New Delhi: Can you trust news articles to correctly predict economic gloom and doom? Should you celebrate when they forecast a period of prosperity?
The answer to those questions when it comes to Indian newsrooms would be mostly yes, according to new research by scholars of the Reserve Bank of India’s Department of Statistics and Information Management (DSIM), published last month in RBI Bulletin.
In their study, titled ‘When a News Story is More Than Just Text: Evidence from Indian Economy’, authors Geetha Giddi, Shweta Kumari, and Sasanka Sekhar Maiti found that in most cases, news articles can also provide real-time signals about the direction of certain economic indicators, and changes that often take longer to reflect in the official data.
However, the authors of the paper, which was released last month, point out that sentiments in news stories “weathered off slightly” during the Covid-19 period, mainly because of the uncertainty the pandemic brought.
Also read: No religious divide: 46% Hindus & Muslims feel ‘Modi govt portrayed too favourably by news media’
How researchers came to their conclusions
The study uses news articles published online from 2015 to 2021 to construct a sentiment index around eight economic indicators.
Five of these — retail inflation, index of industrial production, economic growth, consumption growth, and business expectations and assessments — are from the real sector (relating to the production of goods and services). Two are from external sectors such as export and import growth and one is from the financial sector in the form of bank credit growth.
The authors measured the frequency of words in the published report to identify whether the news contained a positive, negative, or neutral sentiment. They used this to arrive at what they called a Media Sentiment Index (MSI) — a measure calculated on a monthly basis.
Change seen over a year defines the direction of the MSI. For instance, if positive coverage on exports outweighs the negative coverage of the previous year, there’s a strong chance that exports that year are likely to rise — a signal that authors found often reflected in the government’s data released days later.
‘Accuracy is substantially high’
When MSI was compared with the actual official data, the authors found a “strong and significant correlation” between sentiments in online publications and actual statistics. This suggests that news sentiments do have the power to predict the volume and direction of at least these eight indicators.
“The accuracy is substantially high, ranging between 70-90 per cent, for all variables,” the study says, except when it comes to Gross Domestic Product (GDP) growth.
At a time when analysts and policy-makers are trying to ascertain the impacts of various economic shocks, news sentiment data could provide a real-time indication of certain parameters with great accuracy, the authors conclude.
“As the sentiment indicators can be constructed on a high-frequency basis, prior to the release of official statistics and traditional indicators, it is worthwhile to track them regularly for early signals on economic conditions,” the authors said in the study.
“Further extensions can be carried out by incorporating other commonly used measures of confidence, sentiment, and expectations of market agents as available through business tendency surveys, and estimating wholesome empirical models aiming at higher prediction accuracy,” the study added.
(Edited by Uttara Ramaswamy)
Also read: Govt media advisory body gets new name, Central Bureau of Communication. Aim is ‘uniformity’