Study On Multi-factor Statistical Arbitrage Based On Garch Model

Statistical arbitrage is a market trend can be obtained from the steady gains in pairs trading strategy, not only in the application of hedge funds in general, with a short in the market mechanism, such as funds, stock index futures market is also more widely used. With the introduction of stock index futures, short of China's large-scale implementation of the mechanism is a general trend, so the data on China's financial mechanism based on short statistical arbitrage strategy research has a strong practical significance. Statistical arbitrage pairs trading strategy is based on a combination of investment strategies, we must first choose a higher price movements consistent pairs trading portfolio, buying relatively undervalued securities while short relatively overvalued securities, in both price return to normal levels when the opposite operation to get the price difference between the proceeds. Implementation of this strategy has two main points:First, choose the right stock right, the second is to determine the appropriate time to buy or sell transaction, that transaction signals.In previous studies, the trading portfolio selection method is relatively simple. The first approach for the visual method, ie, simple visual stock charts, similar to stock options trading portfolio; The second approach is based on a combination of stock trading sequence, calculate the spread between the two squares, square and select the smallest spread one pair for the trading portfolio; third approach is based on a combination of stock trading sequence, calculate the correlation coefficient matrix, the most relevant for the establishment of trading portfolio;. I believe that the combination of these methods to determine the transactions are too superficial, external focus only on trading portfolio performance, the relevance of that stock, but did not support the stock price on the analysis of the factors discussed in this paper as a breakthrough point, the factors that affect stock price analysis, classification, in order to get the stock on the performance of different factors, and similar factors affecting selection of a trading portfolio of stocks. From different sectors of the stock price performance of different factors, but space is limited, for various industries were discussed and a bit extra, this choice of a representative is the most workable bank stocks. First, stepwise regression analysis of the bank's stock price reached a significant factor, and then select from a bank with similar factors as statistical arbitrage equity trading portfolio, and similar factors but not the other two share the same trend as the comparator group in order to get carried out using stepwise regression multivariate analysis can significantly improve the trading portfolio yield conclusions. This paper selected the Industrial and Commercial Bank, Construction Bank, Bank and Merchants Bank stock analysis showed that four banks are similar to the trend of the stock, which Construction Bank and Bank of maximum correlation coefficient, stepwise regression results also showed Construction Bank and the Bank's shares have similar factors, can be statistical arbitrage, and the Industrial and Commercial Bank and Merchants Bank stock as the comparator group.To determine the trading signals, the method is simple and standard deviation of share price method, this method is that the standard deviation of price change, as long as the price difference between trading portfolio to achieve a particular level, you can trade, this method is simple, easy to operate, sometimes you can get a higher rate of return, return volatility, but great; otherwise a lot of literature as the standard deviation of stock prices is changing over time, and use it as a starting point to implement arbitrage trading signal structure, the basic procedures of this approach is to structure a combination of statistical arbitrage spread sequence, the spread sequence is often subject to AR (1) model, this AR (1) model residuals using time-series model to fit, such as Suppose the spread sequence obey OU processes. As the GARCH model in dealing with financial time series data can be explained fully exploit the residual part of this paper, the advantage of GARCH model, GARCH model to process transactions using a combination of the spread sequence in order to get a more accurate trading signals. Empirical results show that the use of GARCH models can deal with statistical methods sequences spread arbitrage strategy to obtain higher and more stable income.In this paper, the fundamentals associated with the same four selected industry shares, respectively, stepwise regression, stepwise regression were used to determine the support price of the four factors, select a factor of two with similar composition of the trading portfolio, while the remaining the two stock portfolio as a comparison, then use the cointegration test and error correction model to determine the long-term stability of the relationship between the trading portfolio, and use this as a stability factor of the relationship between the proportion of building a portfolio. GARCH model fitted using the residual series of the portfolio in order to build a sequence of time-varying standard deviation, re-use the time-varying standard deviation of the signal sequence to build trade and stop signals. Empirical results show that factor shares of stock with similar support for access to higher and more stable income, rather than stocks with similar factors supporting effective for only two statistical arbitrage opportunities, indicating that the use of stepwise regression can significantly improve the statistical arbitrage effectiveness of success.