Line graphs and change over time

Korea's total fertility rate fell every year from 1.24 in 2015 to 0.72 in 2023, then rose to 0.75 in 2024 and 0.80 in 2025.

When the same quantity is measured again and again over time, the result is a time series, and a line graph is its natural picture. Time runs along the horizontal axis from left to right, the measured value goes up the vertical axis, and the points are joined in order. OpenStax describes the gain plainly: once the points are drawn, trends that are hard to see in a table jump out.

The total fertility rate

The total fertility rate is the average number of children a woman would have between ages 15 and 49 if the birth rates of one year stayed the same for her whole life. It is measured in children per woman. The Korean government's e-Nara Index site, which publishes the series, gives 2.10 as the level at which a population replaces itself. Korea's rate was 4.53 in 1970 and dropped below that level, to 2.06, in 1983. The figures for 2015 to 2025, from the Ministry of Data and Statistics, are 1.24, 1.17, 1.05, 0.98, 0.92, 0.84, 0.81, 0.78, 0.72, 0.75 and 0.80.

Reading slope as speed

On a line graph the steepness of each segment shows how fast the value changed. From 2016 to 2017 the rate fell from 1.17 to 1.05, a drop of 0.12 in one year, the steepest segment in this series. From 2020 to 2021 it fell only 0.03. A segment that tilts upward, as from 2023 to 2024 (+0.03) and from 2024 to 2025 (+0.05), shows a rise. Over the whole stretch from 2015 to 2023 the rate dropped by 0.52, which is about 41.9% of its 2015 value. Reading slopes this way lets you describe a series in sentences like "the fall was fastest around 2017 and slowed after 2020".

What the line between points means

The rate is published once a year. The straight segment between 2019 and 2020 is not a measurement; it only joins two measurements so the eye can follow the order. That is fine for a smooth quantity, but do not read a value "halfway through 2019" off the line as if someone had measured it. For data that are not in time order at all, such as the four US regions, a line is the wrong choice, because it suggests a path from one category to the next.

Even steps of time

The horizontal axis must be a real time scale with even spacing. US census counts come every ten years: 151.3 million in 1950, 179.3 in 1960, 203.2 in 1970, 226.5 in 1980, 248.7 in 1990, 281.4 in 2000, 308.7 in 2010 and 331.4 in 2020 (resident population, April 1 of each year). If a chart placed 1950, 2000, 2010 and 2020 at equal distances, the half century from 1950 to 2000 would look like one decade and the growth rate would be badly misjudged. OpenStax warns against axes that change scale part way across, for example from years to months.

Two lines on one chart

Births and the fertility rate are in different units, so they cannot share one vertical axis. Some charts give each line its own axis, one on the left and one on the right, but then the designer can make the lines cross or run together just by choosing the two scales. A safer way is to index both series to a starting year. Set 2015 = 100 and the births become 100, 92.7, 81.6 and so on down to 52.5 in 2023 and 58.0 in 2025, while the fertility rate becomes 100 down to 58.1 in 2023 and 64.5 in 2025. Now both lines share one scale, and you can see that births fell a little faster than the rate.