See discussion here: https://github.com/pydata/pandas/pull/9848#discussion_r28181084 This would likely help clean up some of the Series/DataFrame discrepancies in arithmetic operations (mostly for datetime types).