This paper further investigates the set-valued information system. First, we bring forward three tolerance relations for set-valued information systems and explore their basic properties in detail. Then the data compression is investigated for attribute reductions of set-valued information systems. Afterwards, we discuss the data compression of dynamic set-valued information systems by utilizing the precious compression of the original systems. Several illustrative examples are employed to show that attribute reductions of set-valued information systems can be simplified significantly by our proposed approach.