Every number has a denominator.
A spike in one year flatters the prints of the next; a collapse haunts them. The honest question about any inflation figure is a short one: compared to what?
The charts on this page draw the consumer price index on its 2012 base, the series that ran to December 2025. From January 2026 the NSO publishes the index on a new base, 2024 = 100. The idea on this page works the same on either ruler; the latest figures are in the Consumer prices guide.
The word, defined
First principlesEvery inflation figure you will ever read is a ratio. It divides what the basket cost in a month by what it cost in the same month a year earlier, subtracts one, and reports the answer as a percentage. Two numbers go in; one comes out. Most readers watch only the top of the ratio, the present. The bottom of it, the year-ago level, is the base, and it does half the work.
When the base was unusually high, the twelve prints that follow are flattered: ordinary prices measured against a spike look like relief. When the base was unusually low, those prints are haunted: ordinary prices measured against a collapse look like a surge. Neither requires anything to happen in the present at all.
The effect is strongest in food, because food is where the spikes live. A failed onion crop or a ruined tomato harvest lifts the level for a season; exactly a year later, that season becomes the base, and the prints turn gentle on their own. The gentleness lasts twelve months. Then the comparison window slides past the spike, and the kindness ends.
One spike, read two ways
Index level, y/y %The cleanest example in recent data sits in the food index. Below, three years drawn twice: the level above, the year-on-year rate below, with the same shaded months marked on both panels. Watch what those months do to each line.
The spike of 2024 became the calm of 2025: same months, two readings
The top line is what food cost, month by month. The bottom line is the same months, each measured against twelve months back. The shaded window is identical on both panels: above, it is the spike itself; below, it is where the spike sits inside the ratio.
Every reading on the bottom panel divides by a month from a year earlier. So the gentle readings of late 2025 are not only about 2025. Their denominators are the shaded months, the most expensive food India had recorded up to then.
In autumn 2024 food prices surged: the level climbed from – in July to a peak of – in October, the highest reading in this window. Twelve months later those months were the denominator. December 2025 printed –, the sixth negative food reading in seven months: not because food had become cheap, but because the comparison was with the most expensive winter on record.
A year earlier the same arithmetic had run the other way, with December 2024 printing – against an ordinary winter base. The spike made 2024 look worse than the present and 2025 look better. Both times, the base was doing part of the talking.
The arithmetic of the base
A worked example, Dec 2024 → Dec 2025Divide – by – and the result is –: the – of the print. Nothing in that division knows about harvests or gluts. It knows two levels and their ratio, and it is exactly as truthful as the month it divides by.
Now change only the denominator. Hold December 2025 where it is and compare it with December 2023, when the food level stood at –. The same month, the same level, prints –. One reading says food is getting cheaper; the other says it has been getting more expensive for two years. Both are correct arithmetic. The difference is entirely the base.
That is why the prints of a spike year, and of the year that follows it, deserve extra scepticism. The number is real. What it means depends on a month you have probably forgotten.
How to misread this concept
- i.
“The number fell, so things improved.”
Not necessarily. Food inflation slid from – to – in twelve months, and part of that slide is the denominator rising between the two readings. Ask what happened to the level: it fell – from its peak, which is real relief, but a smaller story than the prints suggest.
- ii.
“The base doesn't matter.”
Change nothing but the base and the same month prints – or –. A ratio has two inputs. Ignoring one of them is not scepticism about statistics; it is bad arithmetic.
- iii.
“A low print is always good news.”
A low print on a high base can mean prices are pausing at a peak. And a genuinely negative food print carries two stories at once: relief at the shop and distress on the farm. Read the level before you celebrate the rate.