MNI INTERVIEW: New BOE Text Analysis Predicts Sentiment Faster

-Researchers Use Text-to-Data To Create Leading Sentiment Measures

-This Piece Is One Of A Series On The Data Revolution And The BOE

By David Robinson

     LONDON (MNI) - The Bank of England is striving to upgrade its real-time

economic forecasting, developing leading sentiment indicators that predate

commercially available alternatives by up to three months.

     Official growth and inflation data are published with long lags, so

'nowcasting' utilises softer indicators, such as business and consumer surveys,

to give policymakers a fix on what is happening in real time. BOE researchers

analysed newspaper articles touching on financial markets, commodities,

corporations, industry, and the economy to create sentiment indicators that run

well ahead of the familiar GfK and CBI reports.

     The text-based measures boosted the accuracy of simple economic forecasting

models, one of the researchers, Eleni Kalamara of the Data Analytics for Finance

and Macro Research Centre at King's College London, told MNI in an interview,

although she added: "We found we do better on predicting real economy variables

such as inflation and unemployment rather than predicting financial variables."

     DAFM was launched in April by Bank of England Chief Economist Andy Haldane

in a drive to link academic and Bank expertise. Both BOE and European Central

Bank economists were involved in the text-to-data research.

     The BOE's nowcasts have on the whole performed well recently but came

unstuck at the time of the June 2016 Brexit referendum, when, with uncertainty

soaring, they pointed to a slump in activity that never materialised.

     With political and economic uncertainty again high in the UK, the BOE has

been researching how it impacts economic activity.

     The text-based measures did not find strong correlations between

uncertainty mentions and what was going on in the real world.

     "Uncertainty is a very complicated idea. People can be uncertain about

different things and can have positive and negative thoughts," Kalamara said.

     While a business manager may, for example, express uncertainty about the

impact of new technology he or she could view it as broadly positive while

others may view it as negative, with offsetting economic effects.

     "That is why it is more difficult to capture it using text," Kalamara said.

     The researchers are now looking whether they can capture turning points in

the economy through text-to-data analysis, she added.

--MNI London Bureau; tel: +44 203-586-2223; email: david.robinson@marketnews.com

[TOPICS: M$B$$$,M$E$$$,MT$$$$,MX$$$$]