Data Visualization in Management Research: A Primer

Ertug, Gruber, Nyberg, and Steensma (2018) contribute an important editorial titled “A Brief Primer on Data Visualization Opportunities in Management Research” to the Academy of Management Journal. The core objective of this editorial is to educate and inspire management researchers about the array of data visualization techniques that can significantly enhance the communication and understanding of their findings. The authors underscore the foundational principle that “A picture is worth a thousand words,” noting that pictorial representations offer alternative mechanisms for communicating complex information, much like speaking a second language. Research consistently shows that individuals better understand and remember information when it is presented pictorially rather than through words or short sentences.

The editorial systematically introduces both novel qualitative and quantitative data visualization techniques, focusing particularly on those applied to raw data before formal statistical analyses.

Enhancing Qualitative Data Visualization

For qualitative data, which accounts for roughly 20% of submissions to the Academy of Management Journal and often poses challenges for authors in terms of effective presentation, the editorial discusses several compelling techniques that can also help establish trust in findings.

1. Word Clouds: These visualize text by giving greater prominence (larger size) to more frequently used words, thereby providing an overview of main topics, themes, and systematic patterns. They are useful for quickly identifying dominant themes in large amounts of text data and for comparing thematic similarities and differences across various texts. For instance, a word cloud created from the introduction sections of 174 Academy of Management Journal articles illustrated the relative prominence of research themes. A key caveat is the need for preprocessing raw text to remove noise through steps like lowercasing, tokenizing, removing stop words, and stemming. Open-source Python packages like WordCloud offer flexibility in configuring aesthetic features.
2. Word Trees: Invented by Wattenberg and Viégas (2008), this technique places a tree structure on words or phrases that follow a particular word or phrase, arranging them spatially. Similar to word clouds, more frequent words or phrases appear larger. Their visual tree structure makes it easy to spot repetition in contextual words that follow a specific term. An example highlights how word trees explained why certain phrases (e.g., “pledgers will” followed by “receive” in crowdfunding) positively or negatively predicted project success. Raw text preprocessing is also crucial here to avoid noise.
3. Social Graphs: These diagrams illustrate interconnections among people, groups, or organizations in a social network. Nodes represent actors, and edges represent relationships. They can display both qualitative (e.g., color of nodes and edges, labels) and quantitative data (e.g., width of edges, size of nodes). Social graphs are widely used for social network analysis to depict individuals, organizations, and their contacts. They can visualize complex relationships, but for large or dense networks, adjusting node/edge size, luminosity, or trying different layouts (e.g., circular, random, Kamada-Kawai) is necessary to maintain clarity. NetworkX in Python and NetDraw are cited as useful tools for generating these graphs.
4. History Flows: This text visualization method, invented by Wattenberg and Viégas (2010), visualizes the dynamic editing history of text over time. The x-axis indicates versions, and the y-axis represents text length, with unique colors representing different editors and their contributions. They were initially used to visualize Wikipedia editing histories to reveal complex histories of cooperation and conflict among authors. The authors suggest their potential for management research to enhance the visualization of cooperation histories between individuals and firms (e.g., in open-source software development) or to track user reviews and ratings to understand product popularity or user taste dynamics.

Novel Quantitative Data Visualization

Beyond traditional scatter plots, bar charts, histograms, line charts, and pie charts, the editorial explores several novel ways to visualize quantitative data.

1. Multidimensional Scaling (MDS) Plots: The purpose of MDS is to provide a visual representation of proximities (similarities or distances) among a set of objects by reducing multiple dimensions to two or three for easier display. This is useful for comparing differences in multi-dimensional concepts like cultural distance, deviant behavior, or sequences of routines. The number of dimensions needed for an MDS plot is determined by checking “stress,” an indicator of goodness of fit. XLSTAT is mentioned as a commercial software program embedded in Microsoft Excel that can generate MDS plots.
2. Funnel Plots: These are crucial in meta-analytic studies to examine publication bias, where studies with stronger effects are more likely to be published. They plot each study’s effect size against its precision; a skewed or asymmetrical shape indicates bias. They can also be used in literature reviews. The metafor package in R can be used to conduct publication checks, generate funnel plots, and correct for bias.
3. Maps: Maps visualize interconnections among countries or regions (e.g., firm entry, migration patterns) and differences (e.g., total volume of foreign direct investments, R&D investments) using lines, colors, or shades. They can illustrate geographic divisions or global distributions of phenomena, as seen in research dividing Chinese provinces into macro areas or showing subcultural regions in India. Plotly’s Python graphing package is suggested for generating maps.
4. Bubble Plots: Building on scatter plots, bubble plots add a third dimension represented by the size of each dot, allowing the visualization of relationships among three numerical variables simultaneously. For example, the x-axis can depict GDP per capita, the y-axis can depict average life expectancy, and the size of the dots can represent the population of a country. Matplotlib in Python and the scatter command in STATA can generate these plots, with options to adjust transparency in case of overlapping bubbles.
5. Dynamic Plots: These animate visualizations over time (like a GIF) to show dynamic processes, relationships, or distributions. This is particularly useful for longitudinal data, allowing researchers to see how patterns change over time, such as the dynamics of social networks. While any visualization technique can be animated, dynamic plots are especially helpful for understanding network evolution. Matplotlib in Python facilitates the creation of dynamic plots.
6. Tree Maps: Tree maps display hierarchical data as a set of nested rectangles, where the area of each rectangle is proportional to its value. They can convert bar charts or word clouds, with rectangle area reflecting frequency. They are useful for visualizing project portfolios or the influence of variables. The squarify open-source package in Python provides the ability to generate tree maps, allowing users to add labels and colors.

Tools and Aesthetic Considerations

While traditional software like Excel, SAS, STATA, and SPSS are convenient for standard chart types, newer tools popular in data science offer greater flexibility. Tableau provides interactive visualization tools and good data management. R (with packages like ggplot2, plotly, igraph) and Python (with matplotlib, seaborn, networkx) offer extensive customization options, though they require an understanding of programming.

The authors emphasize that “beautiful visualizations reflect the qualities of the data they represent, explicitly revealing properties and relationships inherent and implicit in the data that offer insights and new understanding at a glance”. For a visualization to be truly beautiful, it should be aesthetically pleasing, novel, informative, and efficient. Researchers are encouraged to think not only about the type of visualization but also how to make it beautiful by adjusting aesthetic features like size, proportion, color, and luminosity, making it more appealing, easier to understand, and more pleasurable for readers.

Reference: Ertug, G., Gruber, M., Nyberg, A., & Steensma, H. K. (2018). From the Editors: A Brief Primer on Data Visualization Opportunities in Management Research. Academy of Management Journal, 61(5), 1613–1625. https://doi.org/10.5465/amj.2018.4005