BoG researchers propose AI-powered system to track inflation in real time

Researchers at the Bank of Ghana (BoG) have proposed an artificial intelligence-powered framework to track consumer prices and estimate inflation in real time. The system, called e-Inflation, uses Big Data and AI tools to collect large volumes of price information from websites. It aims to provide timely estimates ahead of official inflation releases.

The proposal is outlined in a working paper by Francis Mawuli Abude, Collins Kwesi Appiah and Julius B. Dasah, published on the central bank’s website. The authors say preliminary results suggest the framework can estimate current inflation rates more accurately and at a lower cost. They also report that it can capture price movements during both stable and unstable economic conditions.

However, the Bank of Ghana notes that working papers reflect the authors’ views and do not necessarily represent those of the bank, its board or management. The papers present research in progress to encourage comments and debate. How the proposed framework will operate The e-Inflation framework draws price data from electronic sources to estimate the Consumer Price Index (CPI), headline inflation, food inflation and non-food inflation.

The authors say the approach could help reduce delays in traditional price collection and publication. According to the paper, official inflation figures are published after the reference month ends. This requires policymakers to estimate current price movements before official data becomes available. Central banks use statistical models, econometric models and machine learning to produce these estimates, a process known as nowcasting.

The researchers argue that existing methods can be costly and do not always deliver accurate, real-time estimates. They therefore argue for “more innovative and cost-efficient methodologies, leveraging Big Data and Artificial Intelligence (AI) techniques, tools, and technologies”. The paper says work involving Big Data and AI has been underway at the Bank of Ghana since May 2021.

This includes identifying alternative sources of consumer prices and comparing their coverage with the official basket of goods and services. The research also involves constructing overall, food and non-food consumer price indices. “The e-Inflation framework is capable of collecting price data every week for monetary policy decision-making and for general economic policy analysis,” the authors state.

They argue that the system could provide more frequent information to support policy decisions. “More importantly, relative to the traditional approach of compiling price data, the e-Inflation framework has the unique advantage of providing real-time CPI statistics as needed.” Challenges and concerns A key concern is whether online prices accurately reflect those charged in physical shops.

The authors acknowledge that “whether online prices are identical to those in physical shops remains the main concern of this research”. However, they say a survey found online and offline prices in Ghana to be statistically identical, supporting their confidence in the data. “Based on this, we are confident in the representativeness of the price data extracted from online sources and in its use in computing the CPI and inflation rates for Ghana.” The framework’s reliability also depends on businesses keeping their online prices current.

“Despite these advantages, the accuracy of e-Inflation statistics depends largely on the availability of current price data on the websites of the selected e-commerce shops and institutions.”