Prepayment decisions bring risks to the process of mortgage-backed securitization (MBS) and make pricing of these assets difficult. With data from a Chinese bank's mortgage-backed security pool, we examine prepayment decisions of the borrowers. Since multiple prepayments are allowed without penalty in China, we also distinguish between the probability and the proportion of prepayment. With discrete choice models and a Type II Tobit model, we conclude that the probability of prepayment is affected by the starting account balance, monthly effects, duration, term of debt, house type, trading type of loan, house value and size of mortgage. The probability is also affected by individual factors, including gender, age, education level, individual income and family income. Interestingly, the female borrowers are more likely to prepay, while older borrowers intend to prepay a fewer number of times but with a larger amount each time. In addition, similar factors as above are also shown to explain the choices of proportions of prepayment, with some reasonable adjustments. Our results imply that the variations in income, expenditure, geographic difference and opportunities in the financial market should be included for consideration of risk in the MBS process in China.