Instead of predicting the exact share of votes a candidate will get in the elections, our models compute the chance that that candidate will win 50% + 1 votes. This is what we call a probabilistic forecast and it’s very different from predicting how many votes a candidate or party will get.
What it means:
- Elections are very complex phenomena. In Ghana, about 15 million people from many different backgrounds, with many different interests, from every part of the country will cast their votes for a presidential candidate.
- Therefore for a party to win, there are many many things that need to go right at the same time. Conversely, a party’s loss is hard to pinpoint to just one thing. It usually takes a confluence of factors for the election to turn out the way it does.
- This means that it’s practically impossible to accurately predict the exact votes a party will get in an election; well, unless you happen to have a hand in rigging the votes!
- Anyone who claims to know the exact share of votes a party will get is kidding or lying to you. You might as well flip a coin or close your eyes and randomly pick a number.
- On the contrary, a probabilistic model like ours looks at a massive number of possibilities and calculates the likelihood that a set of events will occur. In our election’s forecast, we simulate the election at least 10,000 times to generate a distribution of possible outcomes.
- What if voter turnout increases by 5% in an NDC stronghold? What if voter turnout decreases by 20% in a swing region? What if the NPP wins the Volta Region? What if the NDC wins all 16 regions?
- These are just a handful of the kinds of possible scenarios that our forecast models are crunching through. It’s a lot of numbers and a decent amount of code.
- But at the end of the day when we say that Nana Addo has a 68.3% chance of winning, it simply means that out of the 10,000+ simulations our model ran, he won the election in 6,830 of them. Conversely it also means he did not win 3,170 elections.
- This is a simplified view of how we forecast the elections. The reality is a bit more complicated. We are combining artificial intelligence, traditional statistical models and opinion polling data to generate this forecast.
- A probabilistic forecast may seem a little difficult to wrap your mind around. In many ways, it’s the harder way to go about generating the forecast but it’s also the scientifically more rigorous way of doing so. Overtime, it gets easier and my hope is that increasingly our conversations about electoral politics in Ghana will become more data-driven and fact-based.
