Most analysts can recite the Hamada equation. Far fewer can explain why the beta they just plugged into their WACC is wrong — and exactly where the error entered. This guide walks through the full construction of a levered beta: peer selection, unlevering, re-levering, and the judgment calls that determine whether your cost of equity is defensible.
What levered beta actually measures — and why it differs from raw regression output
Levered beta measures the systematic risk of an equity investment, incorporating both the business risk of the underlying operations and the financial risk added by the company’s capital structure. A company with no debt carries only operational risk. Every dollar of debt added increases the volatility of returns to equity holders — and therefore increases beta.
The problem with using a regression beta directly — running a 60-month OLS regression of stock returns against an index — is that you’re capturing the company’s historical capital structure, not the target structure you’re modeling. For M&A work, LBO analysis, or any valuation where capital structure changes at or after close, a regression beta is not the right input. You need to build it from scratch.
Step 1: Select your peer group — this is where most models go wrong
The peer group is the most consequential decision in the entire process. It determines the quality of your asset beta estimate before any formula is applied.
What we see consistently: analysts screen by GICS sub-industry code and stop there. The result is a peer set that includes companies with materially different operating leverage, geographic exposure, or revenue mix — and an unlevered beta that reflects none of the target’s actual risk profile.
A defensible peer group requires three filters, applied in sequence:
Business model alignment: same revenue drivers, same cost structure, not just the same sector label. A SaaS company and a legacy software reseller share a GICS code — they do not share a risk profile.
Size and liquidity: Micro-cap beta estimates are noisy. If you include peers with thin trading volume, their beta observations carry statistical error that your Hamada calculation will not correct for.
Capital structure range: Exclude outliers — peers with negative equity, distressed debt loads, or net cash positions that make their leverage ratios uninformative. You need a usable D/E distribution.
Aim for five to ten peers. Fewer than five gives you an average that a counterparty can challenge on sample size. More than fifteen usually means the screening criteria have become too loose.
Step 2: Collect regression betas and strip out leverage — the Hamada unlever
Once the peer group is confirmed, collect levered betas for each peer. Use two-year weekly returns against the relevant market index for the majority of peers, unless a longer window is needed to capture a full cycle. For peers that have undergone significant capital structure changes in the period, use a shorter window or adjust manually.
The Hamada equation to unlever is:
βU = βL ÷ [1 + (1 − t) × (D/E)]
Where βU is unlevered (asset) beta, βL is the observed levered beta, t is the marginal tax rate, and D/E is the net debt-to-equity ratio at market values.
Three precision points that matter in practice:
Use market value of equity, not book. Book equity in a leveraged company is often suppressed or negative — it produces D/E ratios that are arithmetically unusable.
Use net debt, not gross debt, where cash balances are material. A company sitting on a cash position equal to 40% of its debt is not carrying the leverage its balance sheet implies.
Apply the target company’s marginal tax rate, not the peer’s effective rate. If the peer operates in a jurisdiction with a 15% effective rate but the target is a German entity at 30%, the tax shield adjustment will be materially different.
A common failure point: peers that are currently unprofitable have no functioning tax shield. Applying a positive tax rate to an EBIT-negative company overstates the unlevering adjustment. For those peers, either use a zero tax rate or exclude them from the beta average and note the exclusion.
Step 3: Derive the asset beta — median or mean, and why it matters
Once you have unlevered betas for each peer, you need a central tendency estimate. Whether you use the median or mean depends on the distribution:
If the peer group is tight and no outliers remain after screening, the mean is defensible and picks up the full distribution.
If one or two peers sit significantly above or below the cluster, use the median. The median is more robust to outliers and is harder to challenge in a due diligence setting because it requires no judgment about which peers to exclude — the math handles it.
Document your choice. An advisor or counterparty reviewing your model will ask why you used the median. „Because the distribution was skewed by two outlier observations” is a defensible answer. Silence is not.
Step 4: Re-lever to the target capital structure
With an asset beta in hand, apply the Hamada equation in reverse to re-lever at the target’s capital structure:
βL = βU × [1 + (1 − t) × (D/E)]
The target D/E ratio here is the structure you’re modeling at the valuation date — not the current observed structure if that’s expected to change. In an LBO, for example, the capital structure at entry is materially different from the structure at exit. If you’re building a full LBO model, you may need to re-lever the beta at each exit year to produce a credible cost of equity across the holding period.
Use the target’s own marginal tax rate for the relevering step. This is the rate that determines the value of the tax shield on the incremental debt — which is the economic argument that justifies the adjustment in the first place.
Step 5: Sense-check before it enters your WACC
But here’s what separates a model built to survive scrutiny from one that doesn’t: the output check.
A re-levered beta below 0.5 for an operationally leveraged industrial company is a flag. A re-levered beta above 2.5 for a stable recurring-revenue business is equally suspicious. Neither is automatically wrong — but both require explanation.
Run a cross-check: take your re-levered beta and back-calculate an implied cost of equity using CAPM. Does that cost of equity make intuitive sense against the company’s observed ROIC, against the returns a rational investor would require given the risk profile? If the numbers feel disconnected, the issue is almost always in the peer group or the D/E inputs — not the formula.
The formula is the easy part. The judgment is the work.
Frequently asked questions
What is the difference between levered and unlevered beta? Unlevered beta reflects only the operating risk of a business, stripped of capital structure effects. Levered beta adds the financial risk from debt, making it the correct input for a cost of equity calculation in CAPM.
Why can’t I just use the regression beta directly in my WACC? A regression beta reflects the company’s historical capital structure. If the target structure differs — due to a transaction, refinancing, or different peer leverage — the regression beta produces a cost of equity that does not match the model’s assumptions.
What tax rate should I use in the Hamada equation? Use the marginal statutory tax rate of the entity generating the interest deduction — not the effective rate. The effective rate reflects historical blended outcomes; the marginal rate reflects the value of the incremental tax shield on new debt.
How many peers do I need to build a reliable unlevered beta? Five to ten peers with genuine business model comparability. Fewer than five creates a sample size problem. More than fifteen usually signals that screening criteria have become too broad to produce a meaningful central tendency.
How do I handle peers with negative EBIT in the Hamada unlevering? Apply a zero tax rate for those peers or exclude them and document the exclusion. A tax shield cannot be valued for an entity that generates no taxable income — applying a positive rate overstates the unlevering adjustment.
If you’re building a valuation where the beta estimate will face external scrutiny — a financing process, a board presentation, a fairness opinion — the methodology above is the foundation. For a structured walkthrough of how this integrates into a full WACC build and DCF architecture, the modeling courses at financial-modeling.com cover the complete process in a transaction context.