Latest Research Publications

Emotional expressions have been widely used in online news. Existing research on the perception of online news has primarily focused on the effect of contextual cues on readers’ reasoning and deliberation behavior; the role of discrete emotions such as anger and sadness, however, has been overlooked. This paper addresses this research gap by investigating the influence of angry and sad expressions in online news on readers’ perception of the news. Drawing on the emotions as social information (EASI) theory and the appraisal-tendency framework (ATF), we find that expressions of anger in online news decrease its believability. However, sad expressions do not trigger the same effect. A further test reveals that the effect of angry expressions can be explained by the readers’ perception of the author’s cognitive effort: readers perceive that expressions of anger in the headlines denote a lack of cognitive effort of the author in writing the news, which subsequently lowers the believability of the news. We also show that news believability has downstream implications and can impact various social media behaviors including reading, liking, commenting, and sharing. This research extends current knowledge of the cognitive appraisals and interpersonal effects of discrete emotions (i.e., anger, sadness) on online news. The results also offer practical implications for social media platforms, news aggregators, and regulators that need to manage digital content and control the spread of fake news.

In the standard herding model, privately informed individuals sequentially see prior actions and then act. An identical action herd eventually starts and public beliefs tend to “cascade sets” where social learning stops. What behaviour is socially efficient when actions ignore informational externalities? We characterize the outcome that maximizes the discounted sum of utilities. Our four key findings are: (a) Cascade sets shrink but do not vanish, and herding should occur but less readily as greater weight is attached to posterity. (b) An optimal mechanism rewards individuals mimicked by their successor. (c) Cascades cannot start after period one under a signal logconcavity condition. (d) Given this condition, efficient behaviour is contrarian, leaning against the myopically more popular actions in every period. We make two technical contributions: As value functions with learning are not smooth, we use monotone comparative statics under uncertainty to deduce optimal dynamic behaviour. We also adapt dynamic pivot mechanisms to Bayesian learning.

Video advertisements often show actors and influence agents consuming and enjoying products in slow motion. By prolonging depictions of influence agents’ consumption utility, slow motion cinematographic effects ostensibly enhance social proof and signal product qualities that are otherwise difficult to infer visually (e.g., pleasant tastes, smells, haptic sensations, etc.). Seven studies including an eye-tracking study, a Facebook Ads field experiment, and lab and online experiments—all using real ads across diverse contexts—demonstrate that slow motion (vs. natural speed) can backfire and undercut product appeal by making the influence agent’s behavior seem more intentional and extrinsically motivated. The authors rule out several alternative explanations by showing that the effect attenuates for individuals with lower intentionality bias, is mitigated under cognitive load, and reverses when ads use non-human influence agents. The authors conclude by highlighting the potential for cross-pollination between visual information processing and social cognition research, particularly in contexts such as persuasion and trust, and discuss managerial implications for visual marketing, especially on digital and social platforms.

We introduce the concept of risk entanglement in a preference-free setting to jointly explain the exchange rate volatility, cyclicality, and currency risk premia in the data. Risk entanglement specifies a subset of incomplete market models, in which nondiffusive or nonlog-normal shocks to exchange rates are not fully spanned by asset returns. When risks are entangled, there exist multiple pricing-consistent exchange rates, but none of them are equal to the ratio of the stochastic discount factors (SDFs) or their projections. Decoupling the exchange rate from the SDFs allows us to address key FX market patterns that are puzzling in international finance.

We provide a psychological explanation for the delayed price response to news about economically linked firms. We show that the return predictability of economically linked firms depends on the nearness to the 52-week high stock price. The interaction between news about economically linked firms and the nearness to the 52-week high can partially explain the underreaction to news about customers, geographic neighbors, industry peers, or foreign industries. We also find that analysts react to news about economically linked firms but the 52-week high effect reduces such reactions, providing direct evidence that the 52-week high affects the belief-updating process.

This paper investigates the distinct effects of capital and intermediates imports on firms' productivity growth, and quantifies the importance of tariff structure in trade liberalization in developing countries. Using a large panel of Chinese manufacturing firms, we demonstrate that capital import has a substantially larger productivity effect than intermediates import. On the one hand, while both types of imports exert immediate effects on productivity, only capital import has dynamic productivity effects. On the other hand, we identify significant R&D-capital synergy effect and R&D-inducing effect from capital import, but there is no clear evidence of these effects from intermediates import. Regarding the effects of China's input tariff liberalization following its WTO accession, the change in tariff structure explains 18 percent of the productivity gains.

We provide direct evidence that governments selectively default on debt when they can identify creditors. Analyzing a comprehensive data set of subnational debt, we show that Chinese local governments choose to default on banks with weaker political power. A reduction in a bank's political power relative to other banks increases the likelihood of selective default by local governments. Such default selections are driven by banks’ influence over politician promotion. When local politicians are highly ranked or connected to national leaders, they engage less in selective default as their promotion is less affected by bank loan defaults. Our findings suggest a politics-finance nexus through which government defaults are restrained.

In our target‐initiated theory of takeovers, a target approaches potential acquirers that privately know their standalone values and merger synergies, where higher synergy acquirers tend to have larger standalone values. Despite their information disadvantage, targets can extract all surplus when synergies and standalone values are concavely related by offering payment choices that are combinations of cash and equity. Targets exploit the reluctance of high‐valuation acquirers to cede equity claims, inducing them to bid more cash. When synergies and standalone values are not concavely related, sellers can gain by combining cash with securities that are more information‐sensitive than equities.

A common problem in econometrics, statistics, and machine learning is to estimate and make inference on functions that satisfy shape restrictions. For example, distribution functions are nondecreasing and range between zero and one, height growth charts are nondecreasing in age, and production functions are nondecreasing and quasi-concave in input quantities. We propose a method to enforce these restrictions ex post on generic unconstrained point and interval estimates of the target function by applying functional operators. The interval estimates could be either frequentist confidence bands or Bayesian credible regions. If an operator has reshaping, invariance, order-preserving, and distance-reducing properties, the shape-enforced point estimates are closer to the target function than the original point estimates and the shape-enforced interval estimates have greater coverage and shorter length than the original interval estimates. We show that these properties hold for six different operators that cover commonly used shape restrictions in practice: range, convexity, monotonicity, monotone convexity, quasi-convexity, and monotone quasi-convexity, with the latter two restrictions being of paramount importance. The main attractive property of the post-processing approach is that it works in conjunction with any generic initial point or interval estimate, obtained using any of parametric, semi-parametric or nonparametric learning methods, including recent methods that are able to exploit either smoothness, sparsity, or other forms of structured parsimony of target functions. The post-processed point and interval estimates automatically inherit and provably improve these properties in finite samples, while also enforcing qualitative shape restrictions brought by scientific reasoning. We illustrate the results with two empirical applications to the estimation of a height growth chart for infants in India and a production function for chemical firms in China.