WHY DOES THE NAIVE STOCK SELECTION PORTFOLIO MAKE COUNTRY NOISE BETS?

Having explained how to detect and measure the extent of country noise bets, let us now explain how and why they arise in the naive stock selection portfolio. Both naive stock selection and country-neutral stock selection are intended to compare stocks, while country selection is intended to compare countries. Any comparison is naturally relative to some notion of what is typical within the sample of reference. Crucially, what is typical for a stock is not necessarily the same as what is typical for a country. Naive stock selection falls into the trap of not recognizing this distinction and therefore makes accidental, i.e., noisy, country bets. The primary reason for the difference in what is typical is simply that various countries are not equally represented in the cross-section of stocks. Countries that are over-represented in the cross-section of stocks will dominate a country-agnostic stock average.

This inconsistency in the definition of “typical” is what yields country noise. Let us present a simple example to make these issues more concrete. Consider a hypothetical global market of 90 U.S. stocks, five U.K. stocks, and five Japanese stocks, with book-to-price ratios randomly distributed around a mean of 0.50 for the United States, 0.75 for the United Kingdom, and 1.00 for Japan, respectively. The average country score, taken over the cross-section of countries, is 0.75 (= 1/3 × 0.50 + 1/3 × 0.75 + 1/3 × 1.00). In contrast, the average stock score, taken over the cross-section of stocks, is 0.54 (= (90 × 0.50 + 5 × 0.75 + 5 × 1.00)/100). By design, the average in the space of countries treats the three countries equally, whereas the average over the cross-section of stocks treats the 100 stocks equally. Because there are more U.S. stocks, the stock-based average places greater importance on the United States when determining the typical value for a stock.

The difference in these averages drives the misalignment of resulting country bets in the country selection and naive stock selection portfolios. Continuing our example, country selection makes relative comparisons versus the average of 0.75. Accordingly, the country selection portfolio will deem the United Kingdom to be average and will take a neutral active weight in the United Kingdom. In contrast, naive stock selection makes comparisons versus the average of 0.54. As a result, at the country level, the naive stock selection portfolio will consider the United Kingdom stocks to be cheaper than average and will overweight this country. Consequently, the resulting country bets in the two portfolios will be misaligned.

As our choice of the label “naive” indicates, one can do a lot better than the country-agnostic stock selection approach. However, as our discussion in this section illustrates, to fix this weakness inherently requires a country-aware methodology, such as, a country-neutral portfolio construction framework. In other words, it is impossible to avoid country noise in a stock selection process that ignores country membership.⁹