The Accounting Review, 2023, 98(2), 361-387

This study uses data on earnings targets in annual bonus plans to develop a measure of target difficulty at the beginning of the year. We show that target difficulty is negatively associated with market uncertainty, retention concerns, and CEO entrenchment. We present several key findings about the impact of target difficulty on performance and CEO compensation.

First, when annual bonus targets are more difficult to achieve, this is associated with significantly lower CEO cash compensation and decreases in other compensation awards. Second, moderately challenging targets are associated with abnormally poor fourth-quarter performance, particularly after abnormally favourable third-quarter performance. Finally, greater target difficulty is associated with higher abnormal earnings in the same year, but lower earnings and stock returns in the following year.

Combined, our findings suggest that target difficulty is an important factor for companies to consider when designing incentive plans for CEOs as it affects both company performance and executive compensation.

INFORMS Journal on Computing, 2022, 34(3):1819-1840

Electricity wholesale markets are complex, and renewable energy sources can cause electricity prices to be more volatile because of the intermittent nature of renewable sources. Independent power producers (IPPs) who own both traditional coal-fired thermal generators and renewables face significant challenges when submitting offers to sell electricity.

This paper proposes a new plan called an integrated stochastic optimal strategy to help IPPs participate in both the day-ahead and real-time markets (i.e., two-settlement electricity markets) and sell electricity at the best price possible. This strategy involves the IPP submitting an offer for all periods to the day-ahead market and using a multistage stochastic programming setting to provide real-time market offers for each period.

This strategy has the advantage of achieving maximum profits for both markets over a given operational time horizon. It has been proven to be more profitable than alternative self-scheduling strategies that do not take into account real-time information about renewable energy output and electricity prices.

To make the plan more efficient, we explore using polyhedral structures to derive strong valid inequalities, including convex hull descriptions for certain special cases, which strengthens the proposed model’s formulation. Polynomial-time separation algorithms are then established to speed up the branch-and-cut process. Finally, both numerical and real case studies demonstrate the potential of the proposed strategy to improve profits for IPPs.

Production and Operations Management, advance online publication, 2023

Online platforms are experimenting with various methods such as content screening to moderate the effects of fake, biased, and inflammatory content. However, they face a challenge in using machine learning algorithms to manage online content due to the labelling problem. This means that labelled data used to train the model are limited and expensive to obtain.

To address this issue, we propose a domain-adaptive approach that leverages human judgment to improve fake content detection. We start with a source domain dataset that includes both deceptive and trustworthy general news compiled from a large collection of labelled news sources based on human judgments and opinions.

Next, we use advanced deep learning models to extract distinguishing linguistic features commonly found in the source domain news. We transfer these features associated with the source domain to improve fake content detection in three target domains: political news, financial news, and online reviews. We show that domain-invariant linguistic features learned from a source domain with abundant labelled examples can significantly improve fake content detection in a target domain with very few or highly unbalanced labelled data. We further show that these linguistic features are most useful when there is a high level of transferability between source and target domains.

Our study provides insights into managing online content and resources when applying machine learning for fake content detection. We also outline a modular architecture that can be used to develop content screening tools in a wide range of fields.

Academy of Management Journal, advance online publication, 2023

Pay transparency is an important issue in the workplace as it can promote fairness and help close pay gaps. However, our study reveals some unintended consequences of this practice. We surveyed over 100 medical device firms and discovered that as pay becomes more transparent, managers are more likely to avoid challenging discussions about pay differences by nudging everyone towards a standardised salary (pay compression), which often falls at the lower end of an acceptable range. Consequently, high-performing employees may seek alternative forms of recognition, leading to negotiations with their supervisors for personalised benefits and bonuses (i-deals). Ironically, this shift towards opaque negotiations concerning total compensation packages undermines efforts to reduce pay gaps, as managers may compensate employees in less transparent ways.

Our study implies that while pay transparency can be a step towards a more equitable workplace, it is not a one-size-fits-all solution. Companies need to consider the broader implications of their compensation policies, including non-monetary rewards, and ensure that their efforts towards transparency truly promote fairness and understanding among employees.

MIS Quarterly, 2023, 47(1), 63-96

Unstructured multimedia data (text and audio) provides unprecedented opportunities to derive actionable decision-making in the financial industry, in areas such as portfolio and risk management. However, due to formidable methodological challenges, the promise of business value from unstructured multimedia data has not materialized. In this study, we use a design science approach to develop DeepVoice, a novel nonverbal predictive analysis system for financial risk prediction, in the setting of quarterly earnings conference calls. DeepVoice forecasts financial risk by leveraging not only what managers say (verbal linguistic cues) but also how managers say it (vocal cues) during the earnings conference calls. The design of DeepVoice addresses several challenges associated with the analysis of nonverbal communication. We also propose a two-stage deep learning model to effectively integrate managers’ sequential vocal and verbal cues. Using a unique dataset of 6,047 earnings call samples (audio recordings and textual transcripts) of S&P 500 firms across four years, we show that DeepVoice yields remarkably lower risk forecast errors than that achieved by previous efforts. The improvement can also translate into nontrivial economic gains in options trading. The theoretical and practical implications of analyzing vocal cues are discussed.

MIS Quarterly, 2023, 47(1), 391-422

The COVID-19 pandemic has underscored the urgent need for healthcare entities to develop resilient strategies to cope with disruptions caused by the pandemic. This study focuses on the digital resilience of certified physicians who adopted an online healthcare community (OHC) to acquire patients and conduct telemedicine services during the pandemic. We synthesize the resilience literature and identify two effects of digital resilience—the resistance effect and the recovery effect. We use a proprietary dataset that matches online and offline data sources to study the digital resilience of physicians. A difference-in-differences (DID) analysis shows that physicians who adopted an OHC had strong resistance and recovery effects during the pandemic. Remarkably, after the COVID-19 outbreak, these physicians had 35.0% less reduction in medical consultations in the immediate period and 31.0% more bounce-back in the subsequent period as compared to physicians who did not adopt the OHC. We further analyze the sources of physicians’ digital resilience by distinguishing between new and existing patients from both online and offline channels. Our subgroup analysis shows that, in general, digital resilience is more pronounced when physicians have a higher online reputation rating or have more positive interactions with patients on the OHC platform, providing further support for the mechanisms underlying digital resilience. Our research has significant theoretical and managerial implications beyond the context of the pandemic.

Journal of Financial and Quantitative Analysis, 2023, 58(2), 647-676

We study the importance of peer effects among sell-side analysts who work at the same brokerage house, but cover different firms. By mapping the information network within each brokerage, we identify analysts who occupy central positions in the network. Central analysts incorporate more information from their coworkers and produce better research. Using shocks to network structures around brokerage mergers, we identify the influence of peer effects and the importance of industry expertise on analysts’ performance. A portfolio strategy that exploits the forecast revisions of central analysts earns up to 24% per annum.

Journal of Personality and Social Psychology, 2023, 124(3), 461–482

In the context of COVID-19 government-ordered lockdowns, more individualistic people might be less willing to leave their homes to protect their own health, or they might be more willing to go out to relieve their boredom. Using an Australian sample, a pilot study found that people’s lay theories were consistent with the latter possibility, that individualism would be associated with a greater willingness to violate lockdown orders. Using a longitudinal data set containing location records of about 18 million smartphones across the United States, Study 1 found that people in more individualistic states were less likely to comply with social distancing rules following lockdown orders. Additional analyses replicated this finding with reference to counties’ residential mobility, which is associated with increased individualism. In a longitudinal data set containing mobility data across 79 countries and regions, Study 2 found that people in more individualistic countries and regions were also less likely to follow social distancing rules. Preregistered Study 3 replicated these findings at the individual level: People scoring higher on an individualism scale indicated that they had violated social distancing rules more often during the COVID-19 pandemic. Study 4 found that the effect of individualism on violating social distancing rules was mediated by people’s selfishness and boredom. Overall, our findings document a cultural antecedent of individuals’ socially responsible behavior during a pandemic and suggest an additional explanation for why the COVID-19 pandemic has been much harder to contain in some parts of the world than in others.

Journal of Personality and Social Psychology, 2022, 123(6), 1223–1242

People are excessively confident that they can judge others’ characteristics from their appearance. This research identifies a novel antecedent of this phenomenon. Ten studies (N = 2,967, 4 preregistered) find that the more people believe that appearance reveals character, the more confident they are in their appearance-based judgments, and therefore, the more they support the use of facial profiling technologies in law enforcement, education, and business. Specifically, people who believe that appearance reveals character support the use of facial profiling in general (Studies 1a and 1b), and even when they themselves are the target of profiling (Studies 1c and 1d). Experimentally inducing people to believe that appearance reveals character increases their support for facial profiling (Study 2), because it increases their confidence in the ability to make appearance-based judgments (Study 3). An intervention that undermines people’s confidence in their appearance-based judgments reduces their support for facial profiling (Study 4). The relationship between the lay theory and support for facial profiling is weaker among people with a growth mindset about personality, as facial profiling presumes a relatively unchanging character (Study 5a). This relationship is also weaker among people who believe in free will, as facial profiling presumes that individuals have limited free will (Study 5b). The appearance reveals character lay theory is a stronger predictor of support for profiling than analogous beliefs in other domains, such as the belief that Facebook likes reveal personality (Study 6). These findings identify a novel lay theory that underpins people’s meta-cognitions about their confidence in appearance-related judgments and their policy positions.