This paper uses high-frequency exchange data in five minutes of a day and adopts error-correction model and variance decomposition to study the price discovery between the SMEB-ETF (Small and Medium Enterprise Board Exchange Traded Funds) and its benchmark index, and further explores the information transfer process as well. The conclusions show that co-integration relationship exists between the SMEB-ETF and SMEB-PI (Small and Medium Enterprise Board Price Index) and it gets a long-run equilibrium. As to the ability of price discovery, the SMEB-PI is ahead of the SMEB-ETF. The shock of the SMEB-PI caused by the effects of new information is bigger than that of the SMEB-ETF. The SMEB-PI is better than the SMEB-ETF in explaining the forecasting error variance, and the SMEB-PI is the leading index of information transfer. The effectiveness of ETF market in China remains to be improved.