Three Ways Inflation Happens
A rate, not a level
Inflation is not a price level. It is the rate of change in the price level over a period, usually a year. The level can keep rising forever while inflation falls, because a slower increase is still an increase.
Check it by hand. Suppose the price level goes 100, then 110, then 120. It rose 10 points in year one, which on a base of 100 is 10%. It rose 10 points in year two, which on a base of 110 is 9.09%. The line went up by the same amount both years, and the inflation rate came down. Do it once more: 120 to 130 is 8.33%. Same slope, lower rate, every year.
The left panel is the level — a straight line climbing at a steady slope. The right panel is that same line's rate of change, and it descends. A reader who conflates the two will call slowing inflation "prices falling", and be wrong. Get this first, and the rest of the subject stops being slippery, which is why the distinction between real vs nominal sits at the start of it.
Demand-pull: spending outruns production
The first cause is demand. If total spending grows faster than the economy's capacity to produce, buyers bid against each other for a fixed quantity of goods, and prices rise. Nothing about the goods changed. The money chasing them did.
The arithmetic is unforgiving. Suppose spending rises 8% and the economy can only produce 3% more. The gap, 5 percentage points, cannot become output, so it becomes higher prices. The mirror case is just as clean: 5% more spending with 5% more prices is 0% more real output.
This is the version that shows up when credit is cheap. When borrowing costs little, households and firms spend against future income today, and the spending lands before the supply does. It is also the version a central bank can fight directly, because the tool it holds — the price of credit — acts on exactly this: the willingness to spend borrowed money.
Cost-push: an input gets more expensive
The second cause does not come from the demand side at all. An input gets more expensive — energy, freight, labour, or a component that only one country makes — and firms pass the higher cost through to customers because they can.
Watch the pass-through by hand. If the input is 10% of your cost base and its price doubles, your unit cost rises by 10% of the base: the slice doubles from 10 to 20 points, which is 10 points on the whole. The firm that can pass it on does. The firm that cannot absorbs it in margin, and margins are how cost-push sorts winners from losers.
The cause matters because the response differs. A central bank raising the price of credit does little about a war, a shipping route or a cartel. It can only suppress the second-round wage and price response, and it does that by making the economy weaker. That is why labelling the cause is not academic: demand-pull and cost-push can look identical in the headline and need opposite treatments.
Expectations and the second round
The third cause is the most powerful, because it is self-referencing. Once people expect higher inflation, they act on it: workers ask for higher wages, and firms set higher prices in advance, both to protect themselves from the rise they anticipate. Those very actions deliver the rise. This is the second round, and it is what turns a shock into a regime.
Economists call this inflation expectations, and the feedback is the point. A one-off shock raises prices once. Expectations make the rise repeat, because each round of wage and price setting is written against the last one.
The schematic shows why the label matters. Demand-pull and cost-push are two separate paths into the same outcome; expectations is a third, and the arrow returning from the output back into expectations is the loop. Once that loop is live the original cause can disappear and the inflation carries on without it. That is the difference between a shock, which ends on its own, and a regime, which has to be broken.
Measuring it: CPI, PCE, and what core strips out
Two baskets get quoted constantly and they do not agree. CPI is the consumer's out-of-pocket basket: what households actually pay, weighted to the spending of an urban consumer. PCE covers what all households consume, including things paid for on their behalf — much of healthcare, for instance — so its weights differ and it usually reads a little cooler.
The weights do the work. Take two items, one up 10% and one down 2%. With equal weights the average is 5% minus 1%, which is 4%. Change the weights to 80 and 20 and the same two moves give 8% minus 0.4%, which is 7.6%. Same prices, different basket, different number. That is the whole CPI-versus-PCE argument in one line.
core inflation strips food and energy out of the basket. The reason is not that food and petrol do not matter — they matter more to a household than almost anything. It is that they are the most volatile lines, so they swamp the signal about the persistent trend. Headline inflation tells you what you paid. Core inflation tries to tell you where the trend is heading.
Base effects, and why disinflation is not deflation
Because inflation is a year-over-year comparison, a large move 12 months ago drops out of the arithmetic and changes today's headline without anything happening now. That is a base effect.
Watch it work. A year ago the index sat at 100 and then jumped 10% to 110. This month the index is still 110. The year-over-year rate is (110 minus 110) divided by 110, which is 0%. Nothing fell — the price level is exactly where it was — and yet the headline went from 10% to 0% because the old jump is now inside the base. The same trick works in reverse and manufactures inflation out of nothing.
Which brings the vocabulary. disinflation is a slowing in the rate of increase: prices still rising, just more slowly. Deflation is a falling price level. They are opposite situations and are routinely confused, because both produce a falling inflation number. A single month of disinflation is a base effect. A year of it is a trend. Look at the base before you believe the headline.
Related: CPI · PCE · Core Inflation · Inflation Expectations