By Samuel Chiu October 08, 2016
Sweet Sixteen—it has been sixteen years since we started this project, engaging students in the political process – to be educated and to get involved. It started in year 2000 as a class exercise in MS&E 220, an introductory class in probability within the Management Science and Engineering Department at Stanford University. In MS&E 220, we emphasize putting probability theory into practice to model real world problems.
Now, current participants in the project include both undergraduate and graduate students. Go to the About Us section to see the history of this project starting with the 2000 election.
We process state-by-state polling data to estimate the probability that a particular candidate (e.g., Mr. Trump) will win a state’s electoral votes. These state-by-state win probabilities will be processed to produce a histogram. One can read from the histogram the probability that Mr. Trump will win (for example) 292 electoral votes. We can then compute the probability that Mr. Trump will win the election by winning 270 or more of the electoral votes. The Methodology section contains details of our computation procedure, from polling data input to state-by-state probability estimation and the resultant histogram. Our method is transparent, and we rely on reputable state-by-state polling data; our model accuracy depends on how good the polls are. Some prediction models may tweak polling data to incorporate a subjective belief, or they may use some other secret sauce. We do not pretend to know better than the polling data, which presumably provides an accurate snapshot of the sentiment of likely voters.
Nobody saw it coming. During the primary season, experts scratched their heads as they watched their punditry falling by the wayside. We build prediction models based on assumptions and logic to process information resulting in some output. Modelers use their knowledge and insight to construct models based on their experience. You cannot incorporate what you don’t know. There are unknown unknowns, to borrow the famous words from former Secretary of Defense Donald Rumsfeld. No one needs to apologize for the disastrous predictions, just learn how not to repeat them. We learned from our mistakes in the 2000 election cycle (see the About Us page for the history of our project).
This is an election that defies all conventional wisdom. It will be the subject of studies in political science, sociology, history, statistics, polling science and psychology for decades or even centuries to come; in this country and perhaps all over the world. As a country, I hope that we will collectively reflect to move forward in a positive way. We shall begin with a victory concession speech by Mr. Trump, the Republican Nominee; to begin the healing process and to build a better future for America and for the world.
In the 2016 edition of Stanford Predicts, the undergraduate students are first year Stanford students who took part in SSEA: Stanford Summer Engineering Academy. See this link for a description of the program. Matthew Chen was a SSEA alumnus and he participated in this project in the 2012 election cycle. He is now a master’s student in Computer Science.
As they engage in a current and exciting event, students within the project also learn model building: how to build a bigger model from smaller components and how to make reasonable assumptions about that model.
For a list of participants this year, please visit the team section of this website.
For more information about Stanford Predicts, including inquiries about our methodology or other press inquiries, please contact Matthew Chen.