completely independent. Another possible explanation is that the dearth of large global players reallocating capital across countries means that country-selection opportunities are larger. The data support all of these explanations. Predictability across Global Bond and Currency Markets Using Valuation Measures Now let us turn to the evidence in global bond and currency markets, where valuation effects also exist. We repeat the long/short portfolio construction methodology using the bond and currency markets in the developed world, once again buying the cheapest third and shorting the most expensive third. We measure value in bonds by the slope of the yield curve, and in currencies by trailing five-year excess returns, which is a simple purchasing power parity measure. Our tests include all the independent bond and currency markets in the developed world, and, except for Germany prior to the creation of the euro, exclude the sovereign country bonds and currencies that eventually adopted the common European currency. These amount to nine country markets and 11 currencies. Table 25.1 also reports these results. Consistent with global equity markets, there is a valuation effect in both global bond and currency markets. The information ratio is 0.21 in bonds and 0.50 in currencies, with probability values of 17.9 percent and 1.0 percent, respectively. The fewer number of independent assets within these asset classes probably reduces their information ratios relative to equity markets, consistent with Grinold's Law. Clearly, taken together, the evidence in global asset returns strongly supports predictability from valuation models. Interestingly, in spite of the fact that value measures work across all three primary asset categories as well as within equity markets, the performance on each of these value portfolios constructed from different assets is not highly correlated. In fact, the average correlation is -0.04. We used very simple measures of valuation to demonstrate that significant predictability in country returns exists. When actually implemented, these strategies can be further improved through elementary country-specific adjustments. Clearly addressing differences in taxes, regulation, and economic environment only strengthens the empirical evidence for predictability. Predictability in Stock versus Bond Returns Using Valuation Measures What about timing the market? As it turns out, valuation is also a key driver of expected future total returns on stocks relative to bonds. Taking U.S. data starting in 1926, we simulate a well-known and simple valuation model for timing stocks relative to bonds. We use the "Fed model," which compares the earnings yields on stocks to the interest yield on bonds. In particular, we create an earnings yield gap measure by subtracting yields to maturity on intermediate-term bonds from trailing earnings-price ratios on the S&P 500. In each month, we take a position in stocks equal to five times the earnings yield gap and an offsetting position in intermediate-term bonds. For example, the earnings yield at the end of December 1999 was 3.29