Why percentage change isn't percentage difference (and when to use each)

From 80 to 100 is a 25% increase. From 100 to 80 is a 20% decrease. Same physical change, different percentages. The first is “percentage change.” The second is also percentage change but in the opposite direction. And the symmetric “percentage difference” between them — a third number, 22.2% — is something else entirely.

This is one of math’s most reliable sources of confusion. Let’s straighten it out.

Percentage change (directional)

change% = (new − old) / old × 100

Percentage change uses the starting value (old) as the base. It’s signed: positive for increases, negative for decreases. It’s the answer to “by what percent did this grow/shrink?”

Examples:

• Sales went from $80K to $100K: change = (100 − 80) / 80 × 100 = +25%
• Then sales went from $100K back to $80K: change = (80 − 100) / 100 × 100 = −20%

The two percentages aren’t equal even though the physical change is the same. That’s because the base changed.

Percentage difference (symmetric)

difference% = |a − b| / ((a + b) / 2) × 100

Percentage difference uses the averageof the two values as the base, and takes the absolute value of the difference. It’s always positive and symmetric — the difference between 80 and 100 equals the difference between 100 and 80, both 22.2%.

It’s the answer to “how far apart are these two values, without saying which one is the reference?”

The asymmetry that confuses people

Percentage change is the unit most familiar to most people (“the stock went up 5% today”). Percentage difference is the one statisticians and physicists use for comparison of measurements where neither value is privileged. News reporting almost always means percentage change but sometimes uses the word “difference” loosely.

Worth memorising:

• Percentage change: signed, base = starting value.
• Percentage difference: unsigned, base = average of the two values.
• They’re only equal when one of the values is zero (degenerate).

The mistake that compounds

If something rises 50% and then falls 50%, where does it end up? Most people’s gut says “back where it started.”

Not even close. 100 → 150 (up 50%) → 75 (down 50%). You lost 25% of the original.

Why: the second 50% was applied to the new base (150), not the original (100). 50% of 150 is 75, leaving 75. To get from 150 back to 100 you’d need a 33.3% decrease, not 50%.

This is why investments take longer to recover than to lose: losing 50% requires a 100% gain to recover; losing 80% requires a 400% gain. Underlies a lot of finance writing that sounds obvious until you do the arithmetic.

When to use each

Percentage change

• Year-over-year revenue, profit, traffic, anything with a clear “before” and “after.”
• Stock returns, investment performance, salary increases.
• Population growth (where the previous census is the base).
• Inflation rate (current price vs same period last year).

Percentage difference

• Comparing two measurements where neither is the reference (two lab tests, two surveys).
• Quality control between two production batches.
• Comparing two competitors, two cities, two products — when the question is “how different are they?” not “by what factor did A grow into B?”

Worked example: a sales report that confuses two execs

Q1 sales: $80,000. Q2 sales: $100,000. The CFO reports “Q2 was 25% higher than Q1.” The COO emails back: “Then Q1 was 25% lower than Q2.” Both feel intuitive; one of them is wrong.

The CFO is computing percentage change with Q1 as the base: (100 − 80) / 80 = 25%. Correct.

The COO is reusing the 25% in the other direction. The actual percentage change from Q2 ($100K) back to Q1 ($80K) uses Q2 as the base: (80 − 100) / 100 = −20%. Q1 was 20% lower than Q2, not 25%.

If both executives wanted a single symmetric number — the magnitude of the gap, base-agnostic — the percentage difference is |100 − 80| / ((100 + 80) / 2) = 20 / 90 = 22.2%. That number is neither 25% nor 20%; it’s the only one that reads the same whichever quarter you call “before.”

Picking the right one depends on the audience. Boards usually want directional percentage change (Q2 up 25% on Q1) because growth narrative needs a direction. Cross-team benchmark reports comparing two products or two regions usually want percentage difference because neither side is privileged as “the base.”

Edge cases that trip people up

• Zero in the denominator.Going from $0 revenue to $10K isn’t a “∞%” increase in any useful sense; the conventional move is to report the absolute dollar change instead and note “from a base of zero.” Percentage difference also breaks ((10 − 0) / 5 = 200%, which is mathematically defined but practically useless).
• Negative values.If your “before” is a $5K loss and “after” is a $5K profit, the percentage change formula yields (5 − (−5)) / −5 = −200%, which suggests the business got worse. It didn’t. For sign-flipping series, report absolute change, not percentage change.
• Percentage points vs percent. Interest rate going from 5% to 6% is a 1 percentage point increase, but a 20% relative increase. Both readings are valid; mixing them up is the most common source of newsroom errors. See our percentage vs percentage point guide for the deeper treatment.
• Compounding direction.“Up 10% then down 10%” lands at 99%, not 100% (1.10 × 0.90 = 0.99). The asymmetry is small per step and compounds surprisingly fast over many steps — month-on-month economic series alternating ±2% lose ~0.04% per pair to this effect, ~0.24% over a year.
• Averaging percentages. If a stock returns +50% in year 1 and −50% in year 2, the simple average is 0%. The actual return is −25% (1.5 × 0.5 = 0.75). Use the geometric mean, not the arithmetic mean, for multi-period returns.

When percentage change does NOT apply

Three scenarios where the formula is technically defined but the answer is misleading:

• Bounded ratios (e.g. shares, rates). If unemployment goes from 4% to 5%, “up 25%” is mathematically correct but rhetorically loaded. Most economists report “up 1 percentage point” for bounded quantities. AP Stylebook explicitly codifies this distinction.
• Logarithmic or exponential quantities. A 6.0 vs 7.0 earthquake on the Richter scale isn’t a 17% increase; the scale is logarithmic and a 7.0 releases 10× the energy of a 6.0. Decibels, pH, and stellar magnitudes have the same property.
• Surveys with small denominators. “Support up 50%” from a 4-person to a 6-person sample is statistically meaningless. Always report the underlying counts alongside any percentage change.

Where our calculator stands

Convertitive’s percentage calculatorcomputes percentage change (the more common operation) via its third tab. We don’t compute percentage difference because it’s rarely what people are actually after — when they think they want it, they usually want change, with absolute-value applied. If your job genuinely needs percentage difference, use the formula above directly — it’s simple enough not to need a tool.

The newsroom rule of thumb

AP Stylebook and Reuters both codify a simple rule: report percentage change with one decimal place for values under 10%, no decimal for values 10-100%, and always cite the underlying numbers when reporting a percentage on a survey or sample. Examples from house styles:

• “Inflation rose to 3.2%” — one decimal because under 10%.
• “Tourism is up 47% year-on-year” — no decimal, large value.
• “Approval fell 4 percentage points” — points, not percent.
• “Sales doubled (up 100%)” — “doubled” is shorthand for +100%, “tripled” for +200%, “halved” for −50%.

The deeper editorial discipline: never report a percentage change without exposing the absolute values underneath, at least once in the piece. “A 50% increase from 4 to 6 cases” reads differently from “a 50% increase from 4,000 to 6,000 cases.” Percentages without denominators are the single most common source of statistical misreading in journalism.

Side-by-side: the three answers for the same pair of numbers

For two values a and b, four distinct “percentage” answers are mathematically valid, and they disagree:

The log-return form deserves a mention: it’s signed (unlike symmetric percentage difference) but its magnitude is the same regardless of direction (unlike percentage change). Quantitative finance uses log returns almost exclusively for multi-period series because they add cleanly (the log return over N periods is the sum of the per-period log returns; the simple arithmetic return is not). For one-period reporting to a non-technical audience, percentage change remains the right choice. For internal modelling, log returns are usually better.

Convention summary: news headlines and finance use percentage change. Statistics and physics use percentage difference. Quant trading uses log returns. Mix them deliberately, not accidentally.