The operations you actually use
A percentage is just a fraction of 100: 15% is 15/100 = 0.15. Every operation grows from that. The trick is to identify which question you are asking, because each one has its own formula. The table below gathers the six calculations that answer 99% of everyday cases, the same six the percentage calculator solves, with the steps in view. The one that trips people up most is no accident on this list: it is the percentage change, the engine behind "went up X%", "fell Y%", "returned Z%".
Δ% = (final − inicial) ÷ |inicial| × 100- Δ%
- the percentage change (the "by what percent" it moved).
- inicial
- the starting value, the base you measure against.
- final
- the ending value.
| Question | Formula | Example |
|---|---|---|
| What is X% of Y? | Y × X/100 | 15% of 80 = 12 |
| X is what % of Y? | X/Y × 100 | 45 of 180 = 25% |
| Increase Y by X% | Y × (1 + X/100) | 80 + 25% = 100 |
| Discount X% off Y | Y × (1 − X/100) | 250 − 20% = 200 |
| Change from A to B | (B − A)/|A| × 100 | 80 → 100 = +25% |
| Value before +X% | final / (1 + X/100) | 150 / 1.25 = 120 |
Use the tool below to try each of these calculations with your own numbers before moving on. It shows the substitution into the formula, not just the result, which helps cement which question you are answering.
Two questions that look alike
"What is X% of Y" and "X is what percent of Y" are inverses of each other, and swapping them is the most common mistake of all. In the first you know the percentage and want the value; in the second you have two values and want to find the percentage. The clue is what you are looking for: an absolute number (dollars, grams, questions) or a percent.
- 15% of 8080 × 15/100 = 80 × 0.15 = 12. A 15% tip on an 80 bill is 12.
- 45 is what % of 18045 / 180 × 100 = 0.25 × 100 = 25%. If you got 45 of 180 questions right, that is 25%.
Notice that the second calculation is a pure proportion, "45 is to 180 as X is to 100". It is the same reasoning as the rule of three, which is why the two topics go hand in hand. Whenever a question can be read as "so much is to so much", the rule of three solves it, and the percentage is just the special case where the second term is 100.
Increase, discount and the asymmetry of gains and losses
The smart way to apply an increase or a discount is to multiply by a factor, not to add in parts. Increasing by 25% means multiplying by 1.25; taking 20% off means multiplying by 0.80. This keeps the calculation direct and sets up the reverse operation: finding the original value when you only know the final price.
- Discount: 250 with 20% off250 × (1 − 0.20) = 250 × 0.80 = 200. You pay 200 and save 50.
- Increase: a 25% raise on 8080 × (1 + 0.25) = 80 × 1.25 = 100. The value rises from 80 to 100.
- Original: price with +25% already is 150150 / (1 + 0.25) = 150 / 1.25 = 120. The price before the raise was 120, not 112.50.
original = final ÷ (1 + a)
desconto que desfaz o aumento a = a ÷ (1 + a)
alta que recupera a queda d = d ÷ (1 − d)- a
- the increase applied, as a fraction (25% → 0.25).
- d
- the drop applied, as a fraction (50% → 0.50).
- final
- the value after the change.
This is the asymmetry that fools the beginner investor: an equal-sized drop and gain are not equivalent, because the gain is taken on a smaller base. The bigger the drop, the more brutal the recovery required, an 80% loss needs +400% just to get back to square one. The worked example below shows the arithmetic.
- Worked example: a stock falls 50%You buy at 100 and it drops 50%: 100 × (1 − 0.50) = 50. Now, to get from 50 back to 100, the required change is (100 − 50)/50 × 100 = 50/50 × 100 = 100%. So: it fell 50%, but needs to rise 100%, twice as much, just to break even.
The costliest trap: successive percentages
Here is the mistake that costs real money: applying +X% then −X% does NOT return the original value, and two discounts in a row do NOT add up. This happens because each percentage is taken on a different base. The increase is computed on the smaller value; the next discount, on the already-increased one. Watch the extreme case, +50% then −50%, in the worked example below.
- Worked example: +50% then −50%Start at 100. Up 50%: 100 × 1.50 = 150. Then down 50%: 150 × 0.50 = 75. It ended at 75, not 100, a net loss of 25%. The two "50%" do not cancel because the second is taken on 150, not on 100.
The loss is no coincidence: applying +p then −p (with p as a fraction) always multiplies by (1 + p)(1 − p) = 1 − p², that is, a shrinkage of p². With 20% (p = 0.2), p² = 0.04 = 4%. With 50% (p = 0.5), p² = 0.25 = 25% loss, exactly the 75 in the example. The same logic explains why "cumulative discounts" do not add: two 10% discounts give 0.90 × 0.90 = 0.81, i.e. 19%, not 20%; and a 20% followed by 30% gives 1 − (0.80 × 0.70) = 0.44, i.e. a 44% total discount, not 50%.
fator total = (1 − d₁) × (1 − d₂) × … × (1 − dₙ)- d₁, d₂, dₙ
- each successive discount, as a fraction (20% → 0.20).
- fator total
- the single equivalent multiplier; 1 − factor is the real total discount.
View the data
| x | Value |
|---|---|
| 0 | 100% |
| 10 | 99% |
| 20 | 96% |
| 30 | 91% |
| 40 | 84% |
| 50 | 75% |
| 60 | 64% |
| 70 | 51% |
| 80 | 36% |
| 90 | 19% |
A percentage point is not a percent
When something is already measured in percent, an interest rate, a market share, an approval rating, a dangerous ambiguity appears. If the benchmark rate goes from 10% to 12%, it rose 2 percentage points (the absolute difference), but it rose 20% in relative terms (2 is 20% of 10). Both are true about the same fact, and mixing them distorts the message. The press gets this wrong constantly: a tax going from 5% to 6% becomes a "1% increase" headline when, relatively, it was a 20% jump. Eurostat's statistical style guide separates the two terms precisely to avoid this error in official reporting.
- Percentage point (pp)
- The absolute difference between two percentages. From 10% to 12% is +2 pp. Used to compare rates.
- Percentage change
- The relative change between two values: (final − initial)/|initial| × 100. From 10% to 12% is +20%.
- Relative change
- Another name for percentage change, the difference measured against the base, not in absolute terms.
- Base
- The value a percentage is taken on. Silently switching the base is the root of nearly every percentage mistake.
Percentages and unit conversion are the two calculations where one detail changes everything. If your next problem involves measurements, temperature, distance, weight, see the sibling guide unit conversion without errors. And when the percentage is a return that piles up month after month, the topic becomes compound interest: the guide compound interest, contributions and inflation shows how the same rate behaves over time.
Averaging percentages and the paradox lurking
The fourth error is the subtlest: the average of percentages is not the percentage of the average. If one week you converted 1 visit out of 10 (10%) and the next 80 out of 100 (80%), the naive average of the two rates is (10% + 80%)/2 = 45%. But the real conversion for the period is (1 + 80)/(10 + 100) = 81/110 = 73.6%, much closer to 80%, because the second week had ten times more visits and weighs far more. You may only average percentages directly when the bases are equal; otherwise, sum the numerators and denominators separately. Ignoring this opens the door to Simpson's paradox, where a trend shows up in every group and reverses in the total. The table below sums up the four traps in this guide.
| Trap | Wrong example (intuition) | Correct result |
|---|---|---|
| Increase and discount cancel | +50% then −50% returns to original | 0.75× the original (loses 25%) |
| Percentage point = percent | 10% → 12% is a "2% increase" | +2 percentage points = +20% |
| Add percentages on different bases | 20% + 30% off = 50% | 1 − (0.80 × 0.70) = 44% |
| Average of percentages = percentage of average | average of 10% and 80% = 45% | 81/110 = 73.6% |
Why gains and losses are asymmetric
A percentage drop and a percentage gain of the same size do not balance, because the gain is taken on a smaller base. Falling d (fraction) then rising d leaves you at (1 − d)(1 + d) = 1 − d², always below the start. To truly recover, the required gain is d/(1 − d): −20% needs +25%, −50% needs +100%, −80% needs +400%. That is why protecting capital from big drops is worth more than chasing symmetric gains.
Logarithmic return: how to add changes over time
Simple percentage changes do not add over time, they multiply, because each period starts from the previous one's base. The trick is to switch to the logarithmic return: the natural log of (1 + r). Since the log of a product is the sum of the logs, the log returns of several periods simply add, and the total matches the product of the factors. Example: +50% then −50% gives ln(1.50) + ln(0.50) = 0.405 − 0.693 = −0.288; and e^(−0.288) = 0.75, exactly the 75% from the start of this guide. This is the property Attilio Meucci sums up as "compounded returns aggregate across time". You can work these logs out with the logarithm calculator.
Averaging percentages: when it is a trap
Taking the arithmetic mean of percentages is only honest when they all share the same base (the same number of cases). Conversion, hit, mortality or interest rates usually come from different bases, and there the plain average misleads. The right move is the average weighted by each base's size, or, equivalently, sum the numerators and divide by the sum of the denominators. That is what turned 45% (naive average) into 73.6% (real rate) in the section's example.
Simpson's paradox: the trend that reverses
In Simpson's paradox, described by E. H. Simpson in 1951, a relationship that holds in every subgroup reverses when everything is pooled. Constructed example: across two departments, group B has a higher admission rate than group A in both (80% vs 70% in one, 30% vs 20% in the other), but because B concentrates in the hard department, the total flips, A ends at 72/110 = 65.5% and B at 38/110 = 34.5%. The famous real case is admissions at the University of California, Berkeley (Bickel, Hammel and O'Connell, Science, 1975): aggregate figures suggested bias against women (about 44% of men admitted versus 30% of women), but department by department the bias vanished, women applied more to programs with few openings. The practical lesson: never pool percentages without looking at each group's base.
Frequently asked questions
Why does +20% then −20% not return the original value?
How do I calculate the percentage change between two values?
What is the difference between a percentage point and a percent?
Do a 20% and a 30% discount add up to 50%?
If a stock falls 50%, how much must it rise to recover?
Can I just average two percentages?
How do I add percentage changes over time?
Identify the question before the calculation: X% of Y is Y × X/100; "X is what % of Y" is X/Y × 100; increase and discount are multiplications by (1 ± X/100). And avoid the four errors: +X% then −X% does not return the original (it loses p²), a percentage point is not a percent, discounts on different bases do not add (they multiply), and the average of percentages is not the percentage of the average, to add changes over time, switch to the logarithmic return.
Sources & references
- Eurostat, Glossary: Percentage point (official definition and pp abbreviation)
- Eurostat, Percentage change and percentage points (statistical concept)
- Simpson, E. H. (1951). The Interpretation of Interaction in Contingency Tables. JRSS-B
- Bickel, Hammel & O'Connell (1975). Sex Bias in Graduate Admissions: Data from Berkeley. Science
- Meucci, A. (2010). Linear vs. Compounded Returns, Common Pitfalls in Portfolio Management. SSRN