Most time-of-use (TOU) studies of electricity use in the business sector have found little overall response, as measured by price elasticities, to TOU rates. These studies employed the assumption of weak separability between electricity and all other inputs. Here, we use the generalized Leontief cost function to show that when labor is included in the estimation, the electricity price elasticities are larger, in absolute values, than when labor is erroneously excluded. This result is demonstrated with data on electricity and labor for about 400 Israeli business customers. We also show that the omission of labor from the estimation may cause serious underestimation of the net welfare gains that result from changing a flat electricity price to a TOU rate.