AI Economics Research Podcast

This episode dives into a new Bank of England research paper exploring whether Large Language Models like GPT-3.5 Turbo can form inflation perceptions and expectations similar to human households. Using a clever quasi-experimental design, the study compares LLM outputs to survey data, revealing how AI responds to economic signals and its surprising sensitivity to food inflation. Discover the implications for economic forecasting and social science research, and share your thoughts at feedback@econpod.org. Find the full paper here: https://www.bankofengland.co.uk/-/media/boe/files/working-paper/2026/inflation-attitudes-of-large-language-models.pdf This episode explains a real academic paper in plain English for a general audience. Source paper: Inflation attitudes of large Nikoleta Anesti, Edward Hill and Andreas Joseph - Bank of England Keywords: inflation, large language models, AI, economic forecasting, central banking, macroeconomics

This episode breaks down a research paper from the Federal Reserve Bank of St. Louis, exploring whether different measures of the money supply are useful for forecasting US inflation. Using advanced non-linear techniques, the authors find limited support for monetary aggregates as reliable inflation predictors in the early to mid-2000s. Dive into the specifics of this intriguing macroeconomic study at https://fedinprint.org/item/fedlwp/10440/original and share your thoughts at feedback@econpod.org. This episode explains a real academic paper in plain English for a general audience. Source paper: FEDERAL RESERVE BANK OF ST. LOUIS Does Money Matter in Inflation Forecasting? - FEDERAL RESERVE BANK OF ST. LOUIS https://doi.org/10.20955/wp.2009.030 Keywords: inflation, money supply, forecasting, monetary policy, central banking, macroeconomics

This episode dives into a Federal Reserve Bank of St. Louis research paper that examines how to make economic forecasts more accurate, particularly in an environment of structural change. We break down how combining "recursive" (using all available data) and "rolling" (using only recent data) forecasting methods can significantly improve prediction quality. Learn about this innovative strategy and share your feedback at feedback@econpod.org, or read the full paper at https://fedinprint.org/item/fedlwp/9611/original. This episode explains a real academic paper in plain English for a general audience. Source paper: FEDERAL RESERVE BANK OF ST. LOUIS Improving Forecast Accuracy by Combining Recursive and Rolling - FEDERAL RESERVE BANK OF ST. LOUIS https://doi.org/10.20955/wp.2008.028 Keywords: forecasting, macroeconomics, structural change, central banking, model averaging, forecast accuracy

This episode explains a Bank of England Staff Working Paper, "Non-standard errors," by Albert J Menkveld et al., exploring how variations in researcher choices introduce significant "non-standard errors" into scientific findings. We break down their study of 164 teams testing hypotheses on the same data, revealing these errors are substantial, decrease with peer feedback, and are often underestimated by participants. Have thoughts or questions on how researcher bias impacts economic insights? Send them to feedback@econpod.org. This episode explains a real academic paper in plain English for a general audience. Source paper: Keywords: non-standard errors, research methodology, scientific uncertainty, economics research, central banking, data analysis

This episode breaks down a new Bank of England research paper that explores how economic cycles, like recessions, impact the stability of earnings for UK households. It reveals that during downturns, negative income shocks become more frequent, even if the overall spread of income changes remains similar, and introduces a new model to track these crucial dynamics for policymakers. Tune in to understand why this matters for financial stability and macroeconomic policy, and share your thoughts with us at feedback@econpod.org. This episode explains a real academic paper in plain English for a general audience. Source paper: Modelling income risk dynamics in - Bank of England https://www.bankofengland.co.uk/-/media/boe/files/working-paper/2025/modelling-income-risk-dynamics-in-the-uk-a-parametric-approach.pdf Keywords: Income Risk, UK Economy, Macroeconomics, Financial Stability, Central Banking, Earnings Dynamics

This episode delves into a groundbreaking research paper from the Bank for International Settlements, exploring how artificial intelligence is being harnessed to monitor and predict financial market stress. We break down a novel approach that combines recurrent neural networks (RNNs) to forecast market dysfunction with large language models (LLMs) to explain the underlying drivers. Discover how this dual AI system can identify "canaries in the coal mine" like deviations in currency arbitrage, offering early warnings for financial stability. This episode explains a real academic paper in plain English for a general audience. Source paper: Harnessing artificial - Bank for International Settlements https://www.bis.org/publ/work1291.pdf Keywords: Financial Stability, Forecasting, Artificial Intelligence, Machine Learning, Central Banking, Market Stress

This episode dives into a new Bank for International Settlements (BIS) research paper investigating Artificial Intelligence (AI) adoption across European firms. Learn how AI boosts labor productivity by 4% without displacing jobs in the short term, primarily through 'capital deepening.' We explore the uneven distribution of these gains, the role of complementary investments, and the critical implications for economic policy. This episode explains a real academic paper in plain English for a general audience. Source paper: AI adoption, productivity - Bank for International Settlements Keywords: Artificial Intelligence, Productivity, Employment, Europe, Macroeconomics, Digital Transformation

This episode breaks down a Bank of England research paper that significantly improves economic forecasting methods. It explains how economists use advanced statistical models, factor augmentation, and account for asymmetric loss and volatility to create more accurate predictions. Tune in to understand how these sophisticated adjustments enhance forecast performance across a range of economic variables. This episode explains a real academic paper in plain English for a general audience. Source paper: https://www.bankofengland.co.uk/-/media/boe/files/working-paper/2018/predictive-regressions-under-asymmetric-loss-factor-augmentation-and-model-selection.pdf Keywords: Forecasting, Macroeconomics, Asymmetric Loss, Volatility, Predictive Regressions, Central Banking

This episode dives into a Bank of England Staff Working Paper exploring how machine learning techniques can enhance macroeconomic forecasting. We break down the paper's findings on using diverse data sources, from traditional surveys to text-based indicators from news articles, to predict economic activity. Discover which models and data combinations yield the most accurate predictions and why the biggest gains are often seen in the short term. This episode explains a real academic paper in plain English for a general audience. Source paper: https://www.bankofengland.co.uk/-/media/boe/files/working-paper/2021/forecasting-uk-gdp-growth-with-large-survey-panels.pdf Keywords: Forecasting, Machine Learning, Macroeconomics, Big Data, Economic Prediction, Central Banking

This episode breaks down a Bank of England research paper on using interpretable machine learning for economic forecasting. Discover how advanced AI models can outperform traditional methods by uncovering complex, time-varying relationships in data, specifically demonstrated through forecasting US unemployment. We explain how a new workflow makes these powerful predictions transparent and actionable for decision-makers. This episode explains a real academic paper in plain English for a general audience. Source paper: An interpretable machine learning workflow with an application to economic forecasting https://www.bankofengland.co.uk/-/media/boe/files/working-paper/2022/an-interpretable-machine-learning-workflow-with-an-application-to-economic-forecasting.pdf Keywords: economic forecasting, machine learning, interpretable AI, macroeconomics, central banking, unemployment

This episode explores a Federal Reserve Bank of New York research paper on the economic effects of legalizing mobile sports betting. We unpack how betting and consumer credit distress can spill across state borders, impacting credit scores and delinquency rates, especially for younger individuals. Understand the fiscal asymmetry created for states that haven't legalized, bearing costs without the tax benefits, all explained in plain English. This episode explains a real academic paper in plain English for a general audience. Source paper: Sports Betting Across Borders: Spatial Spillovers - Federal Reserve Bank of New York Staff Reports https://doi.org/10.59576/sr.1184 Keywords: sports betting, consumer credit, spatial spillovers, financial stability, delinquency, state taxation

New research reveals that rising global temperatures are significantly increasing the costs of international trade, a factor largely overlooked in climate impact assessments. This study combines historical trade and weather data to show how climate change, particularly through its effects on sea ports, makes shipping goods more expensive. Discover how ignoring this trade cost channel leads to a substantial underestimation of the true economic welfare impact of a warming planet.

Recovered episode