The surge in artificial intelligence research within the financial technology sector has been remarkable. Banks are strategically cultivating ecosystems through partnerships with universities and international collaborations. Prominent players such as Capital One, JPMorgan Chase, BBVA and ING are embracing open-source collaboration to drive intellectual property development.
European banks are also making significant strides in AI research, with Intesa Sanpaolo taking the lead. Research papers authored by professionals within the banking industry span a wide spectrum, encompassing both theoretical and applied AI research. Applied research in 2022 encompassed areas ranging from quantum machine learning applications to innovative chatbot design and recommendation systems. Intriguingly, there is a noticeable absence of macroeconomics in these research programmes conducted by banks.
The competitive nature of AI research could lead to more secretive approaches, impacting the landscape of public AI research. Tech giants like Google, Meta, Mozilla and Hugging Face have traditionally championed open-source research, but this move towards secrecy is becoming a trend in AI research.
This raises a question of public interest: why should private institutions exclusively profit from public investment in university-driven open-source innovation? Shouldn’t they instead spearhead an unprecedented partnership between banks, universities and public institutions, aimed at jointly investigating the significance of AI in macroeconomic forecasting, addressing both theoretical and practical dimensions?
AI in macro forecasting
Central banks rely on a combination of microfounded econometric models and other tools for macroeconomic forecasts. However, the failure in recent years to predict inflation trends has led to calls for greater co-operation between fiscal and monetary policy and a re-evaluation of forecasting models.
Prominent figures like Janet Yellen, US Treasury secretary, and Mark Carney, former Bank of England governor, have acknowledged shortcomings in understanding inflation dynamics. Claudio Borio, chief economist of the Bank for International Settlements, has questioned the understanding of the inflation process and its root causes.
An International Monetary Fund paper, ‘How Well Do Economists Forecast Recessions?’, reveals the failure of economists to predict 148 of the past 150 recessions. It underscores the complexity of economic prediction, given idiosyncratic human behaviour, technological advancements and geopolitical factors.
Is the issue rooted in data and computational capacity, or is it more about theoretical limitations? The latter hypothesis suggests that AI can enhance macroeconomics but only to a certain extent. Accurate predictions require solid theoretical foundations, and the field is grappling with these challenges.
Policy-makers and major international institutions formulate forecasts by amalgamating inputs from diverse models, economic theories and pragmatic viewpoints. The resulting consensus informs policy decisions. This is where AI can play a greater role. By processing vast datasets, detecting patterns and incorporating multiple variables, AI can refine economic and monetary predictions. Central banks and governments stand to benefit from more accurate forecasts of indicators such as inflation, gross domestic product growth and unemployment rates. These enhanced models can inform policy choices, improving decisions on interest rates, liquidity management and macroprudential measures.
But what about macro theory? Shouldn’t private and public institutions invest more in co-operative research on AI to ensure a harmonious integration of macro and micro finance? If we do not attain improved outcomes at this crucial level, AI algorithms and intelligent systems may excel at predicting immediate and short-term transactions but could prove inadequate in managing medium- to long-term banking and financial risks. This situation is a cause for concern, not only within the banking industry but also for the wider economy, bearing substantial implications.
Revolutionising financial transactions
Without improved macroeconomic forecasts, the potential of AI to manage financial transactions using algorithms and intelligent systems would diminish its medium- and long-term value. In the short term, however, its applications are impressive. AI-driven trading systems have revolutionised trading activities by offering unprecedented speed, accuracy and innovation. These systems analyse vast datasets, identify hidden correlations and execute high-frequency trades.
Machine learning, particularly deep learning, underpins AI-powered trading, unearthing intricate market patterns and making rapid trading decisions. AI solutions excel in handling complexities in financial markets, combining data from technical and fundamental analysis with real-time sentiments from social media. The adaptability and sophistication of AI, seen through methodologies like artificial neural networks and sentiment analysis, enrich trading strategies.
The fusion of AI with quantum computing and emotional data from human-machine interfaces could be transformative. Nonetheless, finding the right equilibrium between AI capabilities and human insights remains essential when navigating the intricacies of financial markets, just as achieving a more refined balance between macroeconomic and financial forecasts is of utmost importance.
The Bank of England’s appointment of Ben Bernanke for an assessment of the models and data supporting its forecasts presents a compelling call to action. This moment provides an opportunity to drive both public and private institutions towards much more ambitious research initiatives in the realm of AI and macroeconomic forecasts.
Edoardo Reviglio is Visiting Senior Research Scholar at Yale Law School and a member of the OMFIF Advisory Board.