State of Play: A Pompous Ass Calls the Election
“Victory has a thousand fathers, but defeat is an orphan.”
-John F. Kennedy
Let me save you time. I, on behalf of AutoPolitic, am projecting the winner in the 2022 Philippines presidential election to be Leni Robredo by at least 4% of the total votes.
My prediction is based on the social intelligence tracking work I have done with my friends at the ADDS group for the last 5 months. We have also used other social media platforms and google to confirm the direction of the campaigns, but Facebook is the primary source of data for the final call and the precision. (You can hear our talks here, here and here.)
We reserve the right to be wrong. If we are right, no one except our competitor will ask us how we did it and we won’t tell.
If we are wrong, it is because of these self-inflicted hubris;
1. Social Intelligence is too sensitive to coordinated inauthentic behavior. In 2022, we relied on the platform to remove trolls and fake accounts and did not adjust the raw number in our reports. We decided not to adjust the number simply because there’s no good framework to do so. While fake accounts do not have votes, but can influence and change votes, we believe this is the biggest causal factor for a failure to accurately predict the votes.
2. Different levels of internet penetration for each class is significantly different from social intelligence over-represent ABC and under-represents DE. We had a strong belief that social intelligence is proportionally representing ABCDE and that the collective internet behavior, if we are wrong, is a strong factor. Perhaps our over-reliance on social data may have doomed our prediction.
3. Social Intelligence measures specific things, and there are many important factors it does not. We could not measure local machinery and vote-buying. We also could not measure the impact of terrestrial broadcasting on voter behaviors. We can only measure what we can get data on. So the general attitude about a specific candidate has to be translated into action that leads to votes, something happened on the way to the polling booth.
4. To a lesser degree, Social Intelligence is blind to geographic-based behavior clustering, local political machinery, party bosses, and barangay captains. While we don’t assume they are trivial, we simply don’t have a good model to predict their impact or a good way to validate the model.
5. 30-40% of the voters make up their mind in the last few days of the campaign and are affected by the recency effect. We will make updates to our prediction based on the latest data as needed. This is especially important in a tight race like what we are observing.
6. Lastly, in a race this close, a last-minute switch of support by Isko, Lacson, or Pacman can make a difference. These contenders can be the king-maker, and the backroom dealing necessary to secure their vote basis can be anticipated but not modeled or predicted by social intelligence. Our prediction assumes none of them will endorse Leni or BBM because of ego. But politics is the art of the possible, and while all the Philippines are waiting for the vote to determine who will lead them through the turbulent post-covid future. As a campaign operator, we are waiting to find out which poll is right.
There is an old Jewish saying, “Man plans. God laughs.” With our incredible hubris of predicting an election, God just fell off the chair, howling with mirth.