Estimation of the PSTRVEC and regime wise granger-causality .

Get Complete Project Material File(s) Now! »

Equilibrium Exchange Rates and Misalignments

The next step is the computation of the exchange rates misalignments. We start by constructing the long-run equilibrium exchange rates as in Kamar & Ben Naceur (2007). The REER is the real eective exchange rate at any time t and is given by loget = ^+ ^ F0 t where F stands for the long-run fundamentals and the corresponding parameters are the estimated regression coecients. Therefore, the misalignments are given by Misalt = log(et)?log(~et). Positive values of misalignments indicate undervaluation while negative values indicate overvaluation of exchange rates. We focus on building the misalignments using the CS-ARDL and the CS-DL with 3 lags as they account for the correlation of the unobserved common factors with the regressors and the use of deeper lags are necessary for the consistency of the ARDL approach. Moreover, the CS-DL is robust to crosssectional dependence. We also provide the medium-run misalignments for robustness.
These are the residuals of the long-run cointegration relationship as for the medium-run equilibrium the fundamentals are kept at their observed values.12 From gures 3.1 and 3.2, both CS-ARDL and CS-DL misalignments follow approximately the same trend, although the misalignments tend to be larger for Brazil, Morocco, Pakistan, Thailand and Turkey under the latter estimator. There is a tendency of misalignment reduction over the years for all countries but Indonesia in real eective terms as the gap between countries narrows over time, except in the years following the 2007 – nancial crisis. The relative productivity seems to play a major role in all the countries in explaining the trend in equilibrium exchange rates.
We also provide the medium-run misalignments using the CS-ARDL with 3 lags under gure 3.3. These are misalignments that are consistent with the observed values of the fundamentals. As expected, the medium-run misalignments are larger than the long-run ones. Only Brazil and Turkey have overvalued currencies prior 1995 and 1999 respectively. All the remaining countries are characterized by an undervaluation trend in the mediumrun.

Robustness Checks

We provide a series of checks in order to ascertain the robustness of our results. These tests use the CS-ARDL up to 3 lags, except for the temporal check, as this estimator allows the computation of the long-run parameters and the speed of adjustment. Besides, we recognize that if the magnitudes of the robustness checks are closer to the ones of the model of interest from the previous section, we will obtain misalignments that are identical.
The rst check uses the RER as the dependent variable. The RER is dened as q = e + p ? p where q represents the nominal exchange rate expressed per unit of a foreign currency, p represents the price level in the foreign country and p the price level in the domestic country; with all the variables in logarithm form. From the results presented under table A-4 of appendix A, RPROD is negative but not signicant in explaining RER behaviour. However, NFA is signicant and negative in all three specications and the magnitudes are similar to those of our model of interest. The speed of adjustment are negative, signicant in all specications and the magnitudes are similar to those of our
model of interest.

READ LESOTHO POLITICAL ECONOMY: INDEPENDENCE TO THE PRESENT

Unit root and cointegration tests

Our measure of EREER will be computed using the long-run cointegration relationship between REER and the two fundamentals. Prior to that, we perform both rst and second generation unit root tests to ascertain the degree of integration of our variables. The rst generation is based on the IPS test of Im et al. (2003) while the second generation is the Pesaran (2007) CIPS test. The latter is assimilated to a generalisation of the Im et al. (2003) test and consists of an augmented Dickey-Fuller regression of the rst dierence of the dependent variable (Couharde & Sallenave 2013). Table 4.1 presents the results. According to the IPS test, we reject the null of unit root for the REER with intercept and the NFA with intercept and trend. However for the CIPS test, all the variables contain a unit root as the null hypotheses are not rejected. We therefore conclude that all the
variables contain a unit root.

1 INTRODUCTION
1.1 Background
1.2 Problem Statement and Signicance of the Study
1.3 Objectives and Research Questions
1.4 Hypothesis
1.5 Methodology and Data Sources
1.6 Conceptual Framework
1.7 Limitations of the Study
1.8 Structure of the Thesis
2 THEORETICAL REVIEW ON MISALIGNMENT
2.1 Introduction
2.2 Equilibrium Exchange Rate
2.3 Estimation of Equilibrium Exchange Rate .
2.4 Undervaluation and Growth
2.5 Transmission Channels
2.6 Summary
3 EQUILIBRIUM EXCHANGE RATES AND MISALIGNMENTS
3.1 Introduction
3.2 Derivation of the Equilibrium Exchange Rate
3.2.1 The theory
3.2.2 The data
3.3 Long Run Cointegration
3.4 Equilibrium Exchange Rates and Misalignments
3.5 Robustness Checks .
3.6 Conclusion .
4 PANEL SMOOTH TRANSITION REGRESSION MODEL
4.1 Introduction
4.2 Equilibrium Exchange Rate
4.2.1 Behavioural equilibrium exchange rate
4.2.2 Empirical results
4.3 PSTR Growth Model
4.3.1 Specication and estimation of PSTR
4.3.2 Variables
4.3.3 Empirical results
4.4 Robustness Checks
4.5 Conclusion and Policy Implications
5 LONG-RUN COINTEGRATION AND GRANGER CAUSALITY
5.1 Introduction
5.2 Equilibrium Exchange Rate Derivation
5.3 Econometric Framewor
5.3.1 Non-linear panel cointegration .
5.3.2 Specication of PSTRVEC
5.3.3 Estimation of the PSTRVEC and regime wise granger-causality .
5.4 Computation of Equilibrium Exchange Rate and Misalignment
5.5 Empirical Results
5.6 Conclusion and Policy Recommendations
6 ON THE POSSIBLE TRANSMISSION MECHANISMS
7 CONCLUSION AND POLICY IMPLICATIONS
BIBLIOGRAPHY