Run regression separately for several periods

What are some reasons that can justify the necessity to run regressions after dividing the data into several periods?
e.g., instead of run reg xxxxx for the whole sample period, I will run reg xxxxxx for the first half 10 years and run the regression again for the latter half separately.