The two measures
Greenblatt's formula uses operating profit, EBIT, in both measures, so that companies with different debt levels and tax rates are compared on the same footing.
Earnings yield = EBIT / enterprise value. This is the price measure. It asks how much operating profit you get for the whole business, including its debt and net of its cash. It is the inverse of EV/EBIT. Our explainer on earnings yield covers the variants, and the one on enterprise value covers how to build the denominator.
Return on capital = EBIT / (net working capital + net fixed assets). This is the quality measure. It asks how much operating profit the business earns on the tangible capital it needs to run. Greenblatt uses tangible capital rather than total assets so that goodwill from past acquisitions does not distort the comparison. It is close to the ROIC measures many analysts use, but it is pre-tax and excludes goodwill.
| Measure | Formula | What a high value says | What can fool it |
|---|---|---|---|
| Earnings yield | EBIT / enterprise value | Operating profit is large relative to the price of the whole business | Peak-cycle EBIT, one-off gains, missing liabilities in EV |
| Return on capital | EBIT / (net working capital + net fixed assets) | The business earns a lot on the tangible capital it uses | Old, written-down assets; negative working capital; a single strong year |
How the ranking works
The mechanics are short:
- Set the universe. Greenblatt's version uses a minimum market capitalization and excludes utilities and financial companies, whose balance sheets make both measures unreliable. Foreign companies listed in the US are often excluded too.
- Rank on earnings yield. The highest earnings yield gets rank 1.
- Rank on return on capital. The highest return gets rank 1.
- Add the ranks. A company ranked 10th on yield and 40th on return scores 50.
- Sort. The lowest combined scores form the list.
- Build the portfolio over time. The book describes adding a few positions every couple of months until the portfolio holds 20 to 30 names, then replacing each one after about a year.
The ranking is relative. A company does not need to clear a threshold; it needs to beat the rest of the universe on both measures at once. That is why the list changes when the market as a whole gets cheaper or more expensive.
Worked example
| Company | EBIT | Enterprise value | Earnings yield | Tangible capital | Return on capital |
|---|---|---|---|---|---|
| Company A | 120 | 800 | 15.0% | 400 | 30.0% |
| Company B | 90 | 450 | 20.0% | 900 | 10.0% |
| Company C | 60 | 1,200 | 5.0% | 100 | 60.0% |
Company A ranks 2nd on yield and 2nd on return, a combined score of 4. Company B ranks 1st on yield but 3rd on return, a score of 4 as well. Company C ranks 3rd on yield and 1st on return, also 4. In a three-company universe, the tie shows the trade-off the formula makes: it rewards balance across the two measures rather than extremes on one. In a universe of 3,000 companies, a name that is merely good on both measures will often beat one that is outstanding on one and poor on the other.
Why the idea makes sense
The logic is the one most value investors would accept. A business that earns a high return on the capital it uses can usually grow without much outside funding. Buying that business at a high earnings yield means paying a low price for that profit. If the market has mispriced it, the price should move toward value over time.
Greenblatt's further point was behavioral. A screen like this produces a list that often looks uncomfortable: companies with bad recent news, out-of-favor industries, or businesses no one is talking about. Sticking with an uncomfortable list through periods of underperformance is hard, and that difficulty is part of why simple screens can keep working at all.
Where this goes wrong
One year of earnings. Both measures use a single year of EBIT. A cyclical company at the top of its cycle looks cheap and high-return just before profits fall. Using normalized earnings over a cycle is the standard correction.
One-off items. A gain on an asset sale or a reversal of a provision inflates EBIT. Reading the income statement and footnotes catches most of these.
Declining businesses. A company in structural decline can show high returns on old, depreciated assets while its revenue shrinks. This is the classic value trap.
Enterprise value gaps. Leases, pensions, minority interests and other claims are often left out of a quick enterprise value. A company can look cheap because the screen missed a liability. Our explainer on off-balance-sheet liabilities lists where to look.
Negative working capital. Some businesses fund themselves with customer prepayments, which makes tangible capital very small and return on capital very large. That can be a real advantage, but it can also make the rank meaningless.
Taxes and costs. A strategy that rebalances every position yearly creates short-term gains in a taxable account and trading costs in thinly traded names. Any backtest you read should say how it handled both.
Magic Formula versus other value screens
The Magic Formula is one of several published screens that turn value investing into rules. They differ in what they treat as cheap and whether they add a quality test.
| Screen | Cheapness measure | Quality measure | Typical result |
|---|---|---|---|
| Magic Formula | EBIT / enterprise value | EBIT / tangible capital | Profitable, capital-light companies at modest multiples |
| Acquirer's Multiple | Enterprise value / operating earnings | None | Deeply cheap companies, often out of favor or cash-rich |
| Net-net (NCAV) | Price below current assets less all liabilities | None | Small, often troubled companies |
| Graham Number | Price below a figure built from EPS and book value | None directly | Companies cheap on both earnings and book |
| Piotroski F-score | Usually applied to low price-to-book stocks | Nine accounting tests | Cheap companies with improving fundamentals |
The choice of screen says something about the investor's view. The Magic Formula assumes high returns on capital tend to persist long enough to be worth paying for. A pure cheapness screen assumes they tend to fade. Net-net and Graham screens lean on the balance sheet rather than the income statement, which helps when earnings are depressed but hurts when assets are overstated. For the asset-based approach, see our explainers on net current asset value and the Graham Number.
Running two screens and comparing the overlap is a cheap way to see which names are cheap for more than one reason. A company that ranks well on the Magic Formula but poorly on F-score, for example, may have strong recent returns and weakening cash flow, which is a question worth answering before going further.
How to use the screen well
Treat the output as a shortlist. A practical routine:
- Check the cycle. Compare this year's operating margin with the last five to ten years.
- Check quality another way. A Piotroski F-score adds balance sheet and cash flow checks the formula lacks.
- Rebuild enterprise value. Add debt-like items the data feed may have missed.
- Read the reason it is cheap. Recent filings and earnings calls usually explain why the market is doubtful, and whether that doubt is temporary.
- Compare with a pure value screen. The Acquirer's Multiple ranks on cheapness alone; names that appear on both lists are cheap without relying on the quality rank.
You can run the screen on Greenblatt's own site, in many of the best stock screeners for value investors, or in a spreadsheet. For the wider set of research tools that help with the follow-up work, see our guide to the best stock research tools for individual investors.
The bottom line
The Magic Formula ranks companies on earnings yield and return on capital and holds those with the best combined rank. It is a disciplined way to find businesses that look both cheap and good on last year's numbers. Its weaknesses come from that same simplicity: one year of data, no view of the cycle and no explanation of why a stock is cheap. Used as a first filter followed by real analysis, it does a useful job.