This chapter was prepared by Kevin C. Cheng (AFR).
External competitiveness has many other aspects that are not directly captured by the real exchange rate. For instance, unit labor costs, labor quality, physical infrastructure, judiciary soundness, political stability, and governance affect a country’s competitiveness.
For details on Kenya’s current trade regime, see Chapter V.
The concept and measurement of an equilibrium exchange rate are a contentious issue in the economics literature. In addition, there are always drawbacks to the various approaches that have been employed by different analysts. Therefore, the results of the econometric analysis presented in this chapter should be interpreted with this caveat in mind.
The sample is truncated at 1980 because REER data were only compiled beginning in 1980.
This approach is one of the most standard approaches used to identify the equilibrium REER for a variety of countries today. For a detailed survey on various estimation method for equilibrium REER, see MacDonald (1995), Montiel (1999), and Rogoff (1996).
These are the variables that are typically used to estimate the equilibrium exchange rate for developing countries. Some papers have also used fiscal and external indicators, which were also initially incorporated in the analysis, but were later dropped owing to either statistical insignificance or non-robustness.
Commodity prices instead of the terms of trade are used because most empirical studies in this area have found that commodity prices are strongly cointegrated with the real exchange rate while finding little link between the real exchange rate and the terms of trade. See, for example, Chen and Rogoff (2002), McDonald (2002).
Productivity refers to labor productivity, calculated as the agricultural output per worker in the agricultural sector. Data were obtained from the World Bank’s World Development Indicators.
While the IMF’s trade restrictiveness index may be a better indicator for openness, the data are only available after the mid-1990s.
Formally, suppose Xt and Yt are two non-stationary and cointegrated stochastic processes, then there exists a θ such that Yt - θXt is stationary. The DOLS of Stock and Watson (1993) estimates θ by running the following regression using the ordinary least squares: