Essays on the Dynamics of Latvian Exchange Rates

Abstract

The thesis examines empirically the sustainability of the Latvian lats' peg to SDR. It aims at identifying fundamental and non-fundamentals risk factors, and the changing structure of the risks that have been instrumental in inducing exchange rate instability. Following the unobserved components model (see Pagan [1996]), the exchange rate is decomposed into a permanent and transitory parts, and risks to the peg's sustainability are investigated via analysing dynamics of each part separately. The risks arising from the deviation of the permanent part from the underlying equilibrium rate driven by economic fundamentals are investigated using the Behavioural Equilibrium Exchange Rate framework, see Clark and MacDonald [1999]. Econometric modelling is based on the system cointegration method of Johansen [1988, 1995] and single-equation methods of two-step Engle-Granger [1987] and Banerjee et al [1993]. Dynamics of three alternative measures of the real effective exchange rates vis-a-vis Latvia's nine major trade partners are investigated using a quarterly data set. For the transitory part, the destabilising impact of foreign shocks is investigated in two steps. Firstly, we adopt Masson's [1998] framework to categorise shocks into monsoonal effects, spillovers, pure contagion. We study susceptibility to foreign currency shocks and assess the relative importance of trade and financial linkages to propagate such shocks using monthly data. The theoretical framework is a standard intertemporal cost minimisation theory. Quadratic cost function is linked to an autoregressive-distributed-lags model, which forms a generic econometric model for our investigation. Secondly, we study the credibility of the peg, the intervention policy of the Bank of Latvia, and the susceptibility to foreign currency, interest rate and market risk shocks using daily data. The theoretical framework is based on the target zone literature developing from Krugman [1991]. The method of empirical target zone modelling is based on Bekaert and Gray [1998]. Unlike Bekaert and Gray, we: (i) do not use a jump specification; (ii) measure the intensity of mean reversion and conditional volatility separately on the strong and weak sides of the currency band; (iii) propose a simple ad hoc method to test for non-linearity. The external susceptibility is analysed by implanting foreign shock variables into the target zone model. Disparity in market response arising from the size of a shock is modelled. Econometric models are augmented by the FIGARCH( 1 ,d, 1) conditional volatility model of Baillie et al [1996].