Estimating Causal Installed-Base Effects: A Bias-Correction Approach