Cost of Capital: the Full Build

Phase 3 · Week 28 · ~20 hours

Where it sits. Third module of the valuation phase. Time value of money built the discounting machine; risk and return priced risk: the equity risk premium, the country risk premium, CAPM's logic, and the clean-INR fix previewed. What follows is the assembly plant. It takes those parts, manufactures the ones still missing (a defensible beta built bottom-up from business fundamentals, a cost of debt built from a rating you can synthesize yourself, market-value weights), and bolts them into the single number every DCF in the two DCF modules that follow will turn on: the weighted average cost of capital.

By the end you can start from a blank sheet, meaning a company name, its filings, and five live data sources, and emerge forty-five minutes later with a WACC you can defend line by line. That is the gate here, and it is the exact skill the phase capstone and every Phase 8 deep dive assume. On the five-questions spine (master map section 2.1), the cost of capital is the other half of question 2. ROIC above what? It is also the denominator of question 5: what the price implies is always relative to a required return.

Suggested hour budget. Core content: 8h · worked examples: 2h · practice set: 5h · mini-project: 3h · flashcards + mastery check: 2h.

App link. The material here powers the Corporate Finance branch's capstone challenges: the WACC-builder (drag inputs from mock filings and data pages into a live build, graded station by station), synthetic-rating drills (coverage ratio → rating → spread against the clock), unlever/relever computation cards, and the error-spotting deck seeded from the ten classic errors below. The valuation sandbox's discount-rate consistency checks (currency match, double-count detection) enforce what you learn here.

Learning objectives

By the end you can:

1. Set a currency-matched risk-free rate for any valuation: state the two conditions that make a rate risk-free, explain why the Indian G-sec yield fails one of them, execute Damodaran's clean-INR fix (G-sec yield minus the sovereign default spread), and apply the rule that the currency of the cash flows dictates the currency of every rate in the stack.
2. Build a bottom-up beta from scratch: unlever a set of peer regression betas with βU = βL / [1 + (1−t)·D/E], average away their noise, and relever to the target's market-value capital structure with βL = βU × [1 + (1−t)·D/E]. Explain, with the standard-error arithmetic, why this beats any single regression beta.
3. Compute and correctly deploy total beta (market beta ÷ correlation with the market) for undiversified owners, and state precisely when it applies and when it must not be used.
4. Assemble a cost of equity from clean Rf + β × (mature ERP + λ·CRP), for domestic firms and exporters, in INR or USD, without double-counting country risk.
5. Estimate a cost of debt three ways (traded-bond YTM, actual rating → spread, and a synthetic rating read off the interest-coverage ratio using Damodaran's lookup table), convert it to after-tax at the marginal rate, and add the country default spread for emerging-market borrowers.
6. Compute market-value weights: convert book debt to market value with the one-bond approximation, include lease liabilities under ASC 842 / IFRS 16 / Ind AS 116, and defend the choice of market over book values.
7. Assemble the full WACC and sanity-check it against market-level ranges, then run the complete blank-sheet build on one Indian and one US company, citing a live source for every input.
8. Spot and fix the ten classic WACC errors in someone else's (or your own) computation.

Prerequisites & connections

Builds on. M3.02 directly. Its mature-market implied ERP, the country risk premium (default spread × relative equity volatility), λ exposure-weighting, CAPM and beta as co-movement all arrive here as finished inputs. Its illustrative mid-2026 values carry over too: US Rf 4.2%, mature ERP 4.3%, India default spread 2.0%, CRP 2.8%, India ERP 7.1%, clean INR Rf 4.5%, every one to be re-pulled live before real use. From M3.01 you need Fisher's real-to-nominal logic and the PV of an annuity, which you'll use to mark debt to market. From M2.02–M2.03 you need coverage and leverage ratios, plus the operating/financing separation. The synthetic rating is really Phase 2's interest-coverage ratio with a price tag on it. M1.08 gave you leases as debt across the three standards, along with effective-interest debt mechanics.

Feeds forward. In M3.04–M3.05 you'll discount FCFF at this WACC and FCFE at this cost of equity, and the never-mix rule gets teeth. M3.06 treats multiples as compressed DCFs, and the r inside P/B ≈ (ROE−g)/(r−g) is the Re you build here. In M3.07, economic profit = (ROIC − WACC) × invested capital. You're building the hurdle in that spread. The model stack in M3.08–M3.09 hard-codes nothing, which means your WACC sheet will cite the sources gathered here. M3.10 gives banks and other financials an equity-only treatment. WACC is meaningless for them, because their debt is raw material, and it isn't financing. You get a preview here and the full treatment there.

The short version. A company's cost of capital is the return its investors could earn elsewhere at the same risk, in the same currency, today. What the company would prefer to pay, and what it paid in the past, have nothing to do with it. Your job is to rebuild that opportunity cost from live market evidence, one auditable piece at a time.

Core teaching content

4.5 Station 2c: total beta: when your investor cannot diversify

CAPM's beta prices only market co-movement, because its marginal investor is diversified and firm-specific risk costs them nothing. You saw why in the risk-and-return work. Now change the investor. Picture a promoter with 90% of her net worth in one unlisted diagnostics chain, or a founder selling his whole company, or a family buying a private business outright. For them, every unit of the firm's volatility is risk they bear. They can't diversify it away.

Total beta rescales beta to total risk. A regression beta is β = ρ × (σfirm/σmarket), correlation times relative volatility. Strip out the correlation, which is the diversification discount, and you price the whole σ:

Total beta = market beta ÷ ρ(firm, market)  =  σfirm / σmarket

A typical single stock correlates with the index at ρ ≈ 0.4–0.5. R² is then 0.16–0.25, so most of any one firm's variance is firm-specific. Total betas therefore run 2–2.5× market betas, and private-buyer costs of equity come out brutally higher. That's correct. Concentration is expensive, and here the arithmetic finally tells you by how much. Damodaran publishes sector total betas alongside the standard tables, on the same page.

So when does it apply? Use total beta when you're valuing a business for an undiversified buyer or owner. That covers private-company sales to individuals, promoter buy-ins, valuing an owner-operator's stake for a settlement, and early-stage ventures where the founder is the marginal investor. Don't use it for listed-company minority stakes. The marginal price-setter there is diversified, so total beta would just make every public stock look overpriced to you. And don't use it when the likely buyer of the private firm is itself diversified, such as a listed strategic acquirer or a PE fund with a portfolio. For them market beta is right. That's why diversified buyers can outbid undiversified ones for the same asset: same cash flows, lower discount rate.

So one business can have two defensible values, because it has two different marginal investors. Keep that in mind when you're negotiating. Worked Example 4 computes a full case.

4.6 Station 3: assembling the cost of equity

Everything now snaps together, and each part came from an auditable source:

Re = Rf(currency-matched, clean) + βL(bottom-up) × [mature ERP + λ × CRP]

• Rf comes from the first station: 4.2% USD / 4.5% clean INR (illustrative).
• βL is the bottom-up build, relevered to this company's market D/E.
• Mature ERP is the implied premium from the risk-and-return work, 4.3% illustrative, refreshed from Damodaran's monthly headline number.
• λ × CRP also comes from the risk-and-return work, where country risk follows where cash flows are earned. λ ≈ share of revenue exposed to the risky economy. Refine it with production or asset location when that differs. CRP(India) = default spread × relative equity volatility = 2.0% × 1.4 = 2.8% illustrative.

Three assembled examples follow, with all inputs illustrative mid-2026:

• The domestic Indian cement leader from the bottom-up beta has all-India cash flows, so λ = 1. Re = 4.5 + 0.97 × (4.3 + 2.8) = 4.5 + 0.97 × 7.1 = 11.4% (INR).
• A TCS-scale Indian IT exporter has ~70% developed-market revenue, which puts λ ≈ 0.3, and a bottom-up βL ≈ 0.9. Firm ERP = 4.3 + 0.3×2.8 = 5.14%. Re = 4.5 + 0.9 × 5.14 = 9.1% (INR). Same country, two very different premiums, because the cash flows are earned in different places.
• The US home-improvement retailer, built in full in the second blank-sheet build: Re = 4.2 + 0.97 × 4.3 = 8.4% (USD).

Install some sanity instincts now. In this rate environment, mature-market (USD) costs of equity for ordinary firms land at roughly 7–12%. For domestic Indian firms, INR costs of equity run roughly 10–15%. A computed Re below the local risk-free rate plus a point, or above ~18% for an established firm, means an input is broken. It's usually the beta or a double-counted CRP. These ranges are era-specific rules of thumb. They aren't constants. Recalibrate them each January with your ERP card.

4.10 Blank-sheet build 1: an Indian cement leader (INR)

This is the full assembly line, run start to finish on an UltraTech-style company. Every figure is illustrative, modeled on ~FY24/FY25 scale. This is a teaching specimen and not a recommendation. Pull the actual annual report, screener.in, and the live Damodaran/RBI/FRED pages, and expect different numbers. Each input's live source is in the table.

Raw inputs, with their sources:

Station 1: Rf. The cash flows are forecast in INR, so you need an INR stack. Rf = 6.5 − 2.0 = 4.5%.

Station 2: beta. βU = 0.93 and D/E = 20,000/3,20,000 = 0.0625, so βL = 0.93 × [1 + 0.75×0.0625] = 0.97.

Station 3: cost of equity. With λ = 1, firm ERP = 4.3 + 2.8 = 7.1%. Re = 4.5 + 0.97×7.1 = 11.4%.

Station 4: cost of debt. ICR = 12,000/1,000 = 12. That's Aaa/AAA even on the stricter small-cap table, and the company is a large cap anyway. The spread is ≈ 0.6%. Rd = 4.5 + 2.0 + 0.6 = 7.1% pre-tax. Now cross-check. Its embedded rate is 1,000/20,000 = 5.0%, which reflects older, cheap borrowings. Fresh NCD coupons in the borrowing note (~7.2–7.5% for a domestic AAA, illustrative) land right on the estimate. Keep 7.1%. After tax, that's 7.1 × 0.75 = 5.3%.

Station 5: weights. V = 3,20,000 + 20,000 = ₹3,40,000 cr. E/V = 94.1% and D/V = 5.9%. Book debt ≈ market here, because the credit is healthy and rates are near issuance levels.

Station 6: assembly and sanity.

WACC = 0.941 × 11.4%  +  0.059 × 5.3%
     = 10.73%         +  0.31%
     ≈ 11.0% (INR)

That passes the sanity check. It's inside the India band, a touch below the median, which is where a lightly levered, large, quasi-AAA cyclical should be. The Phase-2 "~12% hurdle" you've used since the returns-on-capital work now has a construction underneath it.

4.11 Blank-sheet build 2: a US home-improvement retailer (USD)

This one is a Home Depot-scale specimen. It's illustrative, modeled on ~FY24/FY25 scale. Pull the 10-K (SEC EDGAR), FRED, and the live tables.

Raw inputs first. Market cap ≈ $350bn (shares × price, today). Debt is long-term plus current borrowings ≈ $42bn. Add operating plus finance lease liabilities ≈ $8bn and the total is $50bn. Under ASC 842 both lease types are on the balance sheet, so include both. EBIT ≈ $21.5bn and interest ≈ $2.2bn. Marginal tax ≈ 25% (21% federal + state). With revenue ≈ all-US, λ·CRP = 0. The sector βU (retail/building supply, US table, cash-corrected) ≈ 0.88. Rf = 4.2% (FRED DGS10), and mature ERP = 4.3%.

Now the stations, rapid-fire.

1. USD cash flows mean Rf = 4.2%. The Treasury needs no cleaning, which is the luxury of a Aaa-ish sovereign.
2. D/E = 50/350 = 0.143. βL = 0.88 × [1 + 0.75×0.143] = 0.88 × 1.107 = 0.97. Published regression betas are ≈ 1.0, which is consistent. That's comforting, but it doesn't make them authoritative.
3. Re = 4.2 + 0.97×4.3 = 8.4%.
4. ICR = 21.5/2.2 = 9.8, which on the large-cap table means Aaa/AAA with a spread of ~0.6%, for a synthetic Rd = 4.8%. But the actual issuer rating is A-range. Agencies penalize the lease-heavy structure and the buyback appetite in ways one ratio can't see. When an actual rating exists, it wins. The synthetic is the cross-check. The A-range spread is ≈ 1.0%, so Rd = 4.2 + 1.0 = 5.2%, and after tax it's 5.2 × 0.75 = 3.9%. The embedded rate cross-checks at 2.2/50 = 4.4%, with older coupons below today's marginal rate, as you'd expect.
5. V = $400bn, so E/V = 87.5% and D/V = 12.5%. Book equity here is negative from buybacks, so book weights would have been gibberish. It's market weights or nothing.
6. Assembly:

WACC = 0.875 × 8.4%  +  0.125 × 3.9%
     = 7.35%         +  0.49%
     ≈ 7.8% (USD)

It comes in below the US median, which suits a stable-demand, modestly levered giant. Now set it against build 1. The gap between 11.0% INR and 7.8% USD doesn't mean "India is riskier by 3.2 points of judgment". About 2 points of it is inflation differential, ~1 point is λ-weighted country risk, and a little is business mix.

Same machine, different economies, and both builds are defensible.

Common mistakes & how experts think differently

Here are the ten classic WACC errors. Each one has destroyed real valuations, and several turn up in published sell-side work every week. Learn them as a checklist. The mastery check and the app's error-spotting deck test all ten.

1. The currency mismatch. Someone puts a USD risk-free rate under INR cash flows, or the reverse. That smuggles ~2 points of inflation differential into the rate while the forecasts carry the other currency's inflation, and the value ends up wrong by a third at the terminal. Expert reflex. The first question on opening any model is "what currency are the cash flows in?" Then audit the rate stack against it.

2. The dirty risk-free rate. Someone uses the raw G-sec yield as INR Rf while also charging the full CRP, so sovereign default risk gets billed twice (~2 points). Expert reflex. The default spread appears twice and only twice, with opposite signs. It's subtracted from the government yield. Then it's added back in the premium (scaled, λ-weighted).

3. Regression-beta worship. This is pasting a portal's β = 1.47 (±0.3) into CAPM, sometimes without knowing the index, period, or interval behind it. Expert reflex. Betas come from the business, bottom-up. A regression beta is a cross-check, and you reach for total beta when, and only when, the marginal investor is undiversified.

4. Unlever/relever slips. People forget the (1−t), or use book D/E. Some relever at the peer-average D/E when the target's own belongs there, or unlever with the target's D/E when each peer's own is the right one. Expert reflex. Say the formula's story aloud. Strip each peer's balance sheet off with its own numbers, average the naked business risk, dress it in this company's balance sheet.

5. Historical cost of debt. Interest ÷ book debt, or the 2021 coupon, gets used as the marginal rate. After a rate cycle or a credit deterioration, that misses by hundreds of basis points. Expert reflex. The cost of debt is what lenders would charge today, read from a traded YTM, the current rating spread, or a synthetic rating. The embedded rate is only a plausibility check.

6. The wrong tax rate, or a phantom shield. Someone uses the effective tax rate, a polluted average, where the marginal statutory rate belongs. Or they grant the (1−t) shield to a loss-maker that pays no tax. Expert reflex. The shield equals what the next rupee of interest actually saves. That's the statutory marginal rate if the firm is profitable and zero if it isn't. The shield comes back when profitability returns.

7. Book weights, and cash double-counting. Book equity weights are negative for US buyback veterans and ancient history for everyone else. This error's other half is netting cash out of debt and adding it back in the equity bridge. Expert reflex. Use market values only. Pick the gross-debt or net-debt convention once and write it at the top of the model. Cash should appear once and only once.

8. Mismatched rate and cash flow. FCFF (all-capital cash flow) gets discounted at Re, or FCFE (equity cash flow) at WACC. This error is about whose cash flow you're discounting. Currency and real-versus-nominal versions of the same mismatch are error 1, and it's worth learning the two as one family with two axes, claimholder and unit. The first form overvalues levered firms wildly. Expert reflex. The discount rate belongs to whoever owns the cash flow. All claimholders get WACC, and shareholders alone get Re. You'll drill this in the DCF work until it's reflex.

9. The frozen, one-size WACC. One company-wide WACC gets applied to a risky new venture and a bond-like annuity business alike. Or a 2019 WACC is still discounting in 2026. Or a constant WACC runs through a planned de-levering from 40% to 10%. Expert reflex. The rate belongs to the risk of the cash flows being discounted, at that time. That means divisional betas for conglomerates, a rate path when the structure migrates, an annual refresh from the live cards.

10. False precision and WACC-shopping. This is reporting 11.83% as if the third significant figure were knowledge. It's nudging inputs until the DCF hits a pre-chosen price. It's also confusing a 15% aspiration ("our hurdle") with the market's opportunity cost. Expert reflex. Build it carefully, and round it humbly to the nearest half-point. Band it ±1 point in every output, and change it only when an input's source changes. The moment the WACC becomes the tuning knob, the valuation is fiction.

Underneath all ten, novices treat WACC as a number to fetch. Experts treat it as a small audit trail of five inputs, five sources, five dates, assembled by rules so mechanical that two analysts with the same sources should land within half a point of each other.

Save your disagreements about a company for the cash flows, where the thinking gets done.

Worked examples

Worked example 1: Bottom-up beta for a three-business conglomerate (Reliance-style)

Setup. This one is illustrative, modeled on ~FY25 scale, and the segment values are the analyst's estimates, which is the realistic situation. An Indian conglomerate runs three businesses, and no segment is separately listed. On a consolidated basis it has a market cap of ₹18,00,000 crore, total debt of ₹3,00,000 crore and a marginal tax rate of 25%. Its cash flows are predominantly Indian (λ ≈ 0.9 blended, with some export refining revenue).

Step 1: segment values. Estimate each segment's enterprise value from revenue × sector EV/Sales, or from EBITDA × a sector multiple. Use what Phase 2 taught you about which is less distorted for each business:

Step 2: blend the business risk. βU = 0.286×1.05 + 0.428×0.75 + 0.286×0.90 = 0.300 + 0.321 + 0.257 = 0.88.

Step 3: relever at the consolidated structure. D/E = 3,00,000/18,00,000 = 0.1667. βL = 0.88 × [1 + 0.75×0.1667] = 0.88 × 1.125 = 0.99.

Step 4: cost of equity (INR). Firm ERP = 4.3 + 0.9×2.8 = 6.82%. Re = 4.5 + 0.99×6.82 = 4.5 + 6.75 = 11.3%.

The lesson. Whatever the portal says the conglomerate's regression beta is, often 1.1–1.3, it's polluted by index membership and flow effects. You've replaced it with an auditable statement: 43% of this company is a utility-like telecom, and the beta must say so. When the segment mix shifts, as retail grows and energy shrinks, the beta updates by construction, years before five-year regressions notice.

Worked example 2: Synthetic rating and full INR cost of debt for an unrated mid-cap

Setup (synthetic company). Kaveri Precision Components is an Indian auto-ancillary with a market cap of ₹9,000 crore (≈ $1.1bn, so the small-cap table applies). EBIT is ₹240 crore and interest expense ₹60 crore. It's profitable, with a marginal tax rate of 25%, and it has no bonds outstanding and no credit rating. Illustrative market inputs: clean INR Rf 4.5%, India default spread 2.0%.

Step 1: coverage. ICR = 240 / 60 = 4.0.

Step 2: synthetic rating. On the small-cap table (illustrative early-2026 vintage rows), ICR 3.5–4.0 maps to Ba1/BB+ with a spread of ≈ 2.0%. One band up, 4.0–4.5 maps to Baa2/BBB at ≈ 1.6%. An ICR of 4.0 is right on the boundary. Take the conservative Ba1/BB+ at ≈ 2.0%, and carry the Baa2 spread as the optimistic edge of your band. One rupee of EBIT should never move a valuation, and a boundary is the place to show a range, never false certainty. Pull the live sheet before real use.

Step 3: assemble, pre-tax. Rd = 4.5 + 2.0 + 2.0 = 8.5% (INR).

Step 4: cross-check. Embedded: 60 ÷ (avg gross debt ₹600 cr) = 10.0%. That's above the estimate, so read the borrowing note. Half the debt is an old high-coupon term loan from a stress period. For a now-healthier borrower, the marginal rate is lower than the embedded average. The synthetic estimate stands, and the note explains the gap. If the embedded rate had been far below, you'd suspect the reverse, a deteriorated credit that isn't yet repriced in your inputs.

Step 5: after-tax. 8.5 × (1 − 0.25) = 6.4%.

Worked example 3: Marking book debt to market (rates moved)

Setup (synthetic). A firm carries book debt of ₹8,000 crore, with annual interest expense of ₹480 crore (embedded 6.0%) and a weighted-average maturity of 4 years, from the debt note. Its marginal cost of debt today, per the synthetic rating, is 9.5%. Mark the debt to market with the one-bond approximation.

MV = 480 × [1 − 1.095^−4]/0.095  +  8,000 × 1.095^−4
1.095^4 = 1.4377  ⇒  1.095^−4 = 0.6956
Annuity factor = (1 − 0.6956)/0.095 = 3.2045
MV = 480 × 3.2045 + 8,000 × 0.6956
   = 1,538 + 5,565 = ₹7,103 crore        (≈ 89% of book)

Lenders holding 6% paper in a 9.5% world have lost ~11% of face value. So the WACC weights should carry the ₹7,103 crore market claim, and the ₹8,000 crore book entry drops out. Run the direction drill. When rates rise, MV(debt) falls, and so does D/V. For healthy firms in stable rate regimes the adjustment is small enough to skip. But check before you skip it, and always mark it for firms whose credit has moved.

Worked example 4: Total beta: one clinic chain, two buyers, two values

Setup (synthetic). Arogya Diagnostics, an unlisted Indian pathology-lab chain, is for sale. Its sector unlevered beta (diagnostics/healthcare services, EM tables) is 0.80, and the typical sector correlation with the market is ρ = 0.40. After the deal the chain would carry D/E = 0.25. Marginal tax is 25% and λ = 1. Clean INR Rf 4.5%, India ERP 7.1%.

Buyer A: a listed hospital group (diversified shareholders). Market beta applies. βL = 0.80 × [1 + 0.75×0.25] = 0.80 × 1.1875 = 0.95. Re = 4.5 + 0.95×7.1 = 11.2%.

Buyer B: an individual doctor-entrepreneur putting her whole net worth in. Total risk applies. Total βU = 0.80/0.40 = 2.00. βL = 2.00 × 1.1875 = 2.375. Re = 4.5 + 2.375×7.1 = 21.4%.

Same clinic, same cash flows, yet the diversified buyer discounts at 11.2% and the concentrated buyer at 21.4%. So the diversified buyer can rationally pay roughly twice as much for the same growing cash-flow stream. With Gordon at g = 4.5%, the multiples of next-year flow are 14.9× against 5.9×. That explains why strategic and financial buyers routinely outbid founders' own reservation prices, and why promoters diversify by selling stakes. It's also why "the market is overpaying" often just means "the marginal buyer is better diversified than I am." Use total beta only on Buyer-B questions, because Buyer A prices the listed market.

Worked example 5: Blank-sheet WACC for an Indian pharma exporter (INR; bottom-up beta from peers; λ < 1)

Setup (synthetic). Bhavana Life Sciences is an Indian generics-and-API maker with 62% of revenue in regulated US/EU markets and the rest domestic, so λ ≈ 0.38. Market cap is ₹95,000 crore (≈ $11bn, so the large-cap rating table applies). Total debt incl. lease liabilities is ₹9,500 crore, EBIT ₹5,400 crore, interest expense ₹900 crore and marginal tax 25%. Cash flows are forecast in INR, which calls for an INR stack. Illustrative market inputs: clean INR Rf 4.5%, mature ERP 4.3%, India CRP 2.8%, country default spread 2.0%.

Station 1: Rf. INR cash flows call for the clean INR risk-free rate, Rf = 4.5%, which is the G-sec minus the sovereign default spread.

Station 2: bottom-up beta. Don't rely on a single regression. Unlever five listed pharma peers and average away the noise. Then relever at the target's own structure.

Relever at the target's D/E = 9,500/95,000 = 0.10, which gives βL = 0.90 × [1 + 0.75×0.10] = 0.90 × 1.075 = 0.97.

Station 3: cost of equity, with λ. Country risk follows where cash flows are earned, so weight the CRP by exposure: firm ERP = 4.3 + 0.38×2.8 = 5.36%. Re = 4.5 + 0.97 × 5.36 = 9.7% (INR). A λ-blind analyst charging the full 7.1% ERP would get 4.5 + 0.97×7.1 = 11.4%. That's nearly 2 points too high, and it over-discounts cash flows that mostly arise in developed markets.

Station 4: cost of debt. ICR = 5,400 / 900 = 6.0, which falls in the large-cap table's 5.5–6.5 band for A1/A+, with a spread of ≈ 0.8%. The firm borrows in rupees, in India, so it pays the sovereign's default spread in full. λ scales the equity CRP, never the debt's country spread. So Rd = 4.5 + 2.0 + 0.8 = 7.3% pre-tax, and after tax it's 7.3 × 0.75 = 5.5%.

Station 5: weights. V = 95,000 + 9,500 = ₹1,04,500 crore, so E/V = 90.9% and D/V = 9.1%. Book ≈ market here, because the credit is a healthy quasi-AAA and rates are near issuance levels.

Station 6: assembly.

WACC = 0.909 × 9.7%  +  0.091 × 5.5%
     = 8.82%         +  0.50%
     ≈ 9.3% (INR)

Below the India median is where a low-cyclicality exporter belongs, since its developed-market revenue thins the country-risk charge.

The lesson. There are two dials, and you set them independently. Currency is one. INR means the clean INR Rf and the full sovereign spread inside INR debt. The geography of the cash flows is the other, and λ = 0.38 means a shrunken equity CRP. Confuse them, by scaling the debt spread by λ or leaving the equity CRP at full, and the WACC breaks by ~2 points, in opposite directions.

Worked example 6: Blank-sheet WACC for a US industrial (USD; actual rating beats synthetic; debt marked to market)

Setup (synthetic). Cardinal Fasteners Corp is a US industrial-fastener maker with a market cap of $28bn and borrowings plus lease liabilities of $10bn book value. EBIT is $3.2bn, interest expense $520m and marginal tax 25%. All-US revenue means λ·CRP = 0. Rates have risen ~150 bp since most of the debt was issued. Its sector unlevered beta (industrials, US table, cash-corrected) is 0.95, Rf = 4.2% (FRED DGS10) and mature ERP = 4.3%.

Station 1: Rf. USD cash flows mean Rf = 4.2%, and the Treasury needs no cleaning.

Station 4, run first: cost of debt. It goes first because it feeds both the weights and the beta. ICR = 3,200 / 520 = 6.15, which on the large-cap table gives a synthetic A1/A+ with a spread of ≈ 0.8%, for a synthetic Rd ≈ 5.0%. But the actual issuer rating is Baa2/BBB. Agencies penalize the lease-heavy structure and the cyclicality, which one ratio can't see. When a real rating exists, it wins. The synthetic is the cross-check. The BBB spread is ≈ 1.6%, so Rd = 4.2 + 1.6 = 5.8% pre-tax, and after tax 5.8 × 0.75 = 4.35%.

Now mark the debt to market. Embedded coupon = 520/10,000 = 5.2%, and the weighted-average maturity is 6 years. Discount at today's 5.8%:

MV = 520 × [1 − 1.058^−6]/0.058  +  10,000 × 1.058^−6
   = 520 × 4.948  +  10,000 × 0.713
   = 2,573 + 7,130 = $9,703m           (≈ 97% of book)

Rates are up, so the old 5.2% paper is worth less than book. Use $9,703m in both the weights and the relevering D/E, in place of the $10,000m book figure.

Station 2: beta. Market D/E = 9,703/28,000 = 0.35. βL = 0.95 × [1 + 0.75×0.35] = 0.95 × 1.263 = 1.20.

Station 3: cost of equity. Re = 4.2 + 1.20 × 4.3 = 9.4%. It's all-US, so there's no CRP.

Station 5: weights. V = 28,000 + 9,703 = $37,703m, so E/V = 74.3% and D/V = 25.7%.

Station 6: assembly.

WACC = 0.743 × 9.4%  +  0.257 × 4.35%
     = 6.98%         +  1.12%
     ≈ 8.1% (USD)

That's right on the US median, as it should be for a levered, cyclical large-cap.

The lesson. In practice the cost-of-debt station doesn't always come last. Here it had to run first. It's the actual rating that sets Rd, and the flattering synthetic doesn't. Today's Rd is what marks the debt to market, and that market value drives both the weight and the D/E the beta is relevered on. Get Rd wrong and three later stations inherit the error.

Practice set

These problems ramp up. P1–P5 are guided single-station drills, P6–P9 are multi-station and full builds, and P10–P11 are independent and timed. Unless a problem says otherwise, use the illustrative market inputs from the teaching above: US Rf 4.2% and mature ERP 4.3%, clean INR Rf 4.5%, India default spread 2.0%, India ERP 7.1%, and marginal tax of 25% in both markets. In real work, always use live sources.

P1 (guided). The 10-year G-sec yields 7.0%, and Moody's rates India Baa3, which maps to a default spread of 2.2%. (a) Compute the clean INR risk-free rate. (b) Why not just use 7.0%? Answer in a single sentence.

Solution. (a) Rf = 7.0 − 2.2 = 4.8%. (b) A Baa3 sovereign's yield contains compensation for default risk, which means it isn't risk-free, and because the CRP already charges for India risk, using the raw yield bills the same risk twice.

P2 (guided). A peer's regression beta is 1.32 at market D/E of 0.6, tax 25%. Unlever it.

Solution. βU = 1.32 / [1 + 0.75×0.6] = 1.32 / 1.45 = 0.91.

P3 (guided). Build a bottom-up beta for a target at D/E = 0.40 (tax 25%) from three peers: A: βL 1.10, D/E 0.20 · B: βL 1.25, D/E 0.50 · C: βL 0.95, D/E 0.10.

Solution. Unlever each: A: 1.10/1.15 = 0.957 · B: 1.25/1.375 = 0.909 · C: 0.95/1.075 = 0.884. Average βU = (0.957+0.909+0.884)/3 = 0.917. Relever: βL = 0.917 × [1 + 0.75×0.40] = 0.917 × 1.30 = 1.19.

P4. An unrated Indian firm has a market cap of ₹9,500 crore (< $5bn, so use the small-cap table), EBIT of ₹850 crore and interest of ₹250 crore. Illustrative small-cap rows: ICR 3.0–3.5 gives Ba2/BB at a spread of 2.4%, and 3.5–4.0 gives Ba1/BB+ at 2.0%. Compute the ICR, synthetic rating, pre-tax and after-tax INR cost of debt.

Solution. ICR = 850/250 = 3.4, which gives Ba2/BB at a spread of 2.4%. Rd = 4.5 + 2.0 + 2.4 = 8.9% pre-tax (clean Rf + country spread + firm spread). After tax, 8.9 × 0.75 = 6.68% ≈ 6.7%.

P5. Book debt ₹4,000 crore, interest expense ₹240 crore, weighted-average maturity 5 years. Today's marginal cost of debt: 8.5%. Mark the debt to market.

Solution. Embedded coupon = 240/4,000 = 6%. 1.085^5 = 1.5037 ⇒ discount factor 0.6650. Annuity factor = (1−0.6650)/0.085 = 3.9406. MV = 240×3.9406 + 4,000×0.6650 = 946 + 2,660 = ₹3,606 crore, about 90% of book. Use ₹3,606 crore in the weights.

P6 (guided full build, US). Corvid Logistics (synthetic US freight firm): market cap $40bn, total debt incl. leases $20bn, sector βU 0.95, EBIT $4.0bn, interest $1.05bn, all-US revenue. Large-cap table row: ICR 3.0–4.25 gives A3/A−, spread 1.2%. Build the WACC.

Solution. (1) Rf = 4.2% (USD). (2) D/E = 20/40 = 0.50, so βL = 0.95×[1+0.75×0.5] = 0.95×1.375 = 1.31. (3) Re = 4.2 + 1.31×4.3 = 9.8%. (4) ICR = 4.0/1.05 = 3.81, which is A3/A−, so Rd = 4.2+1.2 = 5.4%, or 4.05% after tax. (5) V = $60bn, so E/V = 66.7% and D/V = 33.3%. (6) WACC = 0.667×9.8 + 0.333×4.05 = 6.54 + 1.35 = 7.9% (USD). Sanity check: mid-band for a levered cyclical, so it passes.

P7. An Indian founder is buying out his partner to own 100% of an unlisted logistics firm (his entire net worth). Sector βU 0.90, sector correlation with market ρ = 0.45, post-deal D/E 0.20, tax 25%. (a) Cost of equity for the founder. (b) For a listed acquirer valuing the same firm. (c) One sentence on the gap.

Solution. (a) Total βU = 0.90/0.45 = 2.00, so βL = 2.00×[1+0.75×0.20] = 2.30 and Re = 4.5 + 2.30×7.1 = 20.8%. (b) βL = 0.90×1.15 = 1.035, so Re = 4.5 + 1.035×7.1 = 11.8%. (c) The founder bears total risk while the acquirer's shareholders diversify away the firm-specific part, so the same cash flows are worth far more to the acquirer. Concentration has a price, and this is it.

P8. An Indian pharma exporter earns 65% of revenue in the US/EU (λ = 0.35), bottom-up βL = 1.05, INR forecast. (a) Cost of equity. (b) What would a λ-blind analyst (λ = 1) get, and how large is the error?

Solution. (a) Firm ERP = 4.3 + 0.35×2.8 = 5.28%, so Re = 4.5 + 1.05×5.28 = 10.0%. (b) With λ = 1, Re = 4.5 + 1.05×7.1 = 11.96% ≈ 12.0%. That's about 2 points too high, and it over-discounts cash flows that mostly arise in developed markets.

P9 (independent full build, India). Suvarna Consumer Products (synthetic FMCG): market cap ₹24,000 crore, total debt incl. leases ₹1,000 crore, sector βU 0.72, EBIT ₹900 crore, interest ₹70 crore, all-India revenue. Small-cap table row: ICR > 12.5 gives Aaa/AAA, spread 0.6%. Build the full WACC.

Solution. (1) Rf = 4.5% clean INR. (2) D/E = 1,000/24,000 = 0.0417, so βL = 0.72×[1+0.75×0.0417] = 0.72×1.0313 = 0.74. (3) λ = 1 means an ERP of 7.1%, so Re = 4.5 + 0.74×7.1 = 9.8%. (4) ICR = 900/70 = 12.9, which is Aaa/AAA, so Rd = 4.5+2.0+0.6 = 7.1%, or 5.33% after tax. (5) V = ₹25,000 cr, so E/V = 96% and D/V = 4%. (6) WACC = 0.96×9.8 + 0.04×5.33 = 9.41 + 0.21 = 9.6% (INR). Sanity check: it's at the low edge of the India band, which is right for a near-debtless staples business, and the low-cyclicality βU did the work.

P10 (error hunt). An analyst values an Indian retailer (INR forecasts) as follows: "Rf = 4.2% (US 10-yr Treasury, cleaner market); β = 1.62 (5-yr regression, Yahoo Finance); ERP = 7.1%; so Re = 15.7%. Cost of debt = interest/book debt = 5.1%, after-tax at last year's effective 18% → 4.2%. Weights: book equity ₹6,000 cr, book debt ₹4,000 cr → 60/40. WACC = 0.6×15.7 + 0.4×4.2 = 11.1%." Identify every error and state the direction of each bias.

Solution. There are four errors. (1) Currency mismatch. Putting a USD Rf under INR cash flows and an INR-consistent ERP understates Rf by ~0.3 points here, and it's conceptually broken, because the stack must be one currency. (2) Raw regression beta. It carries a ±0.3 standard error, and nobody knows its index or interval. It should be built bottom-up from retail peers, and 1.62 likely overstates a retailer's business risk, so Re comes out too high. (3) Cost of debt. It uses the embedded book average where the marginal rate belongs, and the effective tax rate where the marginal 25% belongs. That 18% understates the shield, which makes after-tax Kd too high, while the embedded 5.1% likely understates the marginal rate, which pushes it too low. Both are wrong even where they offset. (4) Book weights. A listed retailer's market cap won't equal its ₹6,000 cr book. Book weights overweight debt, because retailers trade far above book, and that biases WACC down. So the 11.1% is a coincidence of compensating errors. It's indefensible even if it comes out numerically close to a clean build.

P11 (timed, 20 minutes, no notes). Build a full INR WACC: clean Rf 4.5%, India ERP 7.1%, λ = 1; sector βU 1.00; market cap ₹8,000 cr, debt ₹2,000 cr; tax 25%; ICR given as 5.0; small-cap table row: 4.5–6.0 for A3/A−, spread 1.2%.

Solution. D/E = 2,000/8,000 = 0.25, so βL = 1.00×1.1875 = 1.19 and Re = 4.5 + 1.19×7.1 = 12.9%. Rd = 4.5+2.0+1.2 = 7.7%, or 5.78% after tax. Weights: E/V = 80%, D/V = 20%. WACC = 0.8×12.9 + 0.2×5.78 = 10.32 + 1.16 = 11.5% (INR). At exam speed the chain runs βL 1.1875 → Re 12.93 → WACC 11.50, and anything in 11.4–11.6 with the correct method passes.

P12 (guided). A firm carries book debt ₹6,000 crore with annual interest expense ₹480 crore (embedded coupon 8%) and a weighted-average maturity of 5 years. Rates have fallen since issuance, and its marginal cost of debt today is 6.5%. (a) Mark the debt to market with the one-bond approximation. (b) What does the result do to the debt weight in WACC, compared with using book? Answer in a single sentence.

Solution. (a) Discount at 6.5%. 1.065^−5 = 0.7299, and the annuity factor = (1 − 0.7299)/0.065 = 4.1557. MV = 480 × 4.1557 + 6,000 × 0.7299 = 1,995 + 4,379 = ₹6,374 crore, about 106% of book. (b) Rates fell below the 8% coupon, so the fixed-rate debt is now worth more than book (₹6,374 > ₹6,000), and the market debt weight rises relative to book, the mirror image of the rates-up case in Worked Example 3.

P13 (guided). Nalanda Renewables (synthetic) is currently loss-making and pays no tax. Its structure: E/V 75% at Re 13.5%, D/V 25% at a pre-tax cost of debt of 10%; marginal statutory rate 25%. (a) Compute its WACC now. (b) What would the WACC be once it is profitable? (c) Which of the ten errors does the naive answer commit?

Solution. (a) With no taxable income there's no interest tax shield, so after-tax Rd = pre-tax Rd = 10%. WACC = 0.75 × 13.5 + 0.25 × 10 = 10.13 + 2.50 = 12.6%. (b) With the shield restored, after-tax Rd = 10 × 0.75 = 7.5%, and WACC = 0.75 × 13.5 + 0.25 × 7.5 = 10.13 + 1.88 = 12.0%. (c) Granting the (1−t) shield to a firm that pays no tax is error #6, the phantom shield. Only once the firm pays tax again do the shield and the drop toward 12.0% come back, and carryforwards can delay that further.

P14 (independent). Deccan Polymers (synthetic Indian specialty-chemicals): market cap ₹15,000 crore (< $5bn, so the small-cap table), total debt incl. leases ₹5,000 crore, sector βU 0.95, EBIT ₹1,200 crore, interest ₹250 crore, all-India revenue (λ = 1), tax 25%. Small-cap table row: ICR 4.5–6.0 gives A3/A−, spread 1.2%. Build the full WACC.

Solution. (1) Rf = 4.5% clean INR. (2) D/E = 5,000/15,000 = 0.333, so βL = 0.95 × [1 + 0.75×0.333] = 0.95 × 1.25 = 1.19. (3) λ = 1 means an ERP of 7.1%, so Re = 4.5 + 1.19 × 7.1 = 12.9%. (4) ICR = 1,200/250 = 4.8, which is A3/A−, so Rd = 4.5 + 2.0 + 1.2 = 7.7% pre-tax, and after tax 7.7 × 0.75 = 5.8%. (5) V = ₹20,000 cr, so E/V = 75% and D/V = 25%. (6) WACC = 0.75 × 12.9 + 0.25 × 5.8 = 9.68 + 1.45 = 11.1% (INR). Sanity check: middle of the India band, a moderately levered mid-cap, so it passes.

P15 (independent). Summit Fasteners Inc. (synthetic US industrial): market cap $20bn, total debt incl. leases $8bn, sector βU 1.10, EBIT $2,400m, interest $600m, all-US revenue, tax 25%. Large-cap table row: ICR 3.0–4.25 gives A3/A−, spread 1.2%. Build the full WACC.

Solution. (1) Rf = 4.2% USD. (2) D/E = 8/20 = 0.40, so βL = 1.10 × [1 + 0.75×0.40] = 1.10 × 1.30 = 1.43. (3) Re = 4.2 + 1.43 × 4.3 = 10.3%. (4) ICR = 2,400/600 = 4.0, which is A3/A−, so Rd = 4.2 + 1.2 = 5.4%, and after tax 5.4 × 0.75 = 4.05%. (5) V = $28bn, so E/V = 71.4% and D/V = 28.6%. (6) WACC = 0.714 × 10.3 + 0.286 × 4.05 = 7.35 + 1.16 = 8.5% (USD). Sanity check: mid-band for a levered cyclical industrial, so it passes.

P16 (independent). Riverside Dermatology Group (synthetic), an unlisted US clinic chain, is for sale. Sector βU is 0.85 and the sector correlation with the market is ρ = 0.50. Post-deal D/E is 0.30 and tax is 25%. It's all-US, in USD, with Rf 4.2% and ERP 4.3%. (a) Cost of equity to a diversified PE buyer. (b) Cost of equity to an individual physician investing her whole net worth. (c) At Gordon growth g = 3%, the multiple of next-year equity cash flow each can pay, and what the gap explains.

Solution. (a) Market beta applies, so βL = 0.85 × [1 + 0.75×0.30] = 0.85 × 1.225 = 1.04 and Re = 4.2 + 1.04 × 4.3 = 8.7%. (b) Total risk applies, so total βU = 0.85/0.50 = 1.70, βL = 1.70 × 1.225 = 2.08 and Re = 4.2 + 2.08 × 4.3 = 13.2%. (c) Multiple = 1/(Re − g). For the diversified buyer it's 1/(0.087 − 0.03) = 17.5×, and for the concentrated one 1/(0.132 − 0.03) = 9.8×. So the diversified buyer can rationally pay about 1.8× as much for identical cash flows. That explains why strategic and PE buyers routinely outbid founders, and why "the market is overpaying" is often just "the marginal buyer is better diversified than I am."

P17 (timed, 20 minutes, no notes). Bharat Precision Forgings (synthetic Indian auto-ancillary exporter): 60% of revenue in developed markets, so λ = 0.40. Build a bottom-up beta from three peers, A: βL 1.10, D/E 0.30 · B: βL 1.30, D/E 0.60 · C: βL 1.00, D/E 0.20 (tax 25% throughout), then finish the full INR WACC. Target: market cap ₹18,000 crore, total debt incl. leases ₹6,000 crore, EBIT ₹1,000 crore, interest ₹200 crore. Small-cap table row: ICR 4.5–6.0 gives A3/A−, spread 1.2%.

Solution. Unlever: A 1.10/1.225 = 0.898 · B 1.30/1.45 = 0.897 · C 1.00/1.15 = 0.870, for an average βU = 0.888. D/E = 6,000/18,000 = 0.333, so βL = 0.888 × [1 + 0.75×0.333] = 0.888 × 1.25 = 1.11. Firm ERP = 4.3 + 0.40×2.8 = 5.42%, so Re = 4.5 + 1.11 × 5.42 = 10.5%. ICR = 1,000/200 = 5.0, which is A3/A−, so Rd = 4.5 + 2.0 + 1.2 = 7.7%, or 5.8% after tax. Weights: V = ₹24,000 cr, so E/V = 75% and D/V = 25%. WACC = 0.75 × 10.5 + 0.25 × 5.8 = 7.88 + 1.45 = 9.3% (INR). At exam speed the chain is βU 0.888 → βL 1.11 → Re 10.5 → WACC ≈ 9.3, and anything 9.2–9.4 with the correct method and the λ-weighted CRP passes. Forget λ and Re inflates to 4.5 + 1.11×7.1 = 12.4%, with the WACC at ~10.7%.

Applied mini-project

Task: two live blank-sheet WACC builds, one Indian and one US, fully sourced. Pick one Indian non-financial with meaningful debt (e.g., UltraTech Cement, Tata Motors, InterGlobe Aviation, Grasim). Get its annual report from the company site or BSE/NSE, and its financials via screener.in. Then pick one US non-financial (e.g., Home Depot, Union Pacific, PepsiCo) and get its latest 10-K via SEC EDGAR. These are teaching specimens and not recommendations. For each one, produce a spreadsheet plus a one-page note covering the following:

1. Inputs page. Every raw input with its source and date: G-sec/DGS10 yields, Damodaran default spread, ERP/CRP, sector unlevered beta (his tables), market cap, debt including lease liabilities (cite the exact note numbers), EBIT, interest, marginal tax rate (tax note), revenue geography for λ.
2. The six stations, computed the same way as in the two blank-sheet builds. Include the synthetic rating even if an actual rating exists. Compare the two, and comment on any gap.
3. Sanity page. Set your WACC against Damodaran's sector cost-of-capital table for that market, and explain any deviation > 1 point.
4. Sensitivity line. WACC at β ± 0.15 and at the neighboring rating notch.
5. A ten-line memo. Put the two WACCs side by side, and explain why the INR-USD gap is mostly not a risk judgment.

Scoring rubric. You pass with ≥ 80% of available points. Log the score and your top miss in the journal.

Case Lab: run case CL-3.1 (Two Yardsticks: an Indian exporter and a US cyclical) in the app's Cases tab for this module. Build both WACCs station by station (an INR exporter with λ < 1 and a bottom-up beta; a USD cyclical whose actual rating beats its synthetic and whose debt must be marked to market), then defend every input against a seeded error before the grader reveals it.

Reading & resources

Primary texts.

• Aswath Damodaran, Investment Valuation (3e), Ch. 7 (Riskless Rates and Risk Premiums) and Ch. 8 (Estimating Risk Parameters and Costs of Financing): the spine of everything above, at full depth. [Paid] [Intermediate]
• Damodaran, The Little Book of Valuation: Ch. 2–3 for a compressed second pass. [Paid] [Beginner]
• McKinsey (Koller, Goedhart, Wessels), Valuation (8e), the cost-of-capital chapter (Part Two): the practitioner-house view; note where it differs from Damodaran (e.g., ERP philosophy) and why both are defensible. [Paid] [Intermediate]
• Pablo Fernandez, "WACC: Definition, Misconceptions and Errors" and "119 Common Errors in Company Valuations": free on SSRN; read after the ten-errors section for the extended horror catalog. [Free] [Intermediate]

Live data. These are your standing sources. Bookmark them all, and refresh each January as part of the ERP-card ritual.

• Damodaran Online (NYU Stern, pages/datasets). The pages you want are Country Default Spreads and Risk Premiums (ctryprem), Levered and Unlevered Betas by Industry (betas, in US/global/emerging sets, where you use the cash-corrected column) and Total Betas by Sector. Add Ratings, Interest Coverage Ratios and Default Spread (ratings) and Cost of Capital by Industry Sector (wacc, US plus emerging). The January "Data Update" posts on his Musings on Markets blog walk through every table. [Free] [All levels]
• FRED series DGS10 (US 10-yr Treasury). RBI Database on the Indian Economy + CCIL (ccilindia.com) for G-sec yields. [Free]
• SEC EDGAR (10-K: debt, lease, and tax notes); screener.in / BSE / NSE for Indian financials, borrowing notes, and shareholding. [Free]

Video & courses.

• Damodaran's NYU Valuation course (free on YouTube + his site). The cost-of-capital sessions (risk-free rates, betas, cost of debt; typically sessions 6–9 of the spring class): watch after reading, with his slides. [Free] [Intermediate]
• Zerodha Varsity, Fundamental Analysis and Financial Modelling modules: the DCF/cost-of-capital sections for India-flavored reinforcement. [Free] [Beginner]
• CFA Level I Cost of Capital reading (in the Corporate Issuers study session), a clean drill set if you have curriculum access. [Paid] [Beginner-Intermediate]

Advanced/reference.

• Pratt & Grabowski, Cost of Capital: Applications and Examples, the exhaustive reference, incl. private-company practice (total beta's home turf). [Paid] [Advanced]
• Damodaran's papers (free, SSRN/his site): "What is the riskfree rate?", "Equity Risk Premiums: Determinants, Estimation and Implications" (annual), "Country Risk: Determinants, Measures and Implications" (annual). [Free] [Advanced]

Going deeper.

• Aswath Damodaran, "The Cost of Capital: The Swiss Army Knife of Finance" (essay + slides on his NYU Stern site and the Musings on Markets blog) [Free] [Intermediate]. His synthesis of every job this one number does (discount rate, hurdle rate, optimization target) and every way practitioners abuse it; the ideal capstone after the ten-errors section. Verify the current link before citing.
• Dimson, Marsh & Staunton, Global Investment Returns Yearbook (UBS) [Paid] [Advanced]. The long-run, multi-country evidence on realized equity risk premiums and real returns that sits underneath the mature ERP you plug into Re; the annual summary is free, the full volume paid. A reality check on how big the ERP has actually been across a century and 20-plus markets.

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

{
  "skill": "The six stations of a WACC build — currency-matched risk-free rate, bottom-up beta by unlever-average-relever, total beta, cost of equity, cost of debt, market-value weights — and the sanity ranges that catch a broken build",
  "use": "A WACC build is a supply chain of eight inputs, each with a source and a date, and its failures are almost always sourcing failures rather than arithmetic ones. That is exactly what a tool can help with. (1) Ask it to build the input register: for each station, the specific series or filing line, its publisher, its as-of date, and the convention it carries. Then fill the register yourself from those sources — a register you can hand to a reviewer is what makes a WACC auditable, and §4.9's sanity net only works if you can trace which input broke. (2) Ask it to assemble the peer set for the bottom-up beta and to pull each peer's market capitalisation, total debt including leases, and marginal tax rate from filings; unlevering is one line, and finding twenty-five peers' capital structures is an afternoon. (3) Ask it to read the segment note and propose the value-weighted business blend for a multi-business firm, with the EV/Sales proxy stated — then argue with the weights, which are the judgment. (4) Ask it to name every place your build could be double-counting country risk, which is §4.9's first suspect and the easiest error to make twice.",
  "tools": ["Chat assistants — Claude, ChatGPT for the input register, the peer-set assembly and the country-risk double-count audit", "Document/PDF AI — for peer capital structures, lease liabilities, tax notes and segment disclosures", "Coding assistants — for the unlever-average-relever loop and the uncertainty decomposition below"],
  "prompt": "For a <SECTOR> company in <COUNTRY>, produce an input register for a WACC build: one row per input (risk-free rate, equity risk premium, country risk premium, unlevered beta, capital structure, marginal tax rate, cost of debt), with columns [Input | Specific series or filing line | Publisher | As-of date | Convention or units | Where double-counting could occur]. Do NOT supply any values, and do NOT compute a WACC — I am doing both. Flag any row where the convention differs between the Indian and US sources.",
  "verify": "Run §4.9's sanity net on your own output before anyone else does: an established non-financial firm outside roughly 6–13% in dollars or 9–16% in rupees has a broken input, and the order of suspects is double-counted country risk, currency mismatch, a raw regression beta, then book weights. Then check the two things a tool will get wrong silently: the risk-free rate must be currency-matched to the cash flows it will discount, and the weights must be market values with lease liabilities treated consistently in both the weight and the beta. Disclose the levering convention — this program uses the (1 − t) Hamada form — and never mix it with Harris–Pringle inside one analysis. Every input carries its date. A model asked for 'the WACC of X' will produce a number in the plausible range with no register behind it, and a plausible WACC with untraceable inputs is worse than an implausible one, because nothing will make you check it (Guardrail 1-5).",
  "diy": "The gate is unaided: you run all six stations from filings and published data, unlever and relever by hand, assemble the WACC, and diagnose a broken build from its output. Nothing in this callout is admissible in that room."
}

Modern Data Analysis

By hand first. You ran the cement build station by station. You unlevered five peers with their own structures and tax rates, relevered at the target, and calibrated the answer against the sector distribution. That sequence is the skill, and the gate keeps you doing it by hand. Code adds two things a hand build can't. One is the peer count that makes the bottom-up beta statistically worth doing. The other is a defensible uncertainty on the output.

Today's workflow. Write the build as a function of a dated input register and you get three properties. First, the peer set can be large. Bottom-up beta wants eight to thirty names, and the √n variance reduction only pays if you actually use them. Regressing thirty peers, unlevering each with its own market debt-to-equity and marginal rate, and taking the median asset beta is a loop. It isn't a project. Second, the register becomes reproducible. Every input is a row with a value, a source and an as-of date, so you can rebuild a WACC months later and a reviewer can find which input moved. A register like that makes the sanity check's diagnostic order usable in practice. Third, the output becomes a range. Once the build is a function, you can push the register's uncertainties through it and read the distribution of WACC instead of a point. That's what the Quantitative lens below does, and it turns the "±1-point WACC sensitivity on every DCF" instruction from a convention into a measurement. Two smaller mechanics are worth automating too. One is a time-varying WACC path for a firm whose leverage is migrating, re-levering beta year by year as the sanity-check station notes. The other is the cash correction to an unlevered beta for cash-heavy firms, which is a single division and is routinely forgotten.

Tools & sources (IN + US). Use pandas for the register and the peer table, statsmodels for the peer regressions, numpy for the propagation. For India, take G-sec yields from FBIL and RBI's DBIE for the rupee risk-free rate. From the annual report's borrowings and lease notes you get the cost of debt and the market-value weights, and the tax note gives the marginal rate. When there's no traded bond, CRISIL/ICRA/CARE rating rationales support a synthetic spread. For a multi-business blend, use the segment note. Re-pull Damodaran's emerging-market beta and cost-of-capital tables (betas, totalbeta, and the cash-corrected column) each January. For the US, FRED (DGS10, DGS30, BAMLC0A0CM, BAMLH0A0HYM2) gives the risk-free rate and credit spreads, and EDGAR gives peer capital structures and lease liabilities. Damodaran's US and global datasets supply sector betas and the cost-of-capital distribution that the sanity table is calibrated against.

# Where WACC uncertainty actually lives — R3: recomputed in-session.
# Illustrative INR inputs chosen to land inside the module's own 9-16% sanity band.
import numpy as np
base = dict(rf=.0700, erp=.0500, bU=.93, DE=.0625, t=.25, rd=.0850, DV=.0588)
sd   = dict(rf=.0025, erp=.0100, bU=.060, DE=.0100, t=.010, rd=.0050, DV=.0100)
def wacc(rf, erp, bU, DE, t, rd, DV):
    return (1 - DV) * (rf + bU * (1 + (1 - t) * DE) * erp) + DV * rd * (1 - t)
rng = np.random.default_rng(303); n = 400_000
draws = {k: rng.normal(base[k], sd[k], n) for k in base}
print(round(wacc(**base) * 100, 2), round(wacc(**draws).std() * 100, 3))      # 11.54  0.995
var = {k: wacc(**{j: draws[j] if j == k else np.full(n, base[j]) for j in base}).var()
       for k in base}                                                          # one at a time
tot = sum(var.values())
print({k: round(v / tot * 100, 1) for k, v in sorted(var.items(), key=lambda x: -x[1])})
#   {'erp': 85.1, 'bU': 8.8, 'rf': 5.6, 'DV': 0.3, 'DE': 0.1, 'rd': 0.0, 't': 0.0}

Verify. Reproduce the cement arithmetic from the teaching before you trust the loop. Your five peers unlever to asset betas of 0.977, 0.977, 0.934, 0.854 and 0.925, for an average of 0.933. Relevering the rounded 0.93 at a market debt-to-equity of 0.0625 with a 25% tax rate gives 0.97. Carry the unrounded average through and you get 0.98, so state which you did. Then verify the sources, which matter more than the sums. A risk-free rate is a specific series on a specific date. A cost of debt is a marginal rate, and the average rate implied by last year's interest expense won't do. Weights are market values, with lease liabilities in the same place in both the weight and the beta. Run the sanity net. If the output is out of band, work through the sanity check's suspect list in order, and don't nudge inputs until the answer looks reasonable. That's the one failure mode a reproducible register makes visible and a spreadsheet hides. On the third leg, a ₹100-face 8.5% preference share at ₹92 costs 9.24%. A build with ₹46,000 crore of equity at 13.2%, ₹2,000 crore of preferred and ₹12,000 crore of debt at 8.6% pre-tax gives a three-leg cost of capital of 11.72%. Dropping the preferred and rescaling the weights gives 11.80%, a gap of 8 basis points.

Quantitative lens

Back at the sanity-check station you met two calibration claims that deserve numbers. First, the cross-firm spread of WACC is narrow. Second, a thesis resting on 40 basis points of discount rate isn't a thesis. Both are right. But neither answers a third question: where inside the build does the uncertainty actually live? That one has a sharp, actionable answer.

Push a defensible uncertainty through each input of the cement-style rupee build and look at the variance contributions. Take one standard deviation of ±25 basis points on the risk-free rate (which G-sec, which maturity, spot or averaged). Take ±1.0 percentage point on the equity risk premium (historical versus implied versus normalised), since the two camps differ by about this much. Then take ±0.06 on the unlevered beta (the eight-peer standard error from the risk-and-return work), ±0.50 points on the cost of debt, ±1 point on the tax rate, and ±1 point on the debt weight. On a base WACC of 11.54%, the resulting distribution has a standard deviation of 0.995 points. Its 5th-to-95th-percentile range runs from 9.93% to 13.20%, a width of 3.27 percentage points, and that spans a 1.66-fold range of perpetuity value at 5% terminal growth.

The decomposition is the finding. Of the summed variance, the equity risk premium contributes 85.1%, the unlevered beta 8.8%, and the risk-free rate 5.6%, together 99.5%. Everything on the debt side of the build adds up to 0.5%. That covers the cost of debt, the tax rate, the debt weight and the peer debt-to-equity ratios put together, and the cost of debt alone contributes essentially nothing. Yet practising analysts spend a disproportionate share of their WACC time on the debt side. They hunt for a traded bond yield. They argue about synthetic spreads, or about whether to use the marginal or the average rate. All that effort goes into half a percent of the variance. That work isn't wrong, and careful cost-of-debt work still matters for a heavily levered firm where the debt weight is large. On a lightly levered company, though, it's polishing a rounding error. Two inputs carry the answer: a macro premium you should be sourcing with a stated policy, and a beta you should be averaging over more peers.

Three consequences are worth carrying away. First, the ±1-point WACC sensitivity you've been told to use isn't conservatism. It's roughly the actual one-sigma uncertainty of a careful build. Second, the highest-value marginal effort in any cost-of-capital exercise is more peers for the beta and a written, dated ERP policy. Those two moves shrink 94% of the variance. Third, the advice to spend your effort on cash flows still holds. The WACC's own uncertainty is real and worth disclosing, but ROIC and growth judgments move value further still, and now you can say by how much instead of just asserting it.

Now the limits. The input standard deviations are assumptions, though defensible ones. The whole decomposition scales with them, so substitute your register's uncertainties and re-run, and don't quote these percentages. This simulation treats the inputs as independent, and they aren't. The risk-free rate and the equity risk premium move together, because an implied ERP is computed against a risk-free rate, so a change in one partly offsets the other. Beta and the debt weight are linked through leverage too. So the true WACC distribution is somewhat tighter than this simulation shows. Normal draws also give beta and the premium symmetric tails, whereas the ERP's plausible range is asymmetric and camp-dependent. And a distribution of WACC isn't a distribution of value. Those same premium and beta uncertainties also touch the terminal-growth cap and the exit multiple. Propagating them into a full DCF is a job for the terminal-value work, and it's a different exercise from propagating them into one rate.

Do it in code. Carry the register as data with an uncertainty column, propagate it, and report the WACC as a central value with its one-sigma band and the variance decomposition beside it. Then check whether your marginal hour is going to the input that owns 85% of the variance. See G1/QM1.01 and G2/DA1.05.

Where this goes next: galaxy cross-links

• G1 · QM1.01 Probability, Distributions & Estimation: error propagation and variance decomposition, which is what turns a list of contestable inputs into a statement about which one matters.
• G1 · FI1.01 Bond Pricing & Yield Measures: the cost of debt as a yield rather than as an interest ratio. The marginal rate is a price in a market, and this is where you learn to read it.
• G1 · FI1.04 Embedded Options & OAS: why a callable or convertible instrument's quoted yield is not its cost of debt, which is the sharpest single trap in station 4.
• G2 · DA1.05 Statistics in Code: the peer regressions, the median asset beta and the bootstrapped band, done at a peer count large enough for the √n gain to be real.
• G1 · AA1.02 Black–Litterman & Risk Budgeting: the disciplined way to hold a view about a market-wide premium instead of importing one, which is the ERP policy decision the sanity check asks you to write down.

Teach it back & journal

Feynman prompt. Explain to a smart 15-year-old why a rupee promised by an Indian cement company must be judged against a different yardstick than a dollar promised by an American hardware chain. Then build both yardsticks for her, out loud, from five ingredients each: a government bond, a subtraction, a family of similar companies, a loan-price table, and a weighing scale. Do it without once saying "CAPM," "beta," or "WACC." Then write the one-page explainer: "A company's cost of capital is what its investors give up, not what it pays." If any station of the assembly line forces jargon without a picture behind it, that station isn't yours yet.

Journal reflection. Ever since the returns-on-capital work you've judged every ROIC against a "~12% hurdle" on faith. Tonight, retire the faith. Record the WACC you built for your mini-project's Indian company next to that 12%. Where does the difference come from? And which of the six stations are you least confident you could defend, line by line, to a skeptical partner? Write down the one input you're most tempted to fudge when a valuation someday disagrees with your prior. Then write down which rule from the ten-errors list you'll invoke against yourself when the temptation arrives.

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