh over 15 years is ~₹18 lakh. Experts price every recurring cost as a share of the terminal outcome, not of this year's balance.

10. Misplacing sequence risk. Believing return order matters for a lump sum, when it does not because multiplication commutes, or believing that it does not matter for SIPs and retirements (it dominates them). Experts ask one question first: are there cash flows? No flows → only the compound rate matters. Flows → the path matters as much as the average.

The meta-difference: an expert runs a sanity-check reflex on every number, checking order of magnitude, doubling count, real against nominal, EAR against quote, and growth against level, in under ten seconds, before believing it or repeating it. The drills below install that reflex.

Worked examples

Worked example 1: An Indian saver's full chain: quote → EAR → after-tax → real

Setup. ₹10,00,000 in a 1-year bank FD quoted at 7.1% nominal, compounded quarterly. Saver's slab rate 30%. CPI inflation 5%. What really happened to purchasing power? (Synthetic but typical numbers; check any large bank's current FD card rates.)

1. EAR from the quote: quarterly rate = 7.1/4 = 1.775%. EAR = (1.01775)⁴ − 1 = 7.291%. Interest earned: ₹72,913.
2. After tax: 72,913 × (1 − 0.30) = ₹51,039 → after-tax nominal return 5.104%.
3. Real, exact Fisher: (1.05104 / 1.05) − 1 = +0.099%.
4. In rupees of purchasing power: 10,51,039 / 1.05 = ₹10,00,990 → real gain ≈ ₹990 on ten lakh.

Read. Nominal +₹72,913; truth ≈ +₹990, so a year's real reward is one good dinner. Nothing was unsafe, and almost nothing was earned. That chain, from quote to effective to after-friction to real, is the same one you will later run on bond yields and company returns.

Worked example 2: CAGR done right, and a CAGR that lies (Asian Paints vs Coca-Cola)

Asian Paints (India). Revenue ≈ ₹12,700 crore (FY2014) → ≈ ₹35,500 crore (FY2024). Illustrative, modeled on actual scale; pull the annual reports or screener.in. Ten intervals:

CAGR = (35,500/12,700)^(1/10) − 1 = (2.795)^0.1 − 1 = 10.8%. Cross-check by doubling: 72/10.8 ≈ 6.7 years per double → ten years ≈ 1.5 doubles → 2^1.5 ≈ 2.83× ≈ the observed 2.80×. Consistent.

Coca-Cola (US). Revenue ≈ $46bn (FY2014) → ≈ $47bn (FY2024). Illustrative, modeled on actual 10-K scale; pull from SEC EDGAR. CAGR = (47/46)^(1/10) − 1 = 0.2%.

Read. Did Coca-Cola stagnate for a decade? No. It refranchised its capital-heavy bottling operations during this period, deliberately shrinking reported revenue while margins and per-share economics improved. The 0.2% is arithmetically correct and analytically misleading: the perimeter changed between the endpoints, so the CAGR mixes growth with corporate surgery. Endpoint CAGRs are only meaningful over a constant perimeter — when the number looks weird, the answer is in what changed in the business, which is exactly where Phase 1's segment notes and MD&A reading will take you.

Worked example 3: Sequence risk: same returns, three lives

Setup. Three-year return set: −25%, +10%, +30% (product: 0.75 × 1.10 × 1.30 = 1.0725, i.e., +7.25% total, 2.36% CAGR). Start ₹100 (think lakhs). Compare orders Bad-first (−25, +10, +30) and Good-first (+30, +10, −25) for three investors.

Investor L (lump sum, no flows): Bad-first: 100 → 75 → 82.5 → 107.25. Good-first: 100 → 130 → 143 → 107.25. Identical, because order is irrelevant without flows.

Investor R (retiree, withdraws ₹6 at each year-end after returns):

Same returns, same withdrawals, and ending wealth differs by ₹6.93. Bad years early forced the retiree to sell depleted units; the recovery had less to work with.

Investor S (SIP-style, adds ₹6 at each year-end): Bad-first: 75+6=81 → 81×1.10+6=95.10 → 95.1×1.30+6 = 129.63. Good-first: 130+6=136 → 136×1.10+6=155.60 → 155.6×0.75+6 = 122.70. The preference reverses: the accumulator wants the bad years early (contributions buy cheap).

Read. One return set, three outcomes. "The market averaged X%" tells you Investor L's story only. Whenever money moves in or out, whether through SIPs, redemptions, retirements, or a company's buybacks (Phase 3 will show the same math decides whether buybacks create value), the path is a first-class variable. Full treatment in Phase 9.

Worked example 4: Basis points with teeth: direct vs regular mutual fund plans

Setup. Same Indian equity fund, two share classes: direct plan expense ratio 0.50% (50 bp), regular plan 1.50% (150 bp), a 100 bp gap (typical order of magnitude; check any AMC's factsheet). Gross portfolio return 12%. Lump sum ₹10,00,000 for 20 years.

1. Net returns: direct 12 − 0.5 = 11.5%; regular 12 − 1.5 = 10.5% (fees accrue continuously against NAV; annual netting is a fair approximation).
2. Direct: 10,00,000 × 1.115²⁰ = 10,00,000 × 8.8206 = ₹88.21 lakh.
3. Regular: 10,00,000 × 1.105²⁰ = 10,00,000 × 7.3662 = ₹73.66 lakh.
4. Gap: ₹14.55 lakh, i.e., 16.5% of the direct-plan outcome, for an input difference of "one percent."

Read. The 100 bp did not cost 1%. It cost a sixth of the terminal wealth, because it was charged on a compounding base every year for twenty years. Reflex to install: convert any recurring bp figure into share of terminal outcome before calling it small. (The identical arithmetic will reappear as "2-and-20" analysis, brokerage drag, and — inverted — as why a durable 100 bp ROIC edge between two businesses is worth a fortune.)

Worked example 5: An Indian home loan: the EMI, the interest bill, and what inflation quietly does to it

Setup. The Sharma family (synthetic) takes a ₹50,00,000 home loan at a 9% annual rate compounded monthly (the standard Indian floating-rate convention) over a 20-year tenure. What is the monthly EMI, how much interest is paid over the life of the loan, and what does the fixed payment really cost in later years? (Illustrative rate; check any bank's current home-loan card.)

The EMI (equated monthly instalment) is the level monthly payment that exactly repays principal P over n months at monthly rate r. It is the annuity-payment mirror of the SIP future-value formula below:

EMI = P × r × (1 + r)^n / ((1 + r)^n − 1)

1. Set up the periods. Monthly rate r = 0.09/12 = 0.0075; number of payments n = 20 × 12 = 240; growth factor (1.0075)^240 = 6.00915.
2. The EMI. EMI = 50,00,000 × 0.0075 × 6.00915 / (6.00915 − 1) = ₹44,986/month.
3. The total interest bill. Total paid = 44,986 × 240 = ₹1,07,96,711 ≈ ₹1.08 crore; principal is ₹50 lakh, so interest ≈ ₹57.97 lakh, more than the amount borrowed. A long tenure is expensive precisely because compounding runs against you: month 1's interest alone is 50,00,000 × 0.0075 = ₹37,500 of the ₹44,986 payment, so the early EMIs are almost all interest and the principal is chipped away only slowly.
4. The true annual cost. The "9%" is a nominal monthly-compounded rate; its EAR is (1.0075)^12 − 1 = 9.38%. You pay 9.38% effective, not 9.00%, so always convert a loan quote to EAR before comparing.
5. What inflation does to the fixed payment. The EMI is fixed in nominal rupees, but prices and incomes are not. Deflate it and the ₹44,986 you pay in the final year is worth 44,986 / 1.05^20 = 44,986 / 2.6533 = ₹16,955 in today's purchasing power at 5% inflation. The nominal burden never changes; the real burden falls by more than half across the loan's life.

Read. Two forces collide here. The interest bill, ₹58 lakh on a ₹50 lakh loan, is the compounding table pointed at you instead of for you. But the deflator is on the borrower's side: a fixed-rate loan is repaid in ever-cheaper rupees, which is why moderate inflation quietly rewards the borrower of fixed-rate debt and punishes the lender receiving those fixed payments. One real-vs-nominal ruler, read from both ends.

Worked example 6: Pricing a risky payoff in dollars: the risk premium is a discount

Setup. A synthetic one-year note, the "Cascade Note (synthetic)," will pay $80, $110, or $140 with probabilities ¼, ½, ¼. A one-year US Treasury bill yields 4% with effectively no doubt. What should the note cost, and where does its "risk premium" actually come from? (This is the dispersion argument turned into arithmetic.)

1. Expected payoff. E = 0.25 × 80 + 0.50 × 110 + 0.25 × 140 = $110. The average outcome is $110, the same number a certain payoff of $110 would give.
2. The risk-free price nobody will pay. If the note were as safe as the T-bill it would be priced to yield 4%: 110 / 1.04 = $105.77. But it is not safe, and it might pay only $80. A risk-averse buyer, offered a dispersed "$110 on average" for $105.77, declines: the pain of landing on $80 outweighs the pleasure of $140.
3. The price falls until buyers appear. Suppose the market demands a 10% expected return to carry this dispersion. Then price = 110 / 1.10 = $100.00, and the realised expected return is 110/100 − 1 = 10%. The risk premium is 10% − 4% = 6 percentage points. Note its mechanism: it did not arrive as a bonus bolted onto the note; it emerged from the price falling from $105.77 to $100. Risk premiums are made of discounts. If fear deepened and buyers demanded 15%, the price would fall further to 110 / 1.15 = $95.65.
4. Unstack it in real terms, using the rate stack. At 2.5% US inflation, the real risk-free rate is 1.04/1.025 − 1 = 1.46% and the note's real expected return is 1.10/1.025 − 1 = 7.32%. The nominal 10% decomposes into roughly real waiting (~1.5%) + expected inflation (~2.5%) + risk premium (~6%), the three layers Phase 3's CAPM and WACC will formalize.

Read. Same $110 average, three different prices ($105.77, $100.00, $95.65), depending only on how much return buyers demand for the spread of outcomes. This is question 5 in embryo: for a given payoff, price is the dial that sets expected return. A lower price mechanically raises the return, which is why "what is the price implying?" is the same sentence as "what return is on offer here?" A guaranteed $110 would command the full $105.77; every dollar below that is the market pricing doubt.

Practice set

Twelve problems in three ramps, guided (worked with you), independent (you first, then check) and timed (clock running), followed by the timed gauntlet, three rapid-fire drill sheets you repeat until fast. Write your answer fully before reading any solution (working agreement 3). Calculator or spreadsheet allowed throughout; no peeking at the teaching sections.

Ramp 1: Guided

P1. A stock falls 40%. What percentage gain returns it to its starting price? Then generalize: why is the recovery gain always larger than the loss?

Solution. Let the start be 100. After −40%: 60. Needed gain = (100 − 60)/60 = 40/60 = +66.67%. General: after loss L, recovery = L/(1 − L); the denominator is the shrunken base, so recovery > loss, and explosively so for large L (−50% → +100%; −90% → +900%). Asymmetry of the base.

P2. ₹50,000 for 8 years at 9%: future value under (a) simple and (b) annual compound interest, and the gap.

Solution. (a) 50,000 × (1 + 0.09×8) = 50,000 × 1.72 = ₹86,000. (b) 50,000 × 1.09⁸; 1.09⁸ = 1.99256 → ₹99,628. Gap = ₹13,628, the interest-on-interest. Bonus check: 1.09⁸ ≈ 2 is the Rule of 72 in action (72/9 = 8 years to double, and the compound leg almost exactly doubled).

P3. Deposit A offers 8.6% compounded quarterly; Deposit B offers 8.8% compounded annually. Which is better, and by how many basis points of EAR?

Solution. A: EAR = (1 + 0.086/4)⁴ − 1 = (1.0215)⁴ − 1 = 8.881%. B: 8.800%. A wins by ≈ 8 bp. Quotes cannot be compared across frequencies; EARs can.

P4. A stylized CPI basket: weights 50% food, 30% housing, 20% transport; price changes +8%, +4%, −2%. Compute inflation and the new index (base 100). Then recompute with weights 20/60/20 and explain the difference in one line.

Solution. Inflation = 0.50(8) + 0.30(4) + 0.20(−2) = 4.0 + 1.2 − 0.4 = +4.8%; index → 104.8. With 20/60/20: 1.6 + 2.4 − 0.4 = +3.6%. Same prices, different basket → different measured inflation; weights are the lens (this is the India-vs-US CPI story in miniature).

Ramp 2: Independent

P5. Compute real returns, exact and approximate: (a) nominal 13%, inflation 6%; (b) nominal 60%, inflation 45%. Comment on when the approximation is usable.

Solution. (a) Exact: 1.13/1.06 − 1 = 6.60%; approx: 7.00% (error 40 bp). (b) Exact: 1.60/1.45 − 1 = 10.34%; approx: 15.00% (error ~4.7 pp, which is useless). The approximation's error ≈ real × inflation; acceptable for quick checks at low single digits, never at high inflation.

P6. A company's revenue: FY21 ₹800 cr; FY25 ₹1,240 cr. (a) Compute the CAGR. (b) Compute the wrong answer produced by dividing by the count of fiscal years shown (5), and state the trap's name.

Solution. (a) Intervals FY21→FY25 = 4. CAGR = (1240/800)^(1/4) − 1 = 1.55^0.25 − 1 = 11.58%. (b) 1.55^(1/5) − 1 = 9.16%, understated by ~2.4 pp. Fence-post error: count gaps, not posts.

P7. Rule of 72 triple-header: (a) doubling time at 9%; (b) India's nominal GDP grows ~10%, so how many years to double (verify the current rate against MOSPI releases); (c) at 6% inflation, how long until prices double, how long until they are ~8×, and what is ₹1's purchasing power at that 8× point?

Solution. (a) 72/9 = 8 years (exact 8.04). (b) 72/10 ≈ 7.2 years. (c) Double in 72/6 = 12 years; 8× = three doublings ≈ 36 years; price level 8× means ₹1 buys one-eighth (12.5 paise) of today's basket. A career is about three inflation-doublings in India, so plan in real terms.

P8. Two index funds track the same index (gross 11%): expense ratios 45 bp and 175 bp. On ₹25,00,000 over 15 years, compute both terminal values and the fee gap in rupees and as a share of the cheaper fund's outcome.

Solution. Net returns: 11 − 0.45 = 10.55%; 11 − 1.75 = 9.25%. Cheap: 25,00,000 × 1.1055¹⁵ = 25,00,000 × 4.5017 = ₹1.1254 crore. Costly: 25,00,000 × 1.0925¹⁵ = 25,00,000 × 3.7698 = ₹94.25 lakh. Gap = ₹18.30 lakh ≈ 16.3% of the cheaper fund's terminal value, from a 130 bp difference.

Ramp 3: Timed (set a clock; write, then check)

P9. (2 minutes.) A consumer-durables loan quotes 18% nominal, compounded monthly. EAR?

Solution. (1 + 0.18/12)¹² − 1 = 1.015¹² − 1 = 19.56%. (The "18%" costs a sixth more than it says.)

P10. (2 minutes.) Your portfolio returned +14% in a year when CPI inflation was 5.7%. Real return, exact, and the approximation for contrast.

Solution. 1.14/1.057 − 1 = +7.85%. Approx: 8.30%, a 45 bp overstatement. Report the exact.

P11. (3 minutes.) "Company X's revenue growth slowed from 24% to 12% this year." Mark each true/false, with one line of arithmetic where needed: (a) revenue is lower than last year; (b) revenue reached a new high; (c) if 12% persists, revenue doubles in ~6 years; (d) this year's absolute revenue addition necessarily exceeded last year's because the base was bigger.

Solution. (a) False, because growth is positive. (b) True, at +12% on the higher base. (c) True, since 72/12 = 6. (d) False: last year's addition = 0.24B; this year's = 0.12 × 1.24B = 0.149B < 0.24B. Deceleration can shrink absolute additions even on a bigger base; check, don't assume.

P12. (3 minutes.) Fund A returns +6% and +6% over two years. Fund B returns +40% and −20%. Compute each fund's two-year multiple and CAGR; state which arithmetic average is higher and which investor is richer.

Solution. A: 1.06² = 1.1236 → CAGR 6%. B: 1.40 × 0.80 = 1.12 → CAGR = √1.12 − 1 = 5.83%. B's arithmetic average (10%) is far higher; A's investor is richer. Volatility drag: dispersion taxes compounding.

The timed gauntlet: rapid-fire drill sheets

Rules: closed notes, calculator allowed, clock visible. A sheet is "passed" at ≥10/12 correct within the time cap; repeat a failed sheet after a day's gap. Do Sheet A around day 3 of the module, B around day 5, C around day 7; log times and scores in your journal. These 36 items are your muscle memory, and the mastery exam assumes this speed.

Sheet A. Percentages, points, basis points (cap: 6 minutes)

1. Express 0.25% in bp.
2. Express 150 bp as a percent.
3. A 7.25% yield rises 40 bp. New yield?
4. Margin moves 18.0% → 21.5%: change in pp, and relative change in %?
5. ₹480 → ₹552: percent change?
6. A stock falls 20% then rises 20%. Net change?
7. What gain recovers a 25% fall?
8. 8% of ₹6,25,000?
9. A bill of ₹1,180 includes 18% tax on the base amount. What is the base?
10. ₹250 stock: +12% then −12%. Final price?
11. Express 0.0007 as bp.
12. A fee falls from 125 bp to 95 bp. Relative reduction in %?

Sheet A key: 1. 25 bp · 2. 1.50% · 3. 7.65% · 4. +3.5 pp; +19.4% · 5. +15% · 6. −4% · 7. +33.33% · 8. ₹50,000 · 9. ₹1,000 · 10. ₹246.40 · 11. 7 bp · 12. −24% (30/125).

Sheet B. Compounding & doubling (cap: 8 minutes)

1. Doubling time at 6% (Rule of 72)?
2. Doubling time at 12%?
3. At 15%, roughly how long to quadruple?
4. ₹1,00,000 at 10% compound, 2 years — FV?
5. ₹1,00,000 at 10% simple, 2 years — FV?
6. 1.05³ ≈ ?
7. 2% per quarter for one year — effective annual rate?
8. Money doubles in 9 years — implied annual rate (rule)?
9. Real GDP at 7% — years to double?
10. ₹10,000 at 12% for 6 years — FV (use the rule, then exact)?
11. At 3% inflation, prices double in how many years?
12. How long to 8× at 12%?

Sheet B key: 1. ~12 yrs · 2. ~6 yrs · 3. two doublings ≈ 9.6–10 yrs · 4. ₹1,21,000 · 5. ₹1,20,000 · 6. 1.158 · 7. (1.02)⁴ − 1 = 8.24% · 8. ~8% · 9. ~10.3 yrs · 10. rule: ≈₹20,000 (one double); exact 1.12⁶ = 1.9738 → ₹19,738 · 11. ~24 yrs · 12. three doublings ≈ 18 yrs.

Sheet C — Real vs nominal & CAGR (cap: 10 minutes)

1. Nominal 9%, inflation 4% — approximate real?
2. Same numbers — exact real?
3. Nominal 25%, inflation 18% — exact real?
4. Sales ₹400 cr → ₹500 cr in 2 years — CAGR?
5. You hold five annual figures, FY21–FY25 — what is n in the CAGR?
6. Fund NAV 20 → 45 in 9 years — CAGR (hint: more than one double)?
7. You need 6% real when inflation is 5% — required nominal, exact?
8. Salary +8%, inflation 6% — real raise (exact, 2 dp)?
9. Revenue growth decelerates 30% → 15% — did revenue fall?
10. ₹40 lakh nominal in 15 years, 5% inflation — value in today's rupees?
11. US fund: 9% nominal, 2.5% CPI. India fund: 12% nominal, 5.5% CPI. Higher real return?
12. −50% then +50% over two years — CAGR?

Sheet C key: 1. ~5% · 2. 1.09/1.04 − 1 = 4.81% · 3. 1.25/1.18 − 1 = 5.93% · 4. √1.25 − 1 = 11.80% · 5. n = 4 (intervals) · 6. (45/20)^(1/9) − 1 = 9.43% · 7. 1.06 × 1.05 − 1 = 11.30% · 8. 1.08/1.06 − 1 = 1.89% · 9. No, still growing at 15% · 10. 40,00,000/1.05¹⁵ = 40,00,000/2.0789 = ₹19.24 lakh · 11. US: 1.09/1.025 − 1 = 6.34% vs India: 1.12/1.055 − 1 = 6.16% → US (despite the lower sticker) · 12. terminal 0.75 → √0.75 − 1 = −13.4%/yr.

Ramp 4: Extension set (loans, real returns, and risk pricing)

Six further problems in the same three modes, two guided, three independent and one timed stretch, pushing the toolkit into EMIs, loans and risk pricing. Same rule: write your full answer before reading the solution.

P13. (Guided.) A ₹8,00,000 car loan at 10% annual, compounded monthly, over 5 years (60 months). Compute the EMI and the total interest paid, using EMI = P × r × (1 + r)^n / ((1 + r)^n − 1).

Solution. r = 0.10/12 = 0.008333; n = 60; (1.008333)^60 = 1.64531. EMI = 8,00,000 × 0.008333 × 1.64531 / (1.64531 − 1) = ₹16,998/month. Total paid = 16,998 × 60 = ₹10,19,858; interest = ₹2,19,858 ≈ ₹2.20 lakh on ₹8 lakh borrowed. Cross-check: month-1 interest = 8,00,000 × 0.008333 = ₹6,667, so the first EMI is ~39% interest. Amortization always pays interest first.

P14. (Guided.) A US investor earns 6.5% nominal, with inflation at 2.8%. (a) Exact real return. (b) $50,000 compounding at 6.5% for 25 years — nominal value, then value in today's dollars.

Solution. (a) 1.065/1.028 − 1 = 3.60% real (the approximation 3.7% overstates by ~10 bp). (b) Nominal: 50,000 × 1.065^25 = 50,000 × 4.8277 = $241,385. Deflate: 1.028^25 = 1.9945, so 241,385 / 1.9945 = $121,027 in today's dollars, equivalently 50,000 × 1.036^25 = $121,027. Both routes agree; the nominal quarter-million is about $121k in today's purchasing power.

P15. (Independent.) ₹2,00,000 for 12 years at 8.5%: future value under (a) simple and (b) annual compound interest, the gap, and a Rule-of-72 cross-check on the compound leg.

Solution. (a) Simple: 2,00,000 × (1 + 0.085 × 12) = 2,00,000 × 2.02 = ₹4,04,000. (b) Compound: 1.085^12 = 2.6617, so ₹5,32,337. Gap = ₹1,28,337 of interest-on-interest. Cross-check: 72/8.5 = 8.47 years per double, so 12 years ≈ 1.42 doublings → 2^1.42 = 2.67× ≈ the observed 2.66×. Consistent.

P16. (Independent.) Your salary rises from ₹12,00,000 to ₹12,84,000; CPI inflation that year was 5.5%. (a) The nominal raise. (b) The exact real raise. (c) At that real pace, how many years to double your real income (Rule of 72, then exact)?

Solution. (a) 12,84,000 / 12,00,000 − 1 = 7.0%. (b) Real: 1.07/1.055 − 1 = 1.42%. (c) 72/1.42 ≈ 51 years; exact ln 2 / ln(1.0142) = 49.1 years. A 7% raise feels large, yet after inflation it doubles real purchasing power only over a near-career horizon. The nominal ruler flatters every pay conversation.

P17. (Independent.) A synthetic one-year claim pays $90, $120, or $150 with probabilities 0.3, 0.4, 0.3. The T-bill yields 4%. (a) Expected payoff. (b) The price if it were priced to yield the risk-free 4%. (c) The price if buyers demand a 12% expected return. (d) In one line, where the risk premium came from.

Solution. (a) E = 0.3 × 90 + 0.4 × 120 + 0.3 × 150 = $120. (b) 120 / 1.04 = $115.38. (c) 120 / 1.12 = $107.14. (d) The 8-point premium (12% − 4%) is the $8.24 discount from the risk-free price to $107.14. A lower price is the only place a higher expected return can come from as the dispersion argument showed.

P18. (Timed, 6 minutes, stretch, five parts.) (i) An Indian FD quotes 7.2% compounded quarterly. EAR? (ii) A US CD quotes 4.6% compounded monthly. EAR? (iii) India CPI is 5.4%, US CPI is 2.6%. Which deposit has the higher real yield? (iv) Years to double at the FD's EAR (Rule of 72, then exact)? (v) ₹5,00,000 in the FD for 10 years, so nominal value, then value in today's rupees at 5.4% inflation.

Solution. (i) (1 + 0.072/4)^4 − 1 = 7.40%. (ii) (1 + 0.046/12)^12 − 1 = 4.70%. (iii) India real 1.0740/1.054 − 1 = 1.89%; US real 1.0470/1.026 − 1 = 2.05% → the US CD, despite its lower sticker rate, because the ruler is matched on both sides. (iv) 72/7.40 = 9.73 years; exact ln 2 / ln(1.0740) = 9.71 years. (v) Nominal 5,00,000 × 1.0740^10 = ₹10,20,660; real 10,20,660 / 1.054^10 = 10,20,660 / 1.6920 = ₹6,03,219 in today's rupees, so the balance more than doubled on paper but grew only about 21% in purchasing power.

Applied mini-project: "The Nominal Illusion Audit"

Task (≈1.5 hours). Run this module's entire toolkit on real data for one Indian and one US company, plus one deposit rate from each country, and write a one-page note titled "What the nominal numbers hid."

Steps and exact sources:

1. Companies: TCS and Microsoft (substitutes allowed: any large, decade-stable Indian and US company, but avoid ones with big M&A or divestitures, or you'll be doing Worked example 2's perimeter forensics).
2. Pull revenue for FY2015 and FY2025 (TCS: annual reports at tcs.com/investors, or screener.in as a convenience, but trace at least one figure back to the annual report itself) and for Microsoft's FY2015 and FY2025 (10-Ks via SEC EDGAR: sec.gov → EDGAR full-text search → MSFT 10-K; stockanalysis.com as convenience, verified against the filing).
3. Compute each 10-year nominal revenue CAGR, writing out the fence-post count explicitly.
4. Pull inflation from primary sources. India: CPI (combined) annual index values from MOSPI (mospi.gov.in) or the RBI Handbook of Statistics. US: CPI-U from BLS or the FRED series CPIAUCSL. Compute each country's decade inflation as the CAGR of the index itself (same formula, which is the point).
5. Convert both companies to real growth via exact Fisher. Compare the nominal ranking to the real ranking; compute both doubling times (nominal and real) with the Rule of 72.
6. Deposit leg: find one current 1-year Indian bank FD rate with its compounding frequency (any major bank's rate card) and one US 1-year Treasury/CD yield (treasurydirect.gov or FRED series DGS1). Compute EARs, then real rates against each country's latest year-on-year CPI print.
7. Write the one-pager: conclusion first (what did nominal numbers hide?), then the three exhibits (growth, inflation, deposits), each with its numbers. Every figure must carry a source and date. File it in your knowledge system, which the toolkit will formalize. For now, one clean document.

Scoring rubric (self-grade honestly; ≥9/12 to consider it passed):

Reading & resources

• Zerodha Varsity, Module on Personal Finance (Part 1), opening chapters on the time value of money, inflation, and returns (varsity.zerodha.com). India-flavored, free, and exactly this module's level. [Free] [Beginner]
• Khan Academy, "Interest and debt" unit (Finance & Capital Markets): simple vs compound interest, Rule of 72 video, inflation basics. Good for a second pass in a different voice. [Free] [Beginner]
• **Morgan Housel, *The Psychology of Money***, ch. 4 "Confounding Compounding" and ch. 5 "Getting Wealthy vs. Staying Wealthy": the behavioral meaning of the long-horizon arithmetic. [Paid] [Beginner]
• **Jeremy Siegel, Stocks for the Long Run (6th ed.)**, ch. 1–2: two centuries of asset-class real returns; the source of the "equities ~6.5–7% real" anchor. Read for the charts; skim the rest for now. [Paid] [Intermediate]
• UBS Global Investment Returns Yearbook (formerly Credit Suisse; Dimson–Marsh–Staunton). Free summary edition published each February: 120+ years of equity/bond/bill returns across countries, the antidote to single-market extrapolation. Search "UBS Global Investment Returns Yearbook summary edition PDF". [Free] [Intermediate]
• Damodaran Online, "Historical Returns on Stocks, Bonds and Bills: 1928–current" (pages.stern.nyu.edu/~adamodar → Data). Updated every January; your permanent US returns dataset. [Free] [Intermediate]
• RBI, Handbook of Statistics on the Indian Economy (rbi.org.in → Publications, annual): interest rates, CPI/WPI series, G-sec yields. Your permanent India dataset. [Free] [Intermediate]
• MOSPI CPI press releases (mospi.gov.in, monthly) and BLS CPI (bls.gov/cpi) or FRED (fred.stlouisfed.org, series CPIAUCSL): the primary inflation sources used in the mini-project. [Free] [Beginner]
• S&P BSE Sensex TRI and Nifty 50 TRI factsheets (spglobal.com/spdji regional pages; niftyindices.com): long-run index levels including dividends. Always prefer TRI for return history. [Free] [Beginner]
• **Gautam Baid, *The Joys of Compounding***, the compounding chapters; an India-rooted preview of the Phase 6/10 reading spine. Optional here. [Paid] [Intermediate]
• Going deeper. Zerodha Varsity, "Personal Finance" module, chapters on loans, EMIs, and amortization (varsity.zerodha.com): reducing-balance vs flat rates, how each EMI splits into interest and principal, and prepayment math. India-flavored and pitched at exactly the level of Worked example 5 and the P13–P18 loan problems. [Free] [Beginner]
• **Going deeper. Howard Marks, The Most Important Thing, the risk chapters ("Understanding Risk," "Recognizing Risk," "Controlling Risk")**. The practitioner's argument that risk is the probability of permanent loss, not mere volatility; it deepens the honest caveat about dispersion and the pricing intuition of Worked example 6. [Paid] [Intermediate]

Tooling note: do every computation here twice, once on a calculator and once in a spreadsheet (^ for powers; later =FV, =RATE, =XIRR). M0.04 builds the full toolkit; arriving there already fluent in (1+r)^n is the goal.

The Modern Analyst's Addendum

Everything above teaches this skill from first principles, by hand. That is how you learn it, and the mastery check still tests it that way. This addendum shows how a working analyst amplifies the same skill today. It adds; it never replaces. (R1/R10)

AI-Augment this skill

``ai-augment-json { "skill": "Percentages and percentage points, CAGR and the fence-post rule, simple versus compound interest and the Rule of 72, APR versus EAR, inflation and the exact Fisher relation, the risk-return stack, the asset-class map, volatility drag and sequence risk", "use": "This module's whole purpose is to install a reflex, and a reflex cannot be installed by watching something else perform it. So the sanctioned uses here are narrower than in any other Phase 0 module, and one of them is unusually valuable. (1) DRILL GENERATION, which is the good one. The timed gauntlet in the practice set is a shape: 'generate twenty more items in exactly this shape — quote-to-EAR conversions, fence-post CAGRs, percentage-point-versus-percent traps — with no answers.' You then solve them by hand and check each one yourself. This is the single highest-value use of a language model anywhere in Phase 0, because the reflex needs volume and volume is expensive to author. (2) STATING A METHOD BACK. 'What is the fence-post rule and why does it bite?', 'why is the geometric mean at or below the arithmetic mean?' — questions about method, answerable from the structure of the thing. (3) TURNING A WORD PROBLEM INTO A FORMULA, then computing it yourself. (4) NAMING WHICH RULER APPLIES: 'this figure is nominal — what would I need to make it real?' is a checklist question, and the checklist is §4.6.", "tools": ["Chat assistants — Claude, ChatGPT — for generating extra drill items in the gauntlet's shape and for restating a method", "Your calculator and your spreadsheet, then a notebook — every answer computed by you, twice, as the module's own tooling note already requires", "The primary series: FRED for US macro, the RBI's DBIE portal and MOSPI for India, AMFI for fund data, and the exchanges' TRI factsheets for index returns"], "prompt": "Generate 20 practice items in exactly the shape of these examples: <PASTE three from the timed gauntlet>. Cover quote-to-EAR conversions at different compounding frequencies, CAGRs over stated year ranges where the fence-post count is the trap, percentage-point versus percent distinctions, and exact-Fisher real returns. Use plausible Indian and US figures. Give me the questions only — no answers, no worked steps, no hints. I am solving and checking these myself.", "verify": "Two bans, and the second is specific to this module. FIRST: never let it do the arithmetic that is the point. A language model is not a calculator; it produces the most plausible-looking continuation of a computation, which for compounding and root-taking is frequently close and wrong, and 'close and wrong' is the worst possible outcome for a reflex you are trying to calibrate. Compute in a spreadsheet or a notebook, where the operation is the operation. SECOND, and this is the one to remember: never take a HISTORICAL RETURN FIGURE from a chat window. §4.8's asset-class table is the most hallucinated content in all of Phase 0, because everybody has a number for 'what do equities return' and almost none of those numbers carry their basis. The four things that must travel with such a figure — price index or total return, nominal or real, which start year, which currency — are exactly what a fluent one-line answer drops. §4.8 already names the sources: Damodaran's historical-returns page, the UBS Global Investment Returns Yearbook, the RBI Handbook, and the exchanges' TRI factsheets. Pull the number, and pull its basis with it.", "diy": "The gate is unaided and timed. You convert a quote to an EAR at any frequency; you compute a CAGR with the correct interval count and sanity-check it against a doubling; you apply the exact Fisher relation rather than the subtraction shortcut; you compute an EMI and a SIP future value; you compute a real return after tax; and you say, without hesitating, whether a stated change is percent or percentage points. Nothing above is admissible in the timed gauntlet, and the gauntlet is the point of the module." } ``

Modern Data Analysis

By hand first. You ran the full chain on paper: a 7.1% quarterly quote becoming a 7.291% EAR, becoming 5.104% after a 30% slab, becoming +0.099% real. You counted intervals instead of data points. You derived the Rule of 72 rather than memorising it. You computed an EMI and watched the real burden of a fixed payment fall by more than half. Every one of those must stay in your hands, because each is a conversion between rulers, and an analyst who cannot convert between rulers cannot notice when someone else has failed to.

Today's workflow. The change is that each conversion becomes a one-line function applied to a series rather than to a point, and that single move fixes the central complaint. The real-versus-nominal argument is that nominal numbers get compared across countries and eras without a deflator; the reason that happens is that deflating one number is a chore. Write real = lambda nom, infl: (1+nom)/(1+infl) - 1 once and every nominal series in your notebook acquires a real twin at the moment it arrives, at zero marginal cost. The same applies to ear, cagr and the deflator: the discipline stops depending on your attention. And once returns are a series rather than a summary, two things that can otherwise only be asserted become measurable: the gap between an arithmetic mean and a CAGR, and how much of a fund's advertised average survives to your wealth.

Tools & sources (IN + US). numpy and pandas for the series, matplotlib for the nominal-versus-real pair. India: MOSPI at mospi.gov.in for the CPI and GDP releases at the statistical source, the RBI's DBIE portal at data.rbi.org.in for downloadable CPI, WPI, money-supply, bank deposit-rate and G-sec-yield series plus the annual Handbook of Statistics on the Indian Economy, AMFI at amfiindia.com for scheme NAVs and expense data, and the BSE and NSE index factsheets in their TRI form, since the price-index version understates equity returns by the dividend yield, which is roughly the size of the whole fee effect you are being taught to respect. US: FRED at fred.stlouisfed.org with the series ids worth memorising (CPIAUCSL for CPI, DGS10 for the ten-year Treasury, FEDFUNDS for the policy rate, TB3MS for bills), Damodaran's "Historical Returns on Stocks, Bonds and Bills: 1928–current" table at pages.stern.nyu.edu, refreshed each January and carrying its own basis notes, and the UBS Global Investment Returns Yearbook for the long international panel that puts the US series in its survivorship context.

``python # The conversion chain as functions, then applied to a series — recomputed in-session (R3) ear = lambda q, m: (1 + q/m)m - 1 real = lambda nom, infl: (1 + nom)/(1 + infl) - 1 cagr = lambda end, beg, yrs: (end/beg)(1/yrs) - 1 # yrs = INTERVALS, not data points print(f"{ear(.071,4):.4%} {real(ear(.071,4)*.70, .05):+.4%}") # 7.2913% +0.0990% print(f"{cagr(255000,95000,10):.4%} {cagr(391,183,10):.4%}") # 10.3778% 7.8879% drag = lambda mean, spread: mean - (((1+mean+spread)*(1+mean-spread))**0.5 - 1) print(f"{drag(.12,.20)*100:.2f} {drag(.12,.30)*100:.2f}") # 1.80 pp 4.09 pp fee = lambda g, f, n: 1 - ((1+g-f)/(1+g))**n # share of TERMINAL wealth lost print(*[f"{fee(.12,.01,n):.2%}" for n in (10,20,30,40)]) # 8.58% 16.42% 23.59% 30.14% ``

Verify. Prove each function against the module's own worked figures before it touches a real series. The chain: a 7.1% quarterly quote gives an EAR of 7.2913%, interest of ₹72,913, ₹51,039 after a 30% slab, and a real return of +0.0990%, or ₹990 on ten lakh. The CAGRs: TCS 10.3778%, Apple 7.8879%, Asian Paints 10.8262%, Coca-Cola 0.2153%. The compounding table: ₹1,00,000 at 8% reaches ₹2,15,892 at ten years and ₹21,72,452 at forty, against ₹4,20,000 simple, a factor of 5.17. The back-loading exhibit: ₹1 at 12% reaches 93.0510 at forty years, with the fourth decade adding 63.09 against the first decade's 2.11. The SIP: ₹98.93 lakh nominal on ₹24 lakh contributed, ₹37.28 lakh in today's rupees at 5% inflation, from a 1% monthly rate whose EAR is 12.6825%. The home loan: an EMI of ₹44,986, a total of ₹1,07,96,711, interest of ₹57.97 lakh on a ₹50 lakh principal, an EAR of 9.3807%, and a final-year payment worth ₹16,955 in today's money. The fee comparison: ₹88.21 lakh against ₹73.66 lakh, a gap of 16.5% of the direct-plan outcome. Sequence risk: the lump sum ends at 107.25 either way, while the retiree ends 6.93 apart and the SIP accumulator ends 6.93 apart in the opposite direction. And the note: $105.77, $100.00, $95.65 at required returns of 4%, 10% and 15%, with real figures of 1.46% and 7.32%.

Quantitative lens

Three results, all exact arithmetic on stated inputs, and the second of them should change how you read the asset-class table for the rest of your life.

Volatility drag has a closed form, and it is about the size of the differences people argue about. For a return that is mean ± spread with equal probability, the compound rate is exactly √((1+m+s)(1+m−s)) − 1. At an arithmetic mean of 12% with a ±20% spread the CAGR is 10.1998%, a drag of 1.80 percentage points, close to the σ²/2 approximation's 2.00. At ±10% the drag is 0.45 pp; at ±30% it is 4.09 pp. The module's own illustration is the extreme case: +40% then −20% averages 10% arithmetically and compounds at 5.83%, a drag of 4.17 pp, which is why the steady 6% fund wins with a four-point-lower headline. Hold that beside the opening comparison, TCS at 10.38% against Apple at 7.89%, a 2.5-point gap the module then largely dissolves in real terms. Notice too that a plausible difference in volatility between two funds is worth about as much as the entire India-versus-US growth differential everyone talks about, and it never appears in a marketing document. The rule that follows is the module's own, now with a magnitude attached: judge a multi-year record by its CAGR, and treat the gap between the arithmetic mean and the CAGR as a measurement of the path rather than a rounding difference.

You cannot measure an expected return from history to the precision the asset-class table appears to offer, and the shortfall is not close. The standard error of a mean is σ/√n. Take a broad equity index at an annual standard deviation of 20%, an order of magnitude, and you should pull the actual figure from the TRI series rather than accept this one, and 99 years of data give a standard error of 2.01 percentage points, so a 95% interval on "US equities returned about 10%" is roughly ±3.94 pp. At σ = 25% it is ±4.92 pp; at 15% it is ±2.95 pp. For a 45-year Indian window, the Sensex sweep quoted earlier, a σ of 20% gives ±5.84 pp and a σ of 30% gives ±8.77 pp, an interval wider than the whole quoted 13–16% band. And the number of years needed to pin a mean return to ±1 pp at 95% confidence is (1.96σ)²: 1,537 years at σ = 20%, 864 at 15%, 3,457 at 30%. There is not that much stock market. This is not a criticism of that table; it is the arithmetic reason the table is written with the disclaimer it carries. "These are orders-of-magnitude historical anchors, not promises" is arithmetically obligatory. It also disposes of a genre of argument you will meet constantly: that one market, one strategy or one decade "outperformed" by two or three points. Two or three points is inside the measurement error of any sample anyone has.

Fee drag is a pure function of the rate and the horizon, and it is independent of how much money you have. The share of terminal wealth a recurring fee consumes is 1 − ((1+g−f)/(1+g))ⁿ, in which no wealth term appears. One hundred basis points against a 12% gross return costs 8.58% of terminal wealth over ten years, 16.42% over twenty, 23.59% over thirty and 30.14% over forty. The module's own worked example lands on 16.5% for a twenty-year horizon, which is that identity. Push the fee instead of the horizon: 50 bp over thirty years costs 12.56%, 150 bp costs 33.27%, 200 bp costs 41.76%. Two consequences travel together. Every recurring cost, whether an expense ratio, brokerage, an advisory fee or tax drag, should be quoted as a share of the terminal outcome before anyone is allowed to call it small. And the mirror is the whole reason Phase 2 exists: a business compounding capital 100 bp faster than another, for the same reason and by the same identity, is worth a fortune.

Honest limits, and the second finding above is the one that most needs them. The standard-error formula assumes annual returns are independent and identically distributed. They are not: real return distributions have fatter tails than a normal, which makes any interval estimated this way an understatement in exactly the years you care about, and there is evidence of mild long-horizon mean reversion, which pushes the other way. The order of magnitude survives both, at a few percentage points rather than a few basis points, but do not quote ±3.94 as though it were measured. The σ values above are stated inputs, not results: what is computed here is the consequence of a standard deviation, it does not estimate one, and the honest move is to re-run it with the σ you get from the actual TRI series. The volatility-drag formula is exact for a symmetric two-state return and approximate for anything else, which is why it sits beside the σ²/2 rule rather than replacing it. The fee identity assumes the gross return is unaffected by the fee, which is generous to the expensive product. And a historical mean, however carefully bounded, is a statement about one realised path of one world, and Phase 6 will spend a module on why extrapolation feels safest precisely when it is least warranted.

Do it in code: write the four conversions as functions and apply them to series rather than points, so every nominal number in your notebook is born with a real twin; report a CAGR and an arithmetic mean side by side and treat their gap as data; quote every recurring cost as a share of the terminal outcome; and put a standard error beside any historical average before you compare two of them. See G1/QM1.01, G2/DA1.05 and G1/QM1.03.

Where this goes next: galaxy cross-links

• Probability, distributions, sampling and estimation (QM1.01). Where the standard error above stops being a formula and becomes a habit: what a sample mean is, what it is not, and why an interval is the honest form of an average.
• Time-series, Monte Carlo and intro ML for finance (QM1.03). Sequence risk generalised. You have seen that the order of returns deciding a retiree's outcome; this is the machinery for simulating paths rather than assuming one, and for asking what fraction of them ruin the plan.
• Statistics in code (DA1.05). Means, geometric means, dispersion and confidence intervals computed rather than quoted, on data you loaded yourself.
• NumPy and vectorized computation (DA1.01). The reason the conversions above are one line each rather than a loop, and the foundation every later engine in this program is built on.
• Failure modes, verification and the primary-source guardrail (AI0.06). The general case of the specific trap set here: a fluent, unsourced historical return figure, delivered without its basis, is the most confidently wrong output you will meet in Phase 0.

Flashcards

This module's flashcards and mastery quiz are wired into the app: see the node's Quiz and Reviews.

Mastery check

Two parallel forms. Timed: 50 minutes per form. Calculator or spreadsheet allowed; notes closed. Numeric answers: rates within ±0.10 percentage points, money within ±0.5%, unless an item states otherwise. Pass threshold: ≥85%, and with twelve equally weighted items that means 11/12 or better. Commit all answers in writing before opening the key (working agreement 3). Passing unlocks the toolkit. A failed attempt: list the misses, restudy only those, wait two days, take the other form.

Form A

A1. (MCQ) A bank's net interest margin moves from 3.2% to 3.6%. Which statement is precise? (a) NIM rose 0.4% (b) NIM rose 40 bp, a 12.5% relative increase (c) NIM rose 4 bp (d) NIM rose 12.5 pp

A2. (Numeric) EAR of 9% nominal compounded monthly?

A3. (Numeric) ₹2,00,000 at 7.5% simple interest for 6 years. Future value?

A4. (Numeric) Same ₹2,00,000 at 7.5% compounded annually for 6 years. Future value?

A5. (MCQ) The Rule of 72 uses 72 rather than 69.3 (= 100·ln 2) because: (a) 72 is exact for annual compounding at all rates (b) ln(1+r) < r makes 69.3/r understate the true doubling time, and 72 both corrects this at typical rates and divides conveniently (c) 69.3 applies only to simple interest (d) it is a rounding convention with no basis

A6. (Numeric) Exact doubling time at 9% (two decimals), using t = ln 2 / ln(1+r)?

A7. (Numeric) Nominal return 11%, inflation 6.2%. Exact real return?

A8. (Short answer) An Indian firm grows revenue 10% (CPI 5%) and a US firm grows 7% (CPI 2.5%). In two sentences with numbers: which grew faster in the sense an analyst should care about?

A9. (Numeric) Revenue ₹1,850 crore → ₹4,070 crore over 7 years. CAGR?

A10. (MCQ) Headline CPI inflation falls from 7.4% to 5.1%. The price level has: (a) fallen 2.3% (b) risen ~5.1% over the latest year, still rising but more slowly (c) been flat (d) fallen unless GDP grew

A11. (Numeric) A portfolio of ₹100 experiences −30% and +40% in some order, withdrawing ₹10 at each year-end after returns. Compute ending wealth for bad-year-first and good-year-first, and the gap. (For reference: with no withdrawals, either order ends at ₹98.)

A12. (Numeric) ₹5,00,000 invested 20 years; gross return 11%, annual fees 125 bp (net 9.75%). Terminal value?

Form A key. A1 (b): 0.4 pp = 40 bp; relative 0.4/3.2 = 12.5%. (a) is ambiguous, (c)(d) confuse units. A2 9.38%: (1 + 0.09/12)¹² − 1 = 1.0075¹² − 1 = 0.09381. A3 ₹2,90,000: 2,00,000 × (1 + 0.075×6) = 2,00,000 × 1.45. A4 ₹3,08,660: 1.075⁶ = 1.54330; the ₹18,660 gap over A3 is interest-on-interest. A5 (b): the derivation is t = ln2/ln(1+r) ≈ 69.3/r%, biased low since ln(1+r) < r; 72 compensates at 6–10% and has many divisors. A6 8.04 years: 0.6931/ln(1.09) = 0.6931/0.08618. A7 4.52%: 1.11/1.062 − 1. (Approximation 4.8% overstates by ~28 bp.) A8 Real growth: India 1.10/1.05 − 1 = 4.76%; US 1.07/1.025 − 1 = 4.39%. The Indian firm grew faster in real terms, but by ~0.4 pp, not the 3 pp the nominal figures suggest. A9 11.92%: (4070/1850)^(1/7) − 1 = 2.2^(1/7) − 1. Cross-check: one double + a bit in 7 years ≈ >10%. A10 (b): disinflation, not deflation. A11 Bad-first: 70 − 10 = 60; 60×1.4 = 84 − 10 = 74. Good-first: 140 − 10 = 130; 130×0.7 = 91 − 10 = 81. Gap ₹7. Order matters only because of the withdrawals. A12 ≈ ₹32.14 lakh: 5,00,000 × 1.0975²⁰ = 5,00,000 × 6.4282. (Gross 11% would have given ₹40.3 lakh, so the 125 bp consumed ~20% of the outcome.)

Form B

B1. (MCQ) A fund's expense ratio is 65 bp. That is: (a) 6.5% (b) 0.65% (c) 0.065% (d) 65% of returns

B2. (Numeric) EAR of 12% nominal compounded quarterly?

B3. (Numeric) $8,000 at 6% simple interest for 3 years. Future value?

B4. (Numeric) Same $8,000 at 6% compounded annually for 3 years. Future value?

B5. (MCQ) At a 36% growth rate, the Rule of 72 says money doubles in 2.0 years. The exact time is 2.25 years. The rule therefore: (a) overstates doubling time at high rates (b) understates doubling time at high rates, breaking down outside roughly 4–15% (c) is exact at 36% (d) only fails below 4%

B6. (Numeric) Exact doubling time at 4.5% (two decimals)?

B7. (Numeric) Nominal return 8.4%, inflation 3.1%. Exact real return?

B8. (Short answer) In two sentences using expected-value language: why must a lender charge a riskier borrower a higher rate even if the lender feels no fear at all?

B9. (Numeric) Revenue $12.4bn → $21.7bn over 6 years. CAGR?

B10. (MCQ) A retailer's same-store sales growth decelerated from 9% to 3%. Store revenue has: (a) fallen 6% (b) fallen 3% (c) risen ~3%, a new high at a slower pace (d) risen 9%

B11. (Numeric) A fund returns +50% then −30% over two years. Compute the two-year multiple, the CAGR, and the arithmetic average of the yearly returns.

B12. (Numeric) SIP of ₹5,000/month for 15 years at 1% per month. Terminal value? (FV = A[((1+r)ⁿ − 1)/r].)

Form B key. B1 (b): 65 bp = 0.65% = 0.0065. B2 12.55%: (1.03)⁴ − 1 = 0.125509. B3 $9,440: 8,000 × (1 + 0.06×3) = 8,000 × 1.18. B4 $9,528.13: 1.06³ = 1.191016; gap over simple = $88.13. B5 (b): the rule says faster than reality at high rates (2.0 vs 2.25); compute exactly outside the band. B6 15.75 years: 0.6931/ln(1.045) = 0.6931/0.044017. B7 5.14%: 1.084/1.031 − 1. (Approximation 5.3% overstates by ~16 bp.) B8 With default probability p, the promised rate must be set so that the expected repayment, meaning (1−p) × full repayment plus p × recovery, at least matches the safe alternative; the higher rate offsets the chance of loss, before any risk aversion is added. Feelings are irrelevant; arithmetic forces the premium. B9 9.78%: (21.7/12.4)^(1/6) − 1 = 1.75^(1/6) − 1. Cross-check: 75% total in 6 years, under one double → below 12%. B10 (c): growth slowed, and the level rose. B11 Multiple 1.50 × 0.70 = 1.05; CAGR = √1.05 − 1 = 2.47%; arithmetic average = 10%. The 7.5 pp gap is volatility drag. B12 ≈ ₹24.98 lakh: 5,000 × [(1.01¹⁸⁰ − 1)/0.01] = 5,000 × 499.58, on ₹9 lakh contributed.

Teach it back & journal

Feynman prompt. Write a one-page explainer for a smart 15-year-old titled "Why the ₹1 lakh in Grandpa's fixed deposit is shrinking while it grows." You must cover: what compounding is (use a doubling story, not a formula), why the bank's quoted rate isn't the real story (compounding frequency, then inflation), and how the Rule of 72 lets you do the whole analysis in your head. No formulas until the final paragraph, and then show FV = P(1+r)^t and real ≈ nominal − inflation and connect each symbol back to your story. If a sentence needs jargon to survive, you don't own the idea yet, so rewrite it.

Journal reflection. Recall one financial decision in your own life, whether an FD, an insurance-cum-investment policy, a salary negotiation, or a purchase on EMI, that you made in purely nominal terms. Rerun it with the tools built here (EAR, real return, fee drag, doubling time). Write: what the nominal frame hid, what the real frame shows, and one sentence on how the five questions would have you look at any number differently now. Date it; you will reread this at week 80.

This module's flashcards and mastery quiz are wired into the app: see the node's Quiz and Reviews.

End of the bootcamp. Next comes the toolkit and knowledge system, where you will set up the spreadsheet, screeners, filing sources, and the knowledge system, and wire these drills into your daily loop.