Online reviews significantly influence consumer purchasing behavior, impacting pricing and sales. However, vendors increasingly manipulate reviews to gain competitive advantages, undermining consumer trust and distorting market competition. While platforms employ various governance measures, their effectiveness remains largely unknown. This study addresses this research gap by exploring the differential effects of platform governance on review manipulation. The increasing prevalence of manipulated online reviews necessitates a deeper understanding of how platforms can effectively mitigate this issue. Previous research has documented the negative consequences of vendor manipulation for various stakeholders, including reputational damage, loss of consumer trust, and distorted market competition. This study contributes to this literature by examining the specific effects of differentiated platform governance, focusing on how interventions targeting manipulation intensity, duration, and perceived product quality affect future manipulation probability. The importance of this research lies in its potential to inform the development of robust mechanisms for detecting and governing review manipulation, ensuring the credibility of online reviews and facilitating informed consumer decision-making. The authors aim to explore the effectiveness of differentiated platform governance measures in reducing online review manipulation through analytical modeling and empirical analysis. This involves developing an analytical model of mixed strategy Nash Equilibrium to capture the interaction between the platform and vendors, and conducting empirical analysis using data from Amazon.com to test the hypotheses derived from the model. The dataset comprises online product review manipulation data from Amazon.com, encompassing a significant number of reviews, a portion of which are identified as manipulative. To the best of the authors' knowledge, this is the first study to explore the impact of differentiated platform governance on review manipulation.

Existing research on online review manipulation broadly focuses on two areas: the consequences of manipulation and the factors influencing it. The first group of studies investigates the negative impacts of manipulated reviews on businesses, including reputational damage and decreased information quality, as well as the harm to consumers through misleading purchasing decisions and eroded trust. The second group identifies factors such as business status, competitive dynamics, and the influence of word-of-mouth. While existing literature highlights the prevalence and consequences of review manipulation, it lacks research on the differential effects of platform governance measures. Although platform governance is discussed as a preventative measure, the impacts of different governance strategies on vendors’ manipulation behavior remain unclear. This study aims to address this gap by analyzing the effects of differentiated platform governance, considering the multifaceted nature of review manipulation and the varying responses of vendors to different governance measures. The literature review also discusses review manipulation dimensions based on information manipulation theory, differentiating between manipulation intensity, duration, and perceived product quality. Finally, the review covers existing platform governance measures, including filtering algorithms, punitive actions, and exhibition rules, while emphasizing the lack of research on differentiated governance based on review manipulation dimensions.

The study employs a mixed-methods approach, combining analytical modeling and empirical analysis. An analytical model using a mixed strategy Nash Equilibrium is developed to capture the strategic interaction between the platform and online vendors. The model incorporates factors such as platform governance costs, the probability of detection, penalties for manipulation, review manipulation intensity, duration, and perceived product quality. The model's assumptions include rational actors, probabilistic strategies, and penalties linked to manipulation intensity and duration. The parameters of the model are carefully defined, including platform costs and benefits, vendor costs, and the impact of manipulation on consumer trust and platform losses. The model yields three propositions regarding the relationship between different aspects of platform governance and review manipulation probability. The empirical analysis utilizes a dataset of online product review manipulation data collected from Amazon.com. The data cleaning process involved filtering products without continuous manipulation, removing meaningless reviews, and creating a final dataset consisting of 50,771 reviews, including 6,062 manipulative reviews. The study defines and measures three key independent variables: platform governance on review manipulation intensity (using recommendation entropy of manipulative reviews), duration (cumulative duration of displayed manipulative reviews), and perceived product quality (using review sentiment). The dependent variable is the percentage of manipulated reviews in the subsequent period (ManDegree). Control variables include star rating, review volume, product price, sales rank, and time since product launch. A dynamic panel data (DPD) model is constructed to analyze the data, addressing potential endogeneity issues by including a lagged dependent variable. The system generalized method of moments (GMM) is employed for estimation. The Arellano-Bond test and Hansen test are used to ensure robust statistical inference.

The empirical analysis supports the hypotheses derived from the analytical model. The results show a significant negative association between platform governance of review manipulation intensity and the probability of future manipulation. Specifically, higher recommendation entropy of manipulative reviews in the previous period leads to a lower probability of manipulation in the subsequent period. This suggests that platforms can effectively curb review manipulation by focusing on the misleading nature of manipulated reviews. Similarly, the study finds a significant negative relationship between platform governance of review manipulation duration and the probability of future manipulation, supporting the hypothesis that shorter durations of manipulative reviews lead to a lower probability of future manipulation. This implies that swift detection and removal of manipulative reviews are key to mitigating manipulation. Regarding platform governance of review manipulation of perceived product quality, the results reveal an inverted U-shaped relationship with the probability of future manipulation. This means that increasing perceived product quality through manipulation initially increases the likelihood of future manipulation, but after reaching a certain point, further increases in perceived product quality result in a decrease in the likelihood of future manipulation. This finding suggests a nuanced approach to platform governance targeting manipulation of perceived product quality is necessary, potentially involving adaptive strategies based on the level of perceived product quality. Correlation analysis provides additional support to the hypotheses and shows significant relationships between the different platform governance measures and review manipulation probability.

The findings of this study support the hypotheses derived from the analytical model and provide valuable insights into the effectiveness of differentiated platform governance in mitigating online review manipulation. The negative relationships between governance of manipulation intensity and duration, and the probability of future manipulation, emphasize the importance of timely detection and removal of manipulative reviews. The inverted U-shaped relationship between governance of perceived product quality and the probability of future manipulation highlights the need for targeted interventions based on the level of manipulation. The results suggest that platforms should adopt differentiated governance strategies tailored to the specific dimensions of review manipulation. This approach would allow platforms to concentrate resources on the most impactful areas and avoid unnecessary interventions in areas with low impact. This study contributes to the literature by providing empirical evidence supporting the effectiveness of differentiated platform governance in mitigating review manipulation. This complements previous research focusing on the consequences and antecedents of manipulation, offering a more comprehensive understanding of the phenomenon.

This study contributes to the literature on online review manipulation and platform governance by demonstrating the effectiveness of differentiated governance strategies. The findings highlight the importance of considering the specific dimensions of manipulation—intensity, duration, and perceived product quality—when designing regulatory policies. The inverted U-shaped relationship discovered concerning perceived product quality suggests adaptive governance is key. Future research could explore other platforms, incorporate vendor competition dynamics, and examine additional forms of review manipulation.

This study has limitations. The empirical analysis is based on data from Amazon.com only, limiting generalizability. The model does not explicitly incorporate vendor competition. Finally, the study focuses on a limited set of manipulation dimensions, excluding others like burstiness and helpfulness manipulation. Future work should address these limitations.