International Choice Modelling Conference, International Choice Modelling Conference 2015

Font Size: 
Accounting for spatial-social influence in environmental management decisions and endogeneity in a panel discrete choice model
Habtamu Tilahun Kassahun, Bo Jellesmark Thorsen, Jette Bredahl Jacobsen

Last modified: 18 May 2015


Extended abstract

Using the richness of information that panel data provides, this article uses a new approach to account for spatial-social interaction among respondents in a stated preference discrete choice experiment (CE). Our definition of  spatial-social interaction  follow the approach suggested by Walker et al. (2011): The proportion of individuals in a peer group who has selected a particular alternatives affects the individual’s decisions. Spatial-social interaction is considered as endogenous for environmental management decision making. This is because unobserved contextual factors that may have influenced the decision of an individual are likely correlated to the decision of individuals who have lived in social and spatial proximity for many years. Our approach is inspired by models in the sorting literature (Allen Klaiber and Phaneuf, 2010; Brady and Irwin, 2011; Abildtrup et al., 2013), capturing that the housing (land) market may be influenced by people choosing to live in a given neighbourhood despite houses in different neighbourhoods are traded on the same market. People in the same neighbourhood are thus more likely to share the same preferences for schooling, green spaces, etc. In our approach, we define a set of spatial-social groups, and use the information of how similar each individual’s choice is to the choice of other individuals within the same spatial-social group as an indicator function in an integrated choice and latent variable approach.  

Data and the CE design

Data for this article is obtained from a CE study which identifies incentives that motivate land users to participate in the management of private and communal lands in the Ethiopian highlands of the Upper Blue Nile Basin. The fact that watershed management practices vary with landscape topography, and regions are poorly integrated in developing countries’, high market segmentation is expected in the preference of watershed management activities (Bennett and Birol, 2010). Our survey and sampling design account for both social and spatial heterogeneity. We selected 4 of the most active erosion hot-spot sub-watersheds (Awulachew et al., 2008) within which 90 socio-spatial organised farmers’ development groups (SSOFGs) were randomly selected. Each SSOFG contains 20 to 30 households and each SSOFG represents both social and spatial units. The selected SSOFGs represent half of all SSOFGs. Once SSOFGs had been identified, 4 or 5 individuals from each SSOFG were randomly selected. The CE was administered for a total of 400 individuals. Each individual received 9 choice tasks. The CE design consists of four groups of attributes. The first group consists of three attributes that are related to the choice of on-farm soil conservation structures on private land. The second group related to common grazing land management. Two attributes are related to the amount and timing of labour contribution for watershed management works.  The last group of attributes consists of capacity building and economic incentives in the form of an additional extension service and subsidized credit facilities respectively. 

Model approach and result

In our analysis we followed an integrated choice and latent variable approach (Ben-Akiva et al., 1999; Bolduc et al., 2005). We treat the percentage of individuals in a spatially organized peer group that has chosen a particular alternative in each sequence of choices as an indirect measure (indicator) of underlying spatial-social interactions (captured as a latent variable).  In turn, the latent variables are used to explain both the choice of environmental management decisions and the indicators. In the choice model, the spatial-social latent variables are interacted with each attributes. Indeed, results show that by doing this we avoid the endogeneity issue and we capture the attribute level impact of spatial-social influence. Our result reveals that the spatial-social effects are highly significant determinants of decision making about common land management relative to private land management. This reflects that common land management requires more collaboration and social interaction. We do not find evidence for spatial-social effect on preference for the different incentives. The overall implication of our results is that watershed management activities that require collaboration of communities have a significant influence on the decision of individuals in the communities. Thus, ignoring the influence of spatial-social interaction leads to a biased model and wrong conclusions of the sensitivity of variables.


J. Abildtrup, S. Garcia, S. B. Olsen and A. Stenger (2013), 'Spatial preference heterogeneity in forest recreation', Ecological Economics 92(0): 67-77.

H. Allen Klaiber and D. J. Phaneuf (2010), 'Valuing open space in a residential sorting model of the Twin Cities', Journal of Environmental Economics and Management 60(2): 57-77.

S. B. Awulachew, M. McCartney, T. S. Steenhuis and A. A. Ahmed (2008), 'A review of hydrology, sediment and water resource use in the Blue Nile Basin', Colombo, Sri Lanka:  International Water Management Institute (IWMI) 81p. (IWMI Working Paper 131).

M. Ben-Akiva, J. Walker, A. Bernardino, D. Gopinath, T. Morikawa and A. Polydoropoulou (1999), 'Integration of choice and latent variable models. Massachusetts Institute of Technology. Cambridge, MA'.

J. Bennett and E. Birol (2010), Choice experiments in developing countries : implementation, challenges and policy implications, Cheltenham [etc.]: Edward Elgar.

D. Bolduc, M. Ben-Akiva and B. Walker (2005), 'Hybrid Choice Models with Logit Kernel: Applicability to Large Scale Models', in Lee-Gosselin, M. E. H. and Doherty, S. T., eds., Integrated Land-Use and Transportation Models: Behavioral Foundations.

M. Brady and E. Irwin (2011), 'Accounting for Spatial Effects in Economic Models of Land Use: Recent Developments and Challenges Ahead', Environmental and Resource Economics 48(3): 487-509.

J. L. Walker, E. Ehlers, I. Banerjee and E. R. Dugundji (2011), 'Correcting for endogeneity in behavioral choice models with social influence variables', Transportation Research Part A: Policy and Practice 45(4): 362-374.



Conference registration is required in order to view papers.