This project focuses on ZocDoc.com – a unique website that integrates physician profiles, patient reviews, and appointment scheduling of physicians onto a single platform. We collected data from the website every day for over a year to construct a novel dataset consisting of profiles, reviews, and ratings for primary care physicians in eight metropolitan divisions. We infer bookings from daily records of appointment availability. ZocDoc displays ratings on a scale of one to five stars, with overall average ratings, rounded to the nearest half-star. We use a regression discontinuity design to identify the causal impact of reviews on patients’ choice of physician. Our preliminary results suggest that patients care quite a bit about quality. However, due to physicians' capacity constraints and the level of demand, 4, 4.5, and 5-star doctors find most of their offered appointments are booked. The main distinction is timing, with lower rating physicians’ appointments booked once the appointments with higher-rated physicians become scarce. We find approximately a doubling in patient volume across the cutoff from 4.5 to 5 stars. We conclude by evaluating the differential impact of ratings, finding that the effects are higher for women physicians, and physicians with more reviews. We find a small, but insignificant difference for hospital affiliate physicians.

In Medicare Advantage, the federal government pays private plans on a per enrollee basis to provide health coverage for Medicare beneficiaries. Previous research has demonstrated that higher plan payments result in greater entry into markets and enrollment of beneficiaries, but that relatively little is passed on the beneficiaries in the form of lower premiums. However, enrollees and insurance companies are not the only ones affected by higher government payments. In this paper, we investigate how higher Medicare Advantage payments affect the transaction prices for medical care. The impact of government payments on provider prices is theoretically ambiguous: if in a strong bargaining position, providers may be able to extract some of this windfall for themselves in the form of higher prices. Alternatively, higher government payments may increase the number of Medicare Advantage beneficiaries enrolled in a plan and increase that plan's bargaining power, allowing them to lower negotiated prices with providers. We investigate this empirically by taking advantage of the urban floor cutoff used in Duggan, Starc and Vabson (2016) and combining this with a price index based on detailed data on transaction prices from Health Care Cost Institute claims data (HCCI) in a regression discontinuity design. We find that Medicare Advantage enrollment is higher but transaction prices for outpatient care are lower in counties that receive higher payments due to the urban floor. In contrast, prices for inpatient care do not appear to be affected by the higher payments, and prices for these services are similar across the threshold. This pattern suggests that the bargaining channel is relevant for determining outpatient prices and that higher government payments can lead to lower prices in health care markets through this mechanism. In ongoing work, we plan to investigate how this effect varies across counties with different levels of provider and insurer concentration.

In this paper, we investigate geographic variation in the prices paid by Medicare Advantage plans across counties. To document price variation, we create novel county-level price indices using itemized transaction-level claims data. In 2016, we find that prices vary substantially across the country, with counties at the 75th percentile of the distribution paying 48.7% more for outpatient procedures and 11.2% more for inpatient care compared to the median. We also examine variation in prices for categories of procedures and admissions, such as radiology and emergency care to uncover the potential drivers of the price variation. Finally, we document persistence in high prices over time and show the relationship between prices and Medicare Advantage penetration rates.

This paper analyzes the value of information for targeting price discounts in shopping applications. It applies methods that combine standard consumer choice models from marketing and economics with matrix factorization techniques from machine learning, whereby users have latent preferences and products have latent attributes, and these are learned from data about consumer choice in a setting where prices vary over time. The paper applies the model to individual panel data from supermarket shopping for a large retailer. The paper analyzes the value of data for increasing profits through personalized price targeting, assessing the relative importance of enriching the model (adding more latent factors) versus more precisely estimating parameters of a fixed model, finding that enriching the model as data grows is an important contributor to improved performance. The paper shows that increasing the length of the history of data used for given set of individuals is substantially more valuable for targeting to those users than adding data about more products or additional users. The results have implications for privacy policy and competition policy.

We ask: To whom does online reputation matter? To analyze this question, we combine two large-scale data sets from novel sources: 1) A large online platform that publishes crowd-sourced reviews about restaurants; And 2) mobile location (ping) data collected from smartphones. We use these data to ascertain the impact of online reviews on restaurant visits. In terms of customers, we compare locals to non-locals and repeat to new customers. In terms of restaurants, we compare chains to non-chains. Our setup allows us to uncover the heterogeneity in treatment effects and thus shed light on the mechanisms behind the impact of star ratings on restaurant visits. We take advantage of the fact that the platform rounds ratings to the nearest “half-star.” As a result, two restaurants with nearly identical ratings can straddle the cutoff to display 4.5 versus 5-stars. These may be viewed as very different by consumers, even if the underlying quality is quite similar. We take advantage of this natural experiment to estimate the causal impact of ratings. We find positive and significant effects of half-star ratings on foot traffic of substantially larger magnitudes than previously documented in Luca (2016) and Anderson and Magruder (2012). The effect is greatest at the extremes; going from 4.5-stars to 5-stars increases foot traffic by over 30-50 percent. We also find that the effect of a half-star increase is largest for new customers, non-locals, and non-chain restaurants.