As the presidential election nears, we are seemingly overwhelmed with data about the candidates. In this post, I take a look at different ways polling results are visualized from various news outlets, analyzing them from a data visualization and accuracy perspective. I’ll highlight a few key items to look for in a polling visualization to help you understand the data, such as:
First up is a visualization from CNBC.com – in an article about how the national political party conventions affected each candidate’s polling numbers in swing states.
This journalist chose a bar chart to represent the sentiment of likely voters towards the candidates in six states, also including the overall figures.
We see in the footnote of the chart that the source is a poll of 4,143 likely voters from six states taken between September 4 and September 6.
Next up is a chart on CNN.com analyzing a similar topic, the sentiment of Republicans and Democrats toward the two Presidential candidates after each of the individual conventions.
This journalist used a slope chart to show the pre-convention to post-convention change in sentiment towards the candidates, segmented by the respondent party.
The poll was conducted from a random sample of 1,000 adults from across the country. The methodology cites leveraging the Lexicoder Sentiment Dictionary to assign positive or negative scores to words that the respondents used when describing each candidate.
The third chart we review comes from fivethirtyeight.com, one of my personal favorite websites for data visualization and analysis. This chart is a continuously updated web page that chronicles the president’s approval rating leading up to the election.
This page uses a scatter/trend chart and table combination to display the president’s approval rating over time. The trend chart shows the aggregate approval rating throughout the year, while also displaying the individual poll results. The table provides detailed information around each included poll.
Each contributing poll is directly cited in the table, along with a link to download the data directly from the polling website. Clicking on the ‘How this works >>’ link gives you an in-depth explanation about how their model is created and the different factors included. This is the gold standard in terms of data transparency.
The fourth visualization we review comes from foxnews.com in an article titled “Voters pick Biden, yet more think their neighbors back Trump.”
The journalist chose to use tables throughout the article to represent polling results. This seems to be a standard practice at Fox News. In fact, it was all I could find when I googled “Fox News polling results.” The table represents the respondent’s choice for president at different points throughout the year.
The table cites the date range of the most recent poll, along with the audience type (registered voters) and the margin of error. Also, just below the article it states that the responses came before and after the announcement of Kamala Harris as Biden’s running mate.
The fifth chart under review comes from the website 270towin.com. It is another web page with a collection of charts displaying polling results.
This page uses small(ish) multiple pie charts to show aggregate polling results by state, including a chart for national results. Similar to fivethirtyeight, links to the individual polls are provided. There are also different sorting options and links to historical voting results for each state.
The provided links to the individual polls provide great context regarding how the polls were conducted, including source, date, sample size, audience type, and margin of error.
The last visualization we analyze comes from The Economist. This is website also sets the bar high for data visualization and analysis standards. This chart is a map from their 2020 forecast page, which is continuously updated based on the latest polling results. The page has numerous graphics and are all worth inspecting.
The authors created a map and distributed bar chart to display the win probability of each candidate by state. The map uses mark size to visualize the state’s electoral votes, while also leveraging color to indicate the win probability percentage. The bar chart below uses the same mark and color encoding, while also sorting by percentage and highlighting the competitive states.
Sources and links to download the source code, model output and individual polling data are provided at the bottom of the page. Similar to fivethirtyeight.com, the source data and methodology are extremely transparent and they even invite you to take a look at the underlying data yourself.
A quality polling visualization is easy on the eyes, uses the appropriate chart type, is clearly labeled, and plainly cites the survey details. If the chart looks overwhelming or confusing, then the colors and design may be lacking. If you have trouble interpreting the data or figuring out how the survey was conducted, then you may be misled by the visualization. Each of these charts had pros and cons, but the pages from fivethirtyeight.com and The Economist stand out as superb examples of polling visualization. In the ‘How this works >>’ page from fivethirtyeight, they even delve into the difference between respondent types (Adults vs Likely Voters vs Registered Voters) and how each affects the validity of the survey.
An effective data visualization should be easy to digest with a clear story, comparison context, and sourced data. If your news provider does not consistently meet those benchmarks, it may be a misleading chart. Always check other sources for validation.
The Data-Informed Voter Series is a 2020 passion project for a team of Evolytics analysts. We aim to be as politically neutral as possible while discussing the data, implications, and interpretations we see in the news. We discuss topics as a pseudo-editorial board with the aim of informing voters on how a professional analyst would interpret data during an election cycle. This project team consists of John Carney, Jay Farias, Liam Huffman, Brian Johnson, Anoush Kabalyan, Laura Sutter, and Krissy Tripp.
Not sure on your next step? We'd love to hear about your business challenges. No pitch. No strings attached.
