Time Horizon: Why a View Without One Cannot Be Judged

A time horizon is the period over which a view can be judged, and a view without one can never be wrong. Hold two reader assumptions still and move only the period: the same illustrative price of Rs 486/- asks for about 33.1 per cent annual earnings growth over three years, about 24.2 per cent over five and about 18.0 per cent over ten. Nothing about the company changed. Somebody chose a period.

Three things arrive from earlier material and none of them is rebuilt below. The valuation application material already settled how a quoted price is run backwards into the growth it carries, and that arithmetic runs below at four periods instead of one. The worked claim was assembled earlier in this sequence, with its case pinned at 6.0 per cent of volume and 3.0 per cent of realisation on top, and those two compound to 9.18 per cent rather than adding up to 9.0. And the treatment of what would break a view settled that an observation which would break a view has to carry a period, or it can be waved away forever. The period is the whole subject from here on. A period is not an administrative detail bolted onto a claim, it is a part of the claim, and changing it changes what was claimed. Published earnings per shareA completed year's profit spread evenly across every share in issue, so a big company and a small one can be lined up side by side. Settled in the accounting layer and taken here as given. of Rs 11.58/- and the price of Rs 486/- at the close of year three are the only two observations in this guide. Everything else is arithmetic or assumption, and each is labelled where it appears.

What is a period doing inside a claim, and what does it actually fix?

Start somewhere ordinary. A neighbour says the small sweet shop at the end of the lane is going to do well. The remark sounds reasonable, so nobody argues. Nothing then happens for four years. The shop is sometimes busy and sometimes empty, and at no point does anybody establish whether the neighbour was right. Had the neighbour said instead that the shop would be running a second counter by next Diwali, that would be a different kind of sentence. The second sentence has a date attached, so the world can eventually answer it. After that Diwali the neighbour has a record rather than an opinion.

Research works the same way. The period fixes three things at once, and it fixes them the moment it is written down rather than later. The period fixes when the claim comes due. Nothing else lets a verdict be delivered at all. A result that lands well inside the period is an instalment rather than an answer, so the period also fixes what counts as evidence along the way. The things that can move in twelve months are not the things that move over a decade, so the period fixes which variables the claim is even about. The same sentence about a company is a different claim over one year and over ten, so the period belongs inside the sentence rather than in a footnote under it.

Why can a view with no period never turn out to be wrong?

The mechanics of that are simpler than they sound. A claim is refuted when a result arrives that the claim said would not arrive. If no period is named, there is no arrival. Every adverse result can be met with the observation that it has not played out yet. Nothing was ever due, so that observation is unanswerable.

The usual reading of a missing period is unfair. Be careful about the motive here. Most of the time this is not a dodge. The absence is usually genuine. Nobody sat down and decided to leave the period out, they simply never wrote one in, and the claim went into the world with a missing part that neither the writer nor the reader noticed. A claim that cannot come due produces no information about the person who made it, whether they withheld the period deliberately or never thought of it. The intention makes no difference at all to the record.

The uncomfortable part is what this does to the analyst's own learning rather than to any credibility with anyone else. If four years of claims cannot be scored, then four years of work have taught the analyst nothing about which of those judgements are good, and the only route anybody has to improving research judgement is scoring it. A period is therefore not a concession made to a sceptical reader. The period is the instrument the analyst applies to their own work.

Why does the same price ask for a different company over a different period?

Here the arithmetic does something people find genuinely surprising the first time, so it is worth going slowly. The price of Rs 486/- at the close of year three against published earnings per share of Rs 11.58/- is 41.9689 times, or about 42.0 times as it is stated below. Turning that into a growth rate requires two things that nobody can observe: a required returnThe yearly return an investor wants before parting with the money. Nobody quotes it, nobody can check it, and two people looking at one share will hold different ones., which is a preference, and an exit multipleWhatever rating the shares are supposed to carry once the period runs out. The exit multiple is a supposition about strangers in the future, not a measurement of the business today., which is a guess about other readers later on. Both are fixed here at 12 per cent a year and 25 times. Both stay fixed from here on, and only the period moves.

Now run it at three years. The price has to reach Rs 486/- grown at 12 per cent for three years, or about Rs 682.80/-. At 25 times, that needs earnings of about Rs 27.31/- a share. Getting from Rs 11.58/- to Rs 27.31/- in three years takes about 33.1 per cent a year. The identical steps at ten years put the price at about Rs 1,509.44/-, the earnings needed at about Rs 60.38/-, and the annual rate that gets there at about 18.0 per cent. The implied growthThe annual earnings growth a price already contains, once a required return, an exit multiple and a period have been stated. Derived in the valuation application material and applied here. nearly halved and not one thing about Sarvani Coatings, an invented paint maker, moved.

The reason is a race between two effects of compoundingApplying a rate repeatedly, so each period grows the result of the one before it rather than the original amount. Ten per cent twice is 21 per cent, not 20. and it is not a close race. Lengthening the period compounds the required return over more years. The earnings needed at the end therefore rise. But a longer period also spreads the journey to those earnings across more years. The annual rate needed to get there falls. The spreading wins, always. The implied rate falls towards the required return of 12 per cent and never below it. At an infinitely long period the price would be asking for nothing more than the return demanded. That is the floor. At one year the same price asks for about 88.0 per cent, at five years about 24.2 per cent, at fifteen years about 15.9 per cent, and it would still be above 12 per cent at sixty. The curve is brutally steep at the short end and almost flat past ten years, and this is arithmetic rather than optimism about anybody.

How far apart do two cases actually get, year by year?

The case underneath the worked claim carries 6.0 per cent of volume and 3.0 per cent of realisation. Volume and realisation compound rather than add, so the case rate is 9.18 per cent and not 9.0, and the rounding to about 9.2 per cent used earlier is the same figure to one place. The extra 0.18 of a percentage pointThe unit for a difference between two percentages. A move from 44.0 per cent to 46.0 per cent is two percentage points, and calling it two per cent would mean something else entirely. looks pedantic until the extra 0.18 is compounded for a decade. Run against the five year implied path of about 24.2 per cent, the two lines separate.

Start from the same Rs 11.58/-. After one year the case reaches about Rs 12.64/- and the implied path about Rs 14.39/-, about 13.8 per cent apart. Notice that this is less than the 15.0 point gap between the rates, and the reason is worth holding on to: a gap in rates gets diluted by one year of the lower path compounding underneath it, so 15.0 points divided by 1.0918 is about 13.8 per cent of actual distance. Now go to year five. The case reaches about Rs 17.96/- and the implied path lands on about Rs 34.26/-, exactly the earnings the price asked for. The two are about 1.91 times apart, a distance of about Rs 16.30/- a share.

An argument about a five year view is genuinely unresolvable in its first year and settles itself almost effortlessly in its fifth: at 13.8 per cent apart, ordinary reporting noise covers the whole distance, and at 1.91 times apart nothing covers it. The paths first stand half again apart in year four. Everything before that is a shouting match, and the mistake most people make is to hold the shouting match anyway.

How is a period chosen instead of inherited?

Most periods in research are inherited rather than chosen, and the commonest source is a spreadsheet. Somebody built the model with five forecast columns because five is what the template had, and from that moment onwards every claim resting on the model is a five year claim, defended in meetings as though a reason had been given for it. A period that came out of a template has been assumed rather than chosen, and the tell is that nobody can say what would have made it four or seven.

There are only two honest bases for the number. The first is how long the named variables would actually take to reveal themselves. The second is how long the analyst would realistically still be watching. A period that outlasts that attention produces a claim nobody will ever mark. Taking the first on the case at hand: the worked claim stands on three variables and no more. Whether gross marginWhat is left of a rupee of sales once the paint inside the tin has been paid for, written as a percentage. Settled in the accounting layer and treated here as a published number. stays at 46.0 per cent. Whether volume keeps running above the field's 4.5 per cent. Whether that 0.13 percentage point step up in share comes round a second time.

None of the three is settleable quickly, and the record shows why for each. A level that survives one year could be a level shift working through, so margin holding needs four consecutive quarters and then a repeat. Call it two published years. A single repeat could be luck, so the share gain needs at least the next published year and the one after it. Call it three. Volume ahead of the field is the slowest of the three. The field's own volume record across five years reads 2.1 per cent, 6.8 per cent, minus 1.4 per cent, 7.2 per cent and 4.5 per cent, a swing wide enough that a single year proves nothing about anybody. The record is a volume record over five years rather than a value figure over one, and volume is the reason the period needs the whole five.

So five years, and the reason is the third variable rather than the template. The reasoned answer lands on the same number the template offered. Name that a coincidence out loud. A coincidence gives no permission to skip the reasoning next time.

What happens when the period quietly moves?

The failure below is quiet precisely because it never feels like a decision. The claim was due at year three. Year three arrives, the result goes the wrong way, and somebody says the story needs longer to play out. Nobody objects. The sentence sounds reasonable, and it is the kind of sentence that gets said in every meeting. The due date moves to year five. The year three result was the test, and it is now reclassified as an interim reading inside a longer claim.

Count what was thrown away. At year three the two paths already stand about 1.47 times apart, a wide enough distance for a published year to speak to. Move the marker to year five and that reading becomes an instalment. Move it again to year seven, where the paths would stand about 2.47 times apart, and the year five reading becomes an instalment too. A period that moves after a disappointing result converts a testable claim into an untestable one at exactly the moment the test was about to run, and it does that whether or not anybody intended it.

The fix costs one line, and it has to be written at the start. Written afterwards the line is worthless. Record the period beside the claim, in the same sentence, and then treat any change to it as a change to the claim rather than as an administrative correction. If the period moves from three years to five, the honest description is that the old claim was settled and a new one has been opened, and the old one goes on the record with whatever verdict year three delivered. The discomfort of that is the entire reason it works.

What does the period do to the evidence collected along the way?

A one year claim is settled by quarters. Four of them make up the whole period, and the fourth one closes it. A five year claim is not settled by quarters at all. One quarter is 5.0 per cent of the period, and reading it as though it answers the claim manufactures noise that feels exactly like information.

Put a number on how little a quarter can carry. One quarter of the worked claim's 9.18 per cent a year is about 2.22 per cent of growth, and one quarter of the implied path's 24.2 per cent is about 5.57 per cent, so the entire distance between the two cases inside a single quarter is about 3.35 percentage points of growth. Ordinary quarterly movement in a paints business, with a festive period and a monsoon in it, covers that distance without anybody's thesis being right or wrong about anything. Matching the evidence to the period is what stops a five year claim being retried every ninety days, and being retried every ninety days is how a long claim gets abandoned before it was ever tested.

Quarters do have a genuine job on a long claim, and earlier material named it. A quarter is where the specific observation that would break the view is looked for. Whether the view has come good yet is a different question. A quarter can show that the margin sat at 46.0 per cent for a fourth time running while input cost for each tin went up, and that is real evidence about a named variable. A quarter cannot show whether a five year claim worked.

The whole thing run once, on one price and four periods

Before the first number, the label. The table below is an exercise built so a period can be seen doing its work, and not a view about Sarvani Coatings Limited or about anything else. The two assumptions belong to an imagined reader and are held identical across all four rows, so the only moving part in the table is the period.

Read the last column and then read the first: the price did not move, the company did not move, and the growth the price appears to assume ranges from about 18.0 per cent to about 33.1 per cent purely because somebody chose a period. That is what the table establishes. Here is what it does not establish. The table does not say which period is correct. Correctness depends on the claim being made and not on the price. Nor does the table say the shares are dear at 41.9689 times or cheap at any other figure. Every figure in the table is a requirement rather than a forecast, so the table says nothing whatever about what will happen.

Put the worked claim beside it and the period stops being an abstraction. At 9.18 per cent a year the claim reaches about Rs 12.64/- after one year against about Rs 14.39/- on the five year implied path, and about Rs 17.96/- after five years against about Rs 34.26/-. So the period on this worked claim is five years, and the reason is the three named variables rather than the convention: margin holding, volume staying ahead of the field and the share gain repeating each need several published years, and no single quarter can settle any of the three.

Which figures here were observed, and which were assumed?

Only two figures in this guide are observations even inside the invented record: the price of Rs 486/- at the close of year three and the published earnings per share of Rs 11.58/- for the same year. Everything else in the four row table was worked out from those two plus a 12 per cent return demanded and a rating of 25 times supposed at the end, neither of which came from anywhere except one imagined reader. A table of four implied growth rates with the assumptions stripped off would read as four facts about the market, when the four are outputs of one person's arithmetic. The distinction matters. What a research writer may publish and must disclose is a matter for the regulator named above.