This paper studies the non-parametric identification and estimation of an empirical rational inattention discrete choice model. Decision-makers do not observe their realized utility perfectly but can obtain a costly signal to inform them about the utility. This model nests the standard discrete choice model as a special case. We characterize the identified set of decision-makers’ prior beliefs based on cross-sectional market-level choice probabilities. We show that the standard discrete choice model is always observationally equivalent to some rational inattention discrete choice models with costly information. We then characterize the moment conditions that can be used to estimate the model parameters. When an exogenous variation of the choice set is available, we show that we can distinguish the standard discrete choice model from the rational inattention discrete choice model with costly information. We apply the model to study the Fox News effect on the 2000 U.S. presidential election. We compare the voters’ prior beliefs with and without the Fox News influence. The result shows that the presence of Fox News has heterogeneous effects for voters with different educational backgrounds.

中文翻译：

---

理性不注意离散选择模型的识别

本文研究了经验理性不注意离散选择模型的非参数识别和估计。决策者并不能完美地观察到他们所实现的效用，但可以获得一个代价高昂的信号来告知他们效用。该模型嵌套了标准离散选择模型作为特例。我们根据横截面市场水平的选择概率来描述已确定的决策者先验信念集。我们证明，标准离散选择模型在观察上总是等价于一些具有昂贵信息的理性不注意离散选择模型。然后，我们描述可用于估计模型参数的矩条件。当选择集的外生变化可用时，我们表明我们可以区分标准离散选择模型和具有昂贵信息的理性不注意离散选择模型。我们应用该模型来研究福克斯新闻对 2000 年美国总统大选的影响。我们比较了受福克斯新闻影响和不受福克斯新闻影响的选民先前的信念。结果表明，福克斯新闻的存在对于不同教育背景的选民具有异质性影响。