Stay informed with free updates
Simply sign up to the Global Economy myFT Digest — delivered directly to your inbox.
Negativity sells newspapers. We’ve long known this statement to be true, with both demand and supply reasons as to why.
The demand side is underpinned by the psychological finding that “bad is stronger than good” — “Bad emotions, bad parents, and bad feedback have more impact than good ones, and bad information is processed more thoroughly than good”.
In our evolutionary past, this was beneficial — you would remember the spot where you heard there was a man-eating lion or you would be eaten.
Today, it translates into more clicks for negative headlines. As our colleague John Burn-Murdoch has previously noted, optimising for engagement in the modern world means meeting the demand for more negativity.
On the supply side is darkness. Journalists provide a vital civic service by shining a light on murky stuff and holding the powerful to account. There are just a lot of scandals to uncover.
Naturally, the FT does not stand apart from other papers in terms of tone. We’ve distilled the last 40+ years of FT content (over 3mn articles) down to a single economic sentiment series, and generally it is negative.
Below is what we’ve called the FT “macro mood”. The figure shows average weekly sentiment since 1982, with sentiment troughs highlighted. Below 0 means the mood was negative:
We’ve constructed this series over at the FT’s Monetary Policy Radar not to make the point that the news as reported in the FT is mostly bad news. Instead, what it shows is how bad the news is in relative terms.
This turns out to be a very useful thing to know. Here is the correlation between daily FT sentiment and Wall Street’s so-called fear gauge, the Vix index. The relationship is strong and statistically significant — you can explain nearly 20 per cent of the daily Vix close price with just variation in daily FT sentiment alone!
FT sentiment is also useful in predicting things. Remarkably, a simple model which uses sentiment to predict US CPI inflation over the next 12 months is better than the Fed’s own staff forecast and a standard benchmark model, as is a model that also includes a breakdown of topics covered in the FT.
(None of these models are good in an absolute sense, it needs to be noted, because predicting inflation over the next year is very hard.)
The macro mood time series is also notable because it tells us that our mood has yet to recover from the 2007-8 global financial crisis. Average sentiment now tends to be far below the pre-crisis, 1982-2006, period (the black horizontal line in the above figure).
Indeed, in only one stretch did average weekly sentiment recover on a sustained basis. This was in the second half of 2017, when the US stock market was mid a bull run. It might also reflect that Theresa May’s bungled electoral effort in June of that year appeared to lower the chances of a hard Brexit, and Trump’s first go at being president was being constrained by adults in the room.
The 2017 sentiment recovery turned out to be a blip, of course, with 2018 and 2019 seeing the start of Trump trade wars and the anointment of Boris Johnson as UK prime minister.
Why has our mood yet to recover since the financial crisis? One immediate answer is that the news has just been that much worse — Covid, inflation, rising populism, et cetera.
The FT’s souring mood might also be seen as part of a wider negative turn in newspaper media since around 1970 that has possibly been driven by increased media competition, as previously reported by Alphaville.
The other answer is that the financial crisis was really so monumental, with the repercussions still playing out in terms of reduced societal trust, extreme politics, lost productivity, and just general malaise. Financial crises have large scarring effects, also it turns out for journalists.
joel.suss@ft.com
How we built the FT macro mood
We’ve done this using some of the latest sophisticated methods for extracting data from text (embeddings, fine-tuned large language models). A general sentiment score is attached to each article, ranging from -1 (the world is going to hell) to 1 (everything is grand). A second model categorises how close each article is to what you might think of as the Platonic ideal of economics articles. That number ranges from 0 to 1 and is used to weigh each article before aggregating to daily or weekly levels. We are updating the macro mood daily over at the FT’s Monetary Policy Radar data page.
