Thesis Risk and Valuation Risk: Two Different Failures

Thesis risk is the business not doing what the analyst assumed. Valuation risk is the business doing exactly what was assumed while the rating applied to those earnings changes. The two fail independently and they need different evidence. On the arithmetic worked below, the rating assumption swings the answer about three times as far as the return assumption does. Valuation risk is the half most often left unnamed.

Splitting the gap between the two, on the analyst's own figures

Both risks live inside one calculation, and the calculation is where they come apart. The panel takes what a quoted price is asking of earnings on one side and what the business case delivers on the other, and shows the distance between them. The distance has to be closed by something. Either the business grows faster than assumed, and that is the thesis half, or the rating applied to those earnings at the end runs above what was assumed, and that is the valuation half. One control moves the gap from one to the other. Two of the fields are not on any document and are marked as the analyst's own: the required returnThe yearly percentage settled on as the analyst's own price for parting with money. The figure is chosen, not published. required each year, and the rating at the end.

At the figures the panel opens on, a price of Rs 486/- and earnings per share of Rs 11.58/-, five years, a 12 per cent required return and a rating of 25 times assumed at the end, the price is asking earnings to grow 24.23 per cent a year. The business case beside it, revenue growing 12 per cent a year with the 46.0 per cent gross margin held and costs growing 10 per cent, carries earnings per share to Rs 23.59/- in year five. Growth in that figure is 15.30 per cent a year. The gap is 8.93 percentage points. With the slider at none of it the whole gap sits on the rating, and the rating would have to run 45.22 per cent above the 25 times assumed for the price to work. With the slider the other way the business carries all of it, which needs revenue growing 15.66 per cent a year rather than 12, with nothing else altered. Neither number predicts anything. The two are the ends of one question about which of the risks is actually being taken.

What is thesis risk, put in one line?

Thesis risk is the risk that the named assumptions written about the business turn out to be wrong. The definition goes no further than that, and thesis risk is set out in full on its own.

A paint shop on a busy road is bought because the buyer has worked out that it will take Rs 40,000/- a week, that the new housing going up behind it will lift that, and that the supplier will keep giving the same terms. Thesis risk is every one of those three sentences being wrong. The housing gets delayed. The supplier tightens terms. Takings come in at Rs 33,000/-. Thesis risk always leaves a mark somewhere a reader can go and look: a number in the till, a letter from the supplier, a hoarding that never came down. The mark is what makes thesis risk researchable, and it is also what makes it the comfortable half of the pair.

On Sarvani Coatings Limited the same three sentences exist and they are written down. The view rests on some part of the recorded gross margin improvement holding, on the company continuing to take share in its field, and on volume. Every one of those has a filing behind it that will eventually say yes or no.

What is valuation risk, put in one line?

Valuation risk is the risk that the rating applied to those earnings changes, even when the business performs. One line again, and nothing in it is about the company at all.

The paint shop again, in the form that surprises people. Every sentence written comes true. The housing arrives on time, the supplier holds terms, takings run at Rs 40,000/- a week exactly as stated. Three years later the shop is up for sale, and the buyer who was going to pay four times a year of takings now pays three. Nothing about the shop changed. The price other people put on a rupee of shop takings changed. Everything researched was right and the owner is still worse off, and there is nothing in the till, the ledger or the supplier file that explains it.

In the listed version this has a name. The same profit priced at a lower multiple than before is a de-ratingMarket shorthand for the same earnings being priced at a lower multiple than they were before. The change is in what buyers will pay, not in the company.. A de-rating does not need a single bad quarter to happen, and it does not need any news at all.

Why do the two fail independently?

Because they live in different places. One lives in the business. The other lives in what other people will pay for a rupee of it. Nothing connects them mechanically. All four combinations occur, and each of them can be pictured.

All four combinations occur. The one rarely pictured, the business disappointing while the rating rises, happens whenever what was already assumed was worse than what actually arrived. The case is anything but exotic. A company reports a soft quarter, the shares go up, and everybody writes a puzzled paragraph about it. There is nothing to puzzle over: the soft quarter was softer than the writer expected and better than the price expected, and the price is what other people are acting on.

Which of the two carries more of the arithmetic here?

The split can be measured rather than asserted. Backward calculation is the tool: the quoted price and the two assumptions together give a reading of the annual compound growthThe one steady yearly rate which, repeated, carries a figure from where it starts to where it lands across a named span. The one rate flattens an uneven path into a single figure. in earnings the price already contains. Then move one assumption at a time and watch the readout move.

First, thesis risk with a number attached. The illustrative record shows a gross marginRevenue less the cost of the materials that went into it, expressed as a percentage of revenue. The measure says nothing about salaries, freight, advertising, interest or tax. of 43.0 per cent in year one and 46.0 per cent in year three, a gain of 3.0 points over those two years. Suppose the assumption fails completely and the whole two year gain goes back. A margin point is one per cent of revenue, so 3.0 points on revenue of Rs 2,415 crore is Rs 72.45 crore of gross profit that never appears. Nothing below the gross line changes, so the whole of it travels down the ladder untouched.

Two notes on that table before it gets used. An effective tax rateThe tax charge actually recorded divided by the profit before tax it was charged on. The rate is a result read off the statement, not a figure anybody sets. quoted to one decimal place is a rounded figure, and rebuilding a rupee amount from it introduces an error that looks derived. So the tax charge in the right hand column is taken at the published ratio of Rs 93 crore on Rs 371 crore rather than at the 25.1 per cent that ratio prints as. And the 3.0 points is a two year move, year one to year three; the one year move from year two to year three is 2.0 points, and the two are never mixed. Reverse the whole two year gross margin gain and earnings per shareThe profit left once tax is paid, divided among the shares in issue. Each share carries the same amount, and that is what lets two companies of different sizes be set beside each other. falls from Rs 11.58/- to Rs 9.32/-, which is thesis risk arriving with a figure attached.

Now valuation risk, worked the other way round. The mechanics take one line: Rs 486/- compounding at 12 per cent for five years reaches Rs 856.50/-, which at a rating of 25 times means Rs 34.26/- of profit per share in year five, and carrying Rs 11.58/- up to Rs 34.26/- takes 24.23 per cent a year. The Rs 856.50/- is not a target and nobody is expecting it. The figure is only what a 12 per cent requirement implies once the arithmetic is done. With the return frozen at 12 per cent and the horizon at five years, only the rating moves. At 35 times the Rs 486/- turns out to hold an expectation of 16.14 per cent a year. At 25 the same Rs 486/- holds 24.23 per cent. At 15 it holds 37.59 per cent. Top to bottom that is 21.45 points of movement, 37.59 less 16.14, out of one assumption about which nothing whatsoever can be researched.

Now the comparison. Somebody chose both ranges, so both are printed beside the swings they produce. Across ratings of 15 to 35 times, holding the required return at 12 per cent, required growth runs from 37.59 per cent down to 16.14 per cent, a swing of 21.45 points. Across required returns of 8 to 14 per cent, holding the rating at 25 times, required growth runs from 19.79 to 26.45 per cent, a swing of 6.66 points. Set against each other, the two swings make the rating assumption worth about 3.2 times what the required return assumption is worth.

A swing quoted without the range it was drawn across means nothing and would travel badly, so both assumption ranges are printed beside the swings they produce. The backward computation itself is ordinary arithmetic belonging to no author. Every percentage figure above follows from the invented ladder: the Rs 72.45 crore is 3.0 margin points on revenue of Rs 2,415 crore, and the tax there was taken at the published Rs 93 crore on Rs 371 crore rather than at the 25.1 per cent it prints as. The reverse arithmetic runs on the per share figure as published, Rs 11.58/-; on the unrounded Rs 11.5833/- behind it every percentage above moves by under a hundredth of a point and the comparison is unchanged.

The threefold answer is a property of this price and this set of assumptions, and of how wide each range was drawn, rather than a general law that carries anywhere. A narrower rating range shrinks the ratio. A wider return range shrinks it further. The honest way to hold this is that the procedure travels and the number does not: on any company, moving each assumption on its own across a range that would actually be accepted shows which one the answer follows. The mechanical test is what sensitivityNudging a single input while everything else stays frozen, then reading how far the answer travelled. Mechanics, not likelihood. work is, and it takes about four minutes.

What evidence bears on each of them?

Thesis risk is tested by things a company publishes. Quarterly results. Segment splits. Volume commentary. Realisation per unit if it is disclosed. The margins of Nandivarman Paints Limited and Kesaria Surface Solutions Limited in the same quarter, showing whether a move was the whole field or one company. The resin and additive prices coming out of a supplier such as Thottam Chemicals Limited. Each of these is a real place a reader can go, and each can contradict something written down. A named place that could contradict a written assumption is what an observableA specific, named thing a reader could actually go and look at to see whether an assumption is holding. A direction of travel or a feeling is not one. is.

Valuation risk is tested by almost nothing a company publishes. There is no filing that says what buyers will pay for a rupee of these earnings in five years. There is no disclosure obligation covering it, no statistic that settles it and no authority that maintains it. The ratings this company and its peers have carried before are available, and they record what has happened, not what will. The risk carrying more of the arithmetic is the risk with almost no evidence available, and those two facts together are the whole discomfort of the split.

Why is valuation risk the one usually left unnamed?

Four reasons, and they are mechanism rather than character. Valuation risk cannot be researched, so no amount of work reduces it and nobody has a method to sell. The risk has nothing to watch, so it cannot be turned into a monitoring line in a note. Naming it embarrasses the writer. A large part of the outcome rewards none of the analysis just completed. And naming it makes the whole exercise look more fragile than the person writing it would like, especially to a reader who came for confidence.

Every one of those four is a reason to write the risk down rather than a reason to leave it out. A risk that cannot be researched is exactly the one a reader needs told, because they cannot find it themselves. A risk with nothing to watch is the one that will otherwise be explained away later as something else. A risk that rewards none of the analysis is the one the analysis will keep failing to catch. And a view that looks fragile because it states its fragile parts is a more useful document than one that looks solid by omission.

What happens when both arrive at once?

Both arriving at once is the ordinary bad case rather than the exotic one, and the arithmetic has a shape people get wrong in both directions. The two failures are the two already computed. Earnings reach Rs 9.32/- against the Rs 11.58/- the view started from, 80.48 per cent of it and a shortfall of 19.52 per cent. The rating settles at 20 times rather than the 25 assumed, 80.00 per cent of it and a shortfall of 20.00 per cent.

The two shortfalls multiply into each other rather than adding, so 80.48 per cent of the earnings meeting 80.00 per cent of the rating leaves 64.39 per cent of what the view assumed, a combined shortfall of 35.61 per cent. Two things follow and only one of them is expected. The combined case is nearly twice as bad as either failure alone, and testing one assumption at a time never finds it. Each case run on its own shows a shortfall of about 20 per cent, and the case where both land shows 35.61 per cent. Nor is the answer 39.52 per cent, the figure adding 19.52 and 20.00 gives. The second shortfall applies to a base the first has already reduced, so adding counts the corner where the two overlap twice, worth 3.90 points.

How does anybody actually use this?

Three readers, three jobs, and not one of them requires a number for the share.

A research analyst uses it to split the post mortem. When a position goes wrong, the first question is which of the two failed, and the answer decides what gets learned. If the business missed, the reading process is what needs work. If the business delivered and the rating moved, the reading process was fine and the assumption that failed was one nobody researched. Getting that split right is the difference between improving and thrashing.

A person managing money for households uses it to size a holding rather than to pick one. Two positions can carry identical business risk and very different valuation risk. One is held at a rating close to what it has carried before, the other at a rating well above it. The second one needs less money behind it, and the reason has nothing to do with the quality of the two companies.

A lender uses the split in reverse. Lending against shares as collateral exposes the lender almost entirely to valuation risk and hardly at all to the business. A borrower can be lent against a perfectly healthy company whose rating halves. Lending against shares is therefore done with a margin over the amount lent rather than with a view on the company. The same two risks that an analyst separates to explain a result, a lender separates to decide how much cushion to hold. And the household version is the paint shop again: the takings are one risk, and what the next buyer will pay for a rupee of takings is a different one entirely.

How is each one written down so it can be watched?

Side by side, in the same four fields. Written out that way, the difference stops being an argument and becomes visible. The business assumption gets what was assumed, where it came from, what would show it moving, and where that would show up. The rating assumption gets what was assumed, where it came from, the range decided in advance as tolerable, and an honest blank.

Writing down a risk that cannot be watched is what stops it being quietly reclassified as a business problem a year later. The blank in that third field is doing real work. The blank is a dated statement that on the day the view was formed, its author already knew one assumption could never be checked against anything. Without it, the only surviving record of the position is a set of business assumptions and a bad outcome, and the two get joined up by whoever reads the file next. The tolerable range does the other half of the job. A range of 20 to 30 times, written on day one, means a rating of 22 times is inside what was accepted and not a surprise. A rating of 15 times sits outside it, and 15 times is the assumption failing rather than the company.

When does the distinction stop mattering?

Three situations, and naming them is the difference between a tool and a habit.

The first is a holding measured in weeks. Earnings barely move over a few weeks, so almost the whole of what happens to the position is the rating changing. The business assumption is still underneath it, but it has had no time to be right or wrong yet. One of the two has not had a chance to arrive, so splitting them here produces a number and no decision.

The second is a holding there is no intention to sell. When the dividend rather than an exit is what is being counted on, no rating is ever applied at the end. The assumption about the business is then the whole of the risk, and valuation risk has nothing to attach itself to. The claim is worth testing honestly before it is accepted. Most people who say they will never sell are describing an intention rather than a rule, and the moment selling is back on the table the second risk is back with it.

The third is a position small enough that the split changes nothing. Where the position would be sized the same way whichever of the two dominates, the separation has cost an hour and bought nothing that will be acted on.

The post mortem never stops mattering, and it never stops mattering in any of the three cases. Even on a holding that ran three weeks, when it goes wrong somebody has to say whether the business did something or the price of a rupee of its earnings changed, because the answer decides what gets tightened next. The distinction can stop being useful for the decision to buy and still be the only thing that makes the outcome legible afterwards.

One case looks like it belongs on this list and does not. When the business misses and the rating falls in the same month, the two feel like one risk with one cause. The two are not one risk, and the panel at the top shows why. The shortfalls came from different inputs and they land on the answer differently, one compounding through five years of earnings and the other dividing the ending figure once. Arriving together is not evidence of being the same thing.