Preference modelling in R: a trial on home buyers’ willingness to pay

• Main Authors: Syahid, Ahmed, Tareq, Mohammad Ali, Nahar, Aizul
• Format: Article
• Published: Pontifícia Universidade Catolica de Sao Paulo 2021
• Subjects: HD Industries. Land use. Labor; HF Commerce
• Online Access: http://eprints.utm.my/id/eprint/96612/; http://dx.doi.org/10.23925/2179-3565.2021v12i2p154-173

Modelling stated preferences is an almost mystical science and as there is no data explaining how the sustainable feature in homes would effectively encourage homebuyers to invest in sustainable housing, it is important to investigate the buyers' willingness to pay (WTP) for sustainable housing. The study of stated preferences often requires the use of specialised software or proprietary programs, which can be difficult and/or expensive to use. This study proposes to re-purpose the support.CEs' package, a program written in the R programming language, from its agronomic roots to measure home buyer preferences for sustainable housing. These are demonstrated through a stated preference discrete choice experiment of choosing model houses with differing levels of energy savings, renewable energy generation, landscaping, soundproofing, ventilation, and price differences. A pilot study was performed using an online survey, constructed using the L-MA design tool provided in the support.CEs' package. The survey was also separated into six blocks of six questions each to reduce the cognitive burden on respondents. The survey was distributed through social media channels. Preliminary results with a limited sample of 20 respondents with mixed income, age, and occupational demographics, analysed using the package's clogit function, that performs conditional logit estimations, have shown that the results have a statistically reliable adjusted rho-squared value and that all coefficients show the expected signs. From this study, it can be concluded that the support.CEs' package can be used to model home buyer preferences and that adequate blocking allows for the measurement of a higher number of variables despite having smaller sample sizes.