Tuesday, November 04, 2008

How Detectable is Voting Fraud?

Today's timely posting is by a guest author, JG Fellow.

Today, we are holding elections. On November 5th, we will likely be debating the existence of fraud. I believe that we have the tools demonstrate the likelihood that such fraud has taken place. Throughout the elections, via Pollster.com, I have watched the results of every single state and national poll. I have been struck by the stability of the results over the last two weeks. To be sure, two polls of the same state may vary, but there are few states where the projected winner is uncertain.

Based on this wealth of data, we know what to expect on November 4th, but we are, by no means, certain to get it. Pollsters occasionally miss biases, or run in herds. We witnessed this in the New Hampshire Democratic Primary this year. But New Hampshire stood out as an exception. Most state's primary results followed the pre-election polling.

So how do we evaluate those results that differ from our expectations? How can we tell the fraudulent from the unexpected? By examining the results at the voting precinct level. It is certainly possible to falsify election results. But while it is possible to lie, it is very difficult to lie well. It is incredibly difficult to generate a fake set of data that continues to look real upon scrutiny. How do I know this? Experience.

Both professionally and personally, I have performed hundreds of Monte Carlo simulations. This is a process whereby you define some rules and roll some dice. If I roll a 6, David Ortiz hits a home run. If I roll a 1, unemployment rises. I have simulated baseball, basketball, football, car accidents, occupational injuries and even terrorist activity. The tragedy of my experience is that I can always find some level on which my generated data fails to replicate reality. While my hitters may meet my expectations, Mariano Rivera ends up as an average pitcher. While the average cost of a car accident is correct, the chance of a $10,000,000 car accident is too high. The first, second and third test of my data may look reasonable, but there will always be a fourth, fifth or sixth.

I have faith in my fellow Americans, and even more faith in checks and balances. But most important, I have faith that such fraudulence could not be done in secret. I would urge major news outlets to work with statisticians to validate the reasonability of the outcome of this election, and I look forward to today with no fear that a crime shall go unwitnessed.

No comments: