Summing over a multiindex level in a pandas series
If you know you always want to aggregate over the first two levels, then this is pretty easy:
In [27]: data.groupby(level=[0, 1]).sum()Out[27]:A B 277 b 37a B 159 b 16dtype: int64