estimate sub-division beta cross-sectional analysis

[qfnuts]

Trying to estimate the divisonal beta, I used quarterly earning retun data of sub-division regressed on the quarterly earning return data of S&P 500.
The results show quite low beta and with low R-square. In other words, it suggests the explantory power of earning of S&P 500 towards the earning of each division is very lower.
Is there a better method for estimating divioional beta?