In order to explain a U-shape pattern of stock returns, Fama and French(1988) suggested the statespace model consisting of I(1) permanent component and AR(1) stationary component. They concluded the autoregression coefficient induced from the state-space model follow the U-shape pattern and the U-shape pattern of stock returns was due to both negative autocorrelation in returns beyond a year and substantial mean-reversion in stock market prices. However, we found negative autocorrelation is induced under the assumption that permanent and stationary noise component are independent in the state-space model. In this paper, we derive the autoregression coefficient based on ARIMA process equivalent to the state-space model without the assumption of independency. Based on the estimated parameters, we investigate the pattern of the time-varying autoregression coefficient and conclude the autoregression coefficient from the statespace model of ARIMA(1,1,1) process does not follow a U-shape pattern, but has always positive sign. We applied this result on the data of 1month returns for all NYSE stocks for the 1926-85 period from the Center for Research in Security Prices.