Do country stock market anomalies have trends? In his March 2018 paper entitled “The Momentum Effect in Country-Level Stock Market Anomalies”, Adam Zaremba investigates whether country-level stock market return anomalies exhibit trends (momentum) based on their past returns. Specifically, he:
- Screens potential anomalies via monthly reformed hedge portfolios that long (short) the equal-weighted or capitalization-weighted fifth of country stock market indexes with the highest (lowest) expected gross returns based on one of 40 market-level characteristics/combinations of characteristics. Characteristics span aggregate market value, momentum, reversal, skewness, quality, volatility, liquidity, net stock issuance and seasonality metrics.
- Tests whether the most reliable anomalies exhibit trends (momentum) based on their respective returns over the past 3, 6, 9 or 12 months.
- Compares performance of a portfolio that is long the third of reliable anomalies with the highest past returns to that of a portfolio that is long the equal-weighted combination of all reliable anomalies.
He performs all calculations twice, accounting in a second iteration for effects of taxes on dividends across countries. Using returns for capitalization-weighted country stock market indexes and data required for the 40 anomaly hedge portfolios as available across 78 country markets during January 1995 through May 2015, he finds that:
- For equal weighting of the initial set of 40 country stock market index hedge portfolios:
- Only 10, 6 or 4 of the 40 hedge portfolios described above generate significant gross returns at 10%, 5% or 1% significance levels, respectively.
- Country stock markets comprised of stocks with high fundamentals-to-price ratios, strong intermediate-term past performance, negative skewness, low debt, low liquidity, low net stock issuance or high value-at-risk tend to outperform other country markets.
- Particularly reliable are earnings before interest, taxes, depreciation and amortization (EBITDA)-to-enterprise value (EV), EBITDA-to-price and sales-to-EV ratios, with average gross monthly returns from 0.69% to 0.96% and gross annualized Sharpe ratios from 0.61 to 0.79. The combination of earnings-to-price ratio and skewness is also highly reliable.
- Results for capitalization weighting of the initial set of 40 country stock market index hedge portfolios are generally similar, except that intermediate-term past performance becomes insignificant.
- Among the 16 modestly to strongly reliable equal-weighted anomalies:
- Average pair-wise correlation of gross returns for equally weighted hedge portfolios is only 0.10, suggesting the value of diversification across anomalies. Average annualized gross Sharpe ratio of the equally weighted portfolio of the 16 anomalies is 1.11.
- Anomalies with good (bad) performance over the past 3, 6, 9 or 12 months tend to outperform (underperform) next month for both average gross return and volatility. Annualized gross Sharpe ratios of the equally weighted top (bottom) third of the 16 anomalies are 0.80 to 0.99 (0.22 to 0.42) across lookback intervals. The 6-month lookback interval is most reliable. Findings are somewhat less reliable for capitalization weighting.
In summary, evidence indicates that country-level stock market return anomalies reliably exhibit momentum for lookback intervals of 6 to 12 months.
Cautions regarding findings include:
- Results are gross, not net. Accounting for monthly turnover of country stock market indexes in anomaly portfolios and in anomaly momentum portfolios would reduce returns, as would costs of shorting country indexes in anomaly portfolios. Shorting may be problematic for some country indexes. [The term “net” used in the paper accounts only for country-level taxes on stock dividends.] Moreover:
- Results ignore costs of creating liquid tracking funds for country stock market indexes. These costs may be high for some indexes.
- Results ignore costs of data acquisition and processing.
- Small country stock markets may lack capacity to absorb large investments. This concern is more acute for equal-weighted portfolios of markets.
- The approach is somewhat in-sample. An investor operating in real time could not have known which anomalies would prove reliable over the full sample.
- Testing a large number of potential anomalies (40 and then 16) on the same data introduces considerable data snooping bias, thereby overstating expectations. Moreover, findings may inherit bias from prior research that employed snooping to discover candidate anomalies.
