Both of these are true of the same two years. £100 in 1970 needs £1,609 today to buy the same things; £100 today would have needed only £6.21 in 1970. They are the same division turned over, and they answer genuinely different questions — so quoting one when you meant the other is the most common way to be wrong by a factor of 259.
Divide 72 by the inflation rate and you have the years until prices double. At 3% that is 24 years, and the exact answer is 23.4. At 7% it says 10.3 against a true 10.2. It is close enough to do in your head, and it turns an abstract percentage into a date.
This converts the cost of a basket, and nothing else. A salary from 1970 run through a price index tells you what it would buy, not what the equivalent job pays now — earnings and prices drift apart over decades, sometimes sharply. Same for house prices, which follow neither. Use a price index for costs and an earnings index for pay.
The further back you go the coarser the source becomes: before 1950 the UK figures are a long-run composite chained onto CPI rather than CPI itself, and the page tells you when you are in that territory. Two significant figures is about all such an answer can carry. Quoting a Victorian conversion to the penny is precision that the data has never had.
A price index records what a fixed basket of goods cost in each year. To move money between two
years you scale by the ratio of the two index values, and that is the entire calculation:
amount × index(to) ÷ index(from). Everything difficult about inflation is in the
index, not the arithmetic.
UK CPI stood at 8.6 in 1970 and 138.4 in 2025. So £100 in 1970 needed
100 × 138.4 ÷ 8.6 = £1,609 by 2025 to buy the same things. Put the other
way, a pound in 1970 did the work of about £16.09 today.
UK prices rose 242% during the 1970s alone. A saver holding cash through that decade lost roughly three quarters of their purchasing power without a single figure on their statement going down. That is the case this tool exists for.
US consumer prices fell between 1920 and 1940 — the index went from 20.0 to 14.0. $100 in 1920 needed only $70 in 1940 for the same basket. Any tool that assumes money always loses value gets the interwar years exactly backwards.
Consumer price inflation — UK CPI (ONS) and US CPI-U (BLS), both as annual averages. Annual figures answer "what was this worth in 1985"; they cannot answer "on 14 March 1985", and a year in which prices moved sharply will not be well described by its own average. The UK series from 1950 to 1988 is the modelled CPI-consistent historical series, because CPI itself does not go back that far; before 1950 even that runs out, and the figures come from the ONS long-run composite price index chained onto CPI at 1950 — the same join the ONS uses. The composite is a coarser, broader measure than CPI, so a Victorian or Edwardian answer here is the right order of magnitude rather than a precise one, and every year taken from it says so on screen.
Because nobody buys the basket. CPI is an average across everything a representative household spends on, and the components diverge enormously: over the last few decades housing, energy and education have risen far faster than the average while clothing and electronics have fallen in real terms. If your spending is concentrated in the fast-moving categories, your personal inflation rate has been higher than the headline for years, and this tool will understate what you have actually experienced.
Not very, in the early series, and the page marks them estimated for that reason.
The index is stored every five years before 1950 and every year from 2015, and anything
in between is filled in assuming inflation ran at one steady rate across the gap. Real
inflation does not oblige. You can measure the cost of that assumption using the page's
own data — drop a stored year, rebuild it from its neighbours, and compare: the
modern annually-anchored years come back within 1%, but 1920 comes back 43%
low and 1970 23% high, because a smooth fill cannot follow a path that spiked
or reversed inside the gap. Those rebuilds span ten years where the tool spans five, so
they overstate the real error — but the shape is right, and it is why an interwar
answer should be read as an order of magnitude rather than a figure. Verified in
inflation_model.py.
It stops being a measurement and becomes an assumption. Beyond the last published year the tool compounds forward at whatever rate you set, defaulting to the 2% both the Bank of England and the Federal Reserve target. The result is clearly marked as projected. Nobody knows next year's inflation, and a tool that presents a forecast in the same typeface as a measurement is lying about how much it knows.
No. Use the source directly. Index-linked contracts, pension revaluation and compensation awards usually specify a named index, a named month and a published table, and the answer has to come from that table rather than from an annual average on somebody's website. This is for understanding, not for settling anything.
Annual average consumer price indices for the UK and US only, with a coarse set of anchor years before 2018 and geometric interpolation between them — a year that is not an anchor is an estimate and the page says so. Not modelled: monthly resolution; RPI, RPIX, CPIH or any index other than the headline consumer one; regional variation within a country; other currencies; exchange-rate movement, so a cross-currency comparison over time is not possible here; asset prices, house prices and wages, none of which follow consumer inflation; tax changes; quality adjustment and substitution effects within the basket itself; the current calendar year as a whole, since it is not finished; and any period before 1900 for the UK or 1913 for the US. UK years before 1950 are not CPI at all — they come from the ONS long-run composite price index chained onto CPI at 1950, which is what the ONS does itself, and they are coarser than anything after them.
Releases in which this page changed, newest last. Derived from the archived copy of every release, not from notes written afterwards — so it reflects what actually shipped. Site-wide passes are left out; they are in the full changelog.