| August 28, 2020
New features allow the comparison of 2020 model predictions to official polls and 2016 projections, with visual highlights of key divergences.
A new feature on our election monitor enables users to compare and contrast monitor forecasts with official polls, and with the same model’s projections in the 2016 election, providing added reference for more sophisticated insights and analysis.
Users can activate the new features using two toggles on the monitor situated below the election map, which highlight how the selected view differs either from official polls or the model’s projections for the 2016 election.
Why is this significant? Research has found that when our news analytics based projections and polls are in agreement they reinforce each other and suggest the result is more likely to happen. When the two contradict each other, however, forecasters should act with caution in placing too much weight on either - although of the two, our model has a better track record at selecting a winner than the polls.
The two screenshots below show how the map looks without the feature enabled and then with it switched on.
Note how in the second picture when the feature is activated certain states on the map become striped. In these states the monitor model projections are indicating a different result from the official polls - suggesting the polls, in particular, may be out.
Running a cursor over a given state brings up an information box with more detailed readings, including the percentage chance of each candidate winning according to the monitor’s projections, and next to it, the poll results.
In the example of North Carolina below, the monitor is predicting Donald Trump will win with a 72.87% probability, whilst the polls are showing a different result, with more support for Biden, who has 47.47%, compared to Trump’s 45.53%.
The divergence indicates an uncertain outcome, although our model has been shown to be the more accurate of the two in the past.
The same features can also be used on the ‘Sentiment’ and ‘Media Exposure’ map viewing settings.
Sentiment quantifies positive or negative news sentiment for each candidate in each state, and Media Exposure measures the percentage of news about a given candidate in each state. Both are tried and tested news analytics gauges for assessing the likelihood of which of two politicians is likely to win. Both Sentiment and Media Exposure contribute to the projections calculated by the model and are also available as separate viewer settings.
Once again, the interpretation is that when the news analytics gauges agree with polls they are mutually reinforcing and when they do not agree they raise doubts, especially about the poll readings.
The other new feature on the monitor is the ‘2016 Election’ toggle. This compares the 2020 to the 2016 model projections and inputs.
Why is this useful? It is useful because it shows how the same model performed during the last election when Donald Trump beat Hillary Clinton, and this is important to consider when interpreting current findings.
This is especially true because the current monitor projections are so one-sidedly in favour of Joe Biden, showing him more than 95% likely to win in most states and in the U.S. overall, which seems to raise the possibility the model may be over exaggerating his lead.
Yet, with the help of the 2016 Election feature, we can see that the same model generated projections that were much less extreme in 2016.
In 2016, the model correctly forecast Donald Trump as the winner, suggesting we should hold faith with the model despite its extreme readings this time round.
The screenshot below shows the new feature in action, with the states where the 2016 and 2020 readings differ highlighted with stars.
Hovering the cursor over one of the highlighted states shows a detailed breakdown of the two sets of stats. In Florida, for example, the model is predicting Joe Biden will win in 2020 with a probability of 72.18% whilst in the previous election it showed Trump as more likely to win at this time, with a 49.02% chance. In the end Trump won in 2016.
Above we have tried to show just some of the key ways in which the new monitor features can be used to provide more nuanced insights and analysis, to find out more why not visit the dashboard and try using the new features for yourself?
is the leading big data analytics provider for financial services. The company’s products allow clients to enhance returns, reduce risk, and increase efficiency by systematically incorporating the effects of public information in their models or workflows. RavenPack’s clients include the most successful hedge funds, banks, and asset managers in the world.
Please use your business email. If you don't have one, please email us at email@example.com.
We will process your personal data with the purpose of managing your personal account on
RavenPack and offering our services. You can exercise your rights of access, rectification,
erasure, restriction of processing, data portability and objection by emailing us at firstname.lastname@example.org. For more information, you can
Your request has been recorded and a team member will be in touch soon.
High inflation has returned in developed markets after decades of lying low. In our latest paper, we show how to build an inflation-based asset allocation strategy using sentiment data and we illustrate that sentiment-based strategies outperform models that depend merely on past observed inflation values.
This year's RavenPack Research Symposium brought two intense days of knowledge sharing in London and New York, from 25 top experts in natural language processing, quantitative investing and machine learning. Together, we explored how firms can leverage new language models to generate alpha, better manage risk and respond to calls for more sustainable investment practices.
Human capital is at the heart of value creation. Our latest research demonstrates how unprecedented workforce insights, sourced from over 200 million job postings, can generate more alpha.