Error-Correction Models Work When Time Series Are Properly Balanced

📌 What This Note Challenges

This article disputes Grant and Lebo's claim that the error correction model (ECM) cannot be applied to stationary data. The contention is that the ECM is entirely appropriate for properly stationary series — but only when stationarity is achieved through a balanced model.

📊 When ECM Is Appropriate

• The ECM can be valid for stationary data.
• Proper stationarity requires a balanced specification of the model; imbalance undermines the conditions that justify ECM use.

⚠️ Limits of Fractional Integration Techniques

• Fractional integration methods can be useful in some settings.
• These techniques have important weaknesses, particularly when applied to many of the time series commonly encountered in political science.

🔎 Two Often-Ignored Complications

• Low power: Standard time-series tests often have little power to detect differences given the sample sizes typical in political science.
• Overfitting: Small samples lead many analysts to overfit models — that is, to model random error or noise rather than underlying relationships.

📉 Why This Matters for Replications

Because of low power and the risk of overfitting in small samples, the results reported in the Grant and Lebo replications could easily be a function of overfitting rather than substantive differences in methods.