Predictive Achievements and Evidence

 from 1980 to Present

My quantitative research into elections commenced in 1980. In 1983, I pioneered the first large-scale electoral opinion survey in Taiwan, an endeavor that was honored with the Golden Tripod Award. This research has continued for over 40 years. In the realm of "vote prediction," whether for single-seat or multi-seat elections, margin of error has consistently been smaller than that of those observed commercial polls. A more significant undertaking has been the dedicated research into "electoral behavior," with the aim of developing long-term predictive models.

Since 1983, I developed an "Image Index" as a tool for vote prediction. This led to the discovery of the "53237 Voter Structure" theory (originally the "7535 Voter Structure"), which has been progressively published since 1986. This theory holds the potential to evolve into a long-term cyclical predictive model. In 1994, this model further indicated several significant events that were likely to occur:

1.The Kuomintang (KMT) would shrink.

2.A three-party competition would emerge. (This subsequently occurred in 2000).

3.Chen Shui-bian would win the presidency by a narrow margin. (This also occurred in 2000).

4.An analysis of floating votes' oscillation suggested that the vote ratio between the KMT and the Democratic Progressive Party (DPP) would shift from approximately 60:40, with one gaining while the other receded, and then vice versa, eventually returning to 60:40 after one cycle. (Was this 2008?)

[Note: The preceding section was written in 2004 when this website was established, using a questioning tone. This was indeed confirmed in 2008.]

Electoral Predictions: Highlights

2004 Presidential Election: When the general public anticipated a landslide victory for Lien-Soong, I predicted that President Chen Shui-bian would win by an extremely narrow margin, which would be accompanied by prolonged disputes. Due to legal restrictions, I was unable to publicly disclose this prediction in my personal capacity. However, I did share my forecast with two scholars of international academic standing and significant academic administrative positions, seeking their insights on potential solutions to the impending societal discord. (This is documented in the "NCCU College of Social Sciences Policy Forum").

2008 Presidential Election: On the eve of the 2008 presidential election, I again presented my predictions to two individuals, inviting them to serve as witnesses. I also noted that the figures for the 2008 election might have been predicted by me more than two decades prior, and this instance merely served as a validation.

My prediction for 2008 yielded an error margin of 1.35%, which, among all presidential election forecasts I have encountered, was the closest to the actual outcome.

2012 Presidential Election: Prior to the 2012 presidential election, I shared my predictions with several friends, inviting their collective observation. My prediction for 2012 again achieved an error margin of 0.43%, once more proving to be the closest to the actual results among all public opinion surveys.

2014 Taipei Mayoral Election: For the 2014 Taipei mayoral election, I conducted a survey commissioned by a third party. Lacking a non-disclosure agreement, I reported the election forecast to all three major political factions (Blue, Green, White) eight weeks before the vote. The final error margin was less than 1%, serving as further evidence of the relative stability of human behavior.

2018 Kaohsiung Mayoral Election: On the eve of the 2018 Kaohsiung mayoral election, I predicted the winners for each administrative district to several friends, as well as groups of National Taiwan University and Chien Kuo High School alumni. This achieved 100% accuracy. More importantly, this might have been the first instance where data and meta-analysis supported the theory that "circumstances create heroes, not heroes create circumstances."

Methodological Distinction

The primary distinction between this research community's approach and other electoral surveys lies in its pursuit of "pattern prediction" and structural analysis through novel quantitative methods, as opposed to conventional rapid-response polling. Regardless of whether it's a single-seat or multi-seat election, the prediction error of this community has never exceeded 2%.

For the 2024 Presidential Election, responding to the needs of election practitioners and the challenge of fully comprehending comprehensive behavioral predictions, I proposed the "Meta-analysis Method." This approach integrates numerous public opinion polls to achieve superior results. For further details, please refer to the article "Comprehensive Review of Presidential Election Polls: Authoritative Scholars Examine Each One!" from the "National Fact-Checking Society - Understanding Polls," which introduces the meta-analysis method.

Adoption Modeling: A Fundamental Theory

Since 1996, I have also conducted a long-term series of studies on the then-emerging behaviors of internet and e-commerce usage. The findings reveal that these two seemingly unrelated fields share a fundamental commonality: the "Adoption Modeling" of human acquisition and selection behavior.

From a rigorous critical perspective, current behavioral research has not yet fully met the conditions to be considered a "science." One reason is the persistent lack of fundamental "laws." It is hoped that Adoption Modeling can provide an opportunity to explore these foundational laws.

The "53237 Voter Structure" and its underlying "Adoption Modeling" appear capable of explaining and predicting long-term electoral behavior over approximately 30-year cycles.

TxNote: Good Theories are Stable

Over 40 years of learning election prediction, I have fortunately never been wrong. However, I constantly remind myself: human behavior is probabilistic knowledge; be careful, this time could be the one I get wrong!

My two main insights are:

1.If fundamental behavioral theories exist, human behavior can be predicted.

TxNote: Good Methods are Dynamic

2.Society is constantly changing. After each prediction, I reflect on newly emerging problems and potential solutions. Current public opinion polling is a thousand times more challenging than when I pioneered election research in Taiwan in 1983. Therefore, each time, it is imperative to consider new issues arising from previous survey methods and to introduce novel, previously unresolved methods. Furthermore, one must not be limited to a single survey method but must integrate the "logic and techniques" of meta-analysis with various other quantitative methods.

Black Pool (Xuan Shui Yuan)

This site is also known as Xuan Shui Yuan, derived from the "Knowledge Wanders North" chapter in the Outer Chapters of Zhuangzi: "Knowledge wandered north above the Mysterious Water."

"Knowledge" (知, Zhi) is a fictitious name, nominally representing an "omniscient one," yet whose actions are those of a "learner." Zhi journeyed north to the "Black Pool" to visit Wu Wei Wei, Kuang Qu, and the Yellow Emperor, seeking wisdom and resolution to doubts. Each responded according to the implied meaning of their names. Based on this allusion, Professor Tung-Hsiung Wu likens "Black Pool" to the Western Delphi, symbolizing a place for seeking wisdom, resolving doubts, and predicting the future.