Building an Economic Scenario: Assumptions Made Explicit

An economic scenario is a stated set of assumptions carried through arithmetic to an outcome. A scenario makes no claim that those assumptions will hold. A forecast claims its central case is what will happen. A scenario claims only that if these assumptions hold, then this follows. The whole difference sits in what is being asserted, and not in the numbers.

Two skills already in hand do most of the work in this guide. One is laying out three cases side by side and giving each of them a weight. The other is taking one variable, moving it by a stated amount, and carrying that move through to a number at the bottom. Neither of those skills is the subject here. The subject is the object those skills produce, and specifically what that object is entitled to claim once it is finished and somebody is reading it.

Start at a wedding, with no economy in sight. An uncle announces that the wedding will cost Rs 3,80,000/-. A cousin writes two lines on the back of an envelope instead: the hall is Rs 2,00,000/-, and three hundred guests eat at Rs 600/- a plate, so the total is Rs 3,80,000/-. Same figure. Same arithmetic. The two are not the same object at all. If four hundred people are expected, the cousin's envelope can be picked up and redone by anybody, and the total comes to Rs 4,40,000/-. Nothing at all can be done with the uncle's sentence except believe him or not. The envelope is worth holding in mind. Everything below is the same envelope, with a country attached.

Forecast vs Scenario: what is each one actually claiming?

Put the two objects side by side and read only the last line of each. A forecast ends with an assertion about the world: this is the number. A scenario ends with an assertion about a chain of reasoning: given these stated inputs, this is what falls out at the bottom. The first is a claim about what happens. The second is a claim about what follows.

Now the part that surprises people. Everything above that last line can be word for word identical. The same three assumptions, the same arithmetic, the same figure at the bottom, rounded the same way, in the same units. A forecaster and a scenario builder can produce the same sheet of working and disagree completely about what they have produced. The two can be built from identical numbers, so the difference is the claim and not the arithmetic.

Which means something uncomfortable about how these things travel. A scenario that gets summarised, forwarded, put on a slide and read out loud arrives at the far end with its conditional stripped off. Nobody removed it deliberately. The word if is simply the easiest thing in a sentence to lose. And at that point a scenario has been presented as a forecast. A claim has been added to it that its construction never supported. The arithmetic did not get worse on the way. The assertion got bigger.

Why does every assumption have to be written down?

Go back to the envelope. The reason the cousin's version is worth more than the uncle's has nothing to do with either of them being right. The cousin's version can be taken apart by somebody who was not there when it was written. Three hundred guests is a line a reader can point at and say no, four hundred. Rs 600/- a plate is a line a reader can point at and say the caterer down the road quotes Rs 700/-. Each line is a handle.

An assumption that was never written down is not a handle. The unwritten assumption is still in there, doing exactly as much work as the written ones, and a reader who disagrees with it has nothing to grip. No part has been named, so the reader cannot say which part they reject. All they can do is take the bottom figure or leave it, and taking it or leaving it is not analysis, it is a vote. Writing the assumptions down is what makes a scenario arguable, and arguable is the entire point of building one.

An assumption list is therefore not a courtesy or a compliance habit. The list is the working part. A scenario is a machine for letting somebody else disagree precisely, and the list is the set of places they are invited to disagree. A scenario nobody can find fault with should worry the person who built it. If a reader cannot locate a line they would change, the scenario has probably not stated enough of itself.

There is a second reason, quieter and just as practical. Six months later, the analyst will not remember what was assumed. The number is what got quoted back, so the number will be remembered. If the list is written down, the one line the year proved wrong can be found and changed, and the rest rerun. If it is not, the whole thing has to be rebuilt from nothing, it will be rebuilt differently, and which of the two versions is on the desk will not be knowable.

What makes a scenario internally consistent?

Here is the test that matters, and it is not the one most people run. The usual instinct is to read a scenario and ask whether it sounds reasonable. Does a five per cent currency move sound plausible? Does a twenty per cent oil move sound extreme? Plausibility questions are interesting, and plausibility questions come second. The first question is whether the assumptions on the list can all be true at the same time.

Macro variables are not independent items on a shopping list, so the assumptions often cannot all hold at once. Some of them are tied to each other by arithmetic that holds whatever anybody thinks. If a list moves two things that are mechanically linked, the list has already decided some of what comes next, and that part is no longer available to be stated separately. Stated anyway, wrongly, it leaves the scenario contradicting itself in writing.

The Republic of Sankhya, an invented economy used throughout, gives the cleanest example there is. The assumption list says the Marut unit rises 5.00 per cent. A rise in the Marut unit is a depreciationthe rupee buying fewer units of the other currency than before. How a currency move splits an exporter's revenue from its imported cost base is worked out separately; the result is taken and used here. of the rupee. The list also says oil rises 20.00 per cent measured in Marut units. Now ask what a barrel costs the company in rupees. Both moves land on the same barrel, one after the other. The rupee cost of oil goes up by 1.05 multiplied by 1.20, compoundedtwo moves stacked rather than laid side by side, so whatever the first one leaves behind is what the second one lands on. Why a pair of macro moves stacks this way instead of adding is worked out separately, and the figure that comes out of it is used here unchanged. to 26.00 per cent. Adding the two moves gives 25.00 per cent instead. Forgetting that the oil is bought in the other currency at all gives 20.00 per cent. Neither figure is what the two stated lines produce together.

So a scenario that states both of those lines and then writes 20.00 per cent into its rupee cost base has contradicted its own assumption list, in ink, in the same document. Nobody needs to wait for the year to end to know that. The contradiction can be checked immediately, by anybody, with a calculator. A scenario combining a large depreciation with unchanged imported costs is not a pessimistic scenario, it is an impossible one.

What does the Sankhya scenario look like with every line stated?

Here is the whole thing, built once, with nothing held back. The company is deliberately small: revenue, costs and one borrowing line, and no more than that. Revenue is Rs 1,000.00 crore, of which Rs 300.00 crore is earned abroad. Costs are Rs 900.00 crore, of which Rs 360.00 crore buys imported inputs, and Rs 90.00 crore of that is oil. Operating profit is therefore Rs 100.00 crore and the operating marginwhat share of every rupee of sales is still standing once the cost base has been paid for, put as a percentage. The margin is read off the trading lines alone, with no reference to a balance sheet. is 10.00 per cent. The base year is the starting point, and no line of it is in dispute.

The assumption list has three lines. The Marut unit rises 5.00 per cent. Oil rises 20.00 per cent measured in Marut units. The policy ratethe headline interest rate a central bank fixes so as to steer what short term money costs across an economy. How it is decided and how it travels are covered separately. Here it is only a line on a list, and the list holds it still. holds, so the borrowing line does not move. Three lines, and every one of them invented for this worked case.

Now run the consistency check before touching the arithmetic. Lines one and two both land on the barrel, so the rupee oil move is fixed at 26.00 per cent by the two lines together. Line one also lands on every other imported input and on the money earned abroad. Line three holds one thing still and does not conflict with anything. The list can hold together, so the scenario is allowed to proceed.

All figures in the table are Rs crore except the last row. Every component is printed so the totals can be rebuilt from scratch: 700.00 plus 315.00 gives revenue, and 540.00 plus 283.50 plus 113.40 gives costs. The profit can be checked a second way to tie it down. Rs 15.00 crore more earned abroad is outweighed by Rs 18.00 crore more spent on imported inputs, so the currency line on its own costs Rs 3.00 crore of profit and leaves Rs 97.00 crore. Then the oil line adds Rs 18.90 crore of cost beyond what the currency alone had already done to the barrel. Taking Rs 18.90 crore off Rs 97.00 crore lands on Rs 78.10 crore. Two routes, one figure.

Look at the last two rows before moving on. One input at 10.00 per cent of the cost base took nearly a quarter of the profit with it. Profit is a thin residualthe remainder after one big figure has been subtracted from another big figure. Small percentage moves in either of the two turn into large percentage moves in whatever is left over, and that is arithmetic rather than anything about the business itself. sitting between two much larger numbers. Revenue rose at the same time as profit fell, so the margin fell further than the profit did. The drop was 2.31 points on a base of 10.00.

How is a scenario judged, if it never claimed it would happen?

Most readers get stuck here, and the stuck feeling is reasonable. If a scenario does not claim to be what happens, then the year ending differently is not evidence against it, and the year ending the same way is not evidence for it. So on what grounds does anybody call one scenario better than another?

Three grounds, and all three are available immediately. First, were the assumptions stated? Not some of them, not the interesting ones, all of them, including the boring line that says something is held still. Second, can the stated assumptions hold together? The consistency test is arithmetic rather than judgement. Third, was the arithmetic carrying the assumptions to the outcome done correctly? Add up the components and see.

A scenario is judged on whether the assumptions were stated, whether they were consistent, and whether the arithmetic carrying them was right, and every one of those three can be checked the day it is written. The year does not need to end. No track record is needed. The working and half an hour are all anybody needs.

Set that against the test people reach for instead, whether it happened. The outcome test arrives years late, and worse, it answers a question the scenario never asked. A perfectly built scenario asserts a conditional, so assumptions that did not come to pass mean the condition never arose and the scenario has failed nothing. A badly built scenario whose outcome happened to land near the eventual print has still got a contradiction sitting in the middle of it. Outcomes are a poor grader of conditionals.

When does attaching a probability turn a scenario into a forecast?

Now the temptation, and it is a strong one because it feels like extra rigour rather than less. There are three cases. Somebody asks which one the analyst thinks. Attaching a number to each feels like an honest answer to an honest question.

Weights on a set are legitimate, and they do a job nothing else does: they say how much of the analyst's thinking each case is carrying, and they force the three to be considered together rather than one at a time. Take the Sankhya set as it was fixed: the base case at Rs 100.00 crore of profit with a weight of 50.00 per cent, an upper case at Rs 130.00 crore with 25.00 per cent, and a lower case at Rs 60.00 crore with 25.00 per cent. The weights sum to 1.00, and the weighted figure works out at Rs 97.50 crore.

Rs 97.50 crore is the most misread number in the set, and it repays a hard look. The weighted figure sits below the base case. Rs 97.50 crore is not the base case, not the upper case and not the lower case, and it happens in none of the three. The weighted figure summarises a set of assumptions together with the weights placed on them, and no year can print it. A reader who quotes it as what the company will earn has taken an average of futures and called it a future.

Probabilities are legitimate on a set and become a forecast the moment one case is singled out as what will happen. The language shows where the line gets crossed. A set with weights on all three cases is still a set. One case described as the likely one, the central one or the base expectation has stopped being a conditional and started asserting an outcome, whatever the other two are still doing beside it. A single case with a probability attached and no set around it has crossed the line already. The only thing that number can be modifying is a claim about the world.

What does a scenario refuse to claim?

Three things, and they are worth saying out loud because a scenario set that does not say them is usually read as claiming them.

A scenario does not claim likelihood. Likelihood is the claim people trip over most.

A scenario does not claim completeness. Three cases were written because three are manageable and three is how a reader holds a range in mind, not because the future has three settings. Every scenario set is drawn from a space of possibilities nobody has counted, and the three that were chosen are three of them. A scenario set is always a subset of possible futures, and presenting three as exhaustive is a claim nobody can support. The honest version of a scenario set says these are three that were built, here is why they were chosen, and here is what would put an outcome outside all three.

And it does not claim that the variables on the list are the only ones that matter. The Sankhya list moves a currency and an oil price and holds a borrowing line still. The list says nothing about volumes, nothing about what a competitor does, nothing about whether the company can pass a cost on. None of the three is assumed to be irrelevant. Each one is assumed to be unchanged, and unchanged is a completely different statement from irrelevant, and a much stronger one. Every line held still is an assumption, and the ones nobody thought to hold still are assumptions made without being noticed.

What makes a scenario survive contact with a reader?

Everything above is about construction. A scenario earns its keep or fails to only after it has been handed to a reader.

A lender sizing a borrowing line reads the assumption list looking for the one line the borrower's ability to pay hangs on. In the Sankhya case that is not the oil line, big as it is. The line saying the policy rate holds is the one touching the interest bill directly, and that is the line the lender reads for. The lender does not have to accept it. The lender pulls it out, puts in a rise of fifty basis pointsa hundredth of a point. Fifty of them added to a rate of 9.00 per cent give 9.50 per cent. Quoting rate moves this way removes the ambiguity buried in a phrase such as half a per cent. Half a per cent could mean half of the rate itself or half a point on top of it. instead, and reruns that one strand: the borrowing is Rs 200.00 crore at 9.00 per cent, so the interest bill lands at Rs 19.00 crore instead of Rs 18.00 crore, and the Rs 1.00 crore of difference amounts to 1.00 per cent of the base profit. Small, checkable, and entirely their own work.

An equity analyst holds a different view on the oil line, so the analyst does the same thing to that line. A treasurer does it to the currency line. And each of them can only do it because the list exists. A reader who can see the assumption list can rerun the arithmetic themselves, and a reader who cannot see it can only accept the whole thing or reject the whole thing. That is the practical difference between a scenario that gets used and one that gets nodded at.

The household version is smaller and exactly the same shape. Somebody planning around a salary works out what the year looks like if the increment lands at one level, and what it looks like at another. The useful part is never the two totals. The useful part is discovering which single line the whole plan hangs on, and that line is the one worth watching. A scenario built and then never disagreed with has produced a number. A scenario built, argued over and rerun has shown where the risk sits, and that second thing is what the exercise was for. Pulling at one line at a time is how a reader produces a sensitivityhow far an outcome moves for a stated one unit move in an input, quoted with the unit move attached. Measuring one is covered separately; it is named here only as the thing a reader produces when they rerun a single line. for the line they care about.