Data skills: reading charts and claims
Percentages and percentage points, cash and real terms, mean and median, axes, revisions and cause, using real figures.
Suits: Maths and statistics, citizenship, media literacy, science. Levels: ages 11–14, 14–16 and 16+. Rebuilt every night from the latest figures.
A week, 2025.
A week, in 2025 prices.
A mean, all employees.
Survey estimate.
Words to know
- Percentage change
- The change divided by the starting value, times 100.
- Percentage points
- The difference between two percentages.
- Real terms
- Adjusted for inflation.
- Mean and median
- Total over count, and middle value.
- Axis
- The scale along the side or bottom of a chart.
- Revision
- A change to a published figure when better data arrives.
Charts
£ a week. Typical weekly pay in cash terms. 2002 to 2025.
Source: ONS, Annual Survey of Hours and Earnings (resident analysis), via Nomis
£ a week. Typical weekly pay in today’s prices. 2002 to 2025.
Source: ONS (Annual Survey of Hours and Earnings and CPI), adjusted by UK Explained
millions of treatments. NHS waiting list, millions. July 2016 to July 2026.
Source: NHS England, Referral to Treatment (RTT) waiting times, England
Typical pay: cash and real terms
| Year | Cash terms | In today’s prices |
|---|---|---|
| 2025 | £767 | £767 |
| 2024 | £728 | £753 |
| 2023 | £687 | £728 |
| 2022 | £642 | £730 |
| 2021 | £610 | £756 |
| 2020 | £586 | £745 |
| 2019 | £585 | £751 |
| 2018 | £568 | £742 |
| 2017 | £550 | £736 |
| 2016 | £539 | £740 |
| 2015 | £527 | £729 |
| 2014 | £518 | £718 |
Real terms use the Consumer Prices Index.
Questions
Read the line chart of typical pay. What was the pay in 2002 and in 2025?
Answer. £391 in 2002 and £767 in 2025.
Before you read a chart, what three things should you check first?
Answer. The title (what is measured), the units (%, £, millions) and the period; then the source and date.
What is the first month and the last month on the waiting-list chart?
Answer. August 2007 and July 2026.
Percentage and percentage points: unemployment was 4.7% and is 4.9%. Give the change in both ways.
Answer. +0.2 percentage points, or +4.3% as a percentage change. They answer different questions.
Cash or real terms? Pay rose from £391 to £767 in cash terms. In prices of 2025 pay was £726 in 2002. Work out the real rise.
Answer. Cash rise: 96%. Real rise: (767 − 726) ÷ 726 × 100 = 6%. The real figure is smaller because prices rose too.
Mean or median? Average weekly earnings were £756 and typical full-time pay £767. Which is a mean and which a median, and why do they differ?
Answer. Average weekly earnings is a mean of all employees; typical full-time pay is a median of full-time workers. A mean is pulled up by a few very high earners and pulled down by part-timers; a median is the middle worker.
Compound growth: the house price rose from £268,843 to £272,611 over a year. Work out the percentage change.
Answer. (272,611 − 268,843) ÷ 268,843 × 100 = 1.4%.
The axis on the house price chart does not start at zero. How does that change how the chart looks, and when is it acceptable?
Answer. Starting above zero stretches the changes and makes them look bigger. It is acceptable for lines showing change over time if the axis is labelled, but a bar chart should always start at zero.
‘Waiting lists are shorter than last year, so the NHS is improving.’ Say what is wrong with that reasoning.
Answer. A fall in one measure does not show the whole picture: it may be due to a change in counting, fewer referrals, or treatments removed from the list. Check other measures, such as how long people wait, and whether the data are revised.
Why are the latest months of many datasets revised later, and how should that change what you say?
Answer. More information arrives after first publication, so first estimates can change. Say ‘provisional’ or ‘early estimate’ and avoid strong claims from one month.
Two lines on a chart move together. Does one cause the other? Use an example from the data.
Answer. Not necessarily. For example, prices and pay both tend to rise over time, but that does not show one causes the other. Look for a mechanism, other explanations and tests, such as changes at different times.
Choose a statistic from this site and write one sentence a journalist could use, with its source, period and one caveat.
Answer. Open question. A good answer names the figure, its unit and date, says who produced it, and includes what it cannot tell you, such as ‘an average that hides differences’.
Ideas for discussion
- Find a chart in a newspaper. What questions from this pack can you ask of it?
- Can a true number still mislead? How?
- What makes you trust a statistic?
Answer sheet: Data skills
Find. £391 in 2002 and £767 in 2025.
Explain. The title (what is measured), the units (%, £, millions) and the period; then the source and date.
Find. August 2007 and July 2026.
Calculate. +0.2 percentage points, or +4.3% as a percentage change. They answer different questions.
Calculate. Cash rise: 96%. Real rise: (767 − 726) ÷ 726 × 100 = 6%. The real figure is smaller because prices rose too.
Explain. Average weekly earnings is a mean of all employees; typical full-time pay is a median of full-time workers. A mean is pulled up by a few very high earners and pulled down by part-timers; a median is the middle worker.
Calculate. (272,611 − 268,843) ÷ 268,843 × 100 = 1.4%.
Explain. Starting above zero stretches the changes and makes them look bigger. It is acceptable for lines showing change over time if the axis is labelled, but a bar chart should always start at zero.
Judge. A fall in one measure does not show the whole picture: it may be due to a change in counting, fewer referrals, or treatments removed from the list. Check other measures, such as how long people wait, and whether the data are revised.
Explain. More information arrives after first publication, so first estimates can change. Say ‘provisional’ or ‘early estimate’ and avoid strong claims from one month.
Explain. Not necessarily. For example, prices and pay both tend to rise over time, but that does not show one causes the other. Look for a mechanism, other explanations and tests, such as changes at different times.
Judge. Open question. A good answer names the figure, its unit and date, says who produced it, and includes what it cannot tell you, such as ‘an average that hides differences’.
Sources: ONS, NHS England, HM Land Registry, HM Treasury. See also: Everything over time · How we calculate this · Words explained.