The Sprint for the Finish: forecasting the US Midterms

Article by Daniel Clemence

The final stage of the US Midterms is underway. On the 6th of November, the US Midterms take place. Considering the trends, the US midterms may become a major political upset. The US Midterms are a way of potentially showing their frustration at Trump. The question that could be asked is twofold. Firstly, what forecast or prediction would I give for the US Midterm election? Secondly, how could potential indicators show on election night mean for how the US Midterms could play out?

What is the difference between forecasting and prediction in elections?   According to Nate Silver in the book The Signal and Noise, the main difference in forecasting and prediction is that prediction usually gives specific outcomes of elections such as predicting that the Conservative Party will have 290 seats at the next UK general election, a prediction on Electoral-calculus based on current polling.  Forecasting an outcome is based on a percentage chance of an outcome, an example of this is the forecasts on Electoral-Calculus of 38% chance of a Conservative majority versus a 33% chance of a Labour majority. It is therefore possible to see that there is a significant difference in outcome, whether it is predicted or forecasting. Predicting the outcome of the US Midterms could lead to largely inaccurate results. This is because political variables such as the so-called “Shy-Tory” effect, where people won’t say who they will vote for but will vote for a certain party. Therefore, forecasting is potentially more accurate than predicting the outcome of elections.

There are some very interesting forecasts and predictions of the US Midterm elections. Sebato Crystal Ball website gives some predictions of the outcome of the Midterms including various House election outcomes; all of these election outcomes predict a Democratic majority but some stronger than others. The website itself suggests that the mean average house change of seats is around 20 since the 1980s, with a median of 10. These averages would leave the Democrats short of taking the House however. A far more interesting forecast by fivethirtyeight.com as of 15th of September (the date the article was written), there is a forecast of an 82.9% chance of the Democrats taking the House of Representatives. This kind of difference between predicting and forecasting is that while the predictions try to calculate the actual seats that the Democrats would win, forecasts give a range of potential outcomes which in the US Midterms is split into two outcomes, given that there are only two parties and that smaller parties are largely irrelevant. Given this evidence, there is a high probability of a Democrat control of the House of Representatives.

The Senate Race does not look as good for the Democrats. The fivethirtyeight.com forecast suggests that the Democrats have an only 33% chance of gaining the Senate this year. Why is this the case? The Democrats have a hard time this year due to the fact that they are defending 8 potentially competitive seats against the Republicans who have just 5. The current makes up of the Senate is there are 51 Republicans and 49 Democrats. Even if the Democrats hold onto all their defensive seats and win one other state, they would still not control Senate because the Vice-President has a vote in case of a tied vote. This makes them fundamentally at a disadvantage compared with the Republicans. There is one caveat here. There is a higher chance of the Democrats taking the Senate than fivethirtyeight.com’s  forecast of Donald Trump winning the presidential election. Indeed, when factoring the states that allowed Donald Trump to win the electoral college, the Democrats have a significantly higher chance of winning the Senate. Fivethirtyeight.com gave Trump a 16% chance of winning Wisconsin; which means that the chance of the Democrats winning the Senate is roughly doubly than that of Trump winning Wisconsin. In Pennsylvania, a state Fivethirtyeight.com gave a 23% chance of winning, this means the Democrats have a 63% improved chance of winning compared with the 2016 Pennsylvanian forecast.

On election night, predicting the outcome of the races maybe determined by turnout and overperformance in various districts. For example, if Democrats overperform in Florida districts 6, 15 and 25, this would indicate that they would be likely to hold their Senate seat. If the Republicans are underperforming in their House races in Texas, they may also be likely to lose their Senate seat also. High turnout may also be a good sign for the Democrats, who do better when there is a higher turnout. If there is a much higher turnout for the Democrats than expected, it is very likely that the Democrats will do even better then expected come election day.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s