Abstract The implied volatility for 50 ETF options in China shows a significant smile pattern across different moneyness. Call and put options on 50 ETFs transacted from February 2015 to December 2018 are obtained. Regression and vector autoregression analyses are employed to investigate the structural relationship between the volatility smile and potential determinants. The determinants considered in this study include liquidity variables, market momentum, features of the underlying asset and time to maturity. The empirical results show that time to maturity and uncertainty of the underlying asset are the main determinants of the structure of the volatility smile. Moreover, realized volatility of the underlying asset and liquidity proxy variables, open interest and the trading volume of options contracts, have significant impact on the structure of the volatility smile.