This paper develops a generalized hedonic model in which an exogenous shock to a single product attribute can affect other attributes, the markets for the product's complements and substitutes, and aggregate quantity produced. These factors are shown to be empirically relevant and to cause bias in traditional approaches. Experimental and quasi-experimental estimators of attribute demand are introduced that address these biases, are transparent, and are straightforward to implement. One of these estimators is applied to measure the marginal military recruit's valuation of educational benefits, which is found to range across packages from -$0.024 to +$0.467 per dollar of benefits.