When Rochdale is in the news it tends to be for the wrong reasons, such as associations with child sex exploitation and urban decay. Yet, in its heyday, Rochdale was one of the most prosperous places on earth. The town hall – a magnificent example of Victorian gothic – exudes civic pride. But it was built a long time ago, when Britain was the workshop of the world, cotton was king and the north of England was more prosperous than the south.
Over the past century or more, the economy’s centre of gravity has shifted and Rochdale faces problems familiar to many places on the wrong side of the geographical divide: how to replace old industries with new ones; how to break out of a low-wage, low-skill pattern of employment; how to persuade the young people leaving for university to come back when they have finished their studies.
The policy response is fairly generic also: improving the transport links; a targeted reduction in business rates to reduce vacancy rates; work on a combined retail and leisure complex; the marketing of the town’s history to attract tourists.
Rochdale is different in one respect, however. Having been at its zenith during one machine age, it is now using another technological advance – machine learning – to measure local wellbeing.
Traditionally, if a local authority wanted to gauge public opinion it would commission an opinion poll, an expensive way of capturing a snapshot of sentiment on any one day.
The Rochdale feelgood factor does away with interviews and harnesses the power of social media instead. Thousands of tweets sent from the town are studied on daily, with algorithms learning which words have a positive or negative context. Whereas the public don’t always respond entirely honestly to opinion pollsters, machine learning detects their true emotional state.
The Bank of England is thinking along the same lines. In a recent speech about the possible uses of big data, Andy Haldane, Threadneedle Street’s chief economist, said researchers were already studying music choices from Spotify to test consumer confidence.
Rochdale’s experiment is being conducted by the economist Paul Ormerod – born and raised in the borough – and his colleague Rickard Nyman, who have used machine learning to compile a London feelgood factor and to study the 2017 general election.
“Recent developments in artificial intelligence and machine learning enable wellbeing to be measured directly from social media,” Ormerod says. “This is not only considerably less expensive than traditional approaches – the algorithms do much of the work – but the results can be generated in real time.”
While the Rochdale feelgood index has only been charted since last Christmas, Ormerod and Rickard have been looking at how people feel in London since 15 June 2016, a week before the EU referendum.
The study of tweets showed there was a sharp fall in the feelgood factor in the immediate aftermath of the Brexit vote, as might have been expected, given that London voted strongly for remain. There was another big drop in November that year when Donald Trump won the US presidential race.
However, on both occasions, sentiment quickly returned to its previous levels, suggesting that most people are not all that bothered by politics and, once the excitement is over, simply get on with their lives. This is consistent with the way in which consumer spending held up much better in the months after the referendum than predicted by most forecasters, including the Bank of England. Had the London feelgood index been available to the Bank in 2016, it might have thought twice about the need for an emergency cut in interest rates.
Ormerod and Rickard used the same real-time Twitter analysis for clients during last year’s general election. They found a strong correlation until just a few days before the poll between the proportion of tweets that mentioned Brexit as a topic and the Conservative lead in the polls.
Theresa May called the election to strengthen her Brexit negotiating hand and it looked as if the Conservatives were heading for a landslide for the first couple of weeks when Brexit was the dominant issue. However, voters decided the election was not only about Brexit and other issues – such as social care – started to show up more regularly in the analysis of tweets.
“The project was not designed to be predictive,” Ormerod says. “But it did nevertheless indicate that the Conservative lead on the day – which was in fact 2% – would be much lower than almost all opinion polls were suggesting.”
Rochdale’s destiny will not be determined by a feelgood factor, even one generated by smart machines, because its problems are deep and of longstanding. Thus far, the only conclusions that can be drawn from it are ones that most people could work out for themselves: Rochdale is happier on a Friday than a Monday, happier at Christmas than in January – when the credit card bills come rolling in – and permanently less happy than London.
It is early days, however. There was a rise in the feelgood factor when the borough was included in the Tour de Manc charity bike ride and the council wants to use the index to work out when it should hold events, how it carries out its services, how the public responds to its regeneration efforts and whether deprived parts of the borough register lower wellbeing.
At a bare minimum, the machine learning experiment will help the council make better-informed judgements about how it spends money. The idea may well catch on. It’s been a while since where Rochdale led others followed.
