Econ Papers Daily / 期刊论文
平均福利变化何时存在？随机系数logit模型中的补偿变差与价格—质量二分
When does the mean welfare change exist? Compensating variation and the price–quality divide in random-coefficient logit
摘要
In random-coefficient logit, compensating variation (CV) is a log-sum difference divided by the marginal utility of income. A log-sum-exp bound shows that pure price-change CV is bounded and has finite moments under every admissible coefficient distribution. For quality changes, finite moments of policy-relevant willingness-to-pay (WTP) coefficients are sufficient. Under independence, finite taste moments, and a nondegenerate policy effect, finite inverse moments of income sensitivity are also necessary. Mean welfare existence therefore depends on the policy and the induced coefficient distribution, not on preference- versus WTP-space parameterization itself. Simulations illustrate this distinction and show why finite sample means cannot establish the existence of a population mean.
中文摘要
在随机系数logit模型中，补偿变异（CV）是log-sum之差除以收入的边际效用。一个log-sum-exp界表明，在每一种可容许的系数分布下，纯价格变化的CV都是有界的且具有有限矩。对于质量变化，政策相关支付意愿（WTP）系数的有限矩是充分条件。在独立性、有限偏好矩以及非退化政策效应的条件下，收入敏感性的有限逆矩也是必要条件。因此，均值福利的存在性取决于政策及所诱导的系数分布，而不取决于偏好空间与WTP空间的参数化本身。模拟结果说明了这一区别，并表明为何有限样本均值无法确立总体均值的存在性。