Exchange Rate Uncertainty and Trade Growth - A Comparison of Linear and Nonlinear (Forecasting) Models

detail in Section 4. The paper ends with conclusions and directions for future research. An
appendix provides further comments on the data used for the empirical analysis.

2 Data and a measure of FX uncertainty

2.1 The imperfect substitute model

The vast majority of the empirical literature analyzes the impact of FX uncertainty on
trade based on some version of the so-called imperfect substitutes model.^15,21 For recent
applications the reader may consult Baum et al.⁹ or Klaassen.¹⁰ In a bilateral version of this
type of model foreign demand for domestic goods is some function

Qt = q (At, A_t, E [et∣Ωt-ι], E [vt∣Ωt-ι]),                                                   (1)

where Q_t is the quantity of (domestic) exports in time t, and A_t (A_t) is the current domestic
(foreign) economic activity. E [e_t∣Ω_t-1] is the expected real FX rate conditional on Ω_t-1 the
set of information available up to time (t — 1), and E [v_t∣Ω_t-1] is a conditional measure of
the uncertainty (risk) associated with the former expectation. The inclusion of domestic
economic activity in (1) is supported by Koray and Lastrapes²² arguing that (at least)
for large countries domestic conditions are likely to be important determinants of export
flows. Moreover, numerous empirical studies^23,24 find domestic economic activity to have
significant explanatory power for observed export patterns. In this paper we will analyze
both perspectives, foreign demand for domestic goods and domestic demand for foreign
goods. For the latter we formalize M_t , the quantity of domestic imports, by means of a
symmetric counterpart of (1) as

Mt = m (At, A_t, E [et∣Ωt-ι], E [vt∣Ωt-ι]).                                                 (2)

As theoretical models (1) and (2) are based on partial equilibrium considerations poten-
tially omitting important macroeconomic transmission channels. To this end, we will specify
vector autoregressive (VAR) systems to start the empirical analysis which are suitable to
embed a rich dynamic structure of the variables of interest. Moreover, VAR models al-
low a respecification as VECMs to cope with potential cointegration between nonstationary
variables.

2.2 Selection of variables

In this paper we analyze sectoral trade flows for a cross section of 15 economies on a multi-
lateral basis. In addition to determining reasonable approximations to the variables entering
the theoretical models (1) and (2) the implementation of multilateral approaches requires
some weighting scheme for the data. Therefore we discuss first the data and, in particular,

<<   1   2   3   4   5   6   7   >>

More intriguing information