Percent vs. Percentage Points: How to Read Statistical News
A change from 20% to 25% is a rise of 5 percentage points and a relative increase of 25%. Learn the two calculations and how to check denominators, periods, and populations.
The answer first
If a rate changes from 20% to 25%, the difference is 5 percentage points (pp), while the change relative to the starting value is a 25% increase. These describe the same change using different reference points. To ask how many percentage points a rate changed, subtract the two rates. To ask how many percent it changed from the start, divide that difference by the starting rate.
Official statistics use this distinction in practice. Eurostat reports that the share of people aged 25–34 in the EU with tertiary education rose from 42% in 2022 to 43% in 2023, an increase of 1 percentage point. This is an example from official statistics that illustrates the concept; it is separate from the worked examples below. Eurostat education statistics
The 20% and 25% in the figure are hypothetical examples created to explain the calculation; they are not results from an actual survey.
What do the two measures tell us?
A percentage (%) expresses how much of a whole a quantity represents, using 100 as the base. Calculate a rate as (part ÷ whole) × 100. For example, if 40 of 200 applicants are approved, the approval rate is 20%. Statistics Korea also defines a percentage as the share of a quantity in the whole. Statistics Korea, Statistical Terms 2020
A percentage point (pp) is the arithmetic difference between two values stated as percentages. 25% − 20% = 5 pp. Because the values are already percentages, use percentage points to describe their difference. The UK Office for National Statistics (ONS) style guide likewise distinguishes a fall from 10% to 9% (1 percentage point) from a 1% relative fall from 10%, which results in 9.9%. ONS style guide
By contrast, the relative percentage change uses the starting value as its reference.
Relative percentage change = (later value − starting value) ÷ starting value × 100
So a rise from 20% to 25% is (25−20) ÷ 20 × 100 = 25%. Saying only “up 5%” can be ambiguous: it might mean 5 percentage points or a 5% relative change. Be precise: say “up 5 percentage points” or “up 25% relative to the starting rate.”
The same difference can look different against a different base
If the starting rate is 20% and the later rate is 25%, the change is 25% − 20% = +5 pp. The relative change is (+5 pp ÷ 20%) × 100 = +25%. In counts, this is like going from 40 approvals among 200 applicants to 50 among 200. The approval rate rose by 5 percentage points, while the number approved increased by 10.
In the other direction, a change from 25% to 20% is 20% − 25% = −5 pp. The relative change is −5 pp ÷ 25% × 100 = −20%. In other words, the rate fell by 5 percentage points, or by 20% relative to its starting value. The relative changes are not symmetrical (+25% on the way up and −20% on the way down) because the starting value used as the denominator is different.
For another example, if an indicator rises from 4% to 6%, the difference is 2 pp and the relative change is 50%. It is also correct to say that it became 1.5 times as large. But “up 2%” can be mistaken for a relative change. The same percentage-point movement can look larger in relative terms when the starting rate is low. Conversely, a rise from 80% to 82% is also 2 pp, but the relative change is 2.5%. Do not infer the absolute size of a change from a large percentage increase in a headline; check both measures.
A relative percentage change is undefined when the starting value is zero. If a rate changes from 0% to 5%, you can say the difference is 5 pp, but (5−0) ÷ 0 cannot be calculated, so you cannot state a relative percentage increase. Calling a change from zero an “infinite increase” is also not justified by the usual relative-change calculation. Report the percentage-point difference and explain why the relative rate cannot be calculated.
First check that the rates mean the same thing
Subtraction and division are simple; whether the two values can be compared is a separate question. Before calculating a difference, check what was counted in the numerator, what was used as the denominator, and when and for whom the measure was collected.
Suppose one year’s figure is the share of all respondents who answered “yes,” while the next year’s denominator includes only valid responses. The difference may then reflect a change in how responses were handled. A rate for the whole population and one for a particular age group, or a rate for all households and one for households that responded to a survey, should not automatically be treated as the same indicator. Even if the indicator has the same name, check the notes and methodology for changes in its definition or coverage.
The periods must match too. A comparison with the same month a year earlier, one with the previous month, and one between annual averages answer different questions. Monthly and quarterly figures may cover different reference periods. The Australian Bureau of Statistics (ABS) notes that annual movements in monthly CPI and quarterly CPI can differ because they measure different periods. ABS CPI FAQs
In a sample survey, a rate is an estimate. If the sample is small or sampling error is large, a numerical difference may be due to chance variation. Check the survey method and sample size, along with any confidence interval or information about uncertainty. A movement of several percentage points alone does not prove that a change is certain or reveal an important cause. Conversely, the absence of a reported significance test does not prove that there was no change. An “observed difference,” a “statistically significant difference,” and a “policy-relevant difference” are different questions. The ONS guidance on uncertainty in survey estimates explains sampling error and confidence intervals.
Remember rounding too. A report may display 20% and 25% even though the original figures were rounded from values with decimal places. The difference calculated from the displayed values may therefore differ slightly from a difference calculated and reported using the underlying figures. If Eurostat or another statistical office gives a change alongside a rounded table, a mismatch with a subtraction from the displayed values is not automatically an error. Check the units, decimal places, and notes.
A checklist for reading an article
When you see a rate change in an article, ask these questions in order:
**What is the rate measuring?** What people or events are in the numerator, and who is included in the denominator?
**Are the definitions the same?** Did the population, survey method, or inclusion and exclusion rules change?
**Which periods are being compared?** The same period last year, the previous period, or annual averages?
**What is the difference in percentage points?** Subtract the starting rate from the later rate.
**What is the relative percentage change?** Divide the difference by the starting rate. If the starting value is zero, mark the calculation as undefined.
**How reliable is the difference?** Check the sample size, error, confidence interval, rounding, and notes about method changes.
**Does the article claim more than the data show?** A rate change is an observed result. Claims about causes or policy effects need separate evidence.
Finally, try rewriting the sentence yourself. If you see “approval rose 5% from 20% to 25%,” you can restate it as “approval rose by 5 percentage points, or 25% relative to the starting rate.” This quick rephrasing reveals whether the headline emphasizes the difference or the relative change.
Remember: A percentage is a share of a whole; percentage points measure the difference between two rates; relative percentage change measures the change against the starting value. When comparing rates, check the denominator, period, population, and uncertainty as well as the arithmetic.