Taiwan Local Elections in Taipei, New Taipei, and Taoyuan: Forecasting based on Computer Simulation
A research team led by Shiping Tang, professor at Fudan University in Shanghai, has just released their computer simulation-based predictions for the Taiwanese upcoming local elections that will be held on November 24, 2018. The forecasting focuses on three municipalities: Taipei, New Taipei, and Taoyuan. Our current forecasting model is limited to predicting the distribution of popular votes between the two competitive camps: the Blue Camp and the non-Blue Camp (including the Green Camp).
This is a purely scientific exercise. The team has no intention to influence the actual elections in Taiwan by any means.
Earlier this month (Nov. 4, 2018), Tang’s team has accurately forecasted the electoral outcomes of two U.S. senator elections (West Virginia and Missouri). For details, see the links at the end of this release.
Our forecasting results
1. Taipei
According to our baseline models, the Blue Camp (candidate Ting Shou-chung) is predicted to win 34.80% (high: 34.84%; low: 34.75%) of the popular votes while other candidates as a whole are predicted to win 65.20% (high: 65.25%; low: 65.16%) of the votes.
The average predictive error as dictated by our baseline models is 1.94%, with its range being [0.24%; 3.66%].
2. New Taipei
According to our baseline models, the Blue Camp (candidate Hou You-yi) is predicted to win 54.19% (high: 54.50%; low: 53.63%) of the popular votes while the Green Camp (candidate Su Tseng-chang) is predicted to win 45.81% (high: 46.37%; low: 45.50%) of the votes.
The average predictive error as dictated by our baseline models is 2.93%, with its range being [0.65%; 4.68%].
3. Taoyuan
According to our baseline models, the Blue Camp (candidates Chen Shei-saint and Yang Li-huan) is predicted to win 32.54% (high: 34.37%; low: 30.71%) of the popular votes while all other candidates is predicted to win 67.46% (high: 69.29%; low: 65.63%) of the votes.
The average predictive error as dictated by our baseline models is 3.14%, with its range being [0.45%; 4.79%].
These predictions are the outcome of a yearlong project conducted by Tang’s team at Fudan University in Shanghai. Prof. Tang is the Fudan Distinguished Professor and Dr. Seaker Chan Chair Professor at the School of International Relations and Public Affairs. He also directs the Center for Complex Decision Analysis (CCDA). For more information on CCDA, check out their website: www.ccda.fudan.edu.cn
The method developed by Tang and his colleagues circumvents the need of relying on polling data for forecasting electoral outcomes. In other words, their whole forecasting exercise does not have any input from public opinion polls. Moreover, their method does not merely predict which candidate is going to win or lose, but their share of votes within a predictive interval dictated by their baseline models.
Over the years, Tang’s team has attempted to develop an agent-based modeling simulation platform that combines micro-level data about the voters and macro-level demographic, economic, political and social data, data-driven methods and theory-driven methods, and forecasting and explanation.
Tang is optimistic that with further refinement and improvement, their approach may one day transform the field of election forecasting and electoral studies worldwide. His team is releasing its predictions to the public domain two days before (Nov. 22, 2018) the actual elections on Nov. 24, 2018, providing the general public and the scholarly community with an opportunity to check the predictions against actual voting results.
Tang’s team will evaluate their forecasting results after the full election results are in and then provide more technical detail regarding their approach.
Links to their 2016 forecasting of the 2016 election in Taiwan
More than two years ago, on Jan. 05, 2016 (Chinese version) and Jan. 10, 2016 (English version), Tang’s team released their forecasting of the election in Taiwan before the actual voting of that election. That forecasting was their first attempt of election forecasting with their new approach. For details, see:
Chinese Version
English Version
Links to their forecasting of U.S. mid-term elections for senate in West Virginia and Missouri
Early this month, on Nov. 04, 2018, Tang’s team released their forecasting of US Senate Elections in two states (West Virginia and Missouri) before the actual voting of that elections. That forecasting was their second attempt of election forecasting with their new approach. Overall, their forecasting has been very accurate. For details, see:
Chinese Version
English Version
For a preliminary comparison of their forecasted results and actual results, see: