Nate Silver’s website FiveThirtyEight just released their final projections for the 2018 midterm elections. The Republicans have a ⅘ chance of keeping the Senate and the Democrats have a ⅞ chance of taking control of the house. These predictions are in large part a continuation of Mr. Silver’s life’s work, using data to predict the future.
After graduating in 2000 from Chicago University with a degree in economics Silver began working in finance which he hated. He has said that that four year period was his only regret. It was during this time when Silver changed paths and began constructing PECOTA, the Player Empirical Comparison and Optimization Test Algorithm, his model for forecasting player performance in the MLB. His algorithm takes in players’ various seasonal statistics and attaches a weighted value to each. Because the game of baseball is exhaustingly analyzed and tediously measured and because baseball has been around for so long practically unaltered, Silver had access to a mountain of good data. By painstakingly sifting through statistic after statistic he was able to construct an effective model that produced consistent correct predictions. Silver sold PECOTA to Baseball Prospectus and stayed on as editor in chief there until 2008.
During the 2008 presidential campaign season, Silver who was always politically engaged and active began blogging under a pseudonym. He quickly decided to transition fully into politics and founded his website, FiveThirtyEight. Baseball to Politics may seem like a stretch but not to Silver who said, “It reminded me of baseball, when you see the same recycled clichés and conventional wisdoms over and over again, some of which isn’t very wise.” Silver and his website caught their first break when they forecasted then presidential candidate Barak Obama would outperform senator Hillary Clinton on Super Tuesday. Silver was off by only a handful of delegates. In the final election, Silvers model predicted the presidential contest to within 5 electoral college votes and correctly predicted the outcome of every senate race.
Silver was able to create a model that seemed entirely new for consumers of political news. Never before could the average reader develop a command of the data and nitty-gritty analysis typically reserved for strategists. FiveThirtyEight was even able to one-up political and polling professionals by systematically unskewing the polls. By weighing the past performance, time and methodology just like baseball stats the algorithm is able to overcome traditional deficiencies in opinion polling. The data-driven philosophy at FiveThirtyEight stems in large part from Nate Silver’s own frustration with the state of political punditry, once saying, “I was just frustrated with the analysis... I saw a lot of discussion about strategy that was not all that sophisticated, especially when it came to quantitative things like polls and demographics”.
2008 propelled Silver into the realm of media stardom, he began a partnership with the NYT and wrote a popular and critically acclaimed blog there for 3 years. His work there helped pioneer the field of data journalism. At the Times he dissected stats from politics, sports, and broader social issues. His models again proved solid, correctly predicting the Republican pick up in the house and Senate.
In 2012 FiveThirtyEight was in truly surpassing form, in its final prediction Barak Obama had a 90.9% of victory. After the dust settled Silver and his model had correctly predicted the outcome in every last state including all 7 swing states. This while some respected polling institutions like Rasmussen only correctly predicted three of their nine swing states. 2012 is when Silver really went from wunderkind to wizard. After dunking on the rest of the field Silver partnered with ESPN. FiveThirtyEight’s monthly users grew steadily during that period, from about 2.8 million unique viewers in April 2014 to 10.7 million unique visitors in January 2016. His site covered a broad range of topics all while keeping an eye on the data.
Silver’s meteoric rise seemed to position him perfectly as a analyst of “Trumpism” during the 2016 election cycle. Instead, FiveThirtyEight dismissed Trump’s candidacy from the top down resulting in an big whiff when predicting the Republican nominee. Silver later apologized for acting like a pundit and passing off educated guesses as facts and failing to take notice of new evidence during the primary (FiveThirtyEight does not have an algorithm for primaries). While predicting the general election Silver took heat for unskewing the polls too far in Trump’s direction, giving Secretary Clinton only a 71% of winning, compared to most of the press, whos predictions put Clinton’s chances far higher. Still, 71% was good enough for many who took that number at face value and took solace in it. While Silver has always defended his prediction, many had their faith in him and polling in general, shaken. Trump’s rise has led to bipartisan mistrust for our institutions including elections and those who predict their results. Silver has had two years to tinker, will it be enough to win back some trust?
While I am fairly familiar with FiveThirtyEight, I never really considered how it brought data to the normal viewer. Many people our age were not following politics before Nate Silver came onto the scene, and, while there are now many convenient ways to see polls, it's strange to think that this was not as accessible as it is now. And that is not to mention his work aggregating things into one, cohesive number rather than having people try to either see only handpicked polls from pundits or trying to wade through them by themselves.
ReplyDelete