基于Netlogo的中国ETF基金套利研究

A study on the arbitrage of Chinese ETF based on Netlogo

首先构建了基于Netlogo的金融资产交易仿真系统,综合考虑套利过程中的ETF基金双重交易机制、冲击成本、平仓成本、交易者类型等因素,通过数理建模量化分析了EFT基金的跨市套利以及期现套利过程。据此利用Matlab对两种套利过程进行算法编程,并对采用历史真实数据与模拟数据获得的套利结果进行了对比分析。从十五只ETF基金日内跨市套利的结果来看,无论是模拟数据还是历史数据,ETF基金跨市套利中反向套利机会要远多于正向套利机会,但模拟期限较短可能会导致仿真结果不理想。此外,跨市套利的瞬时及连续套利收益率的模拟值随着仿真期限的增加偏差率相对较小,这也符合指数型基金收益率波动程度较低的现实特征,同时也恰好说明了本研究构建的Netlogo模拟系统较为理想。对以ETF510300作为现货品种时的股指期货期现套利进行研究,发现在采用融资融券方式套利时,两种数据条件下均存在较高的超界率,且期现套利机会整体上多于跨市套利机会。从套利方向来看,股指期货期现套利中反向套利机会出现的次数、最长持续时间、瞬时套利平均收益和连续套利机会均高于正向套利的。

Exchange-traded index fund plays an important role in the domestic and foreign financial markets,and the number of ETF funds in China reached 167 on August 27,2017.As one kind of innovative financial products in the Chinese fund market,its trading mechanism needs to be improved.This can be explored through experimental financial research.Although the existing literature by computational methods has analyzed arbitrage trading behavior of other financial products,there is no research to study Chinese ETF arbitrage by simulation tools,and most of the arbitrage models neglect the hedging cost,the impact cost of cross-market arbitrage,as well as different types of traders and so on.This study analyzed the data of fourteen ETF funds with different listing time and ETF 510300,which concerns component stocks of the CSI 300 in China.First,of all,we build the trading market simulation model and system based on Netlogo.It considers the hedging cost,the impact cost of cross-market arbitrage,trader type,and other factors simultaneously.Furthermore,it quantitatively analyzes cross-market arbitrage and future-spot arbitrage process and presents a comparative study by history and simulation results.In the first part of the article,after considering differences of investment strategies adopted by different investors,trading tactics in Netlogo will be divided into the following four types:noise investment strategy,the fundamental value investment strategy,flock investment strategy,and momentum investment strategy,a simulation system of Netlogo trading market is constructed accordingly.In the second part of the writing,we develop two main arbitrage models:ETF fund’s cross-market arbitrage and future-spot arbitrage,and calculate the positive and reverse arbitrage boundaries of these two methods through no-arbitrage equilibrium analysis.In the third part of the paper,the simulation results of the return rate of ETF and stock index future by Netlogo are analyzed to verify the accuracy of the simulation method.After the relevant model is previously established,we develop algorithm programming for two kinds of arbitrage process by Matlab.Finally,it makes a comparative analysis of results.The main conclusions are as follows.First of all,based on one-minute high-frequency data of fifteen ETF funds and CSI 300 stock index futures,it shows that the simulation results of Netlogo are close to the distribution characteristic of the ETF fund’s history price.Furthermore,using Netlogo to simulate the average return rate of different types of investors’on the ETF market,this study finds that the highest is fundamental investor,and noise investors the lowest.The finding matches with the real situation of China's securities market,and it also proves the Netlogo system constructed in this paper has a certain accuracy.Second of all,after analyzing the empirical results of fifteen ETF funds’cross-market arbitrage,we find many important findings.First,for the average value of cross-market arbitrage,both the real data and analog data,all ETF funds’reverse arbitrage opportunities increase as trading period increases,and the number of positive arbitrage chance is significantly less than the reverse arbitrage,which indicates that cross-market reverse arbitrage has much more opportunities than positive arbitrage.In addition,the simulation results of quotas,such as the number of arbitrage times of over bounds,the number of instantaneous arbitrage over bounds times,the average number of consecutive arbitrage over bounds,and the average daily return rate of consecutive arbitrage over the upper bounds,are significantly larger than those of historical price.However,the deviation of the historical value decreases as the sample duration increases,which also indicates that a short simulation period will cause poor simulation results.Second,from the average value of intra-market arbitrage and the condition of the full sample,the simulated deviation rate of continuous overbound return rate is minimal in the all quotas.It also shows that the volatility of the continuous arbitrage return rate is relatively small when the simulation period is longer,and the investment income tends to be stable.This finding is also consistent.The realistic characteristics of the low volatility degree of ETF fund show that the Netlogo simulation system built in this study is ideal.Third,for simulation or historical data,cross-market arbitrage result shows that the total number of consecutive arbitrage opportunities in 5 minutes is larger than the total number of consecutive arbitrage opportunities in 10 minutes,followed by the total number of consecutive arbitrage opportunities in 15 minutes,and the total number of consecutive arbitrage opportunities in 20 minutes.In the end,when using the ETF 510300 as a spot in future-spot arbitrage,we find that when using margin way of arbitrage,whether it is based on historical data or analog data by Netlogo,there is always a higher overbound rate.There are more arbitrage opportunities than cross-market arbitrage.From the arbitrage direction,the number of reverse arbitrage opportunities in the future-spot arbitrage,the longest duration arbitrage time,the instantaneous arbitrage average earnings,and the continuous arbitrage opportunities is higher than the positive arbitrage,and it is similar with the cross-market arbitrage conclusions.For both real data and simulation data,the number of arbitrage opportunities continuing for 5 minutes is more than that for 10 minutes.Delaying 10 minutes’arbitrage average return rate is higher than 5 minutes’arbitrage.The reverse arbitrage times is more than the number of positive arbitrage.