Why Simple ADL Trumps ARFIMA for Short Fractionally Integrated Series

🔍 What This Paper Asks

The literature recommends ARFIMA modeling to detect fractional integration and then (a) fractionally difference the data or (b) estimate a fractional error-correction model when variables are cointegrated. However, prior work also shows ARFIMA can give misleading indicators of fractional integration in series with fewer than 1,000 observations. The core question here is whether a simpler approach—the autodistributed lag model (ADL) or its equivalent error-correction model (ECM)—can still recover useful immediate and long-run effects without first diagnosing or correcting for fractional integration.

🔬 Simulation Test of Short Samples and Fractional Integration

• Simulated time series that are fractionally integrated but stationary were used to evaluate estimator performance.
• Performance of the simple ADL/ECM estimators was compared to the implications of the ARFIMA-based recommendation, taking into account the documented problems ARFIMA faces in samples under 1,000 observations.

📌 Key Findings

• The ADL (or equivalent ECM) can, without prior testing or fractional differencing, provide useful estimates of both immediate (short-run) and long-run effects of weakly exogenous covariates in fractionally integrated but stationary data.
• These favorable results hold in settings where ARFIMA diagnostics may be unreliable due to small sample sizes.
• The implication is not that ARFIMA is invalid in large samples, but that the simple ADL/ECM is a robust practical alternative in many applied contexts with limited observations.

✳️ Why It Matters

Applied researchers working with time series that may exhibit fractional integration but have modest sample sizes can often rely on ADL/ECM approaches to estimate short- and long-run relationships without the risk of ARFIMA-induced misdiagnosis. This offers a pragmatic modeling strategy when ARFIMA-based testing is likely to be misleading.