Abstract
Most international financial market studies that compare across countries utilize the US dollar as the common numeraire. We explore the little studied question of the appropriate choice for the base currency and ask if currency choice can affect the final conclusion of whether predictability exists. We provide empirical results for stock return predictability that demonstrate the importance of the numeraire. For example, the existence (absence) of predictability for a US investor does not necessarily imply the existence (absence) of predictability for other foreign investors.
Notes
1 The adjustments are (i) that coefficients with theoretically incorrect signs are set equal to 0 and (ii) if the excess return forecast is negative it is set equal to 0.
2 See, for example, Bossaerts and Hillion (Citation1999), Rapach et al. (Citation2005), Rangvid (Citation2006) and Jordan and Vivian (Citation2011).
3 Jordan et al. (Citation2014) find that market liquidity is related to the forecast performance of fundamental variables and market development is related to the forecast performance of macro variables.
4 The asset demands derived from multiperiod portfolio choice problems in this line of literature are partial equilibrium in nature, meaning the model assumes exogenous return processes. These exogenous return processes are usually calibrated to US data and an assumed set of preferences. There is no attempt to use the implied model of investor behaviour to explain observed asset returns or to model local sentiment effects, as suggested by Rouwenhorst (Citation1999).
5 Some prior literature suggests that currency fluctuations can be hedged and thus the choice of investor perspective is not important (e.g. see Barras, Citation2007); however, currency hedging cannot be perfectly implemented and an investor in real time might not have realized it was necessary to hedge.
6 The restrictions are generally implemented via marketing restrictions under Regulation S (see Wang, Citation2001). Essentially, other than trades on approved foreign physical trading floors, a US citizen must be located outside the US in order to enter into an approved overseas transaction.
7 That is, we exclude France, Canada and Italy. Excluding France and Italy alleviate the dependency issues raised by the fact that since the advent of the euro these countries have the same currency as Germany and thus possess perfect multicollinearity in currency movements over this period.
8 Parameter estimates using fewer than 60 observations could be (highly) affected by estimation error.
9 Note that the OOS R2 is equal to 1 – U2, that is, 1 – Theil’s U2.
10 For OOS R2, either cumulative squared error or cumulative mean-squared error can be used since the number of periods t is constant in both cases and thus cancels out when the ratio is taken.
11 We also implemented the Clark and McCracken (Citation2001) encompassing test (ENC-NEW). Results are qualitatively similar to those for MSE-F. If equal forecast accuracy is rejected, then the regression model forecast is not encompassed by the historical average model. In the interests of brevity, we report only MSE-F.
12 Since our bootstrapped critical values utilize the sample distribution, they control for properties of the sample such as persistence and correlation in the error term (Nelson and Kim, Citation1993), which would affect conventional inference.
13 This could have a similar effect as what we report. The numerator and denominator are usually based on different periods and so currency fluctuations will affect these alternative return measures.