Abstract
Technological progress can dramatically change costs and prices across a market. The two simplest cases are that these costs may be proportional (e.g., upcoming batteries are expected to have twice the watt-hours) or fixed (e.g., a patent with a $2 royalty expires). We show that the linear-logit demand calibration typically used in antitrust is uniquely robust to fixed decreases in prices and costs, but is sensitive to proportional changes. We propose a log-logit demand calibration which is instead uniquely robust to proportional changes, but sensitive to fixed ones. We show that the calibration for log-logit is just as easy to compute as for linear-logit and that it requires the same data.