Standard daily financial time series analyses using autoregressive (AR) models typically
disregard weekends following the trading time hypothesis (TTH) because the relevant assets
of the models are not traded (and thus, their prices are not observed) on weekends. However,
weekends may affect asset prices through time discounting as well as through shocks/news
occurring on weekends. In this regard, we suggest a test for the TTH by using an AR(1)
model, where many asset prices are closely approximated by an AR(1) process. The proposing
test statistics are based upon the differences of AR coefficients and error variances between
Monday and the other weekdays. Asymptotic normality of the suggested test statistics under
the TTH and model stationarity is proved. Under the model of nonstationarity, the test
statistic is asymptotically pivotal/non-standard and the critical values are given from the
Monte Carlo simulations. In an application for the United States S&P 500 data during the
years 2000-2011, we found that the TTH was rejected, particularly during the years of war
and financial crisis. We also confirmed a weakening of the weekend effect as depicted in
Chow, Hsiao and Solt (2003), and Connolly’s (1989) results. It requires us to revise the
dynamic analyses using a time series model of asset prices considering the weekends.