This study revisits the issue of REITs market efficiency for the US having discovered two notable gaps. Noting the complexities, structural changes and nonlinearities in modern financial markets, we employ the fractional integration technique which performs better than other commonly used techniques in the presence of structural breaks, fractional integration, trend-stationarity and regime switching in time series. Summarizing our results, we find that the US REITs market is efficient in the overall sample. However, when the data are splitted, market efficiency only occurs in the pre-crisis period, but becomes less so during the crisis and post-crisis periods. In addition, evidence of mean reverting long-memory behavior is established for REITs volatility, although mean reversion is slower during and after the crisis. These results are robust to different data measurement and have crucial policy implications for potential investors and relevant policy makers.