Hedge Fund

1. Risk Factors Estimation;

Based on each company's main business and the industry background, we estimate the following risk factors as the vital ones influencing the companies' stock return. As for the portfolio which consists of all five companies, we hold the idea that all the factors that affect these companies will also influence the return of the whole portfolio. (see Figure 1)

Figure 1. Risk Factors Forecasting Company Background Corn Crude Gold Nikkei S&P 500 Canadian Yen ABX US Canadian-based Gold Mining Company

√ √

√ √

XOM US US-Based Energy Company

√ √

√

CAT US US-Based Earth-Moving Equipment Firm

√

√

ADM US Corn and Soybean processor √ √

√

KOMA JP Japan-Based Earth-Moving Equipment Firm

√

√

√ PORTFOLIO Combination of all five stocks √ √ √ √ √ √ √

2. Specify statistically significant factors (weekly data);

To verify the estimation, multiple regressions are used to identify statistically significant risk factors. First, we run the regression for each company's log return with all the seven factors' change rates. Then based on the regression outputs, we eliminate the insignificant factors and run the regressions again. The results are showed below. (see Figure 2)

3． Hedge strategy (weekly data);

According to the output above, we clarify the influence of every significant factor. With five million USD invested in each stock, the product of five million USD times the coefficient is the exact amount of its complication. Thus we can derive the quantity to sell or buy by dividing these figures over the respective latest price. To hedge the risk, we can long or short the underlying assets. (See Figure 2)

Figure 2. Weekly Data Regression and Hedging Strategy Company

Corn (Bushel) Oil (Gallon) Gold (Ounce) Nikkei S&P JPY/USD (in JPY) CAD/USD (in CAD) ABX US coefficient 1.465 0.367 -0.566 -0.271 short 4396.3 1394.2 long 2801161.2 17574.6 XOM US coefficient 0.662 -0.503 short 2515.7 2491564.7 long

3. Analysis of the weekly data from July 1986 to Dec 1998.

Respectively, we repeat the regression processes using weekly data from July 1986 to Dec 1998. The statistically significant risk factors are showed in Figure 3 (in Appendix). We can find