Latest Research Publications

Large tensor (multi-dimensional array) data routinely appear nowadays in a wide range of applications, due to modern data collection capabilities. Often such observations are taken over time, forming tensor time series. In this paper we present a factor model approach to the analysis of high-dimensional dynamic tensor time series and multi-category dynamic transport networks. Two estimation procedures are presented along with their theoretical properties and simulation results. Two applications are used to illustrate the model and its interpretations.

Conditional quantile prediction involves estimating/predicting the quantile of a response random variable conditioned on observed covariates. The existing literature assumes the availability of independent and identically distributed (i.i.d.) samples of both the covariates and the response variable. However, such an assumption often becomes restrictive in many real-world applications. By contrast, we consider a fixed-design setting of the covariates, under which neither the response variable nor the covariates have i.i.d. samples. The present study provides a new data-driven distributionally robust framework under a fixed-design setting. We propose a regress-then-robustify method by constructing a surrogate empirical distribution of the noise. The solution of our framework coincides with a simple yet practical method that involves only regression and sorting, therefore providing an explanation for its empirical success. Measure concentration results are obtained for the surrogate empirical distribution, which further lead to finite-sample performance guarantees and asymptotic consistency. Numerical experiments are conducted to demonstrate the advantages of our approach.

The online trading platform Alibaba provides financial technology (FinTech) credit for millions of micro, small, and medium-sized enterprises (MSMEs). Using a novel data set of daily sales and an internal credit score threshold that governs the allocation of credit, we apply a fuzzy regression discontinuity design (RDD) to explore the causal effect of credit access on firm volatility. We find that credit access significantly reduces firm sales volatility and that the effect is stronger for firms with fewer alternative sources of financing. We further look at firm exit probability and find that firms with access to FinTech credit are less likely to go bankrupt or exit the business in the future. Additional channel tests reveal that firms with FinTech credit invest more in advertising and product/sector diversification, particularly during business downturns, which serves as effective mechanisms through which credit access reduces firm volatility. Overall, our findings contribute to a better understanding of the role of FinTech credit in MSMEs.

Police complicity in organized crime is not uncommon, yet it is extremely difficult to examine empirically. Using unique sex transaction data from China, we show that police can be complicit in organized prostitution. Specifically, we document that sauna houses and massage parlors with greater neighborhood police density are likely to be “protected” by police and thus can host higher-risk, higher-penalty sex business. The complicity effect is particularly salient during periods of local prostitution crackdowns, implying selective enforcement. Changes in local leadership and visits of the central government’s discipline teams can attenuate the complicity effect.

Accessibility of Electric Vehicle (EV) charging stations is an important factor for adoption of EV, which is an effective green technology for reducing carbon emissions. Recognizing this, many governments are contemplating ideas for achieving EV adoption targets, such as constructing extra EV charging stations directly or offering subsidies to entice automakers to construct more EV charging stations. To achieve these targets, governments need to coordinate with automakers to ensure that the total number of charging stations is planned optimally. We study this coordination problem by considering the interactions among the government, automakers, and consumers, our equilibrium analysis yields three major results. First, both the government and the automaker should build extra EV charging stations when their construction costs are independent. Simultaneously, the government should offer a per-station subsidy to the automaker only when the adoption target and the construction cost are both high. However, when the construction costs are dependent, the government should delegate the construction to the automaker by offering a per-station subsidy. Second, when the government considers consumer purchase subsidy as an extra lever, we find that the purchase subsidy for consumers is more cost-effective than offering a per-station subsidy to the automaker. Third, the structure of the optimal government policy remains the same regardless of whether the government's goal is to improve EV adoption or consumer welfare. Our results can serve as guidelines for governments when contemplating coordination with automakers for the construction of EV charging stations to improve EV adoption as well as consumer welfare further.

Emergency department (ED) overcrowding and long patient wait times have become a worldwide problem. We propose a novel approach to assigning physicians to shifts such that ED wait times are reduced without adding new physicians. In particular, we extend the physician rostering problem by including heterogeneity among emergency physicians in terms of their productivity (measured by the number of new patients seen in 1 hour) and by considering the stochastic nature of patient arrivals and physician productivity. We formulate the physician rostering problem as a two-stage stochastic program and solve it with a sample average approximation and the L-shaped method. To formulate the problem, we investigate the major drivers of physician productivity using patient visit data from our partner ED, and find that the individual physician, shift hour, and shift type (e.g., day or night) are the determining factors of ED productivity. A simulation study calibrated using real data shows that the new scheduling method can reduce patient wait times by as much as 13% compared to the current scheduling system at our study ED. We also demonstrate how to incorporate physician preference in scheduling through physician clustering based on productivity. Our simulation results show that EDs can receive almost the full benefit of the new scheduling method even when the number of clusters is small.

For a queueing system with multiple customer types differing in service-time distributions and waiting costs, it is well known that the cµ-rule is optimal if costs for waiting are incurred linearly with time. In this paper, we seek to identify policies that minimize the long-run average cost under nonlinear waiting cost functions within the set of fixed priority policies that only use the type identities of customers independently of the system state. For a single-server queueing system with Poisson arrivals and two or more customer types, we first show that some form of the cµ-rule holds with the caveat that the indices are complex, depending on the arrival rate, higher moments of service time, and proportions of customer types. Under quadratic cost functions, we provide a set of conditions that determine whether to give priority to one type over the other or not to give priority but serve them according to first-come, first-served (FCFS). These conditions lead to useful insights into when strict (and fixed) priority policies should be preferred over FCFS and when they should be avoided. For example, we find that, when traffic is heavy, service times are highly variable, and the customer types are not heterogenous, so then prioritizing one type over the other (especially a proportionally dominant type) would be worse than not assigning any priority. By means of a numerical study, we generate further insights into more specific conditions under which fixed priority policies can be considered as an alternative to FCFS.

Using China's 2008 four-trillion-yuan economic stimulus as a setting and proprietary loan data, we study how a large publicly listed state-owned bank responds to the government's countercyclical financing initiative while trying to meet the expectations of bank regulators and public investors. We find that the bank exhibited little changes in the process of setting internal credit ratings of borrowers, and internal ratings remain a valid, albeit weaker, predictor of interest rates in the stimulus period. Interest rates also remain a valid predictor of loan delinquency in the stimulus period. Evidence from analyzing unlisted borrowers is broadly similar. Overall, there is no systematic evidence that loan decisions of the state-owned bank are severely compromised in the stimulus period. The study adds to the limited understanding of how partially privatized state-owned banks balance different objectives in managing credit risk and is relevant to the longstanding debate over the roles of state-owned banks.

We hypothesize that social trust, in mitigating contracting incompleteness, may have an important effect on the activeness and effectiveness of delegated portfolio management. Using a complete sample of worldwide open-end mutual funds, we find that trust is positively associated with the activeness of funds and that trust-related active share delivers superior performance (e.g., approximately 2% per year for cross-border investments). Moreover, “trust in the market” and “trust in managers” play important yet different roles for different types of cross-border delegated portfolio management. Our results suggest that trust acts as a fundamental building block for delegated portfolio management.