A percentage point is a unit for expressing the arithmetic difference between two percentages. A rise from 20% to 25% is an increase of 5 percentage points, but a 25% increase relative to the original value. The distinction is important whenever percentages describe rates, proportions, or shares: percentage points measure the difference on the percentage scale, whereas percentage change measures that difference relative to a specified starting value.
Definition and calculation
If two values are written as and , the signed difference in percentage points is
For example, a decrease from 12% to 9% gives
meaning a fall of 3 percentage points.
If the same values are instead represented as proportions and , with 20% written as , the numerical difference in percentage points is
Thus, one percentage point corresponds to a difference of in a proportion. It is not an additional quantity beyond a percentage; it identifies the scale on which a difference is being reported.
The abbreviations pp, p.p., and ppt occur in tables and technical writing. Spelling out “percentage points” avoids ambiguity, particularly because “ppt” can also mean parts per thousand or parts per trillion.
Percentage points versus percentage change
A relative percentage change divides the change by the original value. For a nonzero starting percentage , its numerical value is
and the result is reported as .
| Initial value | Final value | Percentage-point change | Relative percentage change |
|---|---|---|---|
| 20% | 25% | Increase of 5 percentage points | Increase of 25% |
| 10% | 15% | Increase of 5 percentage points | Increase of 50% |
| 40% | 30% | Decrease of 10 percentage points | Decrease of 25% |
| 0% | 5% | Increase of 5 percentage points | Undefined using the usual formula |
The first two examples have the same percentage-point increase but different relative increases because their starting values differ.
Direction also matters. Moving from 20% to 25% is a 25% relative increase, while moving back from 25% to 20% is a 20% relative decrease. The percentage-point changes are equal in magnitude and opposite in sign.
The wording “increased by 5%” therefore does not mean the same thing as “increased by 5 percentage points.” Applied literally to an initial rate of 20%, a 5% relative increase produces 21%; a 5-percentage-point increase produces 25%.
Relationship to basis points
In finance, small differences between percentage rates are often expressed in basis points. One basis point equals one hundredth of a percentage point:
An interest rate rising from 4.00% to 4.25% has increased by:
- 0.25 percentage points;
- 25 basis points;
- 6.25% relative to its original level.
These describe the same movement on different scales. Basis points are especially useful for discussing interest rates, yields, and financial spreads without confusing changes in rates with relative changes.
Applications
Percentage points are widely used in statistics and economics to compare quantities already expressed as percentages.
- Economic indicators: An inflation rate falling from 6% to 4% has declined by 2 percentage points. This describes a lower rate of price increase, not necessarily a fall in prices.
- Population and survey shares: A group's share rising from 30% to 35% represents an increase of 5 percentage points. The change in the number of people also depends on the population totals.
- Probability comparisons: A probability changing from 0.10 to 0.15 rises by 0.05 on the probability scale, or 5 percentage points, corresponding to a 50% relative increase.
- Model performance: Classification accuracy rising from 90% to 92% improves by 2 percentage points. The error rate falls from 10% to 8%, a 20% relative reduction.
The choice between percentage points and relative percentages depends on the comparison being made. Reporting both can clarify the size of a change.
Interpretation and limitations
Percentage-point arithmetic does not establish that two percentages are substantively comparable. Their definitions, populations, denominators, and measurement periods must also be considered. Identically sized percentage-point changes can represent very different changes in underlying counts.
Rounding can affect reported differences. Subtracting two displayed percentages may give a slightly different result from subtracting the underlying unrounded values.
A percentage-point difference also says nothing by itself about statistical precision or causation. For estimated percentages, a confidence interval may express uncertainty in the difference, also in percentage points. Establishing whether a difference reflects a causal effect requires evidence beyond the arithmetic comparison.