Abstract:
The market microstructure approach has been applied to the three major puzzles of exchange rate
economics: the forward bias puzzle, the excess volatility puzzle, and the exchange rate determination
puzzle. It claims that the imbalances between ‘buyer-initiated and seller-initiated trades’ in foreign
exchange markets are indicative of the transmission link between exchange rates and fundamental
determinants of exchange rates. In the context of the exchange rate determination puzzle, this paper
discusses the market microstructure approach from the stand point of hybrid models that integrate order
flow, fundamentals and non-fundamental variables to establish the determinants of the rand-dollar
exchange rate. Among the non-fundamentals considered is the Economist commodity price index, the
relevance of which is based on Chen and Rogoff (2002). Another non-fundamental variable included is a
proxy for country risk—the differential between the Global Emerging Market Bond Index and the South
African long-term bond.
The paper relies on the autoregressive distributed lag (ARDL) model of Persaran, Shin and Smith (2001)
and as explained in Persaran and Persaran (1997). The ARDL approach to cointegration does not require
pre-testing for the integration properties of the individual series used in the empirical analysis. Instead, it
relies on a bounds testing procedure. In this setting, inference is based on an F-test on the significance of
lagged levels of variables in the error correction form. The results, based on the Schwarz Bayesian
Criterion for choosing a model’s lag length, show that the there is a long-run relationship between the
rand-dollar real exchange rate, nonfundamentals, the fundamentals and the proxy for order flow, which is
the dollar-denominated daily net turnover on the South African markets.
