Handbook of Exchange Rates

11.3 Implementation Choices and Model Characteristics

In the previous section, we introduced the main families of fair value models, including some examples of ongoing applied work by the IMF and Goldman Sachs. Interestingly, both institutions have moved toward using some combination of models. The IMF uses two related UB approaches and a BEER model, while Goldman Sachs uses a BEER model adjusted for Penn Effects. Both institutions also try to use a number of different estimation approaches, combining time series with cross-sectional techniques and, in some cases, completely avoiding econometric estimations. This trend toward combining and comparing models can be observed more broadly across the literature as well. For example, Cline and Williamson (2008b), Dunaway et al. (2006), and Cheung et al. (2009) compare different approaches to estimate fair value for the Chinese renminbi.

The choice of the model and of the implementation technique will typically affect the estimates of fair value in a number of ways. Moreover, fair value modeling is often conducted with the specific aim of using the model systematically as a policy or investment tool, rather than as a one-off exercise. As a result, practical considerations may affect the choice of model as well. In this section, we discuss these implementation choices and model characteristics in more detail.

11.3.1 Horizon/Frequency

A major dimension of fair value models is the time horizon of the analysis. A number of authors have implicitly ...