Stated Choice design comparison in a developing country: recall and attribute nonattendance

Area of Science:

• Experimental Economics
• Health Services Research
• Survey Methodology

Background:

• Experimental designs are crucial for Stated Choice (SC) studies.
• Limited empirical evaluation exists for non-orthogonal constrained SC designs using non-zero prior estimates.
• This study investigates the statistical benefits of efficient SC designs compared to orthogonal designs.

Purpose of the Study:

• To statistically compare the performance of orthogonal and efficient SC experimental designs.
• To evaluate the effect of respondent literacy on Attribute non-Attendance (ANA) across different SC designs.
• To model consumer choice of primary health care providers in rural North India using a real SC design.

Main Methods:

• Comparison of orthogonal and efficient SC designs using D-error, Standardised Akaike Information Criteria (AIC), and coefficient/standard error estimates.
• Controlling for stated and inferred Attribute non-Attendance (ANA).
• Utilizing data from 623 respondents in rural North India.

Main Results:

• Efficient SC designs show statistically better performance when controlling for ANA.
• Efficient designs yield superior model-fit with a 14-day recall period.
• Orthogonal designs perform better in model-fit with longer recall periods (30 days to 12 months).

Conclusions:

• The cognitive demand of efficient designs impacts ANA, more so for illiterate respondents.
• Relaxing orthogonality in SC designs enhances information collection.
• The effectiveness of efficient designs depends on accurate non-zero priors and correct recall period specification.