Measuring the effect of unobservable factors in residential

Measuring the effect of unobservable factors in residential choice behavior Associate Professor Berndt Lundgren, KTH, Stockholm, Sweden Professor Fan Yang Wallentin Uppsala University, Sweden 1

About the authors Associate Professor Berndt Lundgren Division of Real Estate and Construction Management Royal Institute of Technology Stockholm, Sweden Professor Fan Wallentin Department of Statistics, Uppsala University, Sweden Department of statistics, Tianjin University of Finance and Economics, China 2

Problem definition • The heterogeneity in residential choice behavior creates problems since heterogeneity is caused by differences in socio-demographics, attitudes and residential preferences. Traditional econometric models, such as the mixed logit model, latent class analysis have proven not to capture endogenous linkages between latent variables (Hoshino, 2011; Haifeng, et al. , 2014). • How can we measure the effect of endogenous linkages between unobservable factors such as attitudes on residential choice behavior? • By unobservable factors we mean subjectively experienced factors such as safety, perceived noise, apartment standard, functional use of an apartment. • By residential choice behavior we mean actual buying behavior, renting or not renting, intention to accept a detailed development or not, intention to buy or rent an apartment in a newly constructed residential development 3

The purpose of this paper • The purpose of this research is to investigate the effect of unobservable factors on residential choice behavior. • We firstly use a two-step procedure involving a qualitative method to identify factors that may affect potential residents to a new residential development. Secondly we apply statistical analysis to verify the effect of those factors on residential choice behavior. • Structural equation modeling (SEM) and two-group analysis in LISREL are used to verify differences in consumer attitudes between those who didn’t became renters and those who did in relation to value-for-money conclusions. 4

Literature review housing research • Most of the variables that enter into residential consumers’ utility function are latent, unobservable and mostly related to personal experience and values, thus difficult to measure Hoshino (2011), Molin et al (2012) and Walker & Li (2007). • Variables often used in residential choice analysis 1. Price (Dieleman, 2001) 2. Size (Dieleman, 2001) 3. Income (Geist & Mc. Manus, 2008) 4. Location (Lee & Waddell, 2010) 5. Education level (Timmermans et al. , 1992) 6. Employment status (Dieleman, 2001) 7. Age (Lee, Oropesa & Kanan, 1994) 5

Literature review housing research • Latent choice models (LCA) with segment covariates has little ability to tackle the endogenous nature of specific links between sociodemographic status, attitudes and residential location preferences. SEM would be a better method to explore these endogenous linkages, Haifeng, et al. , (2015), • Different methods used for housing choice analysis 1. Traditional housing demand research method 2. Decision plan nets method 3. Meaning structure method (extension of the means-end chain method) 4. Multi-attribute utility method 5. Conjoint analysis method 6. Residential images method 7. Lifestyle method 8. Neoclassical economic analysis 9. Longitudinal analysis ( Jensen et al. , 2011) 6

Literature review marketing research The means-end chain theory, Gutman, (1982) suggest that consumers relate product attributes to personal positive and negative consequences: • What does this attribute do for me or what is this attribute good for? • This knowledge is used to formulate knowledge about why consumers prefer a specific product attribute and to devide consumers into different segments. 7

Method Application of the means-end chain model 8

Method The case study Hornsberg Strand project, Stockholm, Sweden. 9

Method The case study Hornsberg Strand project, Stockholm, Sweden. 10

Method Confirmatory factor analysis, LISREL 11

Method Stage 1: Scale generation and initial purification - Independent interviews (n=20) using the laddering technique. - Analysis and categorization of answers according to the means-end chain theory. - Generate initial pool of items. - Advisory board screening to assess validity. - produced 34 cognitive attitude statements. - Pretest (n=51), reduced 3 items. 31 items remained. Stage 2: Scale purification through exploratory factor analysis - Collection of responses from a representative sample (n=282). -Exploratory Factor analysis, resulted in 7 dimensions. - Assessment of content validity, scales reliability, cumulative variation (65, 7%). -Removed one dimension. - Measured using a 7 point Likert scale. Stage 3: Reliability and validity assessment using SEM - Conformatory factor analysis on 6 dimension. - Assess factor structure. - Assess model fit. - Assess scale and construct reliability. - Asess discriminant validity. - Perform testing of hypothesis. - Validating dimensions that impact marketability according to their effect on value-for-money. 12

Testing of hypothesis Dimensions defined by exploratory factor analysis: Noise Accessability (η 1) Overall impression Relaxation (η 2) Customer perceived value Standard of the apartment Functional use of the apartment Architecture Items from the qualiative study 13

Results group analysis: residential renters, t-values (n=171), LISREL 14

Results group analysis: residential non-renters, t-values (n=111), LISREL 15

Results 16

Conclussion • SEM seem to be a new approach in measuring residential choice behavior in real estate research. The SEM methodology is well known in for example food research. • There is distinct advantages using SEM since we can measure endogenous linkages between latent variables and verify which unobservable factors out of a set of potentiella important one´s that have an effect on residential choice behavior. • This methodology can be used for research in any real estate asset. Knowledge of this methodology is surprisingly unknown to real estate researcher. 17

Thank’s for your attention! Questions? 18