Nate Silver: What I need from statisticians

Exclusive interview: Nate Silver was the guest speaker at this year’s Joint Statistical Meetings in Montreal, Canada where he gave the President’s Invited Address. He is currently the editor-in-chief of FiveThirtyEight blog which will shortly move to ESPN and a Special Correspondent for ABC News but he became a household name when he correctly predicted 49 out of 50 states in the US Presidential election in 2008 by using Bayesian statistics. Four years later, he correctly predicted the winner of all 50 states and the District of Columbia during the 2012 US Presidential Elections between President Barack Obama and Senator Mitt Romney.

In doing so, as ASA President Marie Davidian said in her introduction, ‘Nate’s talents as a statistician put him and statistics into the public spotlight. His statistical skills have brought him great remand but it is his ability to clearly explain his methods so that they are properly understood, his fearlessness in standing by his work, and his transparency in admitting how and why he can do better that have endeared him to the media and the public.”

During his senior year in high school, Silver was the top debater. He obtained a BA degree in Economics with honours from the University of Chicago. With a passion for baseball since his youth, Silver developed PECOTA, a system for forecasting the performance and career development of Major League Baseball players, which he sold to and then managed for Baseball Prospectus from 2003 to 2009. In 2007, writing under the pseudonym “Poblano”, Silver began to publish analyses and predictions related to the 2008 United States presidential election, eventually establishing in March 2008 his own website and blog, which was licensed by the New York Times in 2010.

In September 2012, he published his first book, The Signal and the Noise which reached The New York Timesbest seller list for nonfiction, and was named by as the no. 1 best nonfiction book of 2012. When recently interviewing one of the founding fathers of Bayesian statistics, Professor Dennis Lindley, Silver’s book was by his side. Silver was delighted to hear this when I told him. In July 2013, it was revealed that Silver and his FiveThirtyEight blog will depart The New York Times and join ESPN.

In front of a packed audience of 1,000 statisticians, Silver admitted he did not consider himself a statistician and that it was a thrill and an honour to be there. After his talk, about which you can read about more below, was able to exclusively interview Silver alongside the Editor for Significance, Julian Champkin.

Despite his notable successes, they have not gone to Silver’s head who explained that he regularly reads negative press to keep himself grounded. He came across as shy and genuinely humble with a self-deprecating sense of humour and in a way bewildered by his fame. He talked candidly about how he got introduced to statistics, the first bit of statistics he ever did with his dad aged 8, being a baseball analyst, the Presidential Elections, his book The Signal and the Noise, and how his ability relates to influence after being named one of the world’s Most Influential People according to Time Magazine. Lastly, Nate Silver explains what he needs from statisticians.



For the Address, Silver’s main theme was how journalism and statistics are linked and what journalists can learn from statisticians. At ESPN, Silver explained how he was going to recruit a whole team of statisticians and journalists to cover a wide area, ranging from politics to sports to economics to food. Silver offered several principles for journalists to take on board.

1) Statistics aren’t just numbers. Most people think of a statistic as a batting average or as a number of goods sold. One of his long-time rivals, an editor at Politico, wrote about Silver’s work on as ‘trying to use numbers to prove stuff that can’t be proved by numbers alone’. Most journalists are not aware that statistics is also a scientific discipline and are not using them in the right way. It requires an understanding of the discipline in order to use the numbers correctly.

2) Data requires context. Silver used an example as an article from a writer he admires from the New York Times that China has the world’s second largest economy. Whilst true, China has a very large population and it would have been more appropriate to cite per capita GP instead of average GP.

3) Correlation is not causation. The desire in journalists to tell the story can obstruct the facts and shut down the debate and the critical thinking needed is lost. Journalists who are more aware of data quality issues, do not take statistics at face value and want to ask further questions is the way to go for the future.

4) The average is still the most useful statistical tool ever invented. One of Silver’s main frustrations with the coverage of the election and the polls last year. Journalists do not focus on the average person but prefer a better story about an outlier which tells a better narrative. The bias can be quite explicit. People in politics can be willing to cherry-pick the data and not be apologetic for it. He respects the “lowly average as it performs almost as well as more complex methods but also because it serves as a litmus test for whether the journalist is worth their statistical salt or not”.

Methods can be abused and that is true with any approach, including Bayesian but it offers a more coherent, philosophical way in which to look at the world and…would be useful for journalists to use as well.

5) Human intuition can often cause errors when it comes to making statistical judgements and inferences. Journalists need to be more aware of not only their subject matter but also what their own biases might be.

6) A probabilistic forecast expresses uncertainty instead of trying to conceal it. Journalists are not comfortable with 50/50 and 100% true but are not happy with 75% vs. 25%. Journalists are afraid of saying that most evidence points one way but there is debate that it could move in another direction.

7) Know thy priors. Silver uses the Bayesian approach towards statistics. Methods can be abused but that is true with any approach, including Bayesian but it offers a more coherent, philosophical and sophisticated way in which to look at the world and the Bayesian method would be useful for journalists to use as well, in particular prior beliefs and bias.

8) The word ‘complex’ isn’t always a complement. When a journalist explains that something is complex, he/she may be unintentionally letting the reader know that he/she does not understand. If a statistician did this, it would be even more concerning! In this, Silver sees a lot of parallels between the role of a statistician and that of a journalist. A journalist has to take a complex set of facts and convey some understanding of them to the broader public – which details are most essential and which can be left out?

9) Insiderism is the enemy of scientific objectivity. Silver was fond of some of his reception from critics that nerds are taking over the world but it is an over-simplification. Inside information is expensive and takes a long time to cultivate. Such information may not be very reliable once you receive it. There are very talented reporters who can see through the charade but there are others who ‘do not have a very good BS detector and tend to follow the herd.’

10) Making predictions improves accountability. Towards the end of the election campaign, Silver got into an argument with a colleague and suggested that they place a $2000 bet on who would win, the proceeds going towards those affected by Hurricane Sandy. He was scolded by a New York Times editor that there should be no betting in the newsroom. The notion that his betting on his forecasts would compromise his integrity seemed strange to him. He used to play poker where the whole aim of the game is putting your money where your mouth is. Likewise in the private sector, one ought to have a financial stake in the reliability of their forecasts. In science, predictions are the very core of theory and scientists stake claims that could affect their reputation. What can statisticians learn from journalists? Statisticians need to get out there and spread the gospel. Already many good journalists abide by these principles. There is now a bigger market and demand for a more data-driven approach to journalism and statisticians can help lead the way.

Silver then took part in a lengthy Q&A but was very pleased to answer them and was truthful and patient in his answers. He advised when being contacted by a journalist to check their background first just to check that you are not to be involved in a one sided argument. He advised academia to start blogs to put their own ideas out there to share with the world and this could also help improve communication with journalists. He recommended to statistics students to get hands-on experience where you can be an applied statistician before returning to university to achieve a more advanced degree.

“I think data scientist is a sexed-up term for a statistician.”

When asked which journalists he admires, he replied Bill Haynes for being ahead of his time in sports statistics and his former NY Times colleague, Daniel Kahneman about the world of economics and baseball.Other than journalism, statistics is also needed in education data with collaborations between teachers and statisticians.

When asked that “Data science is the term of the day. Do you think there is a difference between data science and statistics? Silver replied, “I think data-scientist is a sexed up term for a statistician”, the reaction from the audience was for most, one of instantaneous laughter and applause. “Statistics is a branch of science. Data scientist is slightly redundant in some way and people shouldn’t berate the term statistician.”


Copyright: Image copyright of the American Statistical Association