Choice based conjoint analysisBackground
As noted in popular press government policies, the labeling of food products as “genetically modified” is a controversial topic. To help us understand consumer behavior in response to the labels, choice based conjoint data on consumer preferences was collected. The attributes and levels for the study are as follows:
From the raw data as seen above, a design matrix, shown below where salmon and farm_gmo was used as the baseline for a logistic regression, was used to develop 9 profiles to be evaluated by consumers. Consumers offered a binary yes/no decision to the question of would they buy each profile. A sample of 109 consumers completed the study.
Estimates and Probabilities
First, I specified a binary logit model of consumer choice. For the model, I treated “salmon” as the baseline type and “farm raised/genetically modified” as the baseline production method. Using dummy code in finding the utility of each choice, I was able to prepare a model to find the coefficients in my logit model. Using Excel, I built the total log-likelihood function and used solver to find the parameter values for the coefficients that maximize the total log-likelihood.
Then, I computed the predicted probabilities for each individual using the estimates of the model parameters.
Importance
I used Excel to then find the derived importance of each attribute.
Willingness to Pay
Holding production method constant, WTP analysis indicates that the market is willing to pay $3.30 less for tuna relative to salmon and $4.88 less for halibut relative to salmon. This suggests that when setting the price, we should set salmon's price higher than both tuna and halibut, and we can set halibut's price the lowest.
When holding type constant, people would be willing to pay $23.57 more for wild production relative to farm/gmo and $10.02 more for farm relative to farm/gmo. When setting the price, farm/gmo should be the lowest in price while wild should be the highest price.
Share Analysis
In a scenario where there are four different products in the market, a question could be posed:
What would happen to the share of Product 4 if we changed the production method from farm raised to farm raised gmo?
Using the Logit Rule, it can be seen that Product 4’s market share is predicted to decrease.
