I would expect the dtype for a TIMESTAMP column to be time zone aware. ``` In [1]: import pandas_gbq In [2]: df = pandas_gbq.read_gbq('SELECT TIMESTAMP("2004-09-15 05:00:00") AS valid_timestamp') In [3]: df.dtypes Out[3]: valid_timestamp datetime64[ns] dtype: object ``` Note: I believe the behavior used to be correct (include the timezone). [NumPy 1.11 stopped defaulting to UTC and instead uses a naive datetime](https://github.com/numpy/numpy/blob/master/doc/source/reference/arrays.datetime.rst#changes-with-numpy-111). Related: see https://github.com/pydata/pandas-gbq/issues/69 for support for DATETIME columns (which do expect a naive datetime).