Choropleth maps can convey absolute magnitude through the range of the accompanying colour legend

Figures & data

Figure 1. A choropleth map displaying data from an analysis of state-level public support for a federal ban on abortion in the U.S (Fischer and Ali 2021). The colour legend employs a diverging blue-red colour palette, with white in the centre, showing the full range of possible values. The 30% point is marked with a dotted line and labelled to indicate that no state exceeds this level of support. Reproduced with permission.

A choropleth map of the U.S. with the title ‘There is not a single state where support for a federal ban on abortion has more than 30 percent support among the public’. The color legend employs a diverging blue-red color palette, with white in the center, showing the full range of possible values. Each state is colored one of several shades of light blue. The 30 percent point is marked with a dotted line and labeled to indicate that no state exceeds this level of support.

Figure 2. Example stimuli: six choropleth maps showing fictitious pollution data. Four colour legends are displayed below each map, but only one colour legend accompanied the map in each trial. Colour legends with extended ranges have a maximum value equal to double the maximum plotted value (top row: 400; bottom row: 1800). Colour legends with truncated ranges have a maximum value equal to the maximum plotted value in the map (top row: 200; bottom row: 900). During the experiment, all six colour scales were used in conjunction with all maximum values.

A 3x2 grid of choropleth map visualizations, in six different colors. Each has four color legends below. The color legends below the maps in the left column terminate with the same shade of color as the darkest geographical region. The color legends below the maps in the right column terminate with a much darker shade of color. Each color legend is shown with and without numerical labels.

Table 1. Demographic Information.

Sample	Gender			Age (years)		Graph Literacy		Education
	Male (%)	Female (%)	Prefer not to say (%)	Mean	SD	Mean	SD	High school or above (%)
N = 100	59.0	40.0	1.0	30.8	8.8	21.6	4.5	98.0
N = 165	53.9	45.5	0.6	31.8	10.1	21.8	4.5	98.8

Figure 3. An example of a single experiment trial, showing a choropleth map with a truncated colour legend, plus a response marker on the visual analogue scale.

Read the detailed description of this figure

A screenshot with a choropleth map on a white square, against a grey background. The choropleth map shows four shades of red, and the horizontal colour legend starts with a very light pink, and extends to a red, equivalent to the darkest colour in the map. The text above the map reads ‘This map shows the levels of a certain type of pollution, in four geographic regions:’ and the text below reads ‘How urgently should pollution levels in these regions be addressed?’. Below this question is a sliding scale with ‘Not very urgently’ as the left anchor and ‘Very urgently’ as the right anchor. A triangle marker close to the rightmost side of the scale. The text at the very bottom reads ‘Press the spacebar to continue when you have made your response’.

Table 2. CIELab values for colour legends’ start and end colours.

Colour scale	Range	Start colour			End colour
		L*	a*	b*	L*	a*	b*
Reds	Truncated	97.17	2.50	3.58	62.89	53.29	45.11
Reds	Extended	97.17	2.50	3.58	19.69	41.49	24.73
Greys	Truncated	100.00	0.01	−0.01	62.31	0.01	−0.01
Greys	Extended	100.00	0.01	−0.01	0.00	0.00	0.00
Purples	Truncated	98.74	0.69	−0.85	65.59	11.10	−22.97
Purples	Extended	98.74	0.69	−0.85	18.09	49.37	−54.13
Blues	Truncated	98.43	−0.59	−2.38	68.37	−10.35	−26.64
Blues	Extended	98.43	−0.59	−2.38	20.93	11.95	−38.06
Greens	Truncated	98.43	−2.87	2.76	72.70	−40.34	31.54
Greens	Extended	98.43	−2.87	2.76	24.36	−30.10	19.31
Oranges	Truncated	97.05	1.68	5.99	69.77	36.47	59.19
Oranges	Extended	97.05	1.68	5.99	29.34	36.61	39.62

Figure 4. Visual analogue scale responses to the question ‘How urgently should pollution levels in these regions be addressed?’. Distributions for the two conditions are shown using histograms, boxplots, and raw data points representing individual observations. In the ‘Extended Range’ condition, the colour legend’s upper bound was equal to double the maximum plotted value. In the ‘Truncated Range’ condition, the colour legend’s upper bound was equal to the maximum plotted value.

A graphic showing distributions of responses for ‘Extended Range’ and ‘Truncated Range’. Distributions are shown separately for each condition using histograms, boxplots and circles representing individual data points. On the left side is the label "Not very urgently" and on the right "Very urgently". The top histogram, with the label "Extended Range" resembles a Gaussian curve, with its peak roughly in the middle of the axis. The boxplot and raw data also show observations clustered around the middle of the axis. The bottom histogram, with the label "Truncated Range" peaks just before the right hand side of the axis. The boxplot and raw data also show observations are heavily left-skewed.

Figure 5. Mean urgency ratings showing the interaction between colour legend range and numerical label presence, displayed separately for the different samples of participants. Error bars show 95% confidence intervals around the means.

A visualization with eight horizontal axes. The top four are labeled ‘Satisfied Attention Check Criteria (N = 100)’. Of these, the top two are labeled ‘Truncated Upper Bound’. These have means and error bars close to the right hand side of the axis, with the ‘Labels Absent’ condition very slightly further to the right than the ‘Labels Present’ condition. The means and error bars for both ‘Extended Upper Bound’ conditions are near the center of the axis. This entire pattern of error bars in the top four axes is replicated below, in the bottom four axes labeled ‘All Participants (N = 165).

Fischer, J., and A. Ali. 2021. A Federal Ban on Abortion is Wildly Unpopular in All 50 States. Data For Progress.

 Google Scholar

Data availability statement

The pre-registration, plus materials, experiment script, data, analysis code, and Dockerfile are available at https://osf.io/qe9hf/. This repository contains the requisite resources to generate a fully-reproducible version of this paper.