Corporate Action Adjustment: Making Historic Data Comparable

Corporate action adjustment restates historic per-share figures onto today's share count, so a series can be compared with itself. Splits and bonus issues are restated in full, because nothing of value moved. A rights issue takes a factor, because value moved but a participating holder was left whole. Buybacks and placements are not restated at all, because real money left or entered the business.

The whole subject rests on one line established where corporate events were first defined: a per-share figure is one total spread across one count. Change the count and the measurement has changed, even though the total is untouched. One sentence runs through three kinds of action and one price series, and it settles which historic numbers may be set next to each other.

Why can two years of the same company not simply be compared?

Start with a kitchen. One large pot of biryani is cooked for a wedding and the cost works out at Rs 240/- a plate. Next month the identical pot is cooked, same rice, same meat, same gas, but it is served in smaller plates and yields twice as many of them. The cost is now Rs 120/- a plate. Nothing about the cooking has changed. Somebody who looks at those two numbers and reports that the catering costs halved has not measured the kitchen. They have measured the crockery.

Two per-share figures computed on two different share counts are not slightly inaccurate when compared; they are not a comparison at all. They are two different measurements of two different things wearing the same label. And the size of the distortion has nothing to do with the business: it is set entirely by the size of the action. A one into two split misstates a growth rate one way, a one into five split misstates it five times as hard, and the underlying company was identical in both cases.

Sarvani Coatings Limited, an invented issuer, gives a clean instance. Its published history carries two completed actions, and both landed in year one. First a stock split, face valueThe nominal amount printed on a share when it was first issued. It is an accounting label, not a market value, and a split changes it without changing what the holder holds. reduced from Rs 10/- to Rs 2/-, one share becoming five, so the count went from 2.40 crore shares to 12.00 crore. Then, at the end of the same year, a one for one bonus issue took the count from 12.00 crore to 24.00 crore. Neither action put a rupee into the business or took one out. A holder with a hundred shares on the morning of the split had five hundred by the afternoon and a thousand by the end of the year, and held exactly the same share of exactly the same company throughout.

How wrong does an unadjusted series actually get?

Now the numbers side by side. A rule that is merely stated is a rule soon forgotten, and a growth rate watched changing sign is a rule that stays.

Year one delivered profit after tax of Rs 143 crore. Year three delivered Rs 278 crore. Both audited, both published, neither in dispute. Year one earnings per share stood at Rs 59.58/- on the 2.40 crore shares then in issue. Take that Rs 59.58/- and set it beside year three's Rs 11.58/-. The unadjusted pair says earnings per share fell 80.6 per cent over the two years.

Now restate. The split multiplied the count by five and the bonus multiplied it by two, so the cumulative factor is ten. Year one becomes Rs 59.58/- divided by ten, or Rs 5.96/-. Against the same Rs 11.58/-, that is a rise of 94.4 per cent over two years. Spread evenly, the rise works out at 39.4 per cent a year as a compound annual rateA growth rate expressed per year across several years, worked so that applying it each year in turn reproduces the total move. Annual rates multiply across years; they never add.. Same two audited profits, same two years, same company. The reading moves 175.0 percentage points and changes sign.

The middle rung is the most dangerous one on the table, and it is worth a second look. Restating for the split alone and stopping there produces minus 2.8 per cent over two years: a flat, unremarkable, thoroughly believable number that no reviewer would query. The fully wrong answer announces itself. The half-corrected one does not. Only the bottom row, Rs 5.96/-, may stand next to Rs 8.21/- in year two and Rs 11.58/- in year three.

One more thing is worth noticing on that bottom row, and it is the tidiest fact in the whole business. Once year one and year three are on the same count, per-share growth and profit growth are the same number. Dividing both ends by the identical 24.00 crore cannot change a ratio, so Rs 143 crore to Rs 278 crore is a rise of 94.4 per cent and Rs 5.96/- to Rs 11.58/- is a rise of 94.4 per cent. When those two figures disagree, the count moved, and that is the signal to go looking.

Which question decides how an action is treated?

No list has to be memorised. There is one question, asked in two halves, and the three treatments fall out of the answer.

Did value move? And if it did, was a holder who participated left exactly where they started? The whole three-way rule is that single question applied. One idea, not a table to learn. A split and a bonus issue move no value at all, so history is restated in full. A rights issue moves value, but a holder who takes up the entitlementThe right a rights issue hands each existing holder to subscribe for a stated number of new shares at a stated price. It can be taken up, allowed to lapse, or in many markets sold on. ends up neither better nor worse off, so history takes a factor rather than a rewrite. A buyback or a placement moves real money in or out of the business, so history is left exactly as it was filed.

What does restating in full actually involve?

Divide, and divide by the right thing. Every historic per-share figure is divided by the cumulative factor for all the actions that have happened since. Factors multiply across actions and never add, and treating them additively is the quiet, common, entirely invisible error.

Sarvani Coatings had two actions. The split multiplied the count by five. The bonus multiplied it by two. Five times two is ten, and ten is exactly the ratio of 24.00 crore shares to 2.40 crore. The match is a useful check: the cumulative factor is always just the new count divided by the old one. Adding them instead gives seven. Rs 59.58/- divided by seven makes year one Rs 8.51/-, too high by 42.9 per cent. Every ratio built on it, every growth rate, every multiple, every margin per share, is then wrong by that same 42.9 per cent, and none of it looks wrong.

When does a factor do the job instead?

A rights issue is the awkward middle case, and the awkwardness is honest rather than technical. New shares are issued at a price below the market price, so value genuinely moves from the shares that existed to the shares being created. But it moves between the holder's left pocket and their right pocket, provided they take up what they were offered. On Sarvani Coatings' hypothetical one for twenty at Rs 350/- against the illustrative Rs 486/-, a holder loses Rs 6.48/- on each of twenty existing shares and gains Rs 129.52/- on the one new share. The loss totals Rs 129.60/- against a gain of Rs 129.52/-, and the difference is rounding and nothing else.

Because the holder is left whole, history is not rewritten; it is scaled. The theoretical ex-rights priceThe price a share would settle at immediately after a rights issue if nothing else in the world changed. It is the weighted blend of the old shares at the old price and the new shares at the offer price. takes the twenty existing shares priced at Rs 486/- together with the one new share at Rs 350/-, and spreads the lot across twenty one shares. The blend comes to Rs 479.52/-. The factor is Rs 479.52/- over Rs 486/-, or 0.9867, and historic per-share figures are multiplied by it. Year one's restated Rs 5.96/- becomes about Rs 5.88/-.

The rights factor is always less than one, it is usually tiny, and that combination is exactly what makes it dangerous. A single application costs 1.3 per cent. Nobody audits a 1.3 per cent discrepancy; it looks like rounding, or like a slightly different tax charge, or like nothing. But the factor compounds. Ten applications of the same 0.9867 across a long series move it 12.6 per cent away from where it started, and a 12.6 per cent error in a ten year growth series is not rounding at all.

Why is a buyback never restated?

A buyback is the one people get wrong, and they get it wrong for a reasonable-sounding reason: the count changed, so surely the history has to move with it. The history does not move, and the test is the same one that has applied all along.

Under the hypothetical buyback, Sarvani Coatings returns Rs 240 crore by purchasing 0.40 crore shares at Rs 600/-, and the count falls from 24.00 crore to 23.60 crore. The Rs 240 crore left the building. Cash and investments fall away, Rs 312 crore down to Rs 72 crore, and net debtTotal borrowings less cash and liquid investments. Where it comes out negative the business is holding more cash than borrowing, and that state is called net cash. leaves behind net cash of minus Rs 72 crore and settles at plus Rs 168 crore. Year one, year two and year three were all earned by a business that genuinely held that cash. Restating those years onto 23.60 crore shares would not be rebasing history; it would be rewriting it. The bought-back shares actually existed in those years, and that cash actually sat on the balance sheet.

The contrast with the bonus issue is the whole lesson. Both actions changed the count. One of them moved Rs 240 crore and one of them moved nothing at all, and that is the only difference that matters. The buyback's arithmetic also shows why the temptation exists: on unchanged profit of Rs 278 crore, earnings per share moves up from Rs 11.58/- to Rs 11.78/-, a lift of 1.7 per cent that is pure division and no business improvement whatsoever. Meanwhile book value per shareWhat the balance sheet ascribes to each share, being net worth spread across the count in issue. An accounting measure, never a market one. falls from Rs 61.92/- to Rs 52.80/-, a drop of 14.7 per cent. The cash that left was carried in net worth. Restate one of those and not the other, and the series contradicts itself.

Is an adjusted price a price?

The price series needs separate treatment. The arithmetic is identical and the object is not. Earnings per share is a computed statistic; a price is a record that somebody, on a particular morning, paid a particular amount. Adjust it and it stops being that.

Sarvani Coatings' illustrative quoted price is Rs 486/-, and it sits on 24.00 crore shares. The same claim on the business, expressed on the 2.40 crore share basis that applied before the split, would be quoted at Rs 4,860/-. The Rs 4,860/- is Rs 486/- multiplied by the cumulative factor of ten. Run the arithmetic in the other direction and any price actually quoted before the two actions is divided by that same ten to join the current series. An adjusted price history does exactly that to every point in it.

Nobody has ever paid Rs 4,860/- for a share in Sarvani Coatings, and a constructed figure of that kind must never be quoted as a historic price. An adjusted price series is a constructed object built for comparison. The adjusted series is right for plotting a chart, right for computing a return, and wrong for any sentence beginning with the words "the share traded at". A data providerA vendor that collects exchange and filing data and redistributes it in a processed form, usually with its own adjustments already applied. How each one computes those adjustments is its own methodology. will supply the adjusted series by default and will rarely stop to say which it is.

Two sources disagree. Which one is wrong?

Possibly neither, and that is the whole difficulty. A screen shows year one earnings per share of Rs 5.96/-. The issuer's own annual report for year one, printed and filed at the time, shows Rs 59.58/-. Both are correct. One has been rebased onto today's count and the other has not, and neither of them is obliged to say which.

Before either figure is called wrong, establish which basis each one is on, and never build a single series out of two sources. The ban on mixing sources is the operational rule, and it is stricter than it first sounds: it means a five year series pulled from one screen and a current year pulled from a filed figureThe number as the issuer itself reported it at the time, on whatever share count applied then. Filings are not retrospectively rewritten when a later action changes the count. is already a mixed series, even though every individual number in it is right.

What is checked before trusting any series?

Four steps, in order, done once per series rather than once per number.

Every corporate action in the period covered is listed. The cumulative factor is worked by multiplying, and cross-checked against the ratio of the two share counts. The earliest figure and the latest figure are confirmed to sit on the same count. Confirming both ends is the check that catches the half-corrected middle rung. A series adjusted correctly at both ends can still have been adjusted twice somewhere in the middle, so one figure from the middle is checked against the filing it came from. Four minutes, once, against a series that would otherwise have been quietly wrong for the whole time it was in use.