Korea, explained through data
A guide to reading data
The same number can tell different stories depending on its context. Start with these three principles, then explore the topics that interest you.
Three principles for reading numbers
First, what does the number count?
A publication date is not a reporting period. Check the region, whether a figure counts people or events, and whether it is cumulative or weekly. Comparing figures with different scopes or periods can change the conclusion.
Percent is not percentage points
If a hypothetical rate rises from 20% to 25%, the difference is 5 percentage points. The relative increase is (25 − 20) ÷ 20 × 100 = 25%. This formula cannot calculate a relative increase when the starting value is zero.
Moving together does not establish a cause
Falling album sales alongside a higher overseas ranking do not, by themselves, prove that fans changed how they spend. Sales and ranks measure different things. Separate observations from possible explanations and look for evidence that could test those explanations.
Work it out: a 75% rise, but a lower daily average?
Use fictional exhibition attendance counts to see how reporting periods affect a comparison. These numbers are made up for this example, not results from a real exhibition or work. One person may attend more than once, so attendances are not unique visitors.
| Period | Attendances | Days counted | Attendances per day |
|---|---|---|---|
| First period | 120 | 3 | 40 |
| Next period | 210 | 7 | 30 |
Comparing totals gives +75%
(210 − 120) ÷ 120 × 100 = 75%. Total attendance increased, but the reporting period also grew from 3 to 7 days. This alone does not establish a 75% rise in demand.
Comparing daily averages gives −25%
120 ÷ 3 = 40 attendances per day; 210 ÷ 7 = 30 per day. (30 − 40) ÷ 40 × 100 = −25%. Expressing both totals in the same daily unit changes the comparison.
A daily average is still not a final verdict. Weekend shares, opening hours, closure days, and an opening-day effect may differ. This table cannot explain why the average fell or how unique visitor numbers changed. Compare daily records with matching weekdays and opening hours.
Apply this to an article: what was counted → over which dates → whether the calculation uses comparable units → what remains unexplained.
Work it out: percentage points vs. a percentage change
A fictional survey asks the same question of two samples of 200 people each. These numbers are invented for this example, not results from a real survey.
| Wave | Yes answers | Respondents | Share |
|---|---|---|---|
| Before | 40 | 200 | 20% |
| After | 50 | 200 | 25% |
The share rose by 5 percentage points, or 25% in relative terms
50 − 40 = 10 more yes answers, out of 200 respondents in each wave. As a share, that is 25% − 20% = 5 percentage points. Measured relative to the starting share, (25 − 20) ÷ 20 × 100 = 25%. Both numbers describe the same change, but percentage points and a relative percentage change are not interchangeable.
Comparing shares like these does not require equal sample sizes — a 20% share among 200 respondents can still be compared with a 25% share from a different-sized sample. What matters is whether both waves asked about the same target population, used the same question wording and definition of a yes answer, covered a comparable period, and used a similar sampling method. Sample size instead affects how much statistical uncertainty surrounds each share, so these two fictional counts alone cannot establish a trend in the wider population. If the starting share had been 0%, the relative-change formula above would be undefined — but the percentage-point difference could still be reported.
Before trusting a reported rate change, ask not whether the exact same respondents were surveyed each time, but whether both waves measured a comparable population with the same question and method, and whether the change is reported in percentage points or as a relative percentage.
Work it out: a group difference is not proof of cause
Compare the newsletter sign-up rate of readers who attended a fictional event with readers who did not attend. These numbers are invented for this example, not results from a real event or newsletter.
| Group | Signed up | Group size | Share |
|---|---|---|---|
| Event participants | 30 | 100 | 30% |
| Non-participants | 20 | 100 | 20% |
Participants signed up at a rate 10 percentage points higher
30 ÷ 100 × 100 = 30% of participants signed up, versus 20 ÷ 100 × 100 = 20% of non-participants — a gap of 10 percentage points.
This gap does not show that the event caused the extra sign-ups. People who choose to attend an event may already be more interested in the topic than people who do not — this is self-selection, not an effect of the event. The comparison also assumes both groups were measured over the same period and were equally eligible to sign up, which this fictional example does not verify.
To support a causal claim, check whether the two groups were comparable before the event through a credible design, such as random assignment — simply comparing the sign-up rate of each group before the event does not rule out self-selection or other confounding differences between the groups.
Explore by topic
Korea by the numbers
Economic, social, and policy shifts, with the context and comparisons behind the figures.
- Foundation 3.4% of Korea's Export Payments Were Settled in Won — What Is That a Share Of?
- Foundation Welfare ₩148.8tn, National Debt ₩1,519.8tn — What Is the 48.3% Ratio Divided By?
- Application Korea's National Pension Fund: domestic stock target raised, estimated domestic share 26.3% on July 8
- Application North Korea's two-thirds of H1 crypto hacking losses — reading two counts without mixing them
Recent posts
Reading K-culture performance
What film, TV, and music rankings, sales, and viewing metrics can—and cannot—tell us.
Recent posts