Choropleth Classification: Quantile, Equal Interval, Jenks, Std Dev

Why classification matters

A choropleth groups continuous values into a handful of classes, usually about five, and gives each class a colour. Where you put the breaks between classes determines which regions look similar and which look different. Two maps of the same data with different break rules can support opposite readings, so the method is a real analytical choice, not a cosmetic default.

Quantile (equal count)

Quantile classification puts an equal number of regions in each class. With five classes and 50 regions, each colour gets ten regions. Every colour is used and the map always looks balanced and full.

• Good for: ordinal comparison and ranking, and skewed data where you still want every class populated.
• Watch out: class widths vary wildly, so two regions in adjacent classes can be almost identical while two in the same class can be far apart. It can imply differences that are not really there.

Equal interval

Equal interval divides the value range into bands of equal width. If values run from 0 to 100 with five classes, the breaks are at 20, 40, 60, and 80.

• Good for: data spread fairly evenly across its range, and audiences who need to read magnitude directly from the legend.
• Watch out: with skewed data, most regions crowd into one or two classes and the map looks nearly uniform. One outlier can stretch the whole scale.

Natural breaks (Jenks)

The Jenks natural breaks method looks for the groupings that already exist in the data. It sets breaks so that variation within each class is as small as possible and variation between classes is as large as possible, which tends to follow the natural clusters and gaps in the distribution.

• Good for: real, lumpy data such as population, income, or yield, where it usually produces the most faithful-looking map.
• Watch out: the breaks are specific to this dataset, so two maps classified with Jenks are not directly comparable. For a time series or a small-multiple, fix the breaks manually instead.

Standard deviation

Standard deviation classification shades by how far each value sits from the mean, in units of the standard deviation. The middle class straddles the average and classes fan out above and below it, often with a diverging colour scheme.

• Good for: when the question is who is above or below average, or where the anomalies are, rather than the raw level.
• Watch out: it assumes a roughly symmetric distribution. On strongly skewed data the mean is a poor centre and the classes mislead.

How many classes, and how to compare

• Keep to about five classes. The eye struggles to tell more than that apart, so extra classes add false precision.
• Try quantile and Jenks on the same variable and look at what moves. Regions that change class between them are the ambiguous ones.
• For maps that must be compared to each other (over time, or across variables), set the same manual breaks on all of them.
• Pair the method with the right colour ramp: sequential for low-to-high, diverging only when there is a true midpoint such as the mean.

You can flip between all four methods live and watch the map change in Groundtruth. Load a layer, pick a variable, and click through Quantile, Equal, Jenks, and Std dev to feel how much the classification carries. The full mapping workflow is in how to make a choropleth map.