Making Economist-Style Plots in Matplotlib
Pushing Matplotlib to make beautiful and readable plots
You can find the Jupyter notebook for this article on Deepnote.
If you like data visualization, chances are you’ve seen charts from The Economist, either on their website or in their magazine. For decades they have been known for simple and effective visualizations that convey useful information in a simple way. For the past few years, I’ve been constantly inspired by their visualizations. Whether I see them on my Kindle, on the web, or a mobile device, they are always compelling, and they get the point across. This is something all data scientists should aspire to in their own visualizations.
In October 2018, the Economist launched a dedicated blog for their data visualizations called Graphic Detail. If you scroll through their posts you can see that their style remains consistent between chart types. This isn’t because they are a factory making visualizations as fast as they can. Instead, this consistent style is designed to simplify their visualizations and provide the key information to the reader as quickly as possible.
As a Python focused data scientist, in my daily work I almost exclusively use Matplotlib for my data visualizations. I like it because it is incredibly customizable. Every element is under your control. While I’m sure there are data journalists at the Economist using Matplotlib for exploration work or mockups (or ggplot for R), I highly doubt they make their production visualizations using only Matplotlib.
However, I’m a big fan of keeping as much of my workflow in one place. This thinking inspired me to re-create several of the Economists’ chart types in 100% Python and Matplotlib. I have two goals for this project. First, I want to give examples of how to execute great design for charts in Matplotlib. Second, I want to show that a programmatic plotting library can be used to create production plots.
Please note that I won’t be making re-usable functions or a library for this. Instead, if you are interested, I hope you can copy any code you may want and re-use it yourself.
I will be recreating three common plot types the Economist uses regularly: bar charts, line charts, and dumbbell charts. But before we get to that, let’s break down the key elements of the Economists’ visualizations.
Please note that this is being done for educational and informational purposes only.
The Economist’s Style
To get an understanding of the Economist’s visualization style, we can use their very own visual style guide from 2017. Here is the link. Let’s break down the common elements of the Economist’s style.
Title and Subtitle
Charts usually have both a title and subtitle. The title will often be a play on words, like “Unhappy meal” for rising food prices. The subtitle will be a simple statement of what the chart contains. Always located on the top left, they sit flush with the left edge of the chart content. The title is always bold and larger than the subtitle.
Gridlines
Only horizontal grid lines are shown. Minor gridlines aren’t shown at all. Gridlines are spaced in a way that is logical for the data and easy to understand.
Axes and Tick labels
Only the horizontal axis is shown. No other bounding box around the chart data besides this. On the x-axis ticks (the little lines next to the label) are only shown when they are needed to show which data connect to which tick label. On the y-axis, there are no ticks, only the major gridlines. The y-axis tick labels sit on top of the major grid lines inside the chart area.
The exception to this is a horizontal bar chart. In this case, the y-axis is shown but there is no x-axis. Y-axis tick labels sit directly on top of the number.
Colors
The Economist uses two chart palettes, one for the web and one for print. We will focus on the web one. The main color “Econ Red” (#E3120B) is used for the top line and tag box. The grey (#758D99) in the style guide seems to be used for the gridlines. The colors for plotting are:
- Red, #DB444B
- Blue, #006BA2
- Cyan, #3EBCD2
- Green, #379A8B
- Yello, #EBB434
- Olive, #B4BA39
- Purple, #9A607F
- Gold, #D1B07C
There are more color palettes in the style guide for different uses, but this is the main color palette. The color order for each chart type is different, and I will attempt to reproduce that below. All plots will use the web style with a white background.
Other Stylistic Points
A horizontal line in “Economist red” runs along the entire top of the chart. A horizontal rectangle, which they call a tag, sits at the top left corner. Data sources are given at the bottom left in 75% opacity.
We can go over chart-specific styles in the sections below.
Bar Chart
Let’s build a bar chart in the style of the Economist! Here is an example of a bar chart from the Economist. We will use 2020 GDP values by country from a Kaggle dataset (link)¹. The purpose of the chart is to show the top 5 countries by GDP value.
The Economist typically only uses a vertical bar chart where the one or two categorical variables are being compared over time. For the data we are working on, a horizontal bar chart would be more appropriate. This is because the country names are longer, and they wouldn’t fit easily on the x-axis without being rotated. This would make them harder to read. The solution is to push the data over some and have the labels on the y-axis.
Let’s import pandas and matplotlib and read in our data.
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt# This makes out plots higher resolution, which makes them easier to see while building
plt.rcParams['figure.dpi'] = 100gdp = pd.read_csv('data/gdp_1960_2020.csv')
gdp.head()
The GDP numbers here are very long. To make them easier to read we can divide the GDP number by 1 trillion.
gdp['gdp_trillions'] = gdp['gdp'] / 1_000_000_000_000Now we can filter for only 2020 and grab the bottom 9. We do this instead of sorting by descending because Matplotlib plots from the bottom to top, so we actually want our data in reverse order.
gdp[gdp['year'] == 2020].sort_values(by='gdp_trillions').tail(9)Now we can build our chart. The code is here with comments.
And the resulting plot:
Wow, that’s a lot of code! There is a lot there, and it’s worth taking the time to go line by line and understand it with the comments.
Now let’s move on to the line chart.
Line Chart
Line charts are perhaps the most iconic chart from the Economist. Here is an example of a line chart from the Economist.
We will use the same data GDP data as above to create a time series plot of the top 9 countries by GDP. To highlight the path China has taken we will only show the US and China with standout color, and the other 7 as grey.
First, we can get the names of the countries we want to filter from our previous code.
countries = gdp[gdp['year'] == 2020].sort_values(by='gdp_trillions')[-9:]['country'].valuescountries# Output
array(['South Korea', 'Canada', 'Italy', 'France', 'India',
'United Kingdom', 'Germany', 'China', 'United States'],
dtype=object)
Matplotlib does a good job with datetimes for time series, so we can convert the date column to a Pandas datetime Series.
gdp['date'] = pd.to_datetime(gdp['year'], format='%Y')Our data is now ready, let’s get to plotting!
Even more code this time. Typically Economist charts are more square, especially for their print edition. I decided to make it wider to fit better in an article by itself.
Dumbbell Chart
For the last chart, we will make a dumbbell chart. These charts are used to show the differences in a categorical variable, usually over time. Here is an example of a dumbbell chart from the Economist.
We will again use the GDP data. This time to show the difference in GDP from 1960 to 2020 for the top 9 countries.
gdp_dumbbell = gdp[(gdp['country'].isin(countries)) & ((gdp['year'] == 1960) | (gdp['year'] == 2020))].sort_values(by='gdp')gdp_dumbbell.head(5)
Conclusion
While we have seen that it is possible to recreate the chart style of the Economist in Matplotlib, it certainly isn’t easy. For each plot, tens of lines of code are required, and each plot took several iterations of trial and error to get the way you see above.
As you can see in the code, the plotting of the data in Matplotlib is fairly straightforward. However, aligning the text, line, and tag required very specific custom locations to be manually entered. This is one thing that would be good to improve upon in the code above.
This was a great experience, and I learned several things about Matplotlib in the process. The library continues to amaze me with its ability to be customized. Perhaps in the future, I will build a more reproducible method for creating print quality charts in Matplotlib.
Thanks for reading. I mainly write at datafantic.com, so be sure to check it out for more articles from me.
If you like Medium and you would like to directly support my articles you can use my referral link to sign up to become a Medium member.
References
