# The thinking behind BioPharma Dive’s CEO compensation charts

By Nami Sumida • Dec 10, 2020

In September, BioPharma Dive published its second annual CEO compensation story — a deep-dive analyzing compensation data for major biotech and pharmaceutical companies. I worked with the BioPharma Dive journalists to analyze and visualize data on 231 companies’ CEO compensation and median employee salary.

Take a moment to read the article and become familiar with the story and data.

## Data gathering and analysis

BioPharma Dive compiled a list of 231 publicly traded biotech and pharmaceutical companies based on companies included in the SPDR S&P Biotech exchange traded fund, the iShares Nasdaq Biotechnology Index, and several large multinational pharmaceutical companies.

For each company, they collected data on a few key variables: 2019 CEO compensation, 2019 median employee pay, the CEO’s gender, and the number of employees at the company.

Using these variables, we came up with three main insights/trends to report on:

- The wide range of CEO and median employee compensation and the existence of several large outliers
- The non-linear relationship between CEO and median employee compensation, and call attention to companies with particularly small or large CEO to median employee compensation ratios
- The small number of female CEOs and any differences in compensation by gender

## Data visualization

I created an interactive data visualization for each of the three insights. The overall design goal was to communicate the insight while also allowing for exploration of individual data points. I allowed for this level of granularity because I anticipated readers wanting information on specific companies (e.g. their competitors, outliers in the data). Therefore, I included several interactive elements: a search bar to search for company names, filters for company size, and mouse-over tooltips on individual data points.

In two of the charts, I incorporated the “scrollytelling” format — a visual storytelling technique in which content changes as the reader scrolls up or down the page. There were two reasons for using this format: 1) The charts were complex enough that they would benefit from a walkthrough of the interactive elements. 2) I wanted to call out important trends shown in the chart. With scrollytelling, I could overlay the text onto the chart so people could read about a trend and see it in the chart simultaneously.

Next, I’ll go into detail on the reasoning behind certain design decisions for each chart.

### Visualization 1: Histogram

#### Design goal

Show the wide range of CEO and median employee compensation and emphasize several outliers

#### Design solution

A histogram is a great way to show a variable’s distribution. Right away, you can see that CEO compensation has a right tail (or a positive skew), while median employee compensation is more normally distributed. Both have several large outliers, which are also easily identifiable.

I anticipated readers wanting to see compensation details of certain companies (e.g. of their competitors, of the outliers). Therefore, I added a mouse-over tooltip with this information on each rectangle (which represents an individual company).

Mousing-over a rectangle also highlights the corresponding rectangle in the other histogram. This allows readers to see both figures and compare the positions of each in their relative distributions. By comparing the values relative to their overall distributions, readers may be able to draw conclusions about income disparity within a company.

The interval bands at the base of each histogram also make comparisons easier. Each band represents a quartile range. For example, the left-most band in gray represents 0-25% of the data, the next band represents 25-50%, and so on. For a given company, readers can use these bands to compare the quartile that includes its CEO compensation to the quartile that includes its median employee pay.

### Visualization 2: Scatter plot

#### Design goal

Show the ratios between CEO and median employee compensation and highlight companies with particularly small or large ratios

#### Design solution

A scatter plot is one of the best ways to visualize the relationship between two variables. A linear relationship is easily recognizable by dots positioned along a slope, while a non-linear relationship is noticeable in its lack of a defined slope — which is what we have in this chart.

CEO and median employee compensation are not linearly related. In other words, a company with a higher CEO compensation does not necessarily have a higher median employee pay.

When it came to highlighting companies with small or large ratios, I first needed to define a threshold for each. Because there are no industry standards around this, I calculated thresholds based on the variable’s distribution: quartiles. Companies with ratios in the highest quartile have particularly high ratios, and those in the lowest quartile have particularly low ratios.

In the scatter plot, I colored the dots based on their quartile. The gray dots represent companies with ratios between 25:1 to 35:1 — a range close to the median of the dataset. Dots that fall above and to the left of these dots represent companies with higher ratios, while those positioned below and to the right represent companies with lower ratios. As you scroll through the chart, dots with values in the highest and lowest quartiles are highlighted in indigo.

### Visualization 3: Bar chart

#### Design goal

Show the small number of female CEOs in our dataset and shed light on differences in male vs. female CEO compensation

#### Design solution

A simple way to visualize gender differences would have been through summary statistics: the share and average compensation of male vs. female CEOs. However, calculating the average compensation was a bit dubious because there were so few female CEOs (N=20) compared to male CEOs (N=211).

Therefore, I opted for a long bar chart displaying every CEO compensation in the dataset. White bars represent female CEOs, indigo bars represent male CEOs, and the bars are ordered by compensation.

Readers can easily see the small number of white bars in a sea of indigo bars. Moreover, because the bars are ordered, readers can detect biases in compensation amounts — if they exist. For example, if female CEOs made far less than male CEOs, you would see few white bars at the top and most towards the bottom. The bar chart shows no apparent pattern, with white bars somewhat evenly distributed.

–

Here’s the article again. I’ll be writing more blog posts about other data stories, so check back soon!