A high-dimensional dataset created by Koop and Korobilis (2023) that integrates predictive signals from various macroeconomic and financial sources.
inflation_dataA matrix with 245 quarterly observations (rows) and 462 signals (columns):
Transformed target variable: Total CPI (CPIAUCSL)
First and second lag of the target variable
Lagged and transformed signals from the sources listed above
External point forecasts available from 1976-Q1 to 2021-Q4 for quarterly Total CPI (CPIAUCSL), including:
Generated using regression trees, ridge regressions, and elastic nets over expanding and rolling windows
Based on models discussed in Koop and Korobilis (2023) such as Gaussian process regressions (GPR_FAC5), Unobserved Component Stochastic Volatility (UCSV), and Variational Bayes Dynamic Variable Selection (VBDVS_X)
The dataset includes data from the following sources:
FRED-QD dataset (McCracken and Ng, 2020)
Portfolio data (Jurado et al., 2015)
Stock market predictors (Welch and Goyal, 2008)
University of Michigan consumer surveys
World Bank’s Pink Sheet commodity prices
Key macroeconomic indicators from the Federal Reserve Economic Data for Canada, Germany, Japan, and the United Kingdom
The dataset is pre-processed for one-step-ahead forecasts and includes external point forecasts. It spans from 1960-Q3 to 2021-Q4.
Jurado, K., Ludvigson, S. C., and Ng, S. (2015) "Measuring uncertainty." American Economic Review, 105 (3): 1177–1216.
Koop, G. and Korobilis, D. (2023) "Bayesian dynamic variable selection in high dimensions." International Economic Review.
McCracken, M., and S. Ng (2020) “FRED-QD: A Quarterly Database for Macroeconomic Research” National Bureau of Economic Research, Working Paper 26872.
Welch, I. and Goyal, A. (2008) "A comprehensive look at the empirical performance of equity premium prediction." The Review of Financial Studies, 21 (4): 1455–1508.