How to Build Base, Bull and Bear Macro Scenarios

Build a scenario set by fixing the variable list first, then writing each case as a different set of values for those same variables, then carrying every case through identical arithmetic, and attaching probabilities only at the end. Cases built from different variable lists cannot be set beside one another, and that single slip is the commonest way a scenario set falls apart.

Two ideas do their work here without being rebuilt. A macro variable travels along routes to reach a company's revenue, its cost lines and its interest bill, and those routes are set out under macro transmission. A reading discipline settles which published figure a case may start from, and that discipline came out of the work on releases and their later rounds. Both are taken as given from here on.

Where the figures come from: each of the three profit lines below was recomputed from its own published variable values, so a reader can redo any of them independently. The institutions named further down appear for their role alone.

In what order does a scenario set get built?

Seven steps, and the order is the whole method. Each step is an action, and each one leaves behind something written that the next step picks up. Why a currency move reaches a cost line, and why one variable moves an output more than another does, are covered separately and are used here as finished inputs.

The order matters because six of the seven steps are cheap to redo and the first one is not. Change a value at step two and the arithmetic is rerun. Change the variable list after three cases exist and every case has to be written again. A case is only a set of values for a list, and a new list makes the old values an incomplete answer to a different question.

Consider a household deciding whether next year works on one salary. Somebody proposes listing the bills the household does not control, then writing three versions. Rent, the school fee, the electricity bill. Three lines, fixed before anybody argues. Then the good year, the ordinary year and the bad year are each just three numbers against those three lines, and the difference between the three totals is readable because the three lists are identical. The moment one version quietly adds a wedding to the bad year, the totals stop being comparable, and nobody in the room can say how much of the gap is the bad year and how much is the wedding.

Step one: what gets fixed before a single case is written?

The variable list, and nothing else. Three to five macro variables, named, each one a thing the analyst does not control. Nirvi Engineering, an invented manufacturer inside the invented Republic of Sankhya, carries a list of three entries: the exchange rate of the Marut unit, the price of oil in foreign currency, and the policy rateThe interest rate a central bank sets and lends at, which other lending rates in the economy tend to follow. Where it comes from and how it travels are handled elsewhere; here it is simply a number a case names.. The variable list is written once and then closed.

A case is nothing more than a set of values for a list, and two cases can be compared only when the list underneath them is the same one. Fixing the list first is the whole reason three cases can be set beside one another later. Get this wrong and every later step is wasted effort. Get it right and steps two through seven are close to mechanical.

How many variables? Enough that the output can actually move, few enough that a reader can hold them. Three is comfortable. Five is a stretch. Ten is a spreadsheet nobody reads, and the extra seven are usually there because somebody could not decide what mattered rather than because they matter.

Step two: what does the base case actually say?

The base case gives one value to each variable on the list, and every value is one that can be defended by pointing at something rather than by arguing for it. For Nirvi Engineering the base takes the published Sankhya settings as they stand: the exchange rate at 80.00 Marut per foreign unit, the oil price as an indexA series rebased so that one chosen starting point equals a round number, usually one hundred, so later readings are read as movements away from it rather than as levels. at 100.00 in foreign currency terms, and the policy rate at 6.00 per cent.

Base means most assumable, not most likely. Those are different words and they do different work. Most assumable means the source of the number can be named and a reader can go and look at the same place. Most likely is a claim about the future, and a claim about the future is a forecastA statement of what somebody thinks a number will actually turn out to be. Making one honestly, and telling one apart from a scenario, are both covered separately.. A scenario set against a forecast is worked through in Building an Economic Scenario: Assumptions Made Explicit, and the distinction is taken as given here.

The practical test is short. The sentence "this value is here because" has to be finishable. If the ending is a source, the value belongs in the base. If the ending is "it feels about right", it belongs in the bull or the bear, where a stated stretch is exactly what is wanted.

Step three: how do the bull and the bear case get written?

By moving the same three variables and touching nothing else. The bull case moves all three in directions that help this company; the bear case moves all three in directions that hurt it. Notice that the direction which helps has to be worked out for this company rather than assumed. Nirvi Engineering buys more from abroad than it sells there, and macro transmission sets out the routes in full.

A case that introduces a variable the other cases do not carry has left the set, however sensible the new variable is on its own. The rule is not about the quality of the idea. The rule is about what a difference between two outputs is then allowed to mean. Add wages to the bear case and the bear output is now answering a different question from the other two, so putting the three side by side is comparing answers to different questions.

Here are the nine values, three variables against three cases, with nothing hidden. The policy rate row carries its move in basis pointsHundredths of one percentage point. A move of fifty basis points is half a percentage point. Quoting rate changes this way keeps half points and quarter points out of fractions. as well as in per cent, because a rate move is normally quoted that way.

The Marut unit gains in the bull case. A reader who expects an exporter to want a depreciationA currency losing value against another one, so more units of it are needed to buy the same foreign amount. Which businesses it helps and which it hurts is covered separately. will find that surprising. The routes that make it so were laid out in Macro Transmission: How a Variable Reaches a Company's Numbers, and here the direction is simply used.

Step four: what carries all three cases to one output line?

The same arithmetic, run three times. Not similar arithmetic. The same lines in the same order, with only the nine values from step three differing between the runs. The output line for this set is the profit line, and it is defined once so it can be checked: revenue less costs, then adjusted by however much the interest bill has moved away from its base of Rs 18 crore. At base nothing has moved, so the line reproduces the published Rs 100 crore exactly.

The cost baseThe total of what a business spends to produce what it sells, before interest. Named lines let a variable be applied to the part of the cost base it actually touches. splits into three named lines so each variable can be applied to the part it touches: imported inputs other than oil at Rs 270 crore, the oil line at Rs 90 crore, and costs incurred at home at Rs 540 crore, which add to the published Rs 900 crore. The oil line carries both the exchange rate and the oil price, and the two moves compoundTwo percentage moves applied one after the other, so the second acts on the result of the first rather than on the original. Adding two moves instead of compounding them is a separate lesson. rather than add, which the work on commodity prices established and which is taken as given here.

How much each individual variable moves that profit line, expressed per one per cent so two variables can be set beside each other, is sensitivityHow far one output moves when one input moves by a stated amount. Sensitivity is covered separately. Here the arithmetic is simply run, and no single variable is measured on its own., and it is worked through in Macro Sensitivity: Measuring How Much a Variable Moves an Outcome. Step four takes nothing from it. Step four just runs the three columns and writes down what comes out.

Step five: how do probabilities get attached without inventing precision?

One number is written against each case, the three add to one hundred per cent, and that is the end of it. For this set: 50.00 per cent on the base, 25.00 on the bull and 25.00 on the bear. Those three numbers are the whole of step five.

Round numbers are the honest ones here, and false precision is the dishonest one. Saying 50, 25 and 25 says something a reader can check against the stated reasoning: the base carries twice the weight of either side, and the two sides carry the same weight as each other. Saying 47, 26 and 27 says the same thing plus a claim that 26 can be told from 27, and nobody building a scenario set from three macro variables can tell 26 from 27. The extra digits are not extra information. The extra digits are an assertion about the sharpness of the author's own judgement, made silently, in a place where a reader is unlikely to challenge it.

The everyday version: a shopkeeper asked how busy tomorrow will be says "about the same as today, maybe a bit less". Asked to put a number on it he says half. He does not say 47 per cent. Two extra digits would set the listener wondering what the shopkeeper thinks he knows, and the number would be believed less rather than more.

Step six: what is the weighted figure, and what is it not?

Multiply each case output by its probability and add the three. Rs 100 crore at 50.00 per cent gives Rs 50 crore. Rs 130 crore at 25.00 per cent gives Rs 32.50 crore. Rs 60 crore at 25.00 per cent gives Rs 15 crore. The three add to Rs 97.50 crore.

Now the part that matters more than the arithmetic. Rs 97.50 crore sits below the base case and happens in none of the three cases, so it is a summary of the assumptions in the set and not an outcome anybody has described. The three cases put profit at Rs 100 crore, Rs 130 crore and Rs 60 crore. There is no fourth case. No case lands on Rs 97.50 crore, and no fourth set of values exists that would put profit there.

A reader who plans around Rs 97.50 crore has planned for a state of the world that no line of the set describes. A reader covering a shortfall has to cover the bear case at Rs 60 crore. The bear case sits Rs 40 crore below the base, not the Rs 2.50 crore that separates the weighted figure from the base. The weighted figure is Rs 37.50 crore away from the case that would hurt.

Step seven: what would move a probability?

Two or three named conditions per case, written now rather than later, each one a thing that could actually be observed. Not "if things get worse". Something a person could point at and say yes, that happened, or no, it did not. Each condition is written against the three variables the list already carries. A condition about something outside the list has nowhere to attach.

A condition that nobody could ever mark as met or not met is not a condition, and a scenario set full of them is a set whose probabilities can never be revised for a stated reason. That is why the conditions are written against the three variables the list already carries. Conditions written that way keep the whole set closed: three variables, three cases, three probabilities, and every condition pointing back at something already named.

When does the building stop?

When six things are written down and not before: the variable list, three complete value sets, three output figures, three probabilities that add to one hundred per cent, the weighted figure with its label attached, and the conditions for each case. The six written outputs are the finish line, and the finish line is a checklist rather than a conclusion.

A scenario set is not a prediction and cannot be turned into one by finishing it. The building does not stop with a prediction. At the end of step seven what is known is what three stated sets of assumptions do to one company's profit line, and how much weight rests on each set. Nobody knows what will happen, and nothing in the seven steps was ever going to settle it. Honest forecasting is a separate craft, worked through in Forecasting Honestly: Consensus, Range and Probability.

What makes a scenario set worth building rather than decorative?

The three cases differ only in the values of a fixed list, so a reader can see which variable is doing the work. Look back at the bull case: profit rose Rs 30 crore, and Rs 22.50 crore of that came from the oil line while the exchange rate contributed Rs 6 crore and the policy rate Rs 1.50 crore. The decomposition into three variables is readable only because nothing else moved. Change the list between cases and the decomposition disappears, and what is left is three numbers with a story attached.

A lender reads a set this way. The bear case at Rs 60 crore against an interest bill of Rs 19 crore in that same case tells the credit officer how thin the cover gets when all three variables move against the borrower at once, and the officer can then ask which of the three would have to move for that to happen. A household version of the same reading: the bad year in the three column sheet is not there to be believed, it is there to be sized. Somebody can then ask whether the emergency fund covers the gap between the ordinary year and the bad one.

A set is useful when a reader can take it apart. Taking a set apart needs every case built from the same list. A decorative set is one where the three cases each sound reasonable, each carries a number, and none of them can be traced back to a variable, which makes the whole thing an illustration of somebody's mood.

What breaks when one case carries a variable the others do not?