Guosen Securities(SZ:002736) Guoxin Securities·2025-08-12 13:53Quantitative Models and Construction Methods - Model Name: Index Dividend Points Estimation Model Model Construction Idea: This model aims to estimate the dividend points of index futures by considering the dividend impact of constituent stocks, which is critical for accurately calculating the basis and premium/discount levels of index futures contracts[12][37] Model Construction Process: 1. Dividend Points Formula: Dividend points are calculated as: $ \text{Dividend Points} = \sum_{n=1}^{N} \frac{\text{Dividend Amount of Stock}n}{\text{Market Cap of Stock}n} \times \text{Weight of Stock}n \times \text{Index Closing Price} $ Here, $N$ represents the number of constituent stocks, and only stocks with ex-dividend dates between the current date ($t$) and the contract expiration date ($T$) are included[37] 2. Key Steps: - Identify whether the company has disclosed dividend amounts and ex-dividend dates. If disclosed, use the provided data; otherwise, estimate these values[41] - For undisclosed dividend amounts, estimate based on the product of net profit and dividend payout ratio. Net profit is predicted using historical profit distribution patterns, while the payout ratio is estimated using historical averages[46][50] - For undisclosed ex-dividend dates, predict using historical intervals between announcement and ex-dividend dates, applying linear extrapolation if stable patterns exist[50][55] Model Evaluation: The model demonstrates high accuracy for indices like the SSE 50 and CSI 300, with prediction errors generally within 5 points. However, the prediction error for the CSI 500 index is slightly larger, around 10 points[60] Quantitative Factors and Construction Methods - Factor Name: Constituent Stock Weight Adjustment Factor Factor Construction Idea: Adjust the weights of index constituent stocks to reflect daily changes in stock prices and corporate actions, ensuring more precise dividend impact estimation[44] Factor Construction Process: 1. Weight Adjustment Formula: $ W{n,t} = \frac{w{n,0} \times (1 + r{n})}{\sum_{i=1}^{N} w_{i,0} \times (1 + r_{i})} $ Here, $w_{n,0}$ is the weight of stock $n$ at the last disclosed date, and $r_{n}$ is the non-adjusted return of stock $n$ from the last disclosed date to the current date[44] 2. Data Source: Use daily closing weights disclosed by the China Securities Index Company to ensure accuracy and avoid biases from corporate actions like stock splits or rights issues[45] - Factor Name: Net Profit Prediction Factor Factor Construction Idea: Predict annual net profit for constituent stocks based on historical profit distribution patterns, distinguishing between stable and unstable profit distributions[46] Factor Construction Process: 1. Classify companies into two categories: stable and unstable profit distributions 2. For stable companies, predict profits using historical distribution patterns. For unstable companies, use the previous year's profit as the prediction[49] - Factor Name: Dividend Payout Ratio Prediction Factor Factor Construction Idea: Estimate the dividend payout ratio using historical averages, considering the stability of corporate dividend policies[50] Factor Construction Process: 1. If the company paid dividends last year, use the previous year's payout ratio as the estimate 2. If no dividends were paid last year, use the average payout ratio of the past three years 3. If the company has never paid dividends, assume no dividends for the current year[52] - Factor Name: Ex-Dividend Date Prediction Factor Factor Construction Idea: Predict ex-dividend dates using historical intervals between announcement and ex-dividend dates, applying default dates if historical patterns are unavailable[50] Factor Construction Process: 1. If the company has disclosed the ex-dividend date, use the disclosed value 2. If not disclosed, estimate based on historical intervals, applying linear extrapolation for stable patterns 3. If historical data is insufficient, use default dates based on typical dividend schedules[55] Model Backtesting Results - Index Dividend Points Estimation Model: - SSE 50 Index: Prediction error within 5 points[60] - CSI 300 Index: Prediction error within 5 points[60] - CSI 500 Index: Prediction error within 10 points[60] Factor Backtesting Results - Constituent Stock Weight Adjustment Factor: Improved accuracy in daily weight adjustments, ensuring precise dividend impact estimation[45] - Net Profit Prediction Factor: Effective for distinguishing between stable and unstable profit distributions, enhancing dividend amount predictions[49] - Dividend Payout Ratio Prediction Factor: Reliable for companies with stable dividend policies, though less effective for companies with irregular payouts[52] - Ex-Dividend Date Prediction Factor: High accuracy for companies with stable historical patterns, with default dates providing reasonable estimates for others[55]