Doctor of Philosophy (PhD)


Agricultural Economics

Document Type



A large array of literature has revealed the complexity of export performance analysis. Using the case of Ghana, this dissertation, divided into three essays, seeks to provide the methodological guidance, empirical and simulation evidence necessary to analyze a country’s export performance. The choice of Ghana was motivated by the country’s growth experience, strong export and agricultural sectors and implemented reforms and programs since 1983. In the first essay, we created a new trade-weighted Cedi index or real effective exchange rate (REER) that takes into account Ghana’s most relevant patterns of trade and captures the evolution of Ghana’s export price competitiveness overtime. Other factors of export performance have been identified collecting the perspectives of Ghana’s agricultural export sector stakeholders and using grounded theory. This research showed that Ghana’s export price competitiveness, as depicted by the REER, has improved since 1983 but has revealed many additional factors that played a role in the performance of Ghana’s agricultural export sector. Following export demand theory and the procedure of Toda and Yamamoto (1995) and Dolado and Lütkepohl (1996) (TYDL), the second essay estimates causal relationships between exports, the REER and foreign activity over the 1970-2009 period. Two additional models (VAR-GARCH-in-mean) were estimated to investigate the impact of exchange rate volatility on Ghana’s exports. The results support the view that some of the implemented macroeconomic reforms have been the cause of Ghana’s export performance. Additionally, we found that third-country exchange rate volatility has hampered Ghana’s export growth. The third essay tackles methodological shortcomings of the TYDL procedure. In a Monte Carlo experiment, we compared the Schwarz Bayesian criterion (SBC) and the likelihood ratio (LR) tests in terms of their lag order frequency distributions and the finite sample properties of the resulting modified Wald (MWALD) tests. We found that in general, the SBC selects the true lag length more often than the LR tests and that in large samples the choice of the lag selection method does not influence non-causality tests results. This research also revealed that in the presence of moving average terms or in the case of mixed unit-root processes, MWALD tests perform poorly.



Document Availability at the Time of Submission

Release the entire work immediately for access worldwide.

Committee Chair

Zapata, Hector O.