1. Vadali, S., A. Rangaswamy, J. Liechty (2009) Bayesian Approaches to Polyhedral Conjoint Analysis

Abstract

We develop probability models for FASTPACE (Toubia et al. 2003), an approach to Conjoint Analysis via polyhedral optimization techniques. First, we develop a theoretical foundation for FASTPACE by formulating a probability model which can be used to compute estimates identical to FASTPACE estimates. Next, via an appropriate prior for the probability model, we develop Bayesian FASTPACE and further extend it by incorporating shrinkage (i.e., hierarchical version of Bayesian FASTPACE). The resulting model called GENPACE nests FASTPACE, Bayesian FASTPACE, and Hierarchical Bayes models.

We investigate the performance of the competing models in the context of partial profile ratings-based Conjoint Analysis. Our simulations show that Bayesian FASTPACE results in 8% improvement in partworth recovery, on average, as compared to FASTPACE. Bayesian FASTPACE also has a smaller out-of-sample prediction error than FASTPACE in a real-world data set. Finally, our simulations show that GENPACE performance improves with the availability of more information, but it may exhibit poorer relative performance with inadequate levels of information (e.g., few questions asked of each respondent). We conclude by suggesting avenues for further research.

2. Clement, M., A. Rangaswamy, S. Vadali (2009) Consumer Responses to a Legal Alternative to File Sharing. (authors listed in alphabetical order) (Updated: February 2010)

Abstract

In recent years, several free and legal alternatives have become available to college students to download music files. These services may be viewed as “service interventions” that could change the mental models of students toward illegal file sharing, and result in lower downloads of illegal music files. Using the theory of planned behavior to represent mental models, we use cognitive dissonance theory to articulate the hypothesized changes to mental models. We collected data via natural field experiments at a US University and use partial least squares estimation to test our hypotheses.

We find that service interventions reduce the extent of favorable attitudes toward illegal file sharing and weaken the relationship between attitude and intent, whereby favorable attitudes toward illegal file sharing do not necessarily translate into a stronger intent to engage in file sharing. However, the interventions also strengthen the relationship between perceived benefits and attitudes toward illegal file sharing, and reinforce the positive impact on intent to file share for those who had engaged in higher levels of file sharing behavior before the interventions.

3. Vadali, S (2010) An Investigation of the Performance of Competitive Incentive Schemes and Lottery Incentive Schemes vis-a-vis Fixed Fee Incentive Schemes in Improving Conjoint Analysis (Updated: January 2010)

Abstract

Paying a fixed amount of money to participants in choice based conjoint (CBC) studies is the industry standard. Recently, Ding (2007) has shown that a lottery incentive scheme outperformed a fixed fee incentive scheme when predicting out-of-sample choices.

We achieve two research goals in the current paper to extend our understanding of incentive schemes in the context of CBC studies. One, we investigate if a higher fixed fee (e.g., $50 instead of $10) helps improve out-of-sample predictions. Two, the lottery incentive scheme does not induce competition among CBC study participants. Therefore, we investigate the theoretical properties and empirical effectiveness of competitive incentive schemes relative to lottery and fixed incentive schemes.

Our key findings with respect to hit rates for out-of-sample predictions are: (a) offering higher amounts of money is ineffective, and (b) competitive incentive schemes outperform the lottery incentive scheme (Hit Rates of 41 % and 62% for the two proposed competitive schemes vs. 29% for the lottery incentive scheme).