Is a PER of 12 high or low: reading a metric by lining up the market

Say a company has a PER (price to earnings ratio) of 12. Is that high or low? The question cannot be answered. The number 12 alone gives no way of knowing whether it is a common value or a rare one.

Hear that someone is 175 centimeters tall and the sense of it comes right away. Everyone has seen countless people over a lifetime, so they know in their gut roughly where 175 falls. A PER of 12 comes with no such sense. Without ever having seen what to compare it with, there is no way to place it.

So the first step in reading a metric is not memorizing the formula but securing a baseline. Only after knowing roughly what value the metric usually takes in this market does the number in front of you gain meaning.

The problem is where to get that baseline. Many articles present particular numbers as baselines: a PER of around 10, a PBR of 1, an ROE of 15%. These numbers are not entirely wrong, but they come with no source and are never updated as time passes.

The easiest way to get a baseline is to average all stocks. But for financial metrics this does not work well, because the average gets dragged wholesale by a few unusually large values.

Take a concrete example. When a company's profit shrinks to nearly zero, its PER, market cap divided by that tiny profit, reaches the hundreds and sometimes passes 1,000. That is not because the company is especially highly priced, but because the denominator of the division has approached zero.

Mix just five such stocks (1.25%) into the major stocks and the average lifts sharply. Even if all the rest cluster around 10, the average comes out at 30 or 40. Use that average as a baseline and the odd conclusion follows that most stocks are priced lower than average.

The debt ratio has the same problem. A company with almost no equity left shows a debt ratio in the thousands of percent. Revenue growth spikes from the other end as well: if a company had very small sales last year, even a modest increase this year makes growth several hundred percent. Financial metrics are built from division, and division runs wild as the denominator shrinks.

The alternative to the average is simple. Line up all stocks in one row by the size of the metric, then read the value of the stock standing at a given position.

The value of the stock standing in the middle is called the median. If the major stocks are lined up, it is the value of the stock with half in front and half behind. The strength of the median is that extreme values do not move it. Whether or not there is a stock with a PER of 1,000 at the far end, the stock in the middle stays the same, because the median looks only at the order of the row, not at how far it stretches.

The middle value alone is not enough. So the points a quarter and three quarters of the way along the row are read as well: the point 25% from the low end and the point 25% from the high end. Half of all stocks fall between these two values.

Once these three values are known, the number in front of you can be placed immediately. If a PER of 12 is below the point 25% from the low end, it is among the lowest in this market; if it is above the point 25% from the high end, it is among the highest. In between, it is a common value.

Even with the same median, metrics can behave very differently. Two metrics can share a median while in one most stocks cluster tightly around it and in the other they spread widely.

In a tight metric, straying a little from the median already makes a value rare. In a spread-out metric, a value can stray quite far and still be common. That is why the gap between the quarter point and the three quarter point, the spread, matters too.

The smallest and largest values are useful as reference as well, but they are of a different nature. The median and quartiles are numbers that many stocks make together, while the maximum is a number made by a single stock alone. If that one stock has an unusual accounting situation, the maximum shifts wholesale. It is safest used only to confirm that the spread is this wide.

Jupick publishes distributions for Korean and overseas stocks separately. Lining up the two markets in one row would make neither readable.

For market cap the reason is obvious: the currencies differ. Lining up values in KRW and in dollars in one row is like ranking heights while mixing centimeters and inches.

Ratio metrics share units and could be lined up in one row, but splitting them is still better. The two markets differ in accounting standards, in the mix of listed sectors, and in dividend practices. Mix dividend yields from a market with a long-established dividend culture and one without, and the two peaks blur into one, so the position of a stock from either side is misread.

The reference period of the financial data also differs. Korean stocks are based on annual reports and overseas stocks on the latest quarters. Even a metric with the same name measures a different period, so this difference is worth knowing before setting them side by side.

The thing to watch most closely when building a distribution is fill-in values mixed into the data. This is Jupick's internal matter, but it applies equally to the statistics of any service, so it is worth stating.

Jupick's grade calculator requires a number for every one of its nine metrics. But sometimes the source data does not have the value, either because there is no disclosure at all or because the data source did not respond that day. Halting the calculator at that point would leave the stock with no grade at all. So blanks are filled with a neutral fill-in value. It is a value that converts to exactly the middle score, a device that keeps the stock from being either rewarded or penalized.

The problem is that these values look exactly like measured values on screen. Counting them as is to build a distribution creates a peak that was never there, at the position of the filled-in values, and that peak looks like a market statistic. At worst, the median itself becomes a number Jupick made up.

So Jupick records, stock by stock, which metrics hold filled-in values and removes them from the sample when counting the distribution. That is also why the metric page notes, below the distribution table, how many were removed. A statistic that does not state the extent of its sample cannot be checked.

A distribution table answers only one question: is this number a common value or a rare one in this market right now? That is useful enough, but it is no more than that.

Being a rare value means neither good nor bad. A stock at the low end of PER may be there because the market is worried that its earnings are going to shrink. A case where a low PER is a warning rather than a sign that the stock is priced below its worth is called a value trap.

There is a clear limit too: the comparison crosses sectors. A distribution lines up the whole market in one row, which erases structural differences between sectors. A high debt ratio in a sector with heavy capital spending is not strange, and a high ROE in a sector that needs little capital is natural. After checking position in the whole market, looking again within the same sector gives a more accurate view.

It is also important not to draw a conclusion from one metric. Profitability that looks good can be profit made with debt, and a dividend that looks generous can be paid out in excess of earnings. The distribution table exists to put several metrics each in their place, not to make any one metric the decisive basis.

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, and a position anywhere in the distribution is not advice of any kind. Investment decisions and responsibility for them rest with the user.