In this newsletter
Putting our money where our mouths are - Overtone scores and stocks
Economic articles on the Overtone scale
Overtone news
We’re looking for an engineer!
How Overtone data lines up with stock prices
Don’t mention it, but just as the world was emerging from the pandemic, it seems the U.S. is joining several other major economies in slipping performance and a slide towards greater market uncertainty. Against that backdrop of ups and downs, we decided to see what signals Overtone’s service could provide to help us better understand news in a finance context.
There are already many algorithms that financial professionals use for trading stocks a matter of milliseconds after news breaks. But, as financial professionals are not journalists, they do so without the sort of editorial differentiation that Overtone provides. We decided to have a quick look, and we’ve found a relationship between Overtone scores and certain, well-publicised stocks. This means that we can use Overtone scores to assess the impact of business journalism. At least for the stocks we chose, which all come from the Nasdaq.
TLDR: our algorithm shows that selecting only content with high Overtone scores tends to better track the stock prices of the company in that content, rather than taking all the content about that company.
Here’s an outline of our method.
We indexed two stocks that had seen major shifts in stock price over the last 12 months and news coverage related to economic trends such as emerging from the pandemic. One of these was Peloton, the home exercise equipment company which saw sales surge as people stayed home, but also stock dips around recalls. Another was Abbott Laboratories, a health company that received attention for issues such as the baby food shortage during a period of supply chain crunches.
We pulled up all the available articles that contained references to those companies, and calculated the Overtone depth score for each article. Here are examples of articles that scored a 1, such as the Street republishing a press release from Abbott, and a piece from Benzinga that was made by its “automated content engine.”
Abbott Names Robert B. Ford Chairman Of The Board - The Street
Here's How Much $1000 Invested In Abbott Laboratories 10 Years Ago Would Be Worth Today - Benzinga
There are also articles with many journalistic signals that were scored a 5. The first one here is a Bloomberg investigation that used freedom of information laws to obtain a report from health officials about Abbott and the second is a Crain’s Chicago piece.
Tainted Baby-Formula Risk Was Seen Months Before Abbott’s Recall - Bloomberg
Nationwide lawsuits over baby formula head to Chicago court - Crain’s Chicago
As you can see, those two groups of articles are appreciably different editorially, and we added our datapoint on that difference to others it is possible to gather on articles, such as a rough score of the “sentiment” of each piece from very negative to very positive. This resulted in a dataset that contained tens of thousands of articles. We then analysed correlations between Overtone scores, which we weighted with our confidence level, and indexed all these data to the Nasdaq over the period (Summer 2021 – Summer 2022).
Finally, we looked at the same data for all articles (those with any Overtone score, from 1 to 5) and compared those to the scores from only the highest scoring content (those with an Overtone score of 4 or 5). Focusing on these higher scores effectively could act as a “filter” to focus on the most in-depth news, rather than quick updates or aggregation.
High scoring content saw around a 17% better ‘performance’ in terms of tracking stock price.
The conclusion: using Overtone scores as a proxy for a company’s stock price better matches real world data when we use only the highest rated content on the Overtone scale, than when we use content selected at random from all articles about that company. That leads us to believe there may be a benefit in using our technology to sort through news for finance use cases.
We’re expanding our research and we’ll be publishing a white-paper with many more details. We hope this insight is helpful to you, and welcome comments and questions. You can always email us: contact@overtone.ai.
Chart showing indexed stock price (blue) and highest Overtone scores (red) compared to all content (yellow) for September ‘21 – April ‘22:
Share Overtone with your friends who might be interested!
Economic Articles on the Overtone scale
Low-scoring articles from the last week
There is absolutely nothing wrong with articles that receive lower scores on the Overtone scale, though they lack the journalistic signals that higher scoring articles have. As it pertains to economic and business articles, this often means repeating information off of a stock ticker, press release or series of tweets.
Rupee hits 79.81 against US dollar in early trade on fall in crude oil prices - Business Today India
Carrefour Brasil posts small rise in Q2 profit, notes inflation pressures - Reuters
Bill Ackman says Powell must show resolve against inflation, can't signal Fed will reverse course - CNBC
Higher scoring articles from the last week
Articles with higher scores (all these received a 4 or 5) are bringing new information and analysis to the fore. These are not based on the outlet and its reliability (you can see Reuters has both high and low articles) but on what the individual article is accomplishing.
Exclusive: Biden EPA to tackle coal industry carbon with rules on other pollutants - Reuters
Return to coal or reinvent itself? The simmering feud over an Alberta mining town’s future - The Globe and Mail
Why are Louisiana electric bills sky high? ‘This is the most I’ve ever paid in my life’ - The Advocate