Media look to statisticians to increase readership

News

  • Author: Carlos Alberto Gómez Grajales
  • Date: 12 November 2014
  • Copyright: Image appears courtesy of iStock Photo

Traditional media may be in trouble, with decreasing readerships and increasing online competition. Can statisticians help them  attract readers?

thumbnail image: Media look to statisticians to increase readership

Traditional media companies are having problems, that's undeniable. You've certainly read every now and then about massive layoffs in some newspaper or other, such as USA Today [1] and The New York Times [2]. These aren't randomly selected examples, they are just a few that occurred in the last three months. Seeing the ferocious market that grew from the Internet era, it was just a matter of time before media companies started to react with new ideas, looking for new models to bring more readership. I mean, like statistical models.

Several news outlets, particularly in the United States, have started working on their own stochastic models to predict elections, thus providing their readers with a more scientific forecast than the traditional “analysis” of survey results, providing more attractive news than the usual online coverage. For instance, The Washington Post introduced the “Election Lab 2014”, a project that worked on forecasting models to predict who would win midterm elections in the United States [3]. This model, developed by political scientists with training in data analysis, included variables such as presidential approval, growth of Gross Domestic Product, the partisanship of the region and even some of the candidates' information, such as their government experience [4]. The algorithm was also modified to include even more data, such as recent poll results and historical voting results. The modeling process started since may and as the election grew closer, simulation procedures allowed the research team to produce some nice forecasts regarding results.

The Daily Kos had some models running on their own too. These models were devised by Drew Linzer, a political scientist and survey statistician, with notorious expertise in voting. The forecast started by using Bayesian dynamic linear models to detect trends in the results of Opinion Polls for each of the elections [5]. These time series forecasts were used to produce a huge number of Monte Carlo simulations to establish the distribution of likely outcomes on Election Day. Each simulation incorporated a range of different sources of uncertainty and random error, such as survey estimation error and potential changes in voter preferences in the period before Election Day.

Other outlets entered the forecasting competition as well. The New York Times devised LEO, a forecasting model headed by Amanda Cox and Josh Katz [6]. The Huffington Post also published predictions based on polling data [7]. Media companies are heavily investing in such key differentiators. In consequence, we have several complicated models we statisticians can have fun reading about. I know, it is sad what we do for fun.

The coverage these models gathered may have turned them into attractive exercises for other outlets. It is safe to assume that these kind of statistical analysis will become more prevalent in the future, not only in the US but around the world. Just keep in mind, when you see these predictions appearing in your town, that all forecasting algorithms rely on polling information, making them their first and primary source of data. It is therefore important to consider, when analyzing such forecasts, that their results and accuracy depend on the quality of the polls they use as source. Not even Nate Silver's predictions are immune to skewed polling results [8]. This year both the Washington Post and the Daily Kos had slightly better results than 538's forecasts [9].

References

[1] USA Today Undergoes Major Layoffs (Updated). The Huffington Post (US Sept, 2014)
http://www.huffingtonpost.com/2014/09/03/usa-today-layoffs-job-cuts-gannett_n_5760196.html

[2] New York Times plans to Eliminate 100 Jobs in the Newsroom. The New York Times (US Oct, 2014)
http://www.nytimes.com/2014/10/02/business/media/new-york-times-plans-cutbacks-in-newsroom-staff.html?_r=0

[3] Election Lab 2014.The Washington Post (US Nov, 2014)
http://www.washingtonpost.com/wp-dre/politics/election-lab-2014

[4] How Election Lab works. The Washington Post (US May, 2014)
http://www.washingtonpost.com/news/politics/wp/2014/05/05/how-election-lab-works/

[5] Election Outlook: How it works. Daily Kos (US Nov, 2014)
http://www.dailykos.com/election-outlook/how-it-works

[6] Meet Leo, Our Senate Model. The New York Times (US Nov, 2014)
http://www.nytimes.com/newsgraphics/2014/senate-model/methodology.html

[7] Elections 2014 Senate Forecast. The Huffington Post (US Nov, 2014)
http://elections.huffingtonpost.com/2014/senate-outlook/

[8] FiveThirtyEight’s Senate Forecast. FiveThirtyEight's Politics (US Nov, 2014)
http://fivethirtyeight.com/interactives/senate-forecast/

[9] Election Lab on track to forecast 35 of 36 Senate races correctly. The Washington Post (US Nov, 2014)
http://www.washingtonpost.com/blogs/monkey-cage/wp/2014/11/05/election-lab-on-track-to-forecast-35-of-36-senate-races-correctly/

Related Topics

Related Publications

Related Content

Site Footer

Address:

This website is provided by John Wiley & Sons Limited, The Atrium, Southern Gate, Chichester, West Sussex PO19 8SQ (Company No: 00641132, VAT No: 376766987)

Published features on StatisticsViews.com are checked for statistical accuracy by a panel from the European Network for Business and Industrial Statistics (ENBIS)   to whom Wiley and StatisticsViews.com express their gratitude. This panel are: Ron Kenett, David Steinberg, Shirley Coleman, Irena Ograjenšek, Fabrizio Ruggeri, Rainer Göb, Philippe Castagliola, Xavier Tort-Martorell, Bart De Ketelaere, Antonio Pievatolo, Martina Vandebroek, Lance Mitchell, Gilbert Saporta, Helmut Waldl and Stelios Psarakis.