CAJ | 학술논문

本文通过互联网上的产品评论及其回帖的数据,研究了产品网络口碑传播的动态交互过程。我们采用分层贝叶斯选择模型建模,并用马尔可夫链蒙特卡洛（MCMC,Markov Chain Monte Carlo）方法对参数进行估计。结果发现,已有回帖的特征（如正面回帖的比例、负面回帖的比例等）对当前回帖的产品态度有显著影响,并且这种影响在不同的产品评论之间存在很大差异。这种异质性可以通过引入产品评论（即主帖）的特征得到很好的解释。总体而言,已有回帖对产品的态度,以及主帖的特征等均对之后回帖的产品态度有显著影响。此外,本文还发现,在网络口碑传播过程中,正面回帖的影响比负面回帖的影响更大。最后,本文讨论了该研究对营销理论和实践的贡献。
본문통과호련망상적산품평론급기회첩적수거,연구료산품망락구비전파적동태교호과정。아문채용분층패협사선택모형건모,병용마이가부련몽특잡락（MCMC,Markov Chain Monte Carlo）방법대삼수진행고계。결과발현,이유회첩적특정（여정면회첩적비례、부면회첩적비례등）대당전회첩적산품태도유현저영향,병차저충영향재불동적산품평론지간존재흔대차이。저충이질성가이통과인입산품평론（즉주첩）적특정득도흔호적해석。총체이언,이유회첩대산품적태도,이급주첩적특정등균대지후회첩적산품태도유현저영향。차외,본문환발현,재망락구비전파과정중,정면회첩적영향비부면회첩적영향경대。최후,본문토론료해연구대영소이론화실천적공헌。
Online word-of-mouth （WOM） plays a critical role in shaping consumers＇ attitudes toward new products. Consumers usu- ally consult online reviews to obtain others＇ opinions of the new products, and then form their own ones. In a typical online review, one main message initiates the communication, followed by many responses from other consumers to express their attitudes. The in- teraction process reshapes consumers＇ attitudes, and is of special importance for firms. This study investigates whether and how prior responses in a review and the main messages＇ characteristics influence the current responder＇s attitude toward the new products. We collected 26 new product reviews from various websites and kept the first 40 to 50 responses for each review, which results in 1173 responses in total. We specify a Hierarchical Bayesian Ordinal Choice Model to address the research questions. Parameters are estimated by the Markov Chain Monte Carlo （MCMC） method. We find that the proportion of positive and negative responses in a review significantly influence the product attitudes of the following responses. An interesting finding is that positive responses exert stronger influence on product attitudes than negative ones. We also test two psychological effects with this empirical data, namely, re- cency and primacy effects on product attitudes. The results show that both the most recent and the earliest responses exert significant negative effects on the product attitudes of the following responses. This finding provides some evidence for the existence of recency and primacy effects in the setting of online WOM. Besides, the dispersion of prior responses has a positive impact on product atti- tudes, which suggests that the presence of some negative responses help increase the credibility of the review. We also find huge hetero- geneity across reviews, which can be well explained by the charac- teristics of the main messages at the second-level specification. At the end of the paper, the theoretical and managerial implications are also discussed.