I use a computational linguistic algorithm to measure the topics covered in textual descriptions of wine. I ask whether there is information in the text that consumers value. Wine is a prominent example of an experience good. There is substantial product differentiation in the market, and consumers only have limited information on the utility they will receive when consumed. Thus, information is expected to be valuable. Evaluating descriptions of wine produced across the United States, I use a hedonic price regression to explore whether the descriptions provide any new information not already available to the consumer. Initial results suggest that text descriptions are shown to lose their explanatory value when varietal and numerical ratings are included as controls. I then show that once the varietal, region, and numerical ratings are adequately controlled for, there is information in the descriptions that consumers value. (JEL Classifications: C81, D83, L15)

中文翻译：

---

葡萄酒描述提供信息：文本分析

我使用计算语言算法来衡量葡萄酒文本描述中涵盖的主题。我问文本中是否有消费者看重的信息。葡萄酒是体验良好的一个突出例子。市场上存在巨大的产品差异化，消费者对他们在消费时将获得的效用的信息有限。因此，信息被认为是有价值的。在评估美国各地生产的葡萄酒的描述时，我使用特征价格回归来探索描述是否提供了消费者尚未获得的任何新信息。初步结果表明，当将品种和数字评级作为对照时，文本描述会失去其解释价值。然后我展示了一旦​​品种、地区、和数字评级得到充分控制，描述中有消费者重视的信息。（JEL 分类：C81、D83、L15）