Onthida Boonspiamsak

Pab Jotikasthira

Jenogsik Lee

Neal Triplett

JaeKeun Yoon

Overview

Attributes

Diagnostics

Korean Stocks

Results

Optimization

The primary objective of our group was to develop a stock selection strategy for the Korean market using standard financial attributes to differentiate individual stock performance. As the foundation of our research, we used the paper "Stock Selection in Emerging Markets: Portfolio Strategies for Malaysia, Mexico, and South Africa" by Achour, Harvey, Hopkins, and Lang which indicated that significant can be achieved by using an attribute based stock selection methodology in emerging markets. We applied a similar sorting methodology, although we extended the analysis to include bivariate as well as univariate sorts.

Methodology

1. We used data from the Korean stock market from June 1993 to May 1998. The total number of stocks in our data ranged from 106 in June 1993 to 193 in May 1998. While this provided only 59 months of data, we were able to generate striking, consistent results using our sorting methodology. Pushing further back in time would have reduced the number of stocks available below 100. This would have reduced the number of stocks in our bivariate sorts to less than 10.
2. We used the first 48 months (in sample) of data to ascertain which attributes were meaningful, and the last 11 month (out of sample) to verify our results and fine-tune our strategy. In the Results section, you find the find the consolidated 59 months performance data. To examine the breakdown of in sample and out of sample performance, you can download this file.
3. At the start of each period, we sorted the firms by our chosen attribute(s) and divided the stock universe into three (univariate) or nine (bivariate) fractals based on the attribute ranking.
4. We tracked the performance of each fractal for the following period based on a number of Diagnostics, which are outlined in more detail below.
5. We chose seven attributes or combinations of attributes, which demonstrated consistent, meaningful ability to enhance returns without significantly increasing, and in many cases reducing, volatility.
6. To combine the attributes into on overall investment strategy, we utilized an Optimization technique to find the optimal weights for each fractal based on expected returns, volatility, and correlations. This approach could be extended to provide a scoring methodology for each stock based on these weights although this was not included in our project.

The following are lists of the univariate and bivariate attributes we examined in our research:

• Dividend yield
• Earning growth
• Projected Earnings Growth
• Projected P/E
• ROE
• Reinvestment Rate
• Momentum (1mo., 1yr.)
• Cash Flow to Price
• Earnings to Price
• Book to Price
• Revenue Growth

• BV/P and MCAP
• Dividend yield and Earning growth
• Dividend yield and Projected EY

We determined that seven of these attributes provided meaningful and consistent methods for achieving superior stock returns. The specific results for these seven sorts can be found in the Results section.

• Annual and monthly average return and excess returns over market and risk-free
• % Periods > Market
• % Periods with positive return
• Average ranking score for each fractal

• Standard deviations of returns and excess returns
• Year by year ranking score

As you can see from the graph above, the five year period from 6/93 to 5/98 was an extremely volatile one for the Korean market. As a result, most of our focus were on excess returns and excess standard deviations. However, we also examined actual returns, and this provided the same general conclusions. One positive aspect of the tumultuous equity values is that we were able to test our models in up and extreme down periods. Each of our selected models performed at least as well in the down periods as in the up periods.

The following tables summarize the results for each of the significant variables for the period 6/93-5/98. For a complete examination of the diagnostics, Please download the attached file.

Cash Flow to Price

Earnings to Price

Projected Earnings to Price

Return on Equity

Reinvestment Rate

Bivariate: Low fractal Book to Price (Growth Stocks) sorted by Market Cap

Bivariate: Low Dividend Yield sorted by Projected Earnings to Price

In order to use the information provided by each of the sorts in a single investment strategy, we used each fractal as an individual stock portfolio and optimized each one based on expected returns, volatilities, and correlations. For our optimization, we looked at six out of the seven significant attributes. Our goal was to match the Korean stock market's volatility and maximize return. Using this rationale, we derived the following optimal weights for each fractal:

RIR -0.20 -0.06

Using these weightings for the Korean stock market, we achieved the following results:

In-sample performance:

Beat market by 0.96% per month on average.

Beat market 68.75% of the time.

Out-of-sample test on 11-month data:

Beat market by 2.83% per month on average.

Beat market 72.73% of the time