```python In [2]: df = pd.DataFrame({"A": [None, None]}) In [3]: df.describe() Out[3]: A count 0.0 unique 0.0 top NaN freq NaN ``` That way DataFrame[object].describe() is always object dtype.