WACC Explained, and Why the Number Is Mostly an Assumption

The formula, one term at a time

WACC = E/V × Re + D/V × Rd × (1 − T). E is the market value of equity, D the market value of debt, and V their sum, so E/V and D/V are simply the shares of the funding each side provides and must add to one. Re is the return shareholders require, Rd the rate lenders charge, and T the marginal tax rate. The whole expression is an average of two prices weighted by how much of each you use, with one adjustment: the debt term is multiplied by (1 − T) because interest reduces taxable profit and the tax authority therefore pays part of it.

Two conventions matter before you type anything into a spreadsheet. Use market values, not book values, for E and D: the book value of equity records what shareholders paid years ago, while the price they would accept today is what they are actually giving up by leaving the money in. And use the weights you intend to maintain, not last night's accident of share price. WACC is a forward-looking hurdle rate, so it should reflect the capital structure the company will fund itself with, which is why most practitioners use a target ratio and hold it steady rather than recomputing every quarter.

Worked in full: 8.35%

The company has $700 million of equity and $300 million of debt, so V is $1,000 million and the weights are 0.70 and 0.30. Start with equity. CAPM says Re = risk-free rate + beta × equity risk premium. With a 4.0% risk-free rate, a beta of 1.20 and a 5.0% premium, Re = 4.0 + 1.20 × 5.0 = 10.0%. Now debt. The company borrows at 6.0%, but interest is deducted before tax, so at a 25% marginal rate each dollar of interest costs the shareholders only 75 cents: the after-tax cost is 6.0 × 0.75 = 4.5%.

Weight and add. The equity leg contributes 0.70 × 10.0 = 7.00 points and the debt leg 0.30 × 4.5 = 1.35 points, so the WACC is 8.35%. Notice how lopsided the contributions are: equity is 70% of the funding but supplies 84% of the cost, because it is both larger and more expensive. That is worth keeping in mind whenever someone proposes to change the discount rate by renegotiating a loan — the debt leg simply does not have enough weight to move the answer much.

The tax shield is worth exactly 0.45 points

Delete the (1 − T) and recompute: 0.70 × 10.0 + 0.30 × 6.0 = 7.00 + 1.80 = 8.80%. The deduction is therefore worth 8.80 − 8.35 = 0.45 points, and that is not a coincidence you have to trust — it is D/V × Rd × T = 0.30 × 6.0% × 25% = 0.45%. The saving scales with all three: with more debt, a higher borrowing rate or a higher tax rate, the shield grows. Double the debt weight to 60% and the shield doubles to 0.90 points; drop the tax rate to 15% and it shrinks to 0.27.

But the shield is not free money, and the formula hides why. It assumes the company has taxable profits to shield — a loss-making business gets nothing, or at best a deferred benefit. It assumes the deduction is allowed, and many jurisdictions now cap net interest deductibility at a share of taxable earnings. And it assumes Rd stays at 6.0% as you borrow more, which is exactly what does not happen: lenders reprice risk, and beyond some point the equity holders also demand more because their claim has become more volatile. That is why the formula on its own would tell you to fund entirely with debt, and why nobody does.

Re comes from CAPM, and CAPM is made of estimates

Only one of the three CAPM inputs is observable. The risk-free rate is a quoted government bond yield, so reasonable people converge — the argument is only about which maturity and which currency. Beta is a regression coefficient, and its value depends on the window you regress over, the frequency of the returns, the index you regress against and whether the provider applies an adjustment towards one. Two reputable data providers routinely publish betas for the same company that differ by 0.2 or more, and that is not an error in either: they measured different things and both said so in the fine print.

The equity risk premium is worse, because it is not measurable even in principle. It is the extra return investors expect from shares over government bonds in the future, and expectations leave no trace in a database. Estimates are therefore built either from long historical averages, which depend heavily on the start date and the country chosen, or from forward-looking models that back out an implied premium from current prices and forecast dividends, which depend on the forecast. Serious practitioners quote ranges. The range is real disagreement, not sloppiness.

Both estimates hit the answer hard because beta multiplies the premium and the equity leg carries most of the weight. Holding the premium at 5.0%, a beta of 0.80 gives a WACC of 6.95% and a beta of 1.60 gives 9.75% — the same company, the same balance sheet, 2.8 points apart on nothing but a regression choice. Holding beta at 1.20 and moving the premium from 4.0% to 6.0% moves the WACC from 7.51% to 9.19%. Neither input was chosen dishonestly; both were chosen.

What an assumption does to a decision

Discount rates are not published for their own sake; they decide whether projects happen. Take a ten-year project that costs $10 million up front and returns $1.6 million of free cash flow every year. At the base-case WACC of 8.35% the ten-year annuity factor is 6.6054, so the present value of the inflows is $10,569,000 and the net present value is $569,000. The project clears the hurdle and gets built.

Now move only the equity risk premium and leave every other input untouched. At 4.0% the WACC is 7.51% and the net present value is $978,000. At 6.0% the WACC is 9.19% and it is $183,000. Across a range that no reasonable analyst would call extreme, the value of the same project swings by $795,000 — about 8% of the money being committed — purely because of a number nobody can observe. Push the premium to 7.0% and the WACC reaches 10.03%, the net present value turns to −$181,000, and the project is rejected. The project's internal rate of return is 9.61%, so the decision flips at an equity risk premium of about 6.5%.

The effect is far larger in a valuation, because a terminal value divides by a small difference. Value a perpetual $1 million of free cash flow growing at 2% and you get 1 ÷ (WACC − 0.02). At an equity risk premium of 4.0% that is $18.1 million; at 5.0%, $15.7 million; at 6.0%, $13.9 million. A 23.4% swing in the price of a business, from moving one assumption across a range that two competent analysts could hold simultaneously.

Use ranges, not decimals

None of this means WACC is useless. It means it should be reported the way it was produced. Compute the number at three sets of assumptions — pessimistic, central and optimistic — and present the resulting range. Then, instead of asking whether the project clears 8.35%, ask the more useful question: at what discount rate does it stop working? Here that answer is 9.61%, the internal rate of return, and it is a fact about the project rather than a guess about markets. If your central WACC sits comfortably below that threshold, the decision is robust; if it sits within the range, the decision is a judgement call and should be labelled as one.

The second discipline is consistency. A WACC compared against last year's WACC, computed the same way, carries real information about how the cost of funding has moved. The same number compared against a rival's, computed by someone else with a different beta provider and a different premium, carries almost none. Fix your method, write it down, apply it to every project, and treat the absolute level as a convention rather than a discovery. Precision in the method is worth far more than decimals in the output.

Frequently asked questions

Should I use book values or market values for E and D?

Market values, because WACC is an opportunity cost. The book value of equity records what was paid in and retained over the company's history; the market value is what shareholders could get by selling today, and that is what they forgo by leaving the money invested. The gap can be enormous — a company trading at three times book would have its equity weight understated by two thirds if you used the balance sheet. Debt is more forgiving, because loans and bonds usually sit close to face value unless rates have moved sharply or the credit has deteriorated. If your company is unlisted, use a peer-based estimate or a target structure and say clearly in the model which you chose.

Which risk-free rate should I use?

A government bond yield in the same currency as the cash flows, with a maturity close to their horizon. Valuing a ten-year project in dollars means a ten-year Treasury yield; valuing the same project in euros means a euro-area benchmark curve. Mixing currencies is the mistake that does the most damage, because you would import another country's inflation expectations into your discount rate without importing them into your forecast. Some practitioners use a long-run average yield rather than today's, to avoid a hurdle rate that lurches with the bond market; that is defensible as long as you apply the same convention every year and disclose it.

Why do two data providers give different betas for the same company?

Because beta is not a property of the company; it is the output of a regression, and every regression has settings. The estimation window may be two years or five, the returns may be daily, weekly or monthly, the market index may be domestic or global, and many providers shrink the raw estimate towards 1.0 on the grounds that betas historically drift there. Each choice is defensible and each produces a different number. Since a 0.2 difference in beta moves the WACC in our example by 0.7 points, the practical answer is to pick a provider and a convention, state them in the model, and never mix sources within one comparison.

Marginal or effective tax rate in the (1 − T) term?

Marginal, because the shield is created by the tax you would pay on the next unit of profit, and that is what an extra euro of interest actually removes. The effective rate in the accounts mixes in prior-year adjustments, deferred tax movements and foreign rate differences, so it describes history rather than the next decision. Two caveats matter more than the choice itself. If the company has no taxable profit, there is no shield to claim this year at all. And many jurisdictions now cap net interest deductibility as a share of taxable earnings, so a heavily leveraged borrower may be deducting far less than the full 6.0% in our example, and its true WACC is closer to the unshielded 8.80%.

Should every project be discounted at the company WACC?

No, and doing so is one of the most reliable ways to misallocate capital. The company WACC reflects the average risk of everything the company already does. Discounting a much riskier venture at that rate makes it look better than it is, while discounting a safe, regulated, contracted activity at the same rate makes it look worse — so the firm systematically over-invests in risk and under-invests in stability. The correct discount rate belongs to the project, not the balance sheet financing it. In practice you build a divisional rate from the betas of listed companies whose business resembles that division, relever it for the capital structure you intend to use, and apply the company rate only where the project genuinely looks like the company.