Why one company can show several different PERs

Anyone who has checked a company's PER (price to earnings ratio) on two sites has probably seen this. One says 11 times, the other says 14 times. The share price, checked again, is the same. The multiple is not.

The first thought is usually that one side missed an update. That does happen. But a more common cause is something else: the two sites divide different numbers and both call the result PER.

PER is the share price divided by earnings per share. On paper the definition leaves nothing to argue about. But the phrase earnings per share already contains several choices. Only once it is settled which profit, from which period, and the share count as of when, does a single number come out.

This article goes through where those choices split. The conclusion first: seeing a different number does not mean one side is wrong. What needs checking is not which one is right but whether the site states what it divided.

Accounting standards do not leave earnings per share to rough estimation. The international accounting standard on earnings per share fixes the numerator as profit or loss attributable to ordinary shareholders of the parent company. That is not all the profit the company earned but what remains after removing the part that goes to the other shareholders of its subsidiaries.

For a company that does not own its subsidiaries outright, the two numbers diverge. If a subsidiary earned KRW 10 billion and the parent owns 60% of it, the full KRW 10 billion enters the consolidated statements, but KRW 4 billion of it belongs to other shareholders. Which one goes into the denominator changes PER noticeably.

The same standard requires one more thing: basic and diluted earnings per share presented side by side with equal prominence. Diluted earnings per share is recalculated assuming that everything that could later turn into shares, such as convertible bonds and stock options, has done so. Because it assumes more shares, it is usually smaller than basic earnings per share.

The point is that the accounting standard itself requires both numbers. The official figures a company releases on a single day already contain two earnings per share figures. Different PERs across sites are not strange; this is a question that never had a single answer.

Jupick's own situation belongs here too. The profit Jupick uses to calculate PER for Korean stocks is not the profit attributable to owners of the parent, as the standard specifies, but total net income including non-controlling interests, because Jupick takes the total line that appears first on the consolidated income statement. For a company that does not own its subsidiaries outright, if the profit going to non-controlling interests is positive, the denominator is larger by that amount, so Jupick's PER comes out lower than one calculated by the standard. If subsidiaries post losses and that profit turns negative, it comes out higher instead. Jupick is stepping into the very trap just described, and until this calculation is fixed, it seemed right to say so.

Even with the denominator settled, the period remains. There are two main approaches. One uses the annual profit of the most recently completed fiscal year; the other uses the profit of the latest four quarters strung together. The second is usually called the trailing twelve months basis.

The two behave differently. Fiscal year figures are final and audited, so they are stable, but until the next annual report comes out they keep using profit that can be more than a year old. Trailing figures update with each quarterly result and are current, but quarterly figures are not final, can be revised later, and swing widely in seasonal sectors.

Neither is wrong. But the moment values produced by the two methods are set side by side and compared by size, the comparison becomes false. For a company with rising profit, the trailing PER comes out lower than the fiscal year PER, and that gap is a difference of period, not of company.

Jupick is not free of this problem either. PER for Korean stocks is market cap divided by annual net income from the annual report filed on DART, Korea's electronic disclosure system, while PER for overseas stocks is the trailing twelve month value calculated by the data source, taken as is. Two stocks on the same screen can therefore have PERs drawn from different periods. Jupick judged it better to disclose this than to hide it.

Because of this gap, Jupick does not mix Korean and overseas stocks in one table when it publishes metric distributions; it splits them by currency. Different units are one reason, but the bigger reason is that the PERs of the two groups are not measured with the same ruler.

It is not only the denominator that splits. Whether PER is calculated as share price divided by earnings per share or as market cap divided by total net income, the two should be mathematically equal. In practice they sometimes are not.

The Korean market has preferred shares. Preferred shares are listed separately from common shares under a different ticker code and trade at a different price. Measuring the value of the whole company requires adding the market caps of the common and preferred shares, but calculating from one ticker code captures only the common shares' market cap. Net income, on the other hand, is a whole-company number. The multiple changes depending on which side is matched to which.

Treasury shares work the same way. Earnings per share changes depending on whether shares the company has repurchased and holds are subtracted from shares issued. If a company holds a large amount of treasury stock, the difference is not small enough to ignore.

There is also the timing of the price. A site using real-time intraday prices and one using the previous day's close produce different values. The multiple moves even within a day. That is why matching two sites to the decimal place is usually pointless.

Jupick's Korean PER divides the market cap compiled by the Korea Exchange (KRX) by net income from the annual report. If net income is confirmed at zero or below in the run that recalculates this value, the PER is removed. In the investment metrics at the top of the stock page, the whole row disappears; in the sector comparison table further down, only a hyphen remains. The comparison table shows the sector median in the next column, so removing the row would remove the point of comparison too. A loss-making company's PER comes out negative, and that negative number means neither priced low nor priced high, so leaving it as a number would point readers in the wrong direction.

Recalculation runs on a fixed schedule, however. For a company that has just swung from profit to loss, the previously calculated value can stay on screen until the next run. To be sure which number is still valid, there is ultimately no substitute for opening the company's latest disclosure.

Beyond individual stocks there is another layer: numbers that combine many companies, such as a sector PER or a market PER. Most people assume these are averages of the member companies' PERs. In practice they are often not calculated that way.

The corporate finance datasets that Professor Aswath Damodaran of New York University's Stern School of Business publishes free every year state this clearly in their variable definitions. A sector's PER is not a simple average of individual company PERs but the sector's total market cap divided by its total net income. The same documentation says enterprise value multiples are also computed by aggregation in this way.

Working through the arithmetic shows why. A company with very small profit has a PER that spikes into the hundreds. Mix a few such values in and a simple average gets dragged toward them, ending up as a number that resembles no company in the sector. Aggregation has a different property, giving larger companies more weight, but at least it is not shaken wholesale by a few extreme values.

What matters is that the two methods go by the same name. A sector PER of 12 on one site and 18 on another can be the results of different calculations, and the screen rarely says so. For reference, the dataset just cited is updated only once a year and uses trailing twelve month data through the third quarter of the prior year. A table opened this year may in fact describe the world as of the end of the year before last.

This is why Jupick uses the quartiles of the actual distribution instead of a sector or market average. A single average cannot be checked once its calculation method is hidden, but quartiles show exactly how many stocks were lined up and in what order. That is why Jupick always shows the sample size alongside them.

In short, PER splits in at least five places: whether it counts only profit attributable to controlling shareholders or all of it, basic or diluted, fiscal year or trailing twelve months, how preferred and treasury shares are handled, and whether a group figure is an average or an aggregate. Whatever combination is chosen, calling the result PER is not wrong.

So different numbers on two sites are no reason to brand one of them fake. What actually needs checking is whether the site has written down its definition. A number without a stated definition gives no basis for judging whether it is large or small. It cannot be compared with values elsewhere, and even comparing it with yesterday's value is risky, because the definition may have changed quietly.

In practice one habit is enough: compare PERs only within the same site. Comparing two companies on one site at least sets values calculated under the same rules side by side. The moment numbers gathered from different sites are lined up in one row, that row shows differences in calculation method, not differences between companies.

It is also worth being clear about what this article does not say. It does not cover what PER counts as low or high. That judgment needs an understanding of the sector's structure, the nature of the profit, and whether that profit can continue, and a single multiple cannot do that work. The same reason keeps PER out of Jupick's grades. The grade looks at a company's financial condition and does not judge whether the share price is appropriate.

This content is for information only and is not investment advice. Jupick gives no trading advice on any stock and does not forecast future prices or earnings, and a high or low PER is not advice of any kind. Investment decisions and responsibility for them rest with the user.