A machine learning model predicts the number of defects per product using a Poisson regression. A product line introduces a new material that doubles the defect rate. How does this affect the predicted variance of defect counts?