Using Smart Money Signals to Improve Fundamentally-Based Factor Models
Academic research has shown that institutional investors have historically had a significant impact on performance in US equity markets. There has been less research on the relationship between institutional ownership and equity returns in Asia, perhaps because stock markets across the Asia-Pacific region have traditionally been driven more by retail investment than those in the US.
Extensive research by S&P Global Market Intelligence, however, suggests that institutional participation is also a powerful driver of returns in Asia. By extension, this research has demonstrated that when they are supplemented by smart money signals based on a granular understanding of institutional ownership, the performance of both long-only and long-short strategies can be enhanced over time.
Three key checks for validity
Ivan Seah, Head of APAC Capital Markets Product Specialists, S&P Global Market Intelligence, says that this research into the impact of smart money signals in Asia is based on three key checks to confirm their validity. The first is quantification of institutional ownership in Asia. This is considerably below ownership levels of the components of the Russell 3000 index of US stocks, as well as those of the S&P Global BMI Indices, which comprise developed and emerging market equities. Nevertheless, institutional ownership in Asia is substantial enough to be relevant, ranging from over 47.4 percent in Australia to 43.2 percent in Singapore, 33.7 percent in Japan and 28 percent in Hong Kong, based on the constituents of the S&P Global BMI Indices in developed Asian markets.
The second building block for an understanding of the impact of institutional share ownership on portfolio returns in Asia is the granular back-testing of seven smart money signals related to institutional activity in the market.
These are sub-divided into three sub-categories: institutional ownership (IO), IO breadth, and IO dynamics. The IO factor quantifies the total percentage of free float held by institutional shareholders, and the total owned by foreign institutions.
The ownership breadth signal is a granular measure of the average number of institutional investors holding a company’s shares; it also gauges the stability of this ownership based on the ratio of change in breadth to the standard deviation of the change.
Ownership dynamics, meanwhile, measures the concentration of holdings among the five largest institutional shareholders, together with ownership turnover and investment duration expressed as the average length of time institutions have held stock in their portfolios.
Seah says that back-testing each of these factors for all stocks within the universe of the Global BMI Developed Asia Index generates similar results to the back-testing of the same signals for the constituents of the Russell 3000. For each of the signals, stocks are sub-divided into five groupings, with the first quintile made up of the 20 percent with the strongest representation of the factor, and the bottom quintile comprising those with the weakest. “The spread between the returns of the first and last quintiles represents the potential performance of a long/short strategy, while the absolute performance of the first quintile denotes the potential returns of a long-only strategy,” Seah explains.
Applicability to Asian stocks
In both cases, the best performing factors are breadth stability and institutional turnover. As this mirrors the results of the back-testing of the Russell 3000, it validates the applicability of this model to Asian stocks, given the body of evidence pointing to its efficacy in US markets.
A third important test is a correlation check designed to ensure that there are negligible or negative historical correlations between the five best-performing IO signals and the five styles that are most common in multi-factor models. These are momentum, valuation, analysts’ expectations, capital efficiency, and volatility. Once again, the results of this analysis, pointing to zero or negative correlations, were similar to those in the Russell 3000 Index, suggesting that the methodology is as valid in Asian equity markets as it is in the US.
“Using the Japanese equity market as a case study, applying these smart money signals to a multi-factor portfolio generates notable outperformance,” says Seah. He adds that to demonstrate the robustness of the model, S&P Global Market Intelligence compared the returns over a 10-year period between January 2005 and May 2016 of three multi-factor models. The first was a fund based on five styles – the three fundamental factors of value, quality and growth, and two sentiment factors in the form of analysts’ expectations and momentum. The second took each of these styles, but added smart money signals on IO. And the third was based solely on smart money indicators.
The results were striking. The helicopter view was that for the Japanese equity universe researched by S&P, top quintile (long only) excess returns reached 7.2 percent annually, while the spread between the first and fifth quintiles (long-short) was 15.96 percent. The more detailed results of the analysis demonstrate that average Q1 monthly excess returns, the average one-month long-short spread, the Q1 annualized information ratio, and the average one month information coefficient (IC) are all higher for the fundamental and smart money models than they are for the fundamental portfolio excluding the smart money signals. In all cases, this effect is more pronounced among small caps than it is across larger companies.
Equally notable is the hit rate of portfolios using smart money signals. These indicate that both in the case of long-only and long-short strategies, these models outperform on a consistent basis. The average monthly Q1 excess return and long-short spread smart money model hit rates are both 79 percent, compared with 77 percent and 75 percent respectively in the case of the fundamental model.