Mechanically Determined Markups: A New Critique of the Production-Based Approach to Markup Estimation

This paper identifies two critical flaws in the widely used production approach to markup estimation, illustrating them empirically with a new markup series constructed from New Zealand data. The first is the bias that occurs when markups are estimated from revenue (rather than output) data. While well-established in theory, the form and empirical implication of this bias are unclear. We derive an explicit expression for this empirical bias, showing that it can weaken or even invert true markup trends. We then uncover a second flaw: the level of markup estimates is determined mechanically by the researcher’s definition of variable input. We establish this in theory before showing empirically that broad definitions, i.e., aggregations of multiple inputs, depress markup estimates, whereas narrow definitions inflate them. This reduces the estimated markup to an arbitrary value determined by data constraints and researcher choice.