Associate Professor of Economics | Forecasting • Structural Modelling • Macroeconomics

Start here: 💻 MIDAS for Stata · 📺 YouTube · 🌐 Research

Rethinking 50 Years of Empirical Macroeconomics
Recent methodological advances reveal critical, overlooked biases in standard macroeconomic practices:

• Forecasts: Forecasts of period averages need to be tested against the end-of-period no-change, the optimal forecast for a random walk null. Instead, they were tested against the period-average no-change resulting in spurious predictability, calling into question macroeconomic forecasting results from the last 50 years.
• Structural: Working (1960) was wrong: a random walk (RW) does not aggregate to a moving average (MA) process. Consequently, macro model estimated on monthly or quarterly averages potentially mistimed shocks and introduced spurious endogeneity.
• Solution: Halve your forecast error and minimize endogeneity and mistiming through high-frequency techniques.

This page distributes the insights, data, Stata and R code, and techniques to elevate empirical macro.

Mixed-frequency data construction, estimation, and forecasting in Stata.

• bumidas — Bottom-Up MIDAS estimation and model selection.
• mfcollapse — constructs mixed-frequency datasets from daily, weekly, or monthly data.
• umidas — unrestricted MIDAS estimation.
• rmidas — restricted MIDAS with Almon, step, Legendre, exponential Almon, and beta weights.
• Supports direct multi-step forecasts, multiple high-frequency predictors, and BIC/AIC/HQIC selection.
• Includes examples, certification tests, simulations, a foreign-exchange application, and conference materials.

A deep dive into business cycle filtration.

High-frequency and real-time datasets for Econometrics and Machine Learning.

Key Advantage: Using high-frequency and end-of-month (EOM) data enables testing against the random walk hypothesis, reduces shock mistiming, avoids spurious endogeneity, and can improve forecasting accuracy by up to 40% compared to using monthly averages. Moreover, to ensure your empirical results are actually relevant, you must backtest using real-time data, relying strictly on pre-revision data that accounts for publication lags.

Optimized for accurate shock timing and testing the random walk null.

Backtest forecasts using only the information actually available at each point in time.

A streamlined gateway to Canadian Data.

Level up your data analysis and coding from "Zero to Hero."

• 📺 The Stata Economics Masterclass (Complete: 5 Videos + Code)
• 💻 The R Economics Masterclass (In Development: Starter code available now, videos coming soon!)

1. Automated Import & Cleaning: Stop wasting time on manual data formatting.
2. Debug Like a Pro: Identifying AI "slop" and structural errors.
3. Core Skills: Graphing, priors, and predictive modeling.
4. Monte Carlo: Using simulations to verify your results.
5. The "Copy-Paste" Intervention: Full automation of result reporting.