Bank of Ghana uses AI to improve inflation forecasts

The Bank of Ghana (BoG) has deployed artificial intelligence (AI) and machine-learning tools to improve inflation forecasting and strengthen the way it collects and analyses economic data.

First Deputy Governor Dr Zakari Mumuni said the technology was part of the central bank's wider adoption of advanced modelling and big-data tools to provide more accurate information for monetary policy decisions.

Speaking at the 4th Annual Statistics and Data Science Conference in Tamale, Dr Mumuni said AI had helped the Bank improve the accuracy of its inflation predictions, including forecasts made before official economic data are released.

“We also employ machine-learning models to complement standard econometric models in forecasting GDP and performing text-mining analytics,” he said.

The Bank is also using technology in financial supervision, where Dr Mumuni said more detailed and timely data were allowing potential risks to be detected earlier.

He said supervisors had previously relied heavily on static monthly spreadsheets that required manual reconciliation.

“Increasingly granular data can be validated as it arrives, allowing risks to be identified earlier,” he said.

The central bank continues to combine these technologies with traditional economic modelling.

Dr Mumuni said the Bank uses econometric techniques and its Quarterly Projection Model within a Forecast and Policy Analysis System to identify emerging trends, assess risks and examine the likely effects of different policy decisions.

“Through econometric techniques and our Quarterly Projection Model within a Forecast and Policy Analysis System, we identify emerging trends, assess risks and consider the likely outcomes of different policy choices,” he said.

But he cautioned against treating technology as a replacement for human judgement.

“Technology can strengthen our intelligence, but it does not remove the need for human judgment,” Dr Mumuni said.

His comments follow a commitment by the Bank's Governor at his swearing-in in February 2025 to adopt a more proactive approach to managing inflation, including through advanced data analytics and artificial intelligence.

Dr Mumuni said the challenge for policymakers was increasingly about making sense of the huge amount of information now available, rather than simply obtaining data.

“The greatest challenge facing policymakers today is no longer a shortage of data, but rather turning an abundance of data into timely, reliable and actionable intelligence,” he said.

He stressed that data by itself was not enough to produce sound economic policy.

The Bank also continues to gather information directly from communities and businesses across Ghana.

Dr Mumuni said researchers from the Bank's Research Department regularly visit markets, including in Tamale, to track prices and conduct business and consumer confidence surveys.

“Long before a survey appears in a published report, our Research Department staff are in markets across the country including here in Tamale tracking prices and conducting business and consumer confidence surveys,” he said.

He added that this approach was intended to ensure that the Monetary Policy Committee's decisions reflected economic conditions across the country rather than relying primarily on developments in Accra.

“It means staff spending nights away from home so that when the Monetary Policy Committee sits, it reasons from the country’s economic experience, not Accra’s alone,” he said.

Dr Mumuni also urged statisticians and researchers to ensure that new sources of data complement established statistical methods rather than replace them.

“New data should complement not replace properly weighted and nationally representative measures,” he said.

He called for closer cooperation between researchers and policymakers, arguing that both sides needed to challenge and understand the data being used to make economic decisions.

“Researchers should understand the questions confronting policymakers, while policymakers should remain open to researchers who ask uncomfortable questions of the data,” he said.