The fallout at Silicon Valley Bank has shaken the corporate and financing world, with billions impacted before the U.S. government stepped in to shore up deposits. Now the world looks to what just happened. In addition to being a story about investments and banking, what happened at the end of last week and over the weekend is also increasingly being understood as a story about media online. At the time of writing, Signature Bank and Silvergate Bank have also collapsed.
Let’s take a look at what’s going on. Here’s a great summary of why the collapse happened, and why it was so sudden.
SVB was exposed to $80-90Bn worth of low-performing bonds. An attempted sell-off raised eye-brows and some ill-advised messaging rang alarm bells. Investors recommended their network move funds out, sparking a run on cash, meaning the bank didn’t have enough to cover for its depositors. That led to wider panic, a halt in trading, and then the FDIC stepping in. In short, rumour and speculation spread across various media networks.
Spare a thought for the start-up founders who aren’t sure how to make payroll (the US system is bi-weekly), and more generally for anyone with savings there. Are there signs we could have seen? (Assuming banks that start with an ‘S’ will fail isn’t allowed).
Overtone looked at articles from thousands of news sources on everything that mentioned Silicon Valley Bank, from the beginning of March until the clock struck midnight (GMT) on Friday 10 March, by which point the Bank of England had also shuttered SVB’s branch in the UK.
Silicon Valley Bank made an announcement about its books on Wednesday, and outlets began covering it in a straightforward factual way on Thursday (the blue line in the graph below). That night, however, we saw a significant rise in reporting and opinion-reporting signals, such as this article full of reporting signals that cites sources about moves made by Peter Thiel. The opinion-reporting pieces were often related to articles discussing investor panic, which were widely shared and used for follow-up pieces.
Consider the point of view of an investor. If you have stock in the bank itself, or investments in their clients and the bank’s immediate network. You want to safeguard your money as soon as possible. Let’s say you’re tracking SVB news using Overtone’s qualitative data. You could have put together the story Thursday night that this was a likely bank run. More to the point, it was a bank run that was gathering steam. That would likely have led to you taking the decision many others did by the following morning, trying to extract $42Bn from the bank. You would likely have succeeded.
The issue here is that, to some extent, there was no story until the news cycle made it a story. Using Overtone qualitative data, we were able to see that this was the beginning of a serious news cycle, with a press release leading to reporting leading to reaction. Please reach out if you would like to see your own data next to ours for even more understanding of what happened. We should not forget many other factors at play (the bank’s own actions and press releases, social media chatter, private network messaging on Slack and WhatsApp, and others).
But these things happen anyway, and don’t detract from the findings. As shown in the chart below, there was a significant spike in articles carrying reporting signals (and a smaller, noticeable bump in opinion-reporting too) the night before panic struck.
These anomalies allow us to understand a few things.
- Our model can be used to track future bank runs and assess whether investors should cut their losses. We’ll track the ongoing situation to help flag upcoming wobbles and jitters.
- We can more easily see the effect of publications from earlier time zones on business – such as Singapore and India for Europe, and Europe for the US, honing in on individual publications and their impact on the news.
- We can track which types of articles (mostly factual, or reporting, or opinion, say) steer the news cycle.
Most of all, it shows us how the media can throw fuel on fires. When this happens, it seems nearly impossible to stop. Herd mentality steps in: tremors turn to stampedes. Seen from an individual’s perspective – retail investor, startup founder, VC partner – it makes sense to run. Who wants to lose money?
But the news arc is composed of so many perspectives and narratives that to make sense of them – and to see when you could have known what was coming with a degree of certainty – you have to take a step back and look through a wide angle lens. Aggregated over time, these signals show a story that often plays out in the media. The earliest relevant wires receive outsized attention. Engagement rises for content that is referenced more, and those pieces are referenced more when they take a strong position. Reports of investors back-channelling advice to their portfolios do get noticed. Rumours of a bank run do get amplified.
At that moment, there are two stories: the story of what’s happening to Silicon Valley Bank, and the story of which reports and narratives are being picked up.
The key insight here is that it is very hard to tell these two stories apart. To do so, you’d have to read all the news, and have a way of filtering relevant content from everything else.
When the narratives blur, a media outlet feels a professional obligation to include both threads. And so, the hearsay and chatter – it’s irrelevant at this point whether it’s true or not – get a justified platform of their own.
The ball is rolling: no amount of “stay calm” press releases and investor calls can stop an agitated online community from rushing to draw its conclusions.
As we see in the graphic, the rush starts within 12 hours of the initial signals in the press. Soon after that, markets in the US wake up, an overwhelming number of depositors attempt to withdraw funds, and it’s all over for the 40-year-old bank.
The lesson here is: when we see a sharp increase in reporting signals late at night, we would do well to carefully analyse what’s going on (and on and on). Add to that a less pronounced rise in reporting-opinion news – the earliest indicator that news aggregators and re-posters are hastily rewriting earlier copy to suit their own audiences – and you have the ingredients for a media frenzy.
It’s no exaggeration to say that the collective media bears the brunt of the responsibility for the demise of SVB. This is the autopsy of a bank run, clearly played out in real time. The media autopsy is made possible by concentrating on direct content metrics.
These metrics include the type of article that readers, publishers and investors should rely on, as well as which type of article can spark fear. Article type is an example of qualitative data. This highlights the need for qualitative analysis of content in addition to (and in some cases instead of) quantitative metrics.
