Local lockdown support funding

In this blog, academic David Comerford uses behavioural economics to argue that, if the Government provides ungenerous compensation to areas where it imposes local lockdowns, this will reduce cooperation and trust and lead to higher costs in the longer term. Behavioural economics explores the psychology of economic choices and reactions, to examine how people actually react when given real-world options. For example, Comerford reports an experiment that shows that people may not accept an outcome that benefits them if they believe it is unfair. In the experiment, one participant receives £10, and must decide how much to offer to another participant. If the second accepts, both players keep their agreed shares. If not, both players leave with nothing. There is only one round to the game. Results show that a small offer, such as 1p, is rarely accepted—even though it would be a better outcome than the alternative (nothing). Comerford suggests this occurs because “there is a long-term benefit in signalling to others that we cannot be ‘bought’ cheaply. Rejecting a small sum of money sets a precedent”.

Applying this to local lockdowns, Comerford argues that people in the restricted areas will react according to how generous government support is seen to be. He says that if support is not perceived to be fair, this will lead affected populations to be “angry and uncooperative”. This, he says, is “dangerous”, when “cooperation with public health guidance has never been more important”.

Comerford therefore concludes that the Government needs to “act fairly and be seen to act fairly”. He recommends calculating support via a preset and published formula based on need, rather than through individual negotiations with each area. He believes that transparency is a key element of fairness. He also believes this approach would remove “the whiff of suspicion that the Government cares more about some parts of the country than others”.

Read the full article: David Comerford, ‘Local lockdown funding negotiations: What UK government should learn from behavioural economists’, The Conversation, 23 October 2020.

AI and economic forecasting

This article reports experiments that suggest computers using artificial intelligence to analyse newspaper content can improve on standard economic forecasting methods.

The authors state that newspaper content might be valuable in forecasting for three reasons:

  • it is more timely than conventional economic data;
  • it can affect the economic behaviour of readers, so can become self-fulfilling; and
  • it covers aspects of the economy that are not captured in official statistics, including, potentially, the seeds of future crises (“unknown unknowns”).

The experiments analysed more than half a million articles from the Daily Mail, Daily Express and the Guardian.

First, the authors used traditional techniques for assessing sentiment in newspapers, such as considering the balance of positive and negative words. They found the outcomes closely tracked other indicators of sentiment, but did not improve on traditional methods for forecasting key economic variables.

They then used a neural network (a form of machine learning) to allow the computer to decide which words to give greater weight in its analysis. They found that this, when combined with traditional forecasting methods, improved forecasts of variables such as gross domestic product (GDP). The results were similar, regardless of the newspaper analysed.

The models also suggested that the added value of text analysed in this way was greatest at “times of economic change”; for instance, during the financial crisis. The authors say this is “precisely when good economic forecasts matter the most”. They recommended including this type of analysis in economic forecasts, and taking the results seriously if they warned of an “incoming economic storm”.

Read the full article: Economic forecasting Arthur Turrell et al, ‘Machine learning the news for better macroeconomic forecasting’, Bank of England’s Bank Underground blog, 20 October 2020.