MZMi (Jamie) Zhou

Mi (Jamie) Zhou

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Mi (Jamie) Zhou has a total of 14 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Employee Ratings and Reviews Data from Glassdoor

    by Mi Zhou, Yaxuan Li, Zhilei Qiao, Bowen Shi

    This paper presents the employee ratings and reviews data from Glassdoor and the R codes used to collect, clean, and organize the data. We collect three types of information for each Glassdoor review: review metrics, content, and reviewer information. We also calculate some commonly used textual metrics, such as sentiment, readability, the number of uncertainty words, etc. The datasets include necessary identifiers that can connect to other financial data sources. All the variables and metrics are provided at the review level, which enables researchers to aggregate the data from different levels and angles. As a demonstrative example, we use a simple word list to measure how often employees mention COVID-19 in review comments. The R codes provided as an RStudio project are self-contained and can also be modified and applied to other data sources of interest.


  • Tone complexity and analyst forecast behaviors: evidence from earnings conference calls

    by Kyungeun Kwon, Mi Zhou, Tawei Wang, Xu Cheng, Zhilei Qiao

    Purpose: Both the SEC (Securities and Exchange Commission) and the popular press have routinely criticized firms for the complexity of their financial disclosures. This study aims to investigate how financial analysts respond to the tone complexity of firm disclosures.Design/methodology/approach: Using approximately 20,000 earnings conference call transcripts of S&P 1,500 firms between 2005 and 2015, the authors first calculate the abnormal negative tone, the measure of tone complexity; then use such tone measure in econometric models to examine analyst forecast behavior. The authors also test the robustness of the results under different model specifications, tone word lists and alternative tone measure calculations.Findings: Consistent with the notion that analysts respond to the information demand from investors and incur more costs and effort to analyze firm disclosure when the tone is more complex, the authors find that higher tone complexity is positively and significantly associated with more analyst following, longer report duration, more forecast revisions, larger forecast error and larger forecast dispersion. In addition, the authors find that tone complexity has a long-term impact on analyst following but has a limited long-term impact on analyst report duration, analyst revision, forecast error and dispersion.Originality/value: This study complements existing literature by highlighting the information role of financial analysts and by providing evidence that analysts incorporate the management tone disclosed during conference calls to adjust their forecasting behaviors. The results can be used by policymakers as evidence and support for further improving firm communication from a new dimension of disclosure tone.


  • Research in Social Media: Data Sources and Methodologies

    by Roger S. Debreceny, Tawei Wang, Mi (Jamie) Zhou

    This paper examines both the opportunities and limitations in the use of social media for accounting research. Given the dynamic nature of social media and the richness of the context, there are opportunities for researchers to directly observe communication and information exchanges, typically within the context of an observable social network. The paper provides an overview of the characteristics of four commonly used social network sites (SNSs): Facebook, Twitter, LinkedIn, and StockTwits. The data collection details, opportunities, and limitations are set out. The paper also provides illustrative examples of codes that a researcher might employ to extract information from the SNSs. To provide a comparison of accounting-relevant interactions, the paper measures the extent of posts on StockTwits, Twitter, and Facebook for a random sample of corporate announcements.


  • An Examination of the Impact of Tax Avoidance on the Readability of Tax Footnotes

    by Kerry K. Inger, Michele D. Meckfessel, Mi (Jamie) Zhou, Weiguo (Patrick) Fan

    Due to the proprietary nature of tax returns, the tax footnote is the primary source of information for stakeholders about a firm's tax position. However, studies suggest the tax authority acquires information in tax disclosures, creating a trade-off for managers on whether to provide decision-useful information for stakeholders or conceal information from the tax authority. We investigate this trade-off by examining the readability of tax footnotes. We find a positive association between tax avoidance and tax footnote readability for firms with tax avoidance below the industry-year median, consistent with managers highlighting good performance in the form of tax savings with straightforward disclosures. In contrast, we find a negative association between tax avoidance and tax footnote readability for firms with levels of tax avoidance above the industry-year median, consistent with managers concealing tax avoidance from the tax authority. Reinforcing these results, we find that investors place a premium (discount) on tax avoidance when the tax footnote is straightforward in firms with tax avoidance below (above) the industry-year median.


  • Open Access

    A Domain Oriented LDA Model for Mining Product Defects from Online Customer Reviews

    by Zhilei Qiao, Xuan Zhang, Mi Zhou, Gang Alan Wang, Weiguo Fan

    Online reviews provide important demand-side knowledge for product manufacturers to improve product quality. However, discovering and quantifying potential products' defects from large amounts of online reviews is a nontrivial task. In this paper, we propose a Latent Product Defect Mining model that identifies critical product defects. We define domain-oriented key attributes, such as components and keywords used to describe a defect, and build a novel LDA model to identify and acquire integral information about product defects. We conduct comprehensive evaluations including quantitative and qualitative evaluations to ensure the quality of discovered information. Experimental results show that the proposed model outperforms the standard LDA model, and could find more valuable information. Our research contributes to the extant product quality analytics literature and has significant managerial implications for researchers, policy makers, customers, and practitioners.


  • Top Cited Papers
    Open Access

    Project description and crowdfunding success: an exploratory study

    by Mi (Jamie) Zhou, Baozhou Lu, Weiguo (Patrick) Fan, G. Alan Wang

    Existing research on antecedent of funding success mainly focuses on basic project properties such as funding goal, duration, and project category. In this study, we view the process by which project owners raise funds from backers as a persuasion process through project descriptions. Guided by the unimodel theory of persuasion, this study identifies three exemplary antecedents (length, readability, and tone) from the content of project descriptions and two antecedents (past experience and past expertise) from the trustworthy cue of project descriptions. We then investigate their impacts on funding success. Using data collected from Kickstarter, a popular crowdfunding platform, we find that these antecedents are significantly associated with funding success. Empirical results show that the proposed model that incorporated these antecedents can achieve an accuracy of 73 % (70 % in F-measure). The result represents an improvement of roughly 14 percentage points over the baseline model based on informed guessing and 4 percentage points improvement over the mainstream model based on basic project properties (or 44 % improvement of mainstream’s performance over informed guessing). The proposed model also has superior true positive and true negative rates. We also investigate the timeliness of project data and find that old project data is gradually becoming less relevant and losing predictive power to newly created projects. Overall, this study provides evidence that antecedents identified from project descriptions have incremental predictive power and can help project owners evaluate and improve the likelihood of funding success.