Could the music you are listening to right now hold the clue to avoiding the next financial crash? Does the success of historical escapist dramas like “The Crown” reveal a yearning for an end to austerity that fiscal policy-makers should be respecting?
These are not fanciful questions - they are precisely the issues raised in a speech last month given by Andy Haldane, chief economist of the Bank of England. In it, he recognises how much emotion influences behaviour and decisions - often more than reason does - and ponders whether an insight into such moods might be gained from big data sources. “Popular narratives which develop in financial markets and in everyday public discourse have been found to be important empirical drivers of fluctuations in asset prices and in economic activity,” he told an audience at the newly-launched Data Analytics for Finance and Macro Research Centre, part of King's College Business School.
As a specific example, Haldane noted that, “data on music downloads from Spotify has been used, in tandem with semantic search techniques applied to the words of songs, to provide an indicator of people’s sentiment. Intriguingly, the resulting index of sentiment does at least as well in tracking consumer spending as the Michigan survey of consumer confidence.”
To a policy-maker who has decide how to control the flow of money in the economy, such non-traditional data sources and data science techniques are of interest for one very good reason - the old model they relied on failed spectacularly in 2008. As Haldane himself acknowledged: “During the global financial crisis, it is now fairly well accepted that the workhorse Dynamic Stochastic General Equilibrium (DSGE) model of the macro-economy fell at the first fence. It was unable to account for business cycle dynamics, during or after the crisis.”
What they learned from the crash is that another hidden hand was at work - traders filling their own pockets.
Facing their discipline being discredited, economist have begun to look at new ways in which to understand how markets operate, especially the “hidden hand” at work - a concept introduced by Adam Smith in the 18th Century to explain why demand and supply reach equilibrium. What they learned from the crash is that another hidden hand was at work - that of traders filling their own pockets regardless of the risks to their companies or the economy as a whole.
So does streaming data from Spotify or viewing data from Netflix offer a robust alternative? One of the best-known examples of trying to track real-world events through online behaviour does not give much confidence that they can. Google Flu Trends enjoyed several years of spectacular success in beating the US Centre for Disease Control in predicting outbreaks two weeks earlier than the traditional model. Then in 2013 it missed by more than 140% and was eventually axed by the search engine.
Post-analysis revealed at least one of the reasons was that the way data is generated in such platforms is far from pure. Google had developed its auto-search facility that prompted users on terms they were going to input to save them having to type them out. This over-indexed flu-related terms. In the back-end, analysts had seen the correlations and over-fitted their data to the model. All involved also overlooked that on-the-ground medical researchers were having a direct impact on flu outbreaks and their treatment.
With streaming data, the mistake would be to assume that users are leaning into these services and therefore indicating their mood. A friend of mine is a youth worker and musician who routinely asks any young adult using the services he delivers, if they are wearing headphones, what they are listening to. Much to his dismay, the answer is regularly just “Spotify”. No knowledge of what artist, let along song or even its lyrics.
To assume top streams show how people feel not only risks classifying everybody as an Ed Sheeran fan (don’t get me started…), but also that they are active, rather than passive in their listening behaviour. Given the dominant mode of using Spotify Discover or Deezer Flow to be streamed tracks without having to chose them, it seems less a guide to mood and more an indicator of abdicating choice.
Lady Gaga made just $167 for each 1 million plays that “Poker Face” got on Spotify.
Another hidden hand is also likely to be at work - the small proportion of users who not only actively choose, but like tracks and publish their playlists to their social networks. When you realise that a quarter of Spotify’s music inventory has never been played (giving rise to minority-supporting initiatives like forgotify.com), you realise that demand and supply in streaming services is very far from being in balance.
There is one final reason why the Bank of England needs to use caution in relying on these data sources for its big calls on the economy - the fact is that very few of the artists producing tracks see as much benefit from streaming as they do from direct purchases. According to a report in The Guardian early in the life of Spotify, Lady Gaga made just $167 for each 1 million plays that “Poker Face” got on the platform.
With so little to lose, that may be why she decided to leave art-pop behind and move into jazz. But when passive demand combines with under-rewarded supply, who can trust that the economy should be run on the basis of this data?