Many retailers and e-commerce operators (including online platforms) sell products from large and frequently changing assortments, such as sellers of clothes, footwear, bags, fashion accessories, video games, music, movies, books, electronics, and gadgets.

Inferring consumer preference structures in such context is essential for making a variety of marketing actions such as making merchandising decisions and offering personalised shopping recommendations. Yet, it is a very challenging task due to the large and varying nature of the merchandise. In addition, traditional modelling approaches developed in the marketing literature are not well equipped to handle the data.

This requires a new scalable modelling framework of discovering consumers’ preference structures in large and frequently changing assortments at the store/marketplace level. I developed a scalable stock-keeping-unit-level modelling framework of discovering consumers’ preference structures in large and frequently changing assortments at the store/marketplace level. My research proposes a personalised product recommendation system based on individuals’ preference inferred from my model.

One in three Singapore residents is at risk of developing type 2 diabetes mellitus (T2DM) during their lifetime. Over 15 per cent of national healthcare expenditure is spent on diabetes care in Singapore, and these costs are expected to escalate.

In part, this is because of the enormous challenges in delivering healthcare effectively to the multi-ethnic patient population in Singapore. Innovations are needed in tailoring diabetes care to suit the needs of the individual patient due to the heterogeneous nature of T2DM.

The overall goal of the DCOF programme is to develop solutions using big data analytics, behavioural economics and digital technology to personalise diabetes care and improve clinical outcomes among multi-ethnic T2DM population in Singapore.

The programme falls under three tracks:

1. Diabetes treatment advisor
Development of a clinical decision support system using online machine learning, optimisation approaches and data optimisation from the SingHealth Diabetes Registry to support physicians with individualised clinical treatment recommendations.

2. Risk stratification & population insights
Time trends analysis of risk factors control and development of models for micro- and macro-vascular disease risk prediction, as well as disease progression at the individual and population levels.

3. Behavioural modifications
Leverage behavioural science and natural language understanding technologies to optimise the design and delivery of targeted sustainable behavioural interventions.

In addition, the DCOF programme will establish a testing facility, named “Diabetes Sandbox”, within the Diabetes & Metabolism Centre at Singapore General Hospital to validate and evaluate the effectiveness of new products and solutions from researchers, start-ups and industry partners in improving the treatment of patients with diabetes.

These products and solutions, if proven to be successful, have the potential for a nation-wide roll-out. The key value proposition of the DCOF programme is better health outcomes for T2DM patients by decreasing the death rate and burden of the disease, as well as lowering healthcare spending for individuals and healthcare systems.

Premature cessation of antibiotic therapy (i.e., non-adherence) is common in long treatment regimens and can severely compromise health outcomes.

We investigate the problem of designing a schedule of incentive payments to induce socially-optimal treatment adherence levels in an unsupervised setting. The novel elements of this problem stem from its institutional features; there is a single incentive schedule applied to all patients, incentive payments must increase in line with patients’ adherence, and patients cannot be a priori prohibited from any given levels of adherence.

We develop models to design optimal incentives incorporating these features, which are also applicable in other problem contexts that share the same features. In a numerical study using representative data from a tuberculosis epidemic in India, we show that our optimally-designed incentive schedules are generally cost-effective compared to a linear incentive benchmark.

On Twitter, questions to fast food chain Wendy’s official account are answered with a dash of attitude and a side of sass. Unlike the more formal messages that firms typically send, brands like Wendy’s have incorporated distinctively human attributes—including the ability to think and feel, the capacity for creativity, and the presence of a personality—to anthropomorphise the brand through personified content. We predict that personified brands can develop stronger relationships with users by incorporating human-like elements into their digital content. As a result, stronger brand relationships would improve a personified brand’s word of mouth by transforming consumers into advocates.

The Employee Deposit Program in Japan is a form of shadow banking that allows firms to collect demand deposits from their employees. Firms face the threat of runs as the program is not monitored by regulators and is not subject to the protection of deposit insurance. We seek to show that the threat of runs causes firms to use timely loss recognition as a commitment device to mitigate depositors’ concerns regarding firm risk. Such an effect is expected to be more pronounced than that of external bank debt. Furthermore, the timely loss recognition effect of employee deposits may concentrate in firms that (a) are not associated with implicit guarantees by main banks, (b) have a high default probability, (c) have weak bonding between employees and employers, and (d) rely on external financing.

Mediums of exchange, such as money, vouchers, and digital currency, facilitate the exchange of goods and pervade modern economic transactions. By and large, the economics literature assumes that the involvement of mediums does not influence economic and exchange decisions. The present research examines the influence of nominal medium involvement on prosocial resource allocation. In seven pre-registered experiments, the authors find that mediums inhibit prosocial giving: in resource-allocation decisions, prosocial giving is more likely when valuable resources are presented as consumable goods than when presented as mediums that exclusively represent consumable goods, holding fungibility and economic value constant. This medium-friction effect occurs because mediums, compared with the consumable goods they represent, are less likely to trigger imagery processing and associative affective and behavioral responses, such as “warm-glow” giving. These findings challenge assumptions about medium neutrality and shed new light on automatic responses as underexplored drivers of prosocial decisions.

This paper develops a theory of regulatory enforcement in which enforcement and investments are jointly determined by economic fundamentals. Enforcement disciplines misreporting of investment outcomes, which reduces reporting biases as well as market discounts, benefiting entrepreneurship. However, a government is unable to commit to any long-term enforcement policy but discretionarily chooses an enforcement level that is positively associated with the market size. As a result, entrepreneurs collectively have incentives to overinvest to induce excessive enforcement as a public good. However, a large market requires them to coordinately invest more. If the economic environment is not conducive to the coordination, the market can be under-sized and under-regulated.

Robots and automation are transforming the nature of human work. Although human–robot collaborations can create new jobs and increase productivity, the media often warns about the threat of automation: wide-scale replacement of humans with robots and machines, leading to mass unemployment. Although many social commentators see robots as a threat, relatively little research has directly assessed how laypeople react to robots in the workplace. Drawing from cultivation theory, we suggest that employees would generally view the rise of machines negatively. Four studies—including an archival data across 185 metropolitan US areas, two pre-registered experiments conducted in the United States and Singapore, and an experience-sampling study among engineers conducted in India—find that the increasing presence of robots and automation leads to greater job insecurity. Data from the India study also reveals that this robot-related job insecurity, in turn, is positively associated with burnout and workplace incivility. Our findings hold across different cultures and employees, and even in industries not threatened by robots and automation.

A large and growing literature documents rising market concentration, price-cost margins, and measured profit rates since 2000 or earlier, but the relationship between market concentration and average market power remains ambiguous. We document rising market concentration in the retail sector from 2004-2015. We attempt to explain this trend and understand its welfare implications by analyzing households’ consumption behavior across retail firms. We explore how underlying factors from both the supply and demand side drive these trends.

This research focuses on making good decisions based on limited data. The decision-making process involves the conventional wisdom; to prepare for the worst, and hope for the best. That is to say the decision-makers first choose the optimal decision if the world is working against them. Then limited data is used to update the decision to be less conservative. That way, the decision-makers might be able to strike the right balance between risk and reward. The methodology has potential usage in making decisions in a rapidly changing world. For example, it might help construct a portfolio of thousands of stocks with only hundreds of observations.