A Fractional Cointegration Panel Model With Fixed Effect

• Main Author: Fehintola, Olaniran Saidat
• Format: Thesis
• Language: English
• Published: 2023
• Subjects: H1-99 Social sciences (General)
• Online Access: http://eprints.usm.my/60451/1/24%20Pages%20from%20OLANIRAN%20SAIDAT%20FEHINTOLA.pdf; http://eprints.usm.my/60451/

Several authors have studied fractional cointegration in time series data, but little or no consideration has been extended to panel data settings. However, recent economics and financial panel datasets such as portfolio returns across firms, price indices and exchange rates across countries often exhibit long-memory properties. Therefore, this thesis aims to develop a fractional cointegrated panel model with a fixed effect assumption. The first objective was to compare the finite sample behaviour of existing fractional cointegration time-series test procedures in panel data settings. This comparison is performed to determine the best tests that can be adapted to fractional cointegration in panel data settings. Specifically, simulation studies and real-life data analysis were performed to study the changes in the empirical type I error rate and power of six semiparametric fractional cointegration tests in panel settings. The analysis findings revealed that the residual-based tests are useful for adaptation in a panel setting. Secondly, the best two residual-based time series fractional cointegration tests observed were implemented in panel settings using Monte-Carlo simulation experiments. The results of the experiments showed that one of the tests is valid for fractional cointegration order of less than 0.5, the other is generalized and accepts any fractional cointegration order within the range [0, 1] at varying sample sizes. Finally, a fractional cointegrated panel approach was developed for testing the absolute Purchasing Power Parity (PPP) model among 16 West African countries using data that spans 49 years (1971-2019).