Using Technology to Predict the Next Presidential Election – Getting It Right Next Time

This audio was created using Microsoft Azure Speech Services

Boy was the election polling wrong! How in the world could this have happened in this era of technology? We certainly had choices on how to follow election results this year – Old fashioned local TV, National TV, Specialized National TV stations, Web results trackers, HuffPost Elections on Viber, Sidewire, New York Times Election Bot on Slack, Facebook, Twitter etc…  I am sure I missed quite a few.

How many people followed through old fashioned local TV – a lot less than last election.

What process was used for polling you ask?  Old fashioned telephone polling which has been around for a long time and up until now – very accurate. Although, in the 1930s, a magazine called Literary Digest conducted public opinion polls by mail and phone.   This worked because the they had a very large sample size that yielded accurate results from almost every poll. They say history repeats itself. It certainly did in 1936 when Liberty Digest predicted a landslide victory for Alf Landon over Franklin D. Roosevelt.   Boy was the election polling wrong!  The result was flawed because the readership was a specific pool of wealthier people who are disposed to vote republican.

Today, old fashion telephone polling is still used to collect responses from the best representative sample of the American voting public. Mathematical methods and computer analysis are used by the “top” polling organizations to gauge the opinion of the entire nation by asking only a small sample of voters. For this to work, pollsters have to ensure that this small sample group accurately represents the larger pool of eligible voters.  If 55 percent of voters are male for example, then 55 percent of the sample group needs to be male. The same applies to age, race, geographic location, etc to accurately predict the probability of future outcomes.  So what is the scientific technique to do this?  Random dialing software!  That does not sound very scientific to me.   Especially since older Americans tend to answer the phone more often, which throws off the sex and age ratios of the sample and skews the results tremendously.

You could argue that this process has been extremely accurate over the years with the exception of 1936. So why the big miss?  The other main reason was the polling organizations relied upon historical trends of voter behavior.   These voter tendencies changed for the first time but the polling results did not predict them.

Which brings us to what can be done today to improve polling accuracy.  How about on-line surveys? Not surprisingly Google has survey products that it markets to the staffers and operatives who work for presidential and congressional campaigns, as well as the journalists who cover them.  Google claims it can tell a person’s age and gender from their browsing history.   They can also determine a person’s location based on their IP address. Demographic questions are asked in the survey.  The data can be used to model a larger population and drill in on specific trends and tendencies.  Only one presidential campaign used Google’s tool, the company said, while declining to name the candidate.  I think I know the name of the candidate – it was obviously Donald Trump!  How else would he have been so confident going into November 8th?

It’s fair to say that candidates will not rely on traditional polling in 2020.  It will be interesting to see how technology will be used to accurately gauge voter behavior.  The process that can reach the largest representation of the population with a process that is easy for respondents to access and complete will bring polling accuracy into the 21st century.

As the Google example shows, and we all intuitively know, there’s no shortage of data out there. The key is to have the right tools to collect it and make sense of it.

That certainly applies to data centers, which now have power and cooling infrastructure generating about 3 times more data than they did 10 years ago. But we’re putting that data to good use, with tools that use it to help make data centers more efficient and reliable. You can also access all this data remotely, so you can keep tabs on your data center no matter where you may be. And, unlike polls, the data is all highly accurate and informative. To learn more, download the free Schneider Electric white paper no. 237, “Digital Remote Monitoring and How it Changes Data Center Operations and Maintenance.”

Tags: , , , , , , , , ,