摘要：Maintenance and production interact. The ideal way of accounting for this interaction, when estimating production functions, is by picking the temporal length of observations so that they embed integer multiples of the production—maintenance cycles for all inputs. In contrast to labor and land, the production—maintenance cycles of capital sometimes vary tremendously in temporal length, which can make it impossible to implement the ideal method of accounting for the interaction between maintenance and production. This paper empirically tests four second best methods of accounting for maintenance, when the ideal method is impossible. The output elasticities of all inputs (not just the input undergoing maintenance), which emerge from these tests, vary tremendously. This implies that the way that maintenance is incorporated into the analysis (including the standard approach of ignoring maintenance) drastically affects the profit maximizing combinations of inputs derived from production function estimations.
