Columbia Business School Research

How has Wikipedia activity changed for articles with content similar to ChatGPT following its introduction? We estimate the impact using differences-in-differences models, with dissimilar Wikipedia articles as a baseline for comparison, to examine how changes in voluntary knowledge contributions and information-seeking behavior differ by article content. Our analysis reveals that newly created, popular articles whose content overlaps with ChatGPT 3.5 saw a greater decline in editing and viewership after the November 2022 launch of ChatGPT than dissimilar articles did.

The speed of stock price reaction to news exhibits substantial time variation. Higher risk-bearing capacity of financial intermediaries, lower passive ownership of stocks, and more informative news increase price responses to contemporaneous news; surprisingly, these interaction variables also increase price responses to lagged news (underreaction). A simple model with limited attention and three investor types (institutional, noninstitutional, and passive) predicts the observed variation in news responses.

We develop a model of investor information choices and asset prices in which the availability of information about fundamentals is time-varying and responds to investor demand for information. A competitive research sector produces more information when more investors are willing to pay for that research. This feedback, from investor willingness to pay for information to more information production, generates two regimes in equilibrium, one having high prices and low volatility, the other the opposite.

Problem definition: Motivated by ride-hailing platforms such as Uber, Lyft and Didi, we study the problem of matching riders with self-interested drivers over a spatial network.

We consider a finite-horizon multi-armed bandit (MAB) problem in a Bayesian setting, for which we propose an information relaxation sampling framework. With this framework, we define an intuitive family of control policies that include Thompson sampling (TS) and the Bayesian optimal policy as endpoints. Analogous to TS, which, at each decision epoch pulls an arm that is best with respect to the randomly sampled parameters, our algorithms sample entire future reward realizations and take the corresponding best action.

This paper studies product ranking mechanisms of a monopolistic online platform in the presence of social learning. The products’ quality is initially unknown, but consumers can sequentially learn it as online reviews accumulate. A salient aspect of our problem is that consumers, who want to purchase a product from a list of items displayed by the platform, incur a search cost while scrolling down the list. In this setting, the social learning dynamics, and hence the demand, is affected by the interplay of two unique features: substitution and ranking effects.

We developed a predictive statistical model to identify donor–recipient characteristics related to kidney graft survival in the Chilean population. Given the large number of potential predictors relative to the sample size, we implemented an automated variable selection mechanism that could be revised in future studies as more national data is collected. Materials and Methods: A retrospective multicenter study was conducted to analyze data from 822 adult kidney transplant recipients from adult donors between 1998 and 2018.

Emergency Departments (EDs) typically have multiple areas where patients of different acuity levels receive treatments. In practice, different areas often operate with fixed nurse staffing levels. When there are substantial imbalances in congestion among different areas, it could be beneficial to deviate from the original assignment and reassign nurses. However, reassignments typically are only feasible at the beginning of 8-12-hour shifts, providing partial flexibility in adjusting staffing levels.

Background

Equitable protocols to triage life-saving resources must be specified prior to shortages in order to promote transparency, trust and consistency. How well proposed utilitarian protocols perform to maximize lives saved is unknown. We aimed to estimate the survival rates that would be associated with implementation of the New York State 2015 guidelines for ventilator triage, and to compare them to a first-come-first-served triage method.

Methods

The COVID-19 pandemic has caused great disruption to the service sector, and it has, in turn, adapted by implementing measures that reduce physical contact among employees and users; examples include home-office work and the setting of occupancy restrictions at indoor locations.

We develop a model of information and portfolio choice in which ex ante identical investors choose to specialize because of fixed attention costs required in learning about securities. Without this friction, investors would invest in all securities and would be indifferent across a wide range of information choices. When securities' dividends depend on an aggregate (macro) risk factor and an idiosyncratic (micro) shocks, fixed attention costs lead investors to specialize in either macro or micro information.

The composition of natural liquidity has been changing over time. An analysis of intraday volumes for the S&P500 constituent stocks illustrates that (i) volume surprises, i.e., deviations from their respective forecasts, are correlated across stocks, and (ii) this correlation increases during the last few hours of the trading session.

Service systems are typically limited resource environments where scarce capacity is reserved for the most urgent customers. However, there has been a growing interest in the use of proactive service when a less urgent customer may become urgent while waiting. On one hand, providing service for customers when they are less urgent could mean that fewer resources are needed to fulfill their service requirement. On the other hand, using limited capacity for customers who may never need the service in the future takes the capacity away from other more urgent customers who need it now.

The Covid-19 pandemic crisis is ongoing, and it its wake has brought tremendous loss of life and economic loss, disruption and uncertainty. It has simultaneously brought to the surface important challenges about global healthcare systems, political systems and institutions and their response to the multi-faceted crisis, the connectedness and dependencies of modern economies through global supply chains, and issues of inequity, as manifested in segments of the population that have been most affected in terms of health and economic outcomes through the crisis.

In modern equity markets, participants have a choice of many exchanges at which to trade. Exchanges typically operate as electronic limit order books under a "price-time" priority rule and, in turn, can be modeled as multi-class FIFO queueing systems. A market with multiple exchanges can be thought of as a decentralized, parallel queueing system. Heterogeneous traders that submit limit orders select the exchange, i.e., the queue, in which to place their orders by trading off financial considerations against anticipated delays until their orders may fill.

We consider the pricing problem faced by a revenue maximizing platform matching price-sensitive customers to flexible supply units within a geographic area. This can be interpreted as the problem faced in the short-term by a ride-hailing platform. We propose a two-dimensional framework in which a platform selects prices for different locations, and drivers respond by choosing where to relocate in equilibrium based on prices, travel costs and driver congestion levels.

Many service systems are staffed by workers who work in shifts. In this work, we study the dynamic assignment of servers to different areas of a service system at the beginning of discrete time-intervals, i.e., shifts. The ability to reassign servers at discrete intervals, rather than continuously, introduces a partial flexibility that provides an opportunity for reducing the expected waiting time of customers.

Patients whose transfer to the Intensive Care Unit (ICU) is unplanned are prone to higher mortality rates and longer length-of-stay than those who were admitted directly to the ICU. Recent advances in machine learning to predict patient deterioration have introduced the possibility of proactive transfer from the ward to the ICU. In this work, we study the problem of finding robust patient transfer policies which account for uncertainty in statistical estimates due to data limitations when optimizing to improve overall patient care.

Deciding between the use of market orders and limit orders is an important question in practical optimal trading problems. A key ingredient in making this decision is understanding the uncertainty of the execution of a limit order, that is, the fill probability or the probability that an order will be executed within a certain time horizon. Equivalently, one can estimate the distribution of the time-to-fill. We propose a data-driven approach based on a recurrent neural network to estimate the distribution of time-to-fill for a limit order conditional on the current market conditions.

We analyze methods for selecting topics in news articles to explain stock returns. We find, through empirical and theoretical results, that supervised Latent Dirichlet Allocation (sLDA) implemented through Gibbs sampling in a stochastic EM algorithm will often overfit returns to the detriment of the topic model. We obtain better out-of-sample performance through a random search of plain LDA models. A branching procedure that reinforces effective topic assignments often performs best. We test these methods on an archive of over 90,000 news articles about S&P 500 firms.

Auctions are widely used in practice. While also extensively studied in the literature, most of the developments rely on the significant common prior assumption. We study the design of optimal prior-independent selling mechanisms: buyers do not have any information about their competitors and the seller does not know the distribution of values, but only a general class it belongs to.

The Centers for Medicare & Medicaid Services (CMS) and the National Quality Forum have endorsed the 30-day mortality rate as an important indicator of hospital quality. Concerns have been raised, however, as to whether post-discharge mortality rates are reasonable measures of hospital quality as they consider the frequency of an event that occurs after a patient is discharged and no longer under the watch and care of hospital staff. Estimating the causal effect of length-of-stay (LOS) on post-discharge mortality from retrospective data introduces a number of econometric challenges.

Objectives:

To examine whether and how step-down unit admission after ICU discharge affects patient outcomes.

Design:

Retrospective study using an instrumental variable approach to remove potential biases from unobserved differences in illness severity for patients admitted to the step-down unit after ICU discharge.

Setting:

Ten hospitals in an integrated healthcare delivery system in Northern California.

Patients:

Many financial markets are operated as electronic limit order books (LOBs). Over short time scales, seconds to minutes, LOBs can be best understood and modeled as stochastic dynamical systems, and, specifically, ones that exhibit interesting and relevant queueing phenomena. I will offer a brief overview of algorithmic trading in a limit order book, and highlight how queueing phenomena play an important role in trade execution, and as a consequence in market behavior.

In a multi-armed bandit (MAB) problem a gambler needs to choose at each round of play one of K arms, each characterized by an unknown reward distribution. Reward realizations are only observed when an arm is selected, and the gambler's objective is to maximize cumulative expected earnings over some planning horizon of length T. To do this, the gambler needs to acquire information about arms (exploration) while simultaneously optimizing immediate rewards (exploitation). The gambler's policy is measured relative to a (static) oracle that knows the identity of the best arm a priori.

We find that an increase in the "unusualness" of news with negative sentiment predicts an increase in stock market volatility. Similarly, unusual positive news forecasts lower volatility. Our analysis is based on more than 360,000 articles on 50 large financial companies, published in 1996–2014. Unusualness interacted with sentiment forecasts volatility at both the company-specific and aggregate level. These effects persist for several months. Furthermore, unusual news is reflected in volatility more slowly at the aggregate than at the company-specific level.

Motivated by the proliferation of user-generated product-review information and its widespread use, this note studies a market where consumers are heterogeneous in terms of their willingness-to-pay for a new product. Each consumer observes the binary reviews (like or dislike) of consumers who purchased the product in the past and uses Bayesian updating to infer the product quality.

The paper's introduction offers a high-level review of Bitcoin's features, especially its governance by protocol. The paper proceeds to summarize Bitcoin's analysis as a payment system. It pays particular attention to a comparison between Bitcoin and a firm-run payment system.

Admission control and service rate speedup may be used during periods of congestion to minimize customer waiting in different service settings. In a healthcare setting, this can mean sending patients to alternative care facilities that may take more time and/or provide less ideal treatment. While waiting can be detrimental to patient outcomes, strategies used to control congestion can also be costly. In this work, we examine a multi-server queueing system that considers both admission control and speedup.

Maintenance service plans (MSPs) are contracts for the provision of maintenance by a service provider to an equipment operator. These plans can have different payment structures and risk allocations, which induce various types of incentives for agents in the service chain. How do such structures affect service performance and service chain value? We provide an empirical answer to this question by using a unique panel data covering the sales and service records of more than 700 diagnostic body scanners.

We study the dynamic mechanism design problem of a seller that repeatedly auctions independent items over a discrete time horizon to buyers that face a cumulative budget constraint. A driving motivation behind our model is the emergence of real-time bidding markets for online display advertising in which such budgets are prevalent. We assume the seller has a strong form of limited commitment: she commits to the rules of the current auction but cannot commit to those of future auctions.

Firms often finance their inventory through debt and subsequently sell it to generate profits and service the debt. Pricing of products is consequently driven by both inventory and debt servicing considerations. In the present paper, we analyze how debt distorts dynamic pricing decisions and reduces generated sales revenues. We show that debt induces sellers to always price higher than the revenue-maximizing price.

Many service systems have servers with different capabilities and customers with varying needs. One common way this occurs is when servers are hierarchical in their skills or in the level of service they can provide. Much of the literature studying such systems relies on an understanding of the relative costs and benefits associated with serving different customer types by the different levels of service. In this work, we focus on estimating these costs and benefits in a complex healthcare setting where the major differentiation among server types is the intensity of service provided.

Objective:

We sought to build on the template-matching methodology by incorporating longitudinal comorbidities and acute physiology to audit hospital quality.

Study Setting:

Patients admitted for sepsis and pneumonia, congestive heart failure, hip fracture, and cancer between January 2010 and November 2011 at 18 Kaiser Permanente Northern California hospitals.

Study Design:

Unplanned transfers of patients from general medical-surgical wards to the Intensive Care Unit (ICU) may occur due to unexpected patient deterioration. Such patients tend to have higher mortality rates and longer lengths of stay than direct admits to the ICU. A new predictive model, the EDIP2, was developed with the intent to identify patients at risk for deterioration, which in some cases could trigger a proactive transfer to the ICU. While it is conceivable that proactive transfers could improve individual patient outcomes, they could also lead to ICU congestion.

Objective:

To understand the impact of adding a medical step-down unit (SDU) on patient outcomes and throughput in a medical intensive care unit (ICU).

Design:

Retrospective cohort study.

Setting:

Two academic tertiary care hospitals within the same health-care system.

Patients:

Adults admitted to the medical ICU at either the control or intervention hospital from October 2013 to March 2014 (preintervention) and October 2014 to March 2015 (postintervention).

Consumer reviews and ratings of products and services have become ubiquitous on the Internet. This paper analyzes, given the sequential nature of reviews and the limited feedback of such past reviews, the information content they communicate to future customers. We consider a model with heterogeneous customers who buy a product of unknown quality and we focus on two different informational settings. In the first setting customers observe the whole history of past reviews. In the second one they only observe the sample mean of past reviews.

We study a multi-server queueing model of a revenue-maximizing firm providing a service to a market of heterogeneous price- and delay-sensitive customers with private individual preferences. The firm may offer a selection of service classes that are differentiated in prices and delays. Using a deterministic relaxation, which highlights the first-order economic structure of the problem, we construct a solution that is incentive compatible and near-optimal in systems with large capacity and market potential.

In hospitals, Step Down Units (SDUs) provide an intermediate level of care between the Intensive Care Units (ICUs) and the general medical-surgical wards. Because SDUs are less richly staffed than ICUs, they are less costly to operate; however, they also are unable to provide the level of care required by the sickest patients. There is an ongoing debate in the medical community as to whether and how SDUs should be used. On one hand, an SDU alleviates ICU congestion by providing a safe environment for post-ICU patients before they are stable enough to be transferred to the general wards.

In order for a patient to be discharged from a hospital unit, a physician must first perform a physical examination and review the pertinent medical information to determine that the patient is stable enough to be transferred to a lower level of care or be discharged home. Requiring an inspection of a patient's "readiness for discharge" introduces an interesting dynamic where patients may occupy a bed longer than medically necessary.