From the docs:
Another common operation is the use of boolean vectors to filter the data. The operators are: | for or, & for and, and ~ for not. These must be grouped by using parentheses.
http://pandas.pydata.org/pandas-docs/version/0.15.2/indexing.html#boolean-indexing
Try:
alldata_balance = alldata[(alldata[IBRD] !=0) | (alldata[IMF] !=0)]