A1713
Title: Comparative evaluation of open-source and proprietary software for ARDL, EC models, and bounds test for cointegration
Authors: Kleanthis Natsiopoulos - University of Dundee (United Kingdom) [presenting]
Nickolaos Tzeremes - University of Thessaly (Greece)
Abstract: The Autoregressive Distributed Lag (ARDL) and Error Correction (EC) models are widely used in time-series econometrics, particularly for analyzing level (cointegrating) relationships between variables in the presence of mixed levels of integration. Since its release, the ARDL package for R has become a reliable and versatile tool for implementing these models, offering an intuitive framework for estimating ARDL and EC models, automatically selecting optimal lags and conducting bounds tests for cointegration. The package also provides seamless integration with other R packages for post-estimation diagnostics, further enhancing its utility for applied econometric analysis. Additionally, a proof-of-concept (PoC) comparison of the ARDL package with other open-source R packages and proprietary software (EViews, Stata) is presented. The ARDL package demonstrates accurate and consistent results, surpassing other open-source alternatives and offering several advantages over proprietary tools, such as exact sample critical values and additional model representations. The comparison uses the original data from the seminal bounds test paper.
