The error correction coefficients, known as the loading factors, are a key component for price discovery measurement. So far, the loading factors have always been assumed as static. This paper is the first one to investigate the time-varying loading factors for the price discovery measurement. Based on the minute-by-minute data from the S&P 500 cash and E-mini futures markets, this paper reveals that the loading factors are indeed autoregressive of order 1 (AR(1)) which confirms the self-dependence nature of informed trading. Furthermore, we propose three AR(1) processes for the loading factors and assess their performance in price discovery measurement compared to the static loading factor model. Overall, this research confirms the importance of autoregressive loading factors for the price discovery measurement.