Does reducing numerical precision affect real world datasets? | Drawing from Data

Reducing numerical precision is a way to save memory in pandas, but does it make a difference to the conclusions that we might draw from real world datasets?