Conference Sessions - Jesse H. Jones Graduate School of ...
Conference Sessions - Jesse H. Jones Graduate School of ...
Conference Sessions - Jesse H. Jones Graduate School of ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
2 - Data or Structure? Using a Field Experiment to Assess the<br />
Determinants <strong>of</strong> Counterfactual Demand Predictive Performance<br />
Manuel Hermosilla, PhD Student, Kellogg <strong>School</strong> <strong>of</strong> Managment,<br />
Northwestern University, 2001 Sheridan Road, Room 474, Evanston,<br />
IL, 60208, United States <strong>of</strong> America,<br />
m-hermosilla@kellogg.northwestern.edu, Yi Qian, Eric Anderson<br />
We evaluate the roles <strong>of</strong> microeconomic/statistical structure and experimental data as<br />
determinants <strong>of</strong> counterfactual demand predictive performance. The specific<br />
structures under evaluation are those developed by Berry (1994) and Berry et al.<br />
(1995). Experimental demand data is obtained from a large-scale field experiment<br />
that specified several price conditions for various products <strong>of</strong> a retail category.<br />
Counterfactual demand scenarios arose in some <strong>of</strong> these conditions because<br />
experimental prices substantially departed from those observed in a subsample with<br />
non-experimental historical data. By re-estimating structural and structure-free<br />
econometric models with varying amounts <strong>of</strong> the experimental data, we are able to<br />
isolate the contribution <strong>of</strong> each source <strong>of</strong> identification on counterfactual demand<br />
predictive performance. Our results show that both experimental data and<br />
microeconomic/statistical assumptions (as given by the considered models), provide<br />
advantages in counterfactual demand prediction. These results are robust to different<br />
measures, and in general, suggest that the key driver <strong>of</strong> predictive performance in<br />
counterfactual demand scenarios is the quality <strong>of</strong> the data used for estimation.<br />
Results thus suggest that (i) if the goal is to generate predictions <strong>of</strong> counterfactual<br />
demand scenarios, it might not be worth the while to undergo the costly process <strong>of</strong><br />
specification and estimation <strong>of</strong> a structural econometric model <strong>of</strong> demand, and (ii) if<br />
the goal is to understand the fundamentals <strong>of</strong> consumers’ choice, parameter estimates<br />
<strong>of</strong> a structural demand model can be reliably used for counterfactual prediction.<br />
3 - Estimation <strong>of</strong> Willingness to Pay Intervals by Discrete<br />
Choice Experiments<br />
Christian Schlereth, Goethe University Frankfurt, Grueneburgplatz 1,<br />
Frankfurt, 60323, Germany, schlereth@wiwi.uni-frankfurt.de,<br />
Christine Eckert, Bernd Skiera<br />
Knowledge about consumers’ willingness to pay (WTP) is essential for a pr<strong>of</strong>itmaximizing<br />
price management. This willingness to pay has always been regarded as a<br />
point estimate, typically as the price that makes the consumer indifferent between<br />
buying and not buying the product. In contrast, this paper uses discrete choice<br />
experiments and a scale adjusted latent class model to estimate willingness-to-pay as<br />
an interval. The mid value <strong>of</strong> this interval corresponds to the traditional WTP point<br />
estimate and depends on the deterministic utility, while the range <strong>of</strong> the interval is<br />
influenced by the price sensitivity as well as the error variance (scale) that determines<br />
the random utility for a product. This error variance, i.e. the degree <strong>of</strong> uncertainty in<br />
consumers’ choices has strong implications for firms competing in the market. Those<br />
firms with more fa-vorable products should target consumers with low scale (high<br />
certainty), while the other firms should target those with high scale (low certainty).<br />
The results <strong>of</strong> our empirical study demonstrate that such knowledge about individual<br />
intervals <strong>of</strong> willingness to pay enables better segmenting customers. The results<br />
further show that the sizes <strong>of</strong> the will-ingness-to-pay intervals can be large and that<br />
ignoring these sizes may lead to non-optimal pricing decisions.<br />
■ FC02<br />
Legends Ballroom II<br />
UGC-III (Content and Impact)<br />
Cluster: Internet and Interactive Marketing<br />
Invited Session<br />
Chair: Raji Srinivasan, Associate Pr<strong>of</strong>essor, University <strong>of</strong> Texas-Austin,<br />
1 University Station, Austin, TX, 78712, United States <strong>of</strong> America,<br />
Raji.Srinivasan@mccombs.utexas.edu<br />
1 - Bimodal Distribution <strong>of</strong> Emotional Content in Customer Reviews:<br />
Emotional Biases in Online Customer Reviews<br />
Wonjoon Kim, Associate Pr<strong>of</strong>essor, KAIST, Guseong-dong, Yuseonggu,<br />
Daejeon, 305701, Korea, Republic <strong>of</strong>, wonjoon.kim@kaist.ac.kr<br />
Word-<strong>of</strong>-mouth (WOM), most visibly encountered in the form <strong>of</strong> online customer<br />
reviews in recent times, has received considerable attention <strong>of</strong> late by academics and<br />
practitioners alike. While a number <strong>of</strong> studies have examined the phenomenon’s<br />
frequency and distribution patterns to understand its characteristics and the<br />
motivation behind it, WOM contents across its distribution have remained underexplored.<br />
To fill this important gap in the literature, we analyzed WOM contents<br />
using Natural Language Processing (NLP) methods, which are used to study the<br />
intersection <strong>of</strong> computers and human language. Using this approach, we find that<br />
more extreme reviews have a greater proportion <strong>of</strong> emotional content than less<br />
extreme reviews, revealing a bimodal distribution <strong>of</strong> emotional content as in the case<br />
<strong>of</strong> WOM distribution; we refer to this as emotional bias. Second, we find that reviews<br />
have more positive emotional content toward positive extreme ratings than negative<br />
MARKETING SCIENCE CONFERENCE – 2011 FC02<br />
53<br />
emotional content toward negative extreme ratings, falling into a J-shaped<br />
distribution (self-selection bias). Lastly, we find that uncertainty related to product<br />
quality before consumption is associated with more emotional words across different<br />
product categories (uncertainty bias). However, in this context the amount <strong>of</strong> positive<br />
emotional content is higher than that <strong>of</strong> negative content, suggesting a J-shaped<br />
distribution <strong>of</strong> emotional content. Furthermore, we find that search goods are<br />
associated with the lowest amount <strong>of</strong> expressed emotion, followed by experience<br />
goods, and then credence goods. Our findings enhance our understanding <strong>of</strong> the<br />
motivation behind WOM and related consumer behavior in the context <strong>of</strong> product<br />
sales.<br />
2 - Ad Revenue and Content Commercialization: Evidence from Blogs<br />
Monic Sun, Assistant Pr<strong>of</strong>essor, Stanford University,<br />
518 Memorial Way, Stanford, CA, 94305, United States <strong>of</strong> America,<br />
monic@stanford.edu, Feng Zhu<br />
Many scholars are concerned about the impact <strong>of</strong> ad-sponsored business models on<br />
content providers. They argue that content providers, when incentivized by ad<br />
revenue, are more likely to tailor their content to attract “eyeballs,” and as a result,<br />
popular content may be excessively supplied. We empirically test this prediction by<br />
taking advantage <strong>of</strong> the launch <strong>of</strong> an ad revenue-sharing program initiated by a<br />
major Chinese portal site in September 2007. Participating bloggers allow the site to<br />
run ads on their blogs and receive 50% <strong>of</strong> the revenue generated by these ads. After<br />
analyzing 4.4 million blog posts, we find that compared to nonparticipants, the<br />
percentage <strong>of</strong> popular content increases by about 13% on the participants’ blogs after<br />
the program takes effect. More than 50% <strong>of</strong> this increase can be attributed to topics<br />
shifting towards three domains: stock market, salacious content, and celebrities. We<br />
also find evidence that, relative to nonparticipants, the participants’ content quality<br />
increases after the program takes effect.<br />
3 - Does Advertising Affect Chatter? - Assessing the Dynamics <strong>of</strong><br />
Advertising on Online Word-<strong>of</strong>-mouth<br />
Seshadri Tirunillai, University <strong>of</strong> Southern California, Los Angeles,<br />
CA, United States <strong>of</strong> America, tirunill@usc.edu, Gerard J. Tellis<br />
Despite the increased importance <strong>of</strong> consumer media, the factors influencing the<br />
User-Generated Content (UGC) have seen limited research. In this study, we<br />
investigate the effect <strong>of</strong> corporate advertising on UGC using a natural experiment in a<br />
time series setting. We seek to answer the following questions: 1) If there does exist a<br />
relation, can we establish the direction <strong>of</strong> causality? 2) Among the various metrics <strong>of</strong><br />
UGC, which metrics are influenced by advertising and how are they affected? 3)<br />
What are the dynamics <strong>of</strong> such a relationship in terms <strong>of</strong> growth, persistence and<br />
decays? We use big budget corporate advertising campaigns to assess the impact <strong>of</strong><br />
the campaign on the different metrics <strong>of</strong> UGC (e.g. overall volume, negative and<br />
positive UGC) <strong>of</strong> the target firm and the competitors. We try to assess the influence <strong>of</strong><br />
advertising on firms by assessing the differential impact on the UGC metrics. We also<br />
analyze the dynamics <strong>of</strong> advertisement on the metrics UGC using multivariate time<br />
series. We find that after the introduction <strong>of</strong> the advertising campaign, the chatter <strong>of</strong><br />
the target firm increased by about 27% relative to the control firms that did not<br />
undertake any major brand campaign during this period. While we find that the<br />
positive chatter increases markedly as compared to the synthetic control, there was<br />
no systematic decrease in negative chatter during the time period. There is also a<br />
spillover <strong>of</strong> advertising across firms in a market.<br />
4 - Social Influence in the Evolution <strong>of</strong> Online Ratings <strong>of</strong> Service Firms<br />
Raji Srinivasan, Associate Pr<strong>of</strong>essor, University <strong>of</strong> Texas-Austin,<br />
1 University Station, Austin, TX, 78712, United States <strong>of</strong> America,<br />
Raji.Srinivasan@mccombs.utexas.edu<br />
Online consumer review websites prominently display consumers’ online ratings <strong>of</strong><br />
service firms, which influence other consumers’ purchase decisions. Yet there are few<br />
insights on the factors influencing online ratings <strong>of</strong> service firms. The authors develop<br />
hypotheses <strong>of</strong> how other consumers’ online ratings <strong>of</strong> the service firm moderate the<br />
effects <strong>of</strong> a reviewer’s service encounter characteristics – valence <strong>of</strong> the service<br />
encounter, occurrence <strong>of</strong> service failure, and service recovery effort – on the<br />
reviewer’s online rating <strong>of</strong> the service firm. They test the hypotheses using an ordered<br />
probit model with data from 7,499 online reviews <strong>of</strong> hotels in Boston and Honolulu<br />
between 2006 and 2010. The results support the hypotheses <strong>of</strong> the moderating effects<br />
<strong>of</strong> social influence on a reviewer’s online rating <strong>of</strong> the service firm. The authors<br />
decompose the effects <strong>of</strong> service encounter characteristics on online ratings <strong>of</strong> service<br />
firms into ‘service encounter’ and ‘social influence’ components, a novel contribution<br />
to the marketing literature in services. From a managerial perspective, when a service<br />
failure occurs in a service firm with a high online rating, the decrease in the firm’s<br />
long-term rating is more than when it has a low online rating. The opposite is true<br />
when a service recovery effort is attempted.