Date of Award

August 2018

Degree Type


Degree Name

Doctor of Philosophy


Management Science

First Advisor

Purushottam Papatla

Committee Members

Ehsan S Soofi, Xiaojing Yang, Chakravarthi Narasimhan


Bayesian Models, Consumer Choice, Perceived Risk, Sharing Economy, Uncertainty, Willingness-to-pay



The sharing economy for services like Uber and Airbnb has grown significantly. The growth is driven by technology that “whittled down the barriers to the formation and functioning of sharing markets by lowering or eliminating frictions in the identification, search, match, verification, and exchange” (Narasimhan et al 2017).

Reductions in friction in steps to consummate transactions offer two types of savings to consumers. One, monetary savings, results from lower prices typically offered by sharing economy providers (SEP’s) relative to legacy providers (LP’s). The second type of savings results from reduced effort and/or time that consumers need to search, identify, and transact with providers. Thus, a consumer does not have to wait for a taxi to pass by and can instead hail a ride on Uber. A traveler can find an accommodation at a preferred spot in a city easily even in the absence of traditional hotels at that spot. Such reductions in the time and/ or effort needed to locate desired services result in what we label as hassle savings.

While they may not be able to compete on monetary savings, LP’s can still provide hassle savings. For instance, although they may cost more, by being more readily available, traditional cabs in a city like New York can help riders save the time to hail and wait for Uber. Whether consumers weigh monetary or hassle savings more may, however, vary with the consumption context. For instance, avoiding the wait time for an Uber ride by taking a passing by taxi may weigh more if the ride is short and the savings are not substantial. The opposite may be true, however, for long rides where the difference in the cost of Uber and traditional taxis could be quite large. Monetary and/or hassle savings can, therefore, be strategic variables for LP’s and SEP’s. I examine if this is the case empirically in my dissertation through three essays on the sharing economy.

Essay 1: Monetary and Hassle Savings as Strategic Variables in the Ride-Sharing Market

The setting for my first essay is the ride-sharing market where I examine consumers’ choices between Yellow Taxi and Uber in New York City. Specifically, I assume that consumers will weigh monetary savings less than hassle savings if the former is below a threshold but that the opposite will be true for larger savings. I investigate if this is the case using data on paid rides on Yellow Taxi and Uber in New York City. The period of my investigation lies between April 1, 2014 and September 30, 2014, during which data on all rides taken on Yellow Taxi’s and Uber is available from the city.

I focus my investigation on the hundred most frequently occurring latitude, longitude, combinations from where rides on Yellow Taxis originate in the city. I then relate the odds of riders in these neighborhoods choosing Uber over Yellow Taxi for a ride on different days of the week and at different times of the day to my primary variable of interest - the availability of Yellow Taxis. I operationalize availability as a one-week lagged proportion of the total of rides on Yellow Taxis from the neighborhood to the total rides on Yellow Taxi in NYC. I also consider other factors like the intrinsic preference for Uber in that neighborhood and in New York City as a whole, weather, time of day, and type of neighborhood.

If my assumption about the relative importance of monetary and hassle savings is valid, there should be a ride distance below which Yellow Taxis should be preferred for the hassle savings and above which Uber should be preferred for the monetary savings. I find this indeed to be the case at a threshold of 6.64 miles.

Given the potential endogeneity of availability of Yellow Taxis, I take two approaches to assess the reliability of my finding. First, I assume that the availability of Yellow Taxis in each neighborhood could be endogenous with the demand for and availability of paid transportation in the neighborhood. Specifically, I recalibrate my model including two additional covariates as proxies for demand and availability of paid transportation: number of rides taken on subways closest to the neighborhood at the time of the ride and the distance to the nearest subway station. Two, I jointly estimate a supply side equation for the availability of Yellow Taxis in the neighborhood at the time of the ride as a function of a 1-week lagged availability of Yellow Taxis in the same neighborhood at the time of the ride and the demand for and availability of public transportation. I include the residual from this equation as an additional covariate in the log-odds model. Findings from both models are very similar to and consistent with those from the proposed model and confirm that there is a threshold distance below (above) which Yellow Taxis (Uber) is the preferred option.

Essay 2: Variations in the Strategic Value of Hassle Savings

The accommodation sharing market is the setting for my second and third essays. Accommodations are experience goods because amenities and the quality of services may vary from provider to provider, increasing consumers’ uncertainty. Consumers, therefore, seek information on the features of accommodations before choosing one. Standardization mostly provides this information in the case of legacy providers like branded hotels. Sharing economy providers, however, cannot rely on standardization since the rented personal accommodations do vary across providers. Consumers, therefore, need to rely on alternative sources of information like user-generated ratings and reviews. Ratings and Reviews thus provide hassle savings by reducing uncertainty and can, therefore, be a strategic variable in the accommodation market. I investigate its effect in my second essay.

In the first essay, I examined variations in the relative value of monetary and hassle savings with consumption context. In this essay, I investigate whether the value of hassle savings itself varies with consumption context. If it does, the strategic role of features that provide hassle savings to sharing economy customers will also vary for providers. Providers should then invest more in features that provide hassle savings in contexts where they are valued more but can reduce such investments in other contexts.

Specifically, my goal is to understand if hosts obtain price premiums for receiving higher ratings from guests and how those premiums vary across consumption contexts, which I operationalize as different types of accommodations and regions within the city. Airbnb guests realize hassle savings by relying on ratings provided by other guests to reduce uncertainty about the features and services of listings. The value of the savings should, therefore, be higher in consumption contexts with greater uncertainty.

I hypothesize that uncertainty is likely to be higher under two consumptions contexts. One, where the number of listings in a location is very large. Two, where the number of listings and hence the number of ratings is small. I investigate if these are indeed the patterns by estimating a hedonic model of rental prices for Airbnb listings between April 2016 and October 2017 in the five boroughs of New York City for three types of accommodations: (1) entire – a house or apartment rented in its entirety (2) private – one room in an apartment and (c) shared – an accommodation shared by multiple guests. In each of the borough-type combinations, I assume that listings that receive an average rating of 5.0 are the treatment group and those with ratings of 4.0 – 4.99 are part of the control group. I then use propensity score matching to identify the treatment and control samples for each of the combinations. Estimates of the effect of a higher rating on the price premium are consistent with my hypotheses. Premiums are higher in combinations that have fewer listings or have a large number of listings.

Essay 3: Social Relationships as Strategic Variable in the Accommodation-Sharing Market

In addition to reviews and ratings (as in Essay 2), an additional source that sharing economy providers have been offering is information on whether the host or any previous buyers of a shared accommodation are acquaintances of a prospective renter. Airbnb, for instance, offers this through a feature called social connections that allows visitors to see only those accommodations reviewed by their friends or friends of friends on Facebook. The feature thus provides hassle savings by reducing uncertainty (perceived risk) and can, therefore, be a strategic variable in the accommodation market. I investigate its effect in my third essay.

My empirical analysis involves data on the search and time to the first purchase of a sharing accommodation by those who register on the Airbnb site. I examine two outcomes: (1) whether or not a purchase occurs (2) time to purchase if one occurs. The data includes Airbnb consumer prospects who registered between January 2014 and June 2014. I select consumer prospects who have used social connection feature at least once and use a proportional hazards model to relate time to first purchase to my primary variable of interest – social connections. I operationalize social connections as the number of times that a registered user uses the social connections feature before making the first purchase or terminating the search without a purchase. I also control for the effects of demographics (gender and age), how a registered user first arrived at the Airbnb site (e.g., via a link on Facebook or a search engine), and the number devices she uses for accessing the Airbnb site. I model the occurrence of the purchase/non-purchase of an accommodation as a binary logit related to the same variables and model the two outcomes jointly. My findings indicate a significant effect of social connections in reducing the time to, and increasing the likelihood of, the first purchase.

The social connections variable could, however, be endogenous with search time. Those who have friends on Facebook may be more experienced online users and hence, faster in searching and more willing to purchase, online. Additionally, they may be using the social connections feature only because it allows them to see which of their friends may be hosts or had used accommodations they are also considering. I take two approaches to investigate whether these are alternative explanations for my findings. First, I use propensity score matching with visitors who use the social connections feature on Airbnb as the treatment group matched with those who do not use this feature and re-estimate my models on the pooled sample. I use signup method which indicates whether people used Facebook/Google to set up an account on Airbnb before searching for accommodations. I also use age as a matching variable as a proxy for experience with- and interest in- using social media and learning about friends’ activities. Results from this re-estimation are consistent with my findings and indicate that social connections are indeed reducing search time and increasing the likelihood of a purchase.

Second, I exploit possible geographic differences in the hassle savings’ value of social connections to validate my findings. Specifically, I hypothesize that the value of hassle savings should be larger when someone is searching internationally rather than domestically in the US since uncertainty should be higher with the former. I therefore re-estimate my model with geographic-specific estimates of the effects of social connections. I do find that the effects are larger both on the time to make the first purchase and on the likelihood of the first purchase for international listings than domestic ones.

Available for download on Monday, December 14, 2020

Included in

Marketing Commons