Time-frequency domain analysis of exchange rate market integration in Southern Africa Development Community: A Hilbert-Huang Transform approach

Abstract

The desire of most African economic communities to introduce a common currency has persisted for years. As postulated by the Optimum Currency Area hypothesis, coordination of policy indicators among member countries is desirable for stable monetary union. In this regard, the integration of exchange rate markets has been studied and cited as one of the key indicators that could signal economic integration. Therefore, analysis of similarities, interdependence, and information transfer across exchange rate markets in Southern African Development Community (SADC) is a necessity to measure the extent of integration in the region. However, the intrinsic complexity of exchange rate data generation and its stylised characteristics of non-stationarity and non-linearity influence the modelling of such data in terms of the accuracy of the analysis and the embedded policy direction. In response, this thesis proposes empirical mode decomposition-based market integration analysis to address the limitations of the existing literature which fails to recognise the heterogeneity of market participants and data generation of the exchange rate in SADC. The data employed for the thesis are the daily real exchange rates from 15 out of 16 member countries of the SADC from 3rd January, 1994 to 7th January 2019. The choice of study window and countries was based on the availability of adequate and consistent data for robust analysis and the period after South Africa, the largest economy, joined SADC. Based on the criteria, Zimbabwe was excluded from the analysis. To achieve the purpose of this thesis, a four-step approach was used. The first step reviewed and explored the non-stationarity and non-linearity stylised facts about the data and observed that exchange series in SADC are non-stationary and non-linear. The second stage compared the performance of two Hilbert-Huang Transforms (EMD and EEMD) to decompose SADC exchange rate markets of which EEMD emerged superior. The components of the decomposed series were examined for dominance and ability to define the exchange rate trajectory in SADC. The residue of all the markets explained over 80% of the variation of the original series except Angola. The short- and long-term comovement was analysed through the analysis of the characteristics of IMFs and residues. The analysis of the IMFs and residues obtained from EEMD showed that exchange rate markets in SADC are driven by economic fundamentals and 12 out of 15 countries examined showed some level of similarity in the long-term trend. In the third stage, EEMD-DCCA based multifrequency network was introduced to study the dynamic interdependence structure of the exchange rate markets in SADC. This was done by first decomposing all series into intrinsic mode functions using EEMD and reconstructing the series into three frequency modes: high, medium, and low frequency, and residue. The DCCA method was used to analyse the cross-correlation between the various frequencies, residues and original series. These were meant to address the non-linearity and non-stationarity in observed exchange rate data. A correlation network was formed from the cross-correlation coefficients to reveal rich information iii than would have been obtained from the original series. The results showed similarities between the nature of cross-correlation between high-frequency series mimicking the original series. There was also a significant cross-correlation of long-term trends of most SADC countries’ exchange rate markets. The final stage proposed EEMD-Effective transfer entropy-based model to study exchange rate market information transmission in SADC at various frequencies. The combination of Ensemble Empirical Mode Decomposition (EEMD) and the Rényi effective transfer entropy techniques to investigate the multiscale information transfer helped quantify the directional flow of information at four frequency domains, high-, medium-, and low-frequencies, representing short-, medium-, and long-terms, respectively, in addition to the residue (fundamental feature). This revealed a significant positive information flow in the high frequency, but negative flow in the medium and low frequencies. Based on the findings of this thesis we recommend that EEMD based method be used in the analysis of financial data that susceptible to non-linearity and non-stationary to elicit the time-frequency information. In terms of policy towards monetary formulation, we recommend a stepwise approach to monetary integration in SADC.