Communicating

Data visualization and tables are crutial tools in applied economics, to explore, diagnosize and communicate.

Objective

This session aims to helping you develop essential skills to communicate the outputs of your econometrics analyses, with a strong focuses on producing effective data visualization.

Summary

A significant portion of communicating economic research results-whether in presentations or written papers-relies on graphs, visuals and tables. This session aims to help you develop essential skills for effective data visualization and table-making and sharpen your interest and awareness of the topic. It discusses the importance of good visualization and tables, describes a conceptual approach that can help us design our graphs and tables and suggests general “best” practices.

Session Outline

1. Why good data viz matters?
   • What do we use data viz for?
   • Why make good data viz?
2. What are the characteristics of good data viz?
   • Tell a story
   • Cognitive load
   • Data viz theory
3. How to make an ok data viz?
   • Non-negotiable rules
   • “Best” practice
   • Choosing a type of graph
4. Communicating in economics
   • Exploratory data analysis
   • Regression tables
5. Summary

Take-aways

Main take-aways

• Identify the message or story you want to convey
  • Choose the appropriate communication device as a consequence (eg the appropriate type of graph)
• Guiding principle: reduce the cognitive load of your audience
• Make sure to produce legible and understandable outputs
• Nice aesthetics go a long way and can shield from some implicit cognitive biases

Materials

Open slides in html

Open slides in PDF

Specific resources for this lecture

Handbooks

Here is a short list of useful applied data visualization handbooks:

• Fundamentals of Data Visualization provides a great code-free introduction to data viz.¹
• Modern Data Visualization with R and Data Visualization - A practical introduction both discuss key data viz concepts and teaches you how to apply them in R and ggplot.
• R for Data Science provides a nice introduction to ggplot, THE package to build plots in R. To go further you can read this book.

Additional resources on data viz

There are massive amount of resources on data viz. Here is a subset, based on the lecture outline:

• Why good data viz matters?

  • What do we use data viz for?
    • Allows to synthesize large amount of data
    • Explore, explain and make inference
      • Why looking at data?
      • Existing tools for exploratory data analysis in R
  • Why make good data viz?
    • Visualizations can be deceptive
      • Defense against dishonest charts
    • Credibility
      • Credibility and typography
      • Reasons to make beautiful graphs on credibility and aesthetics
      • In defense of simple charts

• What are the characteristics of a good data viz?

  • Tells a story
    • Storytelling and data visualisation
  • Limits the cognitive load
    • Cognitive Load as a Guide
    • More on cognitive load for data viz
  • Leverages principles from data viz and perception theory
    • More on Gestalt principles (plus a research article)
    • How to find and create good color palettes on
    • Data-to-ink ratio
      • A critique of the data-to-ink ratio

• How to make an ok data viz?

  • Non-negotiable rules
  • Good practices
    • Concrete recommendations
      • For color
      • For text
    • Do’s and don’ts
      • Statistics oriented
      • A collection of dataviz caveats
      • Sometimes it’s ok to break the rules
  • Choose a plot type
    • Know what exists: graph galeries
      • Data Viz Project
      • Flowing Data - Chart type
      • Dataviz Inspiration
      • R graph gallery
    • What questions to ask when creating charts
    • Decision trees for a graph type
      • Financial Times tree
      • Data viz catalogue
      • Datawrapper guide
      • Data to viz graph

• Communicating in economics

  • Papers on data viz in economics
    • Schwabish (2014)
    • Cattaneo et al. (2024)
    • Korting et al. (2023)

• Main take-away points

Useful R Packages

• ggplot: THE package to build plots in R.
• modelsummary: a package to summarize data and present regression results in nice tables and plots.
• sf: THE package to work with spatial data in R.

References

Cattaneo, Matias D., Richard K. Crump, Max H. Farrell, and Yingjie Feng. 2024. “On Binscatter.” American Economic Review, May. https://doi.org/10.1257/aer.20221576.

Korting, Christina, Carl Lieberman, Jordan Matsudaira, Zhuan Pei, and Yi Shen. 2023. “Visual Inference and Graphical Representation in Regression Discontinuity Designs.” The Quarterly Journal of Economics 138 (3): 1977–2019. https://doi.org/10.1093/qje/qjad011.

Schwabish, Jonathan A. 2014. “An Economist’s Guide to Visualizing Data.” Journal of Economic Perspectives 28 (1): 209–34. https://doi.org/10.1257/jep.28.1.209.

Footnotes

1. Although R code can be found on the book’s GitHub if needed.