Hotelling Meets Wright: Spatial Sorting and Measurement Error in Recreation Demand Models
本文指出传统休闲需求模型存在遗漏变量和测量误差问题，个体旅行成本因居住地选择和观测困难导致估计偏误，提出工具变量法修正，模拟和案例显示福利成本估计降低12%。
Conventional applications of recreation demand models likely suffer from two standard challenges with demand estimation, namely, omitted variables bias and measurement error. Idiosyncratic prices in the form of individual-level travel costs can exacerbate these two challenges: the potential for nonrandom selection into travel costs through residential sorting and the difficulty of observing individual-level travel costs both work to bias traditional model estimates. I demonstrate the magnitude of this potential bias in conventional estimates of recreation demand models. I provide a relatively simple instrumental variables approach to address these two empirical challenges that substantially outperforms traditional estimates in numerical simulations. Replicating English et al., I find that accounting for potential selection into travel costs and measurement error through the instrumental variables approach decreases estimates of the welfare costs of the 2010 Deepwater Horizon oil spill by 12%.