Earnings Per Share (EPS)

What is Earnings Per Share (EPS)?

EPS is the last line of the income statement, and it answers a narrow but important question: after every cost, tax, and interest payment has been settled, how much profit is attributable to each share you own? Net income by itself tells you the size of the profit pool. EPS tells you the slice attached to your claim on the business. A company earning $2 billion sounds impressive until you learn it has four billion shares outstanding, at which point each share represents fifty cents of annual profit.

Two versions appear in every filing. Basic EPS divides profit by the plain weighted average share count for the period. Diluted EPS assumes every stock option, restricted share unit, warrant, and convertible security that could reasonably turn into common stock has already done so, which pushes the share count up and the per-share figure down. Diluted is the conservative number and the one most analysts default to, particularly for technology companies where stock-based compensation creates a steady drip of new shares. The share count is a weighted average rather than a year-end snapshot, because a company that issued shares in November should not have those shares counted against profits earned in March.

One more distinction matters. Profit attributable to common shareholders is net income after preferred dividends have been paid, since preferred holders get their cut first. Companies also publish "adjusted" or "non-GAAP" EPS alongside the reported figure, stripping out items management considers one-off: restructuring charges, acquisition costs, impairments, sometimes stock compensation. Adjusted EPS can be genuinely more useful for judging the underlying business, and it can also be a place where recurring costs quietly get labeled as exceptional year after year. Screening data conventionally uses reported figures on a trailing twelve months (TTM) basis, summing the last four reported quarters so the number stays current between annual reports. For companies that report in another currency, including foreign businesses with US listings and ADRs, values are converted to USD so comparisons hold across the universe.

Formula

The numerator is profit belonging to common shareholders, which means net income with preferred dividends removed. Most companies in a screening universe have no preferred stock, so the adjustment often changes nothing, but it matters for banks and utilities where preferred issuance is routine. The denominator is the weighted average share count over the reporting period, adjusted for the dilutive effect of options and convertibles. Both halves are measured over the same window, and on a trailing twelve months basis that window is the four most recent reported quarters.

Share counts get retroactively adjusted for stock splits so that historical EPS remains comparable. A two-for-one split doubles the share count and halves EPS, and every prior period is restated to match, which is why a company's EPS history does not show a sudden cliff on the split date. Backtests need to respect the timing of when each figure became public rather than the period it describes, since a quarter that ended in December is not knowable until the filing lands weeks later.

How to Interpret EPS

Comparing raw EPS across two companies is close to meaningless, and this is the single most common mistake made with the metric. Share counts are an accident of corporate history. A company that never split its stock can post EPS of $40 while an equally profitable peer that split repeatedly posts $2. Neither number says anything about which business is better or which stock is cheaper. EPS only becomes informative when you put it next to something: the share price, giving you the P/E ratio or its inverse, the earnings yield; or the company's own EPS in prior periods, giving you an earnings growth rate.

The trend is where most of the signal lives. Steadily rising EPS over many years is the signature of a business that both grows profits and refrains from diluting its owners. Flat EPS against rising net income is a warning that shareholders are being diluted as fast as the business is growing, which is common among serial acquirers and companies that pay heavily in stock. The reverse also happens: EPS can climb while net income is flat because aggressive buybacks are shrinking the denominator. Neither is inherently bad, but a shareholder should know which engine is doing the work.

Negative EPS means the company lost money over the period, which is normal for early-stage biotech and unprofitable growth companies and a red flag almost everywhere else. Cyclical companies swing between strong and negative EPS across a cycle, so a single trailing figure can badly misrepresent normal earning power. And because EPS is an accrual accounting output, it can drift away from the cash a business actually collects. Checking EPS against operating cash flow per share is a quick test of earnings quality.

Why EPS Matters for Investors

EPS is the number the market trades on quarter to quarter. Analyst estimates are published in EPS terms, earnings surprises are measured in cents of EPS against consensus, and a stock's short-term reaction to results often tracks the gap between reported and expected EPS more closely than anything else in the release. Longer term, EPS growth is one of the two engines of equity returns, alongside changes in the multiple investors are willing to pay. A business that compounds EPS at ten percent a year for a decade will do most of the work for its shareholders even if the P/E ratio never budges. For income investors, EPS also sets the ceiling on what a dividend can sustainably be, which is exactly what the payout ratio measures.

Using EPS in Stock Screening

The most common use is as a quality gate rather than a ranking factor. Requiring positive EPS over the trailing twelve months removes loss-making companies before any other filter runs, which sharpens value screens considerably, since a low price-to-book on an unprofitable company is often a value trap rather than a bargain. Benjamin Graham's defensive investor criteria went further and demanded positive earnings in each of the previous ten years, a demanding filter that survives today in quality screens.

A practical starting screen: EPS (TTM) greater than zero, EPS growth over the last three years above ten percent annually, and a P/E ratio below the sector median. That combination looks for companies with real profits, a track record of growing them, and a price that has not already accounted for the growth. Peter Lynch's approach was a variant of this, pairing EPS growth against the multiple through the PEG ratio, on the reasoning that a company growing earnings at twenty percent could justify a P/E of twenty while a company growing at five percent could not. Screening on EPS growth alone is riskier, because a company recovering from a depressed base can post enormous percentage growth from an almost meaningless starting point.

Backtesting with EPS

Earnings-based screens have a long research record, and the broad finding is that buying profitable companies at reasonable earnings multiples has beaten the market over long horizons, with painful stretches of underperformance when growth stocks are in favor. Testing any EPS strategy honestly depends on two things. The first is point-in-time data. A December quarter is not public until the filing appears, often six to eight weeks later, so a backtest that assigns December's EPS to a January rebalance is trading on information nobody had. SledgeKey ties every observation to the date it became publicly available, so a historical screen reflects only what an investor could genuinely have known.

The second is survivorship. Companies with collapsing EPS tend to get acquired, delisted, or go bankrupt, and a universe that quietly drops them makes every earnings screen look better than it was. Including delisted companies in the historical universe is what keeps the result honest. Restatements are the subtler trap: when a company revises prior EPS, the revised figure is the one sitting in most databases today, even though the original was what the market saw. Any earnings-based backtest should be read with that in mind.

Related Metrics