Method and System for Consistent Estimation from Self Selected User Feedback by Providing Financial Incentives to Users
Publication Date: 2014-Sep-04
The IP.com Prior Art Database
Ayman Farahat: INVENTOR [+2]
AbstractA method and system is developed for consistent estimation of a population response from a subset of self selected respondents or users who provide feedback.
A method and system is developed for consistent estimation of a population response from a subset of self selected respondents or users who provide feedback.
The method and system focuses on varying the probability of providing feedback by utilizing financial incentives including products / goods as instruments where different users are randomly offered different financial incentives for providing feedback. The random variation in probability of providing feedback satisfies the exclusion restriction needed to consistently estimate the models using a Heckman style model.
Alternatively, the method and system makes donations to charities that the users select from a set of pre-selected list of charitable organizations, thus reducing the transaction cost and making it feasible to efficiently collect feedback.
Heckman developed a sample selection model, shown below, that is used for identifying the factors that impact how users respond to surveys. The challenge in estimating these models is that while the goal is to estimate the response of the population, data is only available for survey respondents. Heckman showed that direct estimation without taking into account the selection process leads to biased estimates. To address this issue, he proposed to jointly consider whether users respond to a survey and how they respond to the survey. The sample selection utilizes two equation models, such as a probit selection model (eq. 2) that determines whether users respond to the survey and a response model (eq. 1) that determines how users respond to the survey (for example by giving a numerical rating). Further, error for the outcome is specified by error eq. 3 and eq. 4 specifies the error term for the probit model. Additionally, the error term in both the equations is correlated and represented using eq. 5.
Consistent estimation of the two equation models requi...