We employ a model of utility maximization with household production to derive empirical hybrid production functions and interpretations of the e↵ects of observable inputs on household products when unobserved inputs are optimally chosen by households. Instrumental variable approaches cannot yield consistent estimates of marginal e↵ects of observed inputs due to unobserved inputs being consumer choice variables. A correctly specified hybrid model can sometimes provide bounds for true marginal products. We further derive how conditioning on information about household consumptions of non-production goods can increase the accuracy of the bounds. Most studies in the existing household production literature ignore these issues.