A Rate, Not a Level

Inflation is a rate of change. That single fact accounts for most of the confusion in public argument about it, and it is worth sitting with before anything else.

When the rate falls from nine per cent to three, prices are not falling. They are rising more slowly. Everything that got more expensive is still expensive, and the shopping that cost a hundred pounds two years ago costs about a hundred and twelve today regardless of how encouraging the statistic is. Which is why a government announcing that it has halved inflation is so reliably met with disbelief: it is telling the truth about a derivative, and the electorate is looking at a level.

The second thing to get straight is why inflation is bad, because the obvious answer is wrong. It is not that things cost more: in a steady inflation, wages rise alongside prices, and a uniform doubling of every price and every income leaves everything real exactly as it was. The harm is not in the general rise.

It is in the unevenness, and above all in the surprise.

Contracts are written in nominal terms, against an expected path. A mortgage, a wage agreement, a pension, a bond: each embeds a guess about inflation. If the outturn is above the guess, borrowers repay in cheaper money and lenders lose, and the transfer bears no relation to anybody's productivity or to anything either party agreed. If it comes in below, the transfer runs the other way and is equally arbitrary. Unexpected inflation is a redistribution that nobody designed.

Fully anticipated inflation is comparatively cheap. There are menu costs (the real expense of changing prices) and shoe-leather costs, from the effort of holding less cash, and distortions where the tax system is not indexed. Real, and small, next to the redistribution and the loss of information when nobody can tell whether a price has risen because the thing is scarcer or because everything has.

Which is why stability of the rate matters more than its level, and why the argument about whether the target should be two per cent or four is much less important than whether people believe it.

The historical episode that established all of this is worth knowing, because it is the standard case of an empirical regularity destroyed by the attempt to exploit it. In 1958 A. W. Phillips documented an inverse relationship in British data between wage inflation and unemployment: lower unemployment came with higher inflation. This was read as a policy menu. A government could choose a point on the curve and buy lower unemployment with a little more inflation.

Milton Friedman and Edmund Phelps said, separately and before the event, that this would not survive being used. The trade-off, they argued, depends on the inflation being unexpected: workers accept a wage that turns out to be worth less than they thought, employers hire more, unemployment falls. Once the policy is anticipated, workers price it in, real wages return to where they were, and all that remains is the inflation.

The 1970s duly produced high inflation and high unemployment together, the combination the original curve said was impossible. The relationship was reformulated with expectations in it, and the episode became the founding example of what is now called the Lucas critique: a statistical relationship observed under one policy regime cannot be assumed to survive a change in the regime, because the relationship was partly an artefact of what people expected the policy to be.

What about the opposite direction? Falling prices sound like relief and are not. Deflation raises the real value of debts, squeezing borrowers exactly when incomes are falling. It rewards waiting, since anything you want will be cheaper next month, which reduces demand and reinforces the fall. Nominal wages are notoriously resistant to being cut, so the adjustment arrives as unemployment instead. And conventional interest rates cannot go far below zero, which means the central bank's main instrument loses traction precisely when it is needed most. Japan spent the better part of two decades demonstrating each of these in turn.

Finally, the mechanism, because "the central bank raised interest rates" describes something narrower than most people picture. The policy rate applies to a small set of transactions between the central bank and the banking system; the rate paid on reserves, the rate at which the bank lends. Everything else moves in response, because that rate anchors the alternative to any other short-term investment: mortgage rates, corporate borrowing costs, the exchange rate, the discount rate applied to every future cash flow and therefore every asset price.

The transmission is indirect, partial, and slow. The usual estimate is that the peak effect on inflation arrives somewhere between twelve and twenty-four months later, which means policy is always being set on a forecast of conditions that do not yet exist, using a tool whose effects will land after the situation has changed.

Which is the awkwardness in the most common modern case. A war disrupts energy supply, prices rise, output falls. This is a supply shock, and the central bank's instrument works by suppressing demand, so any response makes the output side worse while addressing the price side. The argument for acting anyway is about second-round effects: if a temporary jump becomes embedded in expectations and settles into wage agreements, removing it later costs far more. That is a judgement about credibility rather than a mechanical rule, and it is where most of the genuine disagreement among central bankers lives.