题　目：Voice Of Respondents: Attitude Measures In A/b Tests

嘉　宾：厉行  副教授  北京大学光华管理学院

主持人：才凤艳  教授  上海交通大学安泰经济与管理学院

时　间：2023年4月26日（周三）13:30-15:00

地　点：上海交通大学 徐汇校区 安泰经济与管理学院A305室

内容简介:

This paper proposes a framework for measuring the attitudinal effects of branding ads using ran- domized control trials (RCTs). Subjects are randomly assigned to treated and control groups, with only those in the treated group being exposed to branding ads through algorithms. Unlike widely-used performance-based ads that use behavioral metrics as outcomes, we use surveys to measure subjects’ attitudes. However, the standard ITT-based estimates used to evaluate performance-based ad effects are not directly applicable due to two main challenges. One is the existence of non-response bias that not all subjects respond to the survey; their response propensities are affected by the treatment status. In this case, the difference in response values between treated and control groups may come from the changes in the composition of respondents, not their attitudes. Two is the data sparsity when trying to impute the response values for the non-response subjects, caused by the low response rates. Our methods deal with two challenges and produce an estimate of the attitudinal effects of branding ads. We collaborate with one large ad platform to implement this method. Empirically, subjects from the platform exhibit non-response bias and our framework produces a different estimate than the naive ITT-based calculation. Ad platforms and branding managers can use our framework to measure the effects of branding ads and balance their spending between branding and performance ads.

演讲人简介：

厉行，北京大学光华管理学院市场营销系副教授、博士生导师、中国高校市场学研究会理事。主要研究领域为营销模型和产业经济学，研究问题包括在线广告的效果衡量、线上线下渠道融合、社交网络的传播、版权保护与产品定价等。在Rand Journal of Economics, The Economic Journal, Management Science, Journal of the Academy of Marketing Science, 《经济学（季刊）》《营销科学学报》等国内外重要学术期刊上发表论文。

欢迎广大师生参加!