New model incorporates dynamics of country and industry factors

• Pages: 133

Page 133

Traditionally, a standard method has been used to identify the relationship between stock returns and country and industry factors. A simple cross-sectional regression of stock returns is run for a large number of firms and countries on a set of dummy variables that capture a variety of country and industry characteristics. Then monthly variations in these dummy variable coefficients are averaged over time, and their respective contribution in explaining total stock return volatility is computed.

While appealingly simple, this methodology has significant drawbacks, say Catão and Timmerman. First, to compute country and industry contributions, the traditional model uses either arbitrary fixed subperiods-an approach that invariably introduces sample-selection biases-or rolling moving averages that may result in artificial U-shapes or curves. Second, this procedure implicitly assumes that changes in country and industry factors are very gradual. In reality, policies that influence country risk often display distinct changes, and the emergence of new technologies, such as information technology, can rapidly and drastically alter the dynamics of industry factors. Third, the linear structure of the standard model simply cannot account for periods of sustained increases in volatility.

To overcome these drawbacks, Catão and Timmerman developed a new two-stage model. In the first stage, crosssectional regressions of individual firms' stock returns are run on a set of country and industry dummy variables to form country-specific and industry-specific portfolios.

Each...