Guide to 2020 Taiwan Presidential Election: Interactive Maps, Polling Results, and Bot-generated Voter Analysis

Category: Best news application

Country/area: Taiwan

Organisation: CommonWealth Magazine

Organisation size: Big

Publication date: 30/12/2019

Credit: Yu-Hsin Lee, Wei-Han Chi, Yi-Wen Lin, Chia-Yu Peng, Ya-Wen Pai, Shih-Yen Chen, Chen-Hua Chen, Li-Hsun Tsai, Kai-Ting Tsao

Project description:

2020 election played a crucial role in Taiwan’s democracy. The race between Presidential Tsai and KMT candidate Han Kuo-Yu reflected people’s decision to vote for pro-China or pro-America stance.

The project covers four sections, including Presidential Election Interactive Map, Bot-generated Voter Analysis, Policy Watch, and Polling Results Explained to provide comprehensive election reportage, to provide comprehensive data analysis: Which constituencies elected pro-China candidate? These constituencies were leaning towards the KMT party (blue) or DDP party (green)? What about the voter demographics?

Impact reached:

Since published in December 2019, the project generated 3.6 million pageviews within two weeks, bounce rate was lower than 20%, and pages per session reached 2.6 pages. Those numbers suggested that readers had engaged highly with innovative news content.

Among other things, Bot-generated Voter Analysis compiles all the demographic data (such as age, education, median income, and gender ratio) in 22 cities/counties, 368 towns, and 7761 villages as well as cross-compares election data of party vote share and voter turnout. It totally produced 8151 short pieces of data analysis and generated more than 2 million pageviews within 24 hours after poll results released.



Techniques/technologies used:

Data source: Central Election Commission, Ministry of Finance, Ministry of the Interior, Central Weather Bureau (all the data come from government’s open data)

Data processing: php, python, nodejs, QGIS (for shp files)

Visual design: d3.js, jQuery

Infra: Google Cloud Platform

To give journalists easy access to maps, the developer created an additional function: users can key in “Konami Code” (↑↑↓↓←→←→BA) in the desktop version of Presidential Election Interactive Map to download SVG files and then key in “gg” to close the function.

More information:http://bit.ly/2RXKJSK

What was the hardest part of this project?

During 2020 election, Taiwan saw a surge of disinformation and fake news. The public lost trust in news outlet, even show obvious hostility to them. The hardest part of the project was that how we can rebuild trust between media and readers.

We decided to hand over the right of news analysis and interpretation to readers. This was the first time we mainly focused on data-driven interactive map and bot-generated analysis rather than journalists reporting since Commonwealth Magazine was founded 38 years ago. By means of data visualization and bot-generated analysis, readers were capable of perceiving and interpreting the implication of those hard data. 

With the help of CrowdTangle, we observed that readers would apply these content and share their own thoughts about election results on Facebook, Instagram, and Twitter. For example, Han Kuo-Yu Supporters rushed to explain in which constituencies KMT had lost. Tsai Ing-wen supporters surprisingly found that their neighbors didn’t vote for Han Kuo-Yu.

What can others learn from this project?

With interactive maps and bot-generated data analysis, readers could recognize immediately the changes of pan-blue and pan-green vote share and voter demographics; moreover, they were allowed to identify the shifting constituencies in 2020 election with one click.

2020 election also included legislative election. Policy Watch section scraped 78-thousand social media posts from 433 candidates’ social media accounts (Facebook, Instagram, and YouTube). We selected 318 keywords based on 15 critical issues (such as same-sex marriage, energy, and cross-strait relations) in hope of presenting candidates’ arguments and stance on these issues and providing guiding reference to voters.

Project links: