Inflation Expectations: Why What People Believe Matters

Inflation expectations are the working assumptions ordinary people carry about what prices will do next. Expectations matter because acting on an assumption helps bring it about: a shopkeeper who expects costs to climb prints a higher price now, and a supplier who expects the same asks for more in the contract. A belief reaches the price level through ordinary decisions, not through anything mysterious.

Two things sit underneath that sentence, and both come from earlier in these notes. The first is that a price is set by two sides deciding, one offering and one accepting, so anything that changes what either side decides changes the price. The second is that the general price level is nothing more grand than a weighted total of many such prices, so a change spread across enough of them shows up in the index. Put those together and everything that follows falls out: a belief is not a force acting on prices from outside, it is something that enters the ordinary decision a price is made of. Everything worked here runs on the Republic of Sankhya, an invented country used across these notes.

What is an inflation expectation, if it is not a published forecast?

Start by throwing out the usual picture, a document with a number printed on it. The picture is not wrong so much as it points at the wrong object. An inflation expectation, for the purpose of understanding prices, is the assumption about future prices sitting inside a decision that somebody is making right now. Nobody writes it down. Nobody publishes it. The assumption usually has no number attached in the mind of the person holding it, only a direction and a rough size, along the lines of things are going to be dearer next year so I had better allow for it.

Think about a woman who runs a tea stall outside a bus depot in Sankhya. In March she has to decide what to charge for a cup from April. She is not forecasting anything and would laugh at the word. She is remembering that milk went up twice last year, guessing that it will keep going up, and deciding whether to move from Rs 12/- to Rs 14/- now or wait and move later. Whatever she decides, an assumption about future prices has gone into a price that people will actually pay. An inflation expectation has just done its work, and it never appeared in any document.

The same object shows up in bigger decisions and gets more formal as it goes. A worker deciding what to ask for at a pay review is carrying one. A parts supplier signing a three year wage agreementA pay arrangement fixed in advance for a stated stretch of time, so the number written in it cannot be changed when conditions change. with its staff is carrying one, and has now written it into a document that binds it for three years. A wholesaler deciding whether to hold stock or sell it down is carrying one. The place to look for an expectation is never a report. A price list, a contract, a pay demand and a purchase decision are where it lives, and looking anywhere else is why the effect is so hard to see and so difficult to argue with.

How does a belief about prices actually reach a price?

An account that says beliefs matter and then waves at the rest has taught nothing. There are three ordinary routes, and each of them can be watched happening in a market street in an afternoon.

The first route is pricing ahead. A seller who reprints a price list, a menu or a rate card is committing to a number for some stretch of time, and during that stretch the costs behind the number will move. Any seller who has been caught once by that will build the expected rise into the number now rather than eat it later. The Sankhya tea stall going to Rs 14/- in April rather than Rs 13/- is doing exactly this, and the extra rupee is not greed, it is a cost that has not arrived yet being charged for in advance.

The second route is contracting ahead. Where two parties fix a rate for years rather than months, the expected rise gets negotiated into the rate. Neither side will knowingly agree to a number that only works if prices stand still. A three year supply agreement, a rental agreement with a stated annual step up, a pay settlement covering two years: each of them takes a belief about future prices and turns it into a legally binding number. Some contracts skip the guesswork by writing in indexationWriting a payment so it moves automatically with a published index, instead of the two sides having to agree a new number each time. instead. Indexation does not remove the expectation so much as hand it to a published series.

The third route is buying ahead. A buyer who expects a price to rise brings the purchase forward. A Sankhya household that normally buys its year of grain in November buys it in August instead; a builder who expects cement to be dearer after the monsoon stocks up before it. Nothing about the total quantity wanted has changed, but the timing has, and demand that arrives earlier meets the same supply and pushes the price up now. Not one of these three routes requires anybody to be irrational, and not one requires anybody to coordinate with anybody else. Each is a sensible individual decision taken alone, and that is precisely why the effect is so hard to talk anyone out of.

What happens when enough sellers price for the same number?

Here is where the subject turns from interesting to slippery, so it is worth working rather than asserting. Take the extreme case first because it is the clearest. Suppose every seller in the Sankhya core basketThe part of the consumer basket left over once food and energy are taken out of it. Core is a slice of the same basket, not a separate one. decides that prices will rise 6.00 per cent over the year ahead, and each one prices accordingly. Ask what the core index then does. The index rises 6.00 per cent. Nothing else is in it: the core index is what those sellers charged, added up with weights. There was no separate test the belief had to pass. The belief was the mechanism.

Now read the year end note somebody writes about it. Expected, 6.00 per cent. Outcome, 6.00 per cent. Difference, none. The natural sentence to write next is that the sellers read conditions well, and that sentence is the trap. The belief helped produce the very number that then appeared to confirm it, so being correct is not the same as understanding conditions, and nothing in the outcome can separate the two stories. A group that genuinely understood cost pressures and a group that simply decided on a number together would leave exactly the same trace in the data, which is no trace at all.

None of this is a reason to sneer at anyone who was right. The arithmetic is a reason to be careful about what being right proves. A belief checked against something the believers could not themselves move would teach something. When they could move it, and did, the check has quietly gone missing.

Can one Sankhya belief be traced all the way to an index reading?

One belief can be traced that far, and the arithmetic is short enough to do in the head once it has been seen. Everybody holding the same belief is a clean case but an unrealistic one, so the sellers divide instead. In the Sankhya year already recorded, the core part of the consumer basket rose 2.00 per cent. Every core seller pricing for 2.00 produces exactly that. The year ahead is a different case, one where opinion among core sellers has divided.

Sellers accounting for 60 per cent of the core basket by index weightHow much of a basket a category takes up. The bigger the share, the more that category's own price move counts towards the total. price for a 6.00 per cent rise. The remaining 40 per cent price for 2.00 per cent, as before. Neither share was counted anywhere. Both are assumed figures. The core index for that year is then the share weighted blend of what the two groups charged.

Read the arithmetic once slowly. Take 0.60 times 6.00 for 3.60, add 0.40 times 2.00 for 0.80, and the total is 4.40 per cent. Now do the interesting thing and swap the shares over. Put 40 per cent on the higher belief and 60 per cent on the lower one, and the same arithmetic gives 0.40 times 6.00 for 2.40, plus 0.60 times 2.00 for 1.20, a total of 3.60 per cent. Nothing about costs, harvests, wages or conditions changed between those two calculations. Only the shares of belief changed, and the index moved 0.80 points. One line of arithmetic carries the whole claim.

The blend also carries through to the headline, with everything else held still. Core carries half the consumer basket and rose 2.00 per cent, so in the recorded Sankhya year it contributed 1.00 point of the 6.70 per cent headline. Had core come out at 4.40 per cent instead, it would have contributed 2.20 points, and if food and energy had repeated their recorded contributions of 4.80 and 0.90 points exactly, the headline would have read 7.90 per cent rather than 6.70. The last step assumes food and energy repeat themselves precisely. Real components never repeat exactly, and holding them still is what lets the core arithmetic be seen on its own.

Push the slider to 100 and the panel shows the case worth remembering. Every seller prices for the higher number, the outcome equals that number exactly, and the group is perfectly accurate. Now pull it back to 55 and the larger group is still nearer than the smaller one, by a narrower margin. The group with more weight behind it is nearer at every setting above half. The arithmetic of a weighted average produces that ranking, not anybody's judgement. Watch the panel say so at every position, because that sentence is the whole reason accuracy proves so little here.

What does it mean for expectations to be anchored?

Anchoring is a word about the longer run, not about today. Expectations are described as anchored when a price shock arrives, moves the reading for the year in front of everybody, and leaves what people believe about the longer run sitting where it was. A Sankhya household watching onions triple after a bad harvest, shrugging, and still assuming that prices over the next five years will do roughly what they always did, is holding an anchored belief. The same household concluding that this is how things are now and that everything will keep climbing has an unanchored one.

Here is the part that catches people. The test of anchoring is not what people expect now. The test is whether the longer run expectation moves when a shock arrives. A stable reading through a quiet stretch therefore says almost nothing. During a calm period an anchored belief and an unanchored one look identical, because nothing has happened that could have separated them. The two beliefs are candidates that have not yet been asked the only question that distinguishes them. Reading calm as evidence of anchoring is like calling a bridge sound because no lorry has crossed it.

Why is a shifted belief harder to deal with than a one off price rise?

Because the two do completely different things to the years afterwards. A one off rise is an event: onions triple, the index jumps, and twelve months later the comparison is against the already high price, so the rise drops out of the reading on its own. Nobody has to do anything. Waiting works. Few problems have that property.

A shifted belief is not an event, it is a change to the assumption sitting inside every decision taken from that day forward. Every price list reprinted, every contract signed, every pay demand made now carries the new number, and each of those decisions creates a price that feeds the index that people look at when forming the next belief. A shock leaves once it has passed through. A shifted belief has to be dislodged rather than waited out, and separating the two is worth far more than it looks.

Say that back in household terms. A wedding in the extended household costs a lot this year and next year it does not, and the budget recovers by itself. A household that concludes from that wedding that everything is dearer now starts padding every single line of its monthly budget. The padding now sits in every decision rather than in one, so it will not unwind on its own.

How would anyone measure what people expect?

There are two broad answers in wide use, and both of them work. Neither is the winner. Every way of measuring an expectation carries a known fault, and a measurement whose fault stays hidden teaches false confidence instead.

The first is to ask. Put a question to a sample of households or of professional forecasters and count the answers. Asking produces a survey measureA reading built by asking a sample of people a question and counting their answers, so it records what they said rather than what they did., and its fault is exactly what the label says. A survey records what people said, and what people say is not always what they will do. People answer with the price of the thing they bought most recently, they round, they answer a slightly different question from the one they were asked, and the wording of the question changes the number that comes back. None of that makes the reading useless. The fault makes a survey a reading of stated belief rather than of acted belief.

The second is to read the belief out of prices. Where two otherwise similar instruments exist, one of which compensates the holder for price rises and one of which does not, the gap between what they pay is a break-even rateThe gap between what two otherwise similar instruments pay, where one compensates the holder for price rises and the other does not., and it looks like a clean market judgement about future prices. The gap is not clean. Inside that gap sits a risk premiumExtra return a holder asks for bearing something uncertain. The premium is mixed into a market price and never printed as a separate line., because a holder taking the uncertain side wants paying for it, and no method can pull the two apart with confidence. A gap of 4.00 points, to take an illustrative number, might be four points of expected price rise, or three and a half points plus half a point of compensation for uncertainty, and the market prints one number for both.

Why can nobody give this subject a clean causal story?

Because the arrow runs both ways and there is no first step. What people expect feeds into the prices they set now. The index counts those prices. The published index is one of the main things people look at when deciding what to expect next. Then round it goes again. Expectations move prices and prices move expectations, so no clean causal story exists to be had, and the loop is more useful stated plainly than dressed up as a chain.

The loop is not a hedge and not an admission of ignorance. The loop is a structural feature of the subject, and it has a practical consequence worth carrying away. Any sentence of the form prices rose because expectations rose is at best half of what happened, and the same is true of the sentence with the two halves swapped. The honest form is longer and less satisfying: expectations and prices moved together, each feeding the other, and the share belonging to either is not recoverable from the outcome.

What does a lender actually do with any of this?

A lender cannot observe what a borrower believes and does not try. A lender can read how long the borrower's contracts run on both sides, and a contract is where a belief about prices has already been written down and signed. A business whose selling prices are fixed for three years while its costs are not has taken a position on future prices whether it meant to or not, and the size of that position is readable straight off the two contract lengths.

Take a Sankhya bus operator on a municipal route. The fare is fixed by the route agreement at Rs 3,000/- a day, three hundred operating days a year, so revenue is Rs 9,00,000/- a year and cannot move. Diesel, tyres and wages start at Rs 2,400/- a day, or Rs 7,20,000/- a year, and are not fixed by anything. Run those costs forward at 6.00 per cent a year, the higher of the two beliefs used above, and watch what the fixed fare does to the margin.

The margin falls from Rs 1,80,000/- to Rs 42,468/- over three years, and the last cost figure is rounded to the nearest rupee. Nothing went wrong. Nobody was careless. The operator simply signed away the ability to reprice while keeping every cost floating, and a 6.00 per cent belief coming true is enough to take nearly four fifths of the margin. The contract length question therefore comes before any question about the borrower's opinion. Opinions are not lendable against, and a signed rate that runs for three years is. A lender reading this operator would want to know what happens in year four, whether the agreement carries any step up written into it, and what the operator would do if costs ran at 6.00 per cent rather than at the 2.00 per cent recorded in the year before.

The Republic of Sankhya is invented.
Educational material. Not advice on any investment, tax, budget or market position.