Soda map of Florida. Photo: Matthew T. Campbell.

Political analysts have developed many tools for predicting and analyzing election results. Yet, many well respected analysts are often wrong with their predictions even when close to an election. Most data comes from polls that you may notice, are often wrong also.

So many analysts were predicting a win for Mitt Romney only days before the South Carolina GOP primary, and even though Gingrich gained a significant lead closer to voting day, nobody predicted such a wide victory. Gingrich won South Carolina with 12.6% of the vote over Romney, while the polls leading up to it saw Gingrich leading by an average of 5%.

The poll average for tonight’s Florida Republican primary show Romney leading by 13% while the most recent poll from Public Policy Polling (PPP) shows Romney leading by 8% over Newt Gingrich.

Let’s use a different method.

Recently brought to my attention was a map designed by Matthew T. Campbell from the Spatial Graphics and Analysis Lab of the Department of Cartography and Geography of East Central University in Ada, Oklahoma. This map, although from 2003, utilized research data to geographically represent what words Americans preferred to use for a soft drink. Traditionally, people say “pop” in the Midwest, “soda” in the Northeast and West Coast, and “Coke” throughout the South. Down South, when you ask a waiter or waitress for a “Coke,” most likely they will ask you: “what kind?” It’s not strange if the answer is “Sprite.” This anomaly is probably because Coca-Cola was invented in Atlanta, Georgia; like Kleenex for tissues, it became the generic term. If you look at Florida in detail, you will see that there is a lot of diversity. This is because Florida, is now home to many transplants – stereotypically, retirees from the Northeast.

Why is this significant? Well, if you’ve conducted interviews with voters at exit polling locations during an election, you’ll often discover surprisingly illogical answers. Not to downplay the importance of suffrage, but it seems to me that most voting decisions are instinctual rather than intellectual. Most often, we choose the party we support because of ingrained beliefs.

The problem with actual polls is that there are too many variables beyond a pollster’s control. Some people don’t like answering polls; some lie or justify their answers; many respondents don’t feel confident that a poll is actually anonymous as opposed to the actual vote they cast – even if there is no place to write their name – causing them to respond in ways different than their actual vote in order to avoid society’s disapproval.

Knowing this, I’m interested to see whether something as apolitical as generic names for soft drinks, can give a more accurate result than political polling. People are hesitant to answer political questions, often lacking the proper knowledge, but at least they would be more likely to honestly answer what word they use for soft drinks.

Using this map to predict the result of tonight’s primary is risky, but I couldn’t prevent my curiosity in seeing how accurate of a predictor it will be.

The process is rather simple. Focusing only on Florida, I begin by assuming that since the South traditionally uses the word “coke,” the counties colored in red and dark red will be more traditional and would have fewer northern transplants. Assuming that northeastern Republicans are more moderate than Republicans in the South, the counties in green will have a more moderate electorate. Since Mitt Romney has shown to be the favorite among moderates, and Newt Gingrich popular with the hardliners, I divided the counties by giving Romney the counties in green, and Gingrich the counties in red and dark Red. The light red counties, where respondents are split almost equally on the word they use, I gave to Romney because of his recent momentum. Counties with no data, or those under the “other” category, were not allocated to either candidate.

After the counties were given to their respective candidate, I added the populations of all registered Republicans in each county and came up with a total of possible votes for Gingrich and Romney. I used all registered Republicans since the actual turnout rate is not needed for this purpose.

Finally, I divided each candidate’s total with the total amount of registered Republicans in the State of Florida, including the small portion within the counties not given to either candidate.

The final result gave Mitt Romney a lead of 55% of the vote, and Gingrich 45%. Surprisingly, this 10% margin falls between PPP’s latest 8% and Real Clear Politics’ 13% lead for Romney.

No doubt, there are many problems with this system. First, the map is from 2003. If it is accurate today, it can be said that tonight’s result could have been predicted then. Many things have changed in Florida since the map was created, but assuming most transplants move into areas with other transplants, and not into tightly traditional communities, the problems are alleviated using current GOP voter population statistics. New residents will only add to the population of counties that already say “soda.” Second, not having been part of the original study that led to the creation of the map, I cannot vouch for the accuracy of their process.

Tonight will be a test for this risky experiment. If Romney wins by 10% over Newt Gingrich based on what Floridians use to refer to their soft drinks – rather than PPP’s 8% lead – I’ll be glad to have spared myself the trouble reading their lengthy statistical analysis. If it is correct, one can safely say that campaigning and debates don’t make much difference in the end unless something outrageous happens during them.