EM Country Rotation Based On A Stock Factor Model

This study is part of our efforts to test the feasibility of building an Emerging Market (EM) country rotation model. In this study, we look at how effectively our internally-developed EM stock selection model can guide country overweights/underweights. Back testing shows that stock-level factor alpha can indeed be captured at the country level. With the rapid development of single country ETFs, capturing factor alpha at the country level may prove to be an efficient, practical alternative to individual stock selection.

The Approach

This study has three distinctive features compared to our previous country rotation studies. First, in the prior studies, all factor signals were calculated at the index level, hence, when using the cap-weighted MSCI EM country index, the momentum factor signal, for example, was heavily influenced by abnormal price movement of large companies. In this study, the momentum factor is calculated at the stock level and a Z-score is applied to minimize outliers. The factor Z-scores of individual stocks are then cap weighted to generate a country-level factor signal, essentially rolling the stock level signal to the country level.

The second feature of this approach is that an array of fundamental and market factors are employed to generate a composite score for each country. Previously, we only studied single factor categories. Our EM stock selection model includes the following factor categories.

  • Valuation: Five separate valuation metrics are employed, and for each, both trailing and forecast readings are used.
  • Growth: Growth rates are computed for both top-line and bottom-line; the rate is based on trend growth, trailing, and forecasted rates.
  • Sentiment: A total of seven factors are used to gauge investor sentiment toward a particular stock (and toward a country when the score is rolled up at the country level).
  • Quality: Several measures are used to rank company profitability; Quality factors are also used for the ranking within this factor category.
  • Price-based: This includes Momentum and Risk readings based on stock volatility.

Each factor category is assigned a proprietary weighting, and a composite score/country ranking is calculated by rolling up the Z-score of each to the country level.

The third feature of this approach is that factor signals are based on a much broader stock universe. Leuthold EM stock universe covers more than 4,000 stocks, versus 2,600 and 800 stocks underlying MSIC EM IMI and MSCI EM Index, respectively. In our opinion, aggregate factor readings across a broader stock universe could better capture the country effect. In addition, we applied this model to 27 EM countries/markets, including two frontier countries (Argentina and Vietnam), and we split China into foreign market and domestic market.

Leuthold EM Country Index Back-Test Outcome

Good performance spreads between the top and bottom quintile countries and between quintile 1 (Q1) and the benchmark.