nep-dcm New Economics Papers
on Discrete Choice Models
Issue of 2019‒01‒07
eight papers chosen by
Edoardo Marcucci
Università degli studi Roma Tre

  1. Adoption and Cooperation Decisions in Sustainable Energy Infrastructure: Evidence from a Sequential Choice Experiment in Germany By Oberst, Christian; Harmsen - van Hout, Marjolein J. W.
  2. Estimating Consumer Inertia in Repeated Choices of Smartphones By Grzybowski, Lukasz; Nicolle, Ambre
  3. Liquidity constraints, risk preferences and farmers’ willingness to participate in crop insurance programs in Ghana By Abdulai, Awudu; Goetz, Renan; Ali, Williams; Owusu, Victor
  4. Benefit Transfer and Commodity Measurement Scales: Consequences for Validity and Reliability By Robert J. Johnston; Ewa Zawojska
  5. Multitask Learning Deep Neural Network to Combine Revealed and Stated Preference Data By Shenhao Wang; Jinhua Zhao
  6. Influence of High-Speed Railway System on Inter-city Travel Behavior in Vietnam By Tho V. Le; Junyi Zhang; Makoto Chikaraishi; Akimasa Fujiwara
  7. Does Random Consideration Explain Behavior when Choice is Hard? Evidence from a Large-scale Experiment By Victor H. Aguiar; Maria Jose Boccardi; Nail Kashaev; Jeongbin Kim
  8. Using Deep Neural Network to Analyze Travel Mode Choice With Interpretable Economic Information: An Empirical Example By Shenhao Wang; Jinhua Zhao

  1. By: Oberst, Christian (E.ON Energy Research Center, Future Energy Consumer Needs and Behavior (FCN)); Harmsen - van Hout, Marjolein J. W. (E.ON Energy Research Center, Future Energy Consumer Needs and Behavior (FCN))
    Abstract: In this paper, we propose and apply the design of a sequential discrete choice experiment to examine homeowner preferences regarding the adoption of micro-generation systems and willingness to cooperate in sustainable energy infrastructure. Adoption and cooperation decisions of private households in the energy sector are complex, interlinked, and assumably sequential. A common design with single choice tasks reflecting both adoption and cooperation decisions is assumed as cognitively too burdensome for survey respondents. The objective of the proposed sequential choice task design is twofold. Firstly, reducing complexity for respondents. Secondly, reflecting a step-wise decision process as is appropriate for the studied decisions. Our application from the energy sector is motivated by the need for innovative business models for non-industrial prosumers providing flexibility services in (local) distribution grids, due to an increasing amount of volatile and decentrally generated electricity. Results indicate that respondents reveal more pronounced preferences when dealing with their decision in sequential steps and that the task design has a lasting effect on respondents’ choices. By estimating latent class logit models, five consumer classes are identified and labeled by their distinguished motivational foci: costs (1), climate protection (2), self-supply (3), local reference (4), and other (5).
    Keywords: Choice Experiment; Micro-generation; Renewable Energy; Community Energy; Energy Transition
    JEL: C25 D12 Q42
    Date: 2017–10
  2. By: Grzybowski, Lukasz; Nicolle, Ambre
    Abstract: In this paper, we use a unique database on switching between mobile handsets in a sample of about 5,000 subscribers using tariffs without commitment from a single mobile operator on monthly basis between March 2012 and December 2014. We estimate discrete choice model in which we account for disutility from switching to a different operating systems and handset brands and for unobserved time-persistent preferences for operating systems and brands. Our estimation results indicate presence of significant state-dependency in the choices of operating systems and brands. We find that it is harder for consumers to switch from iOS to Android and other operating systems than from Android and other operating systems to iOS. Moreover, we find that there is significant time-persistent heterogeneity in preferences for different operating systems and brands, which also leads to state-dependent choices. We use our model to simulate market shares in the absence of switching costs and conclude that the market share of Android and smaller operating systems would increase at the expense of the market share of iOS.
    Keywords: Smartphones,Consumer Inertia,Switching Costs,Mixed Logit,iOS,Android
    JEL: L13 L50 L96
    Date: 2018
  3. By: Abdulai, Awudu; Goetz, Renan; Ali, Williams; Owusu, Victor
    Abstract: This paper analyzes smallholder farmers’ decisions to participate in crop insurance programs, using cross-sectional data from cocoa farmers in the Ashanti, Brong-Ahafo and Western Regions of Ghana. Given the significance of output uncertainty and imperfect capital and insurance markets, we develop a theoretical framework to show how risk preferences and liquidity constraints influence farmers’ crop insurance participation decisions. We use a stated preference approach to obtain information on farmers’ willingness to participate in crop insurance programs, and a discrete choice model to examine the factors that influence their participation decisions. We find that risk preferences and liquidity constraints influence farmers’ willingness to participate in crop insurance programs. The results also show that the probability of participating in crop insurance programs is higher for males, the more educated, and those who trust others. The levels of fertilizer and pesticide expenditure and the access to credit are also found to significantly influence the decision to adopt the programs.
    Keywords: Crop Production/Industries, Risk and Uncertainty
    Date: 2018–12–20
  4. By: Robert J. Johnston (Clark University); Ewa Zawojska (University of Warsaw, Faculty of Economic Sciences)
    Abstract: Non-market goods can be measured on cardinal or relative scales. Consider a marsh of two hundred acres, of which twenty acres would be affected by a policy. The same affected area can be measured in cardinal terms (twenty acres) or as a relative proportion (ten percent of the marsh). This seemingly inconsequential transformation can have significant implications for benefit transfer across sites—a simple observation that remains unacknowledged by the literature. This article provides the first theoretical and empirical evaluation of variable measurement conventions within benefit transfer, deriving conditions under which different types of measurement scales are expected to enhance validity and reliability. Theoretical results are illustrated using an application of discrete choice experiments to coastal flood adaptation in two Connecticut (USA) communities. Empirical findings validate expectations from the theoretical model, with both suggesting that transfers over goods measured in relative units may substantially outperform transfers over goods measured in cardinal units.
    Keywords: Benefit Transfer; Flood Adaptation; Measurement; Scale; Reliability; Stated Preference; Validity; Willingness to Pay
    JEL: Q51 Q54
    Date: 2018
  5. By: Shenhao Wang; Jinhua Zhao
    Abstract: It is an enduring question how to combine revealed preference (RP) and stated preference (SP) data to analyze travel behavior. This study presents a new approach of using multitask learning deep neural network (MTLDNN) to combine RP and SP data and incorporate the traditional nest logit approach as a special case. Based on a combined RP and SP survey in Singapore to examine the demand for autonomous vehicles (AV), we designed, estimated and compared one hundred MTLDNN architectures with three major findings. First, the traditional nested logit approach of combining RP and SP can be regarded as a special case of MTLDNN and is only one of a large number of possible MTLDNN architectures, and the nested logit approach imposes the proportional parameter constraint under the MTLDNN framework. Second, out of the 100 MTLDNN models tested, the best one has one shared layer and five domain-specific layers with weak regularization, but the nested logit approach with proportional parameter constraint rivals the best model. Third, the proportional parameter constraint works well in the nested logit model, but is too restrictive for deeper architectures. Overall, this study introduces the MTLDNN model to combine RP and SP data, relates the nested logit approach to the hyperparameter space of MTLDNN, and explores hyperparameter training and architecture design for the joint demand analysis.
    Date: 2019–01
  6. By: Tho V. Le; Junyi Zhang; Makoto Chikaraishi; Akimasa Fujiwara
    Abstract: To analyze the influence of introducing the High-Speed Railway (HSR) system on business and non-business travel behavior, this study develops an integrated inter-city travel demand model to represent trip generations, destination choice, and travel mode choice behavior. The accessibility calculated from the RP/SP (Revealed Preference/Stated Preference) combined nested logit model of destination and mode choices is used as an explanatory variable in the trip frequency models. One of the important findings is that additional travel would be induced by introducing HSR. Our simulation analyses also reveal that HSR and conventional airlines will be the main modes for middle distances and long distances, respectively. The development of zones may highly influence the destination choices for business purposes, while prices of HSR and Low-Cost Carriers affect choices for non-business purposes. Finally, the research reveals that people on non-business trips are more sensitive to changes in travel time, travel cost and regional attributes than people on business trips.
    Date: 2018–12
  7. By: Victor H. Aguiar; Maria Jose Boccardi; Nail Kashaev; Jeongbin Kim
    Abstract: We study population behavior when choice is hard because considering alternatives is costly. To simplify their choice problem, individuals may pay attention to only a subset of available alternatives. We design and implement a novel online experiment that exogenously varies choice sets and consideration costs for a large sample of individuals. We provide a theoretical and statistical framework that allows us to test random consideration at the population level. Within this framework, we compare competing models of random consideration. We find that the standard random utility model fails to explain the population behavior. However, our results suggest that a model of random consideration with logit attention and heterogeneous preferences provides a good explanation for the population behavior. Finally, we find that the random consideration rule that subjects use is different for different consideration costs while preferences are not. We observe that the higher the consideration cost the further behavior is from the full-consideration benchmark, which supports the hypothesis that hard choices have a substantial negative impact on welfare via limited consideration.
    Date: 2018–12
  8. By: Shenhao Wang; Jinhua Zhao
    Abstract: Deep neural network (DNN) has been increasingly applied to microscopic demand analysis. While DNN often outperforms traditional multinomial logit (MNL) model, it is unclear whether we can obtain interpretable economic information from DNN-based choice model beyond prediction accuracy. This paper provides an empirical method of numerically extracting valuable economic information such as choice probability, probability derivatives (or elasticities), and marginal rates of substitution. Using a survey collected in Singapore, we find that when the economic information is aggregated over population or models, DNN models can reveal roughly S-shaped choice probability curves, inverse bell-shaped driving probability derivatives regarding costs and time, and reasonable median value of time (VOT). However at the disaggregate level, choice probability curves of DNN models can be non-monotonically decreasing with costs and highly sensitive to the particular estimation; derivatives of choice probabilities regarding costs and time can be positive at some region; VOT can be infinite, undefined, zero, or arbitrarily large. Some of these patterns can be seen as counter-intuitive, while others can potentially be regarded as advantages of DNN for its flexibility to reflect certain behavior peculiarities. These patterns broadly relate to two theoretical challenges of DNN, irregularity of its probability space and large estimation errors. Overall, this study provides a practical guidance of using DNN for demand analysis with two suggestions: First, researchers can use numerical methods to obtain behaviorally intuitive choice probabilities, probability derivatives, and reasonable VOT. Second, given the large estimation errors and irregularity of the probability space of DNN, researchers should always ensemble either over population or individual models to obtain stable economic information.
    Date: 2018–12

This nep-dcm issue is ©2019 by Edoardo Marcucci. It is provided as is without any express or implied warranty. It may be freely redistributed in whole or in part for any purpose. If distributed in part, please include this notice.
General information on the NEP project can be found at For comments please write to the director of NEP, Marco Novarese at <>. Put “NEP” in the subject, otherwise your mail may be rejected.
NEP’s infrastructure is sponsored by the School of Economics and Finance of Massey University in New Zealand.