Abstract
Suppose you have the possibility to choose to adopt one of a number of different behaviors or to choose to buy one of a number of different products, and suppose your choice is influenced by your individual perception of the average choices made by others. Economists Brock and Durlauf (in Am. Econ. Rev. 92(2):298, 2002; The Economy as an Evolving Complex System III. Oxford University Press, New York, 2006) have derived seminal theoretical results for the equilibrium behavior of the multinomial discrete choice model with social interactions, assuming homogeneous decision-makers, global interactions and laws of large of numbers. The research presented in this paper extends Brock and Durlauf’s model to allow for unobserved preference heterogeneity between choice alternatives by studying the nested logit model. Next, by drawing on the computational possibilities permitted through social simulation of multi-agent systems (MAS), this paper relaxes the assumption of global interactions and considers instead local interactions within several hypothesized social and spatial network structures. Additional heterogeneity is thus hereby induced by the influence on a given decision-maker’s choice by the particular network connections he or she has and the particular perceived percentages, for example, of the agent’s neighbors or socio-economic peers making each choice. Discrete choice estimation results controlling these heterogeneous individual preferences are embedded in a multi-agent based simulation model in order to observe the evolution of choice behavior over time with socio-dynamic feedback due to the network effects. The MAS approach also gives us an additional advantage in the possibility to test size effects, and thus relax the assumption of large numbers, as well as test the effect of different initial conditions. Finally an extra benefit is gained via the MAS approach in that we are not confined to study only the equilibrium behavior, and have the possibility here to observe the time-varying trajectories of the choice behavior. This is important since smaller network sizes are revealed to be associated with higher volatility of the choice behavior in this model, and consequently stochastic cycling between equilibria. Averaged over time, the emergent behavior in such case yields a quite different picture than the theoretical results predicted by Brock and Durlauf. Furthermore being able to observe the emergent behavior allows us to see the subtle role of the unobserved heterogeneity in the nested logit model in breaking the symmetry of the multinomial logit model. We can see the temporal patterns by which theoretically predicted dominant equilibria emerge or not according to different social and spatial network scenarios. With an eye towards application in the context of transportation mode choice, we conclude highlighting limitations of our present study and recommendations for future work.
Similar content being viewed by others
Notes
We thank an anonymous reviewer for suggestions regarding interpretation of the alternative nesting structures. A multinomial probit model would have allowed us to capture unobserved heterogeneity without presuming a particular nesting structure, however in this paper we specifically want to be able to draw on the theoretical results made possible by the analytical closed form of the nested logit model.
In our companion work to this paper (Dugundji and Gulyás 2012a), also other socioeconomic characteristics of the decision-makers such as gender, age, and income and respondent-specific attributes of their trip to work such as travel time used from this table for computing systematic utilities in the iteration phase.
Transportation modal split in the greater Amsterdam region for home to work: 1947—20 % bicycle, 75 % transit, 5 % car; 1960—18 % bicycle, 67 % transit, 16 % car; 1971—13 % bicycle, 41 % transit, 46 % car; 1991—18 % bicycle, 22 % transit, 60 % car (Bertolini 2007); in this paper, end of 1990’s: 26.7 % bicycle, 23.7 % transit, 49.6 % car.
In the words of one anonymous reviewer: “Network effects lead to path-dependency … if you replace biking or transit with cars, it can be very difficult to go back, even if there are a lot of bike and transit infrastructure investments.”
References
Aoki M (1995) Economic fluctuations with interactive agents: dynamic and stochastic externalities. Jpn Econ Rev 46(2):148–165
Arentze TA, Timmermans HJP, Hofman F (2008) Creating synthetic household populations: problems and approach. Transp Res Rec 2014:85–91
Arentze TA, van den Berg P, Timmermans HJP (2012) Modeling social networks in geographic space: approach and empirical application. Environ Plan A 44(5):1101–1120
Barrett CL, Beckman RJ, Khan M, Marathe MV, Stretz PE, Dutta T, Lewis B (2009) Generation and analysis of large synthetic social contact networks. In: Rossetti MD, Hill RR, Johansson B, Dunkin A, Ingalls RG (eds) Proceedings of the 2009 winter simulation conference. IEEE Press, New York
Ben-Akiva M (1973) Structure of passenger travel demand models. Dissertation, Massachusetts Institute of Technology, Cambridge
Ben-Akiva M, Lerman SR (1985) Discrete choice analysis: theory and application to travel demand. MIT Press, Cambridge
Berry S, Levinsohn J, Pakes A (1995) Automobile prices in market equilibrium. Econometrica 63:841–889
Berry S, Levinsohn J, Pakes A (2004) Differentiated products demand systems from a combination of micro and macro data: the new vehicle market. J Polit Econ 112(1):68–105
Bertolini L (2007) Evolutionary urban transportation planning: an exploration. Environ Plan A 39:1998–2019
Bhat CR, Pendyala RM (2005) Modeling intra-household interactions and group decision-making. Transportation 32:443–448
Bierlaire M (2003) BIOGEME: a free package for the estimation of discrete choice models. In: Proceedings of the 3rd Swiss transportation research conference, Ascona, Switzerland
Blume LE, Durlauf SN (2003) Equilibrium concepts for social interaction models. Int Game Theory Rev 5(3):193–209
Blume LE, Brock WA, Durlauf SN, Ioannides Y (2011) Identification of social interactions. In: Benhabib J, Bisin A, MO Jackson (eds) Handbook of social economics. Elsevier, North Holland
Brock WA, Durlauf SN (2001) Discrete choice with social interactions. Rev Econ Stud 68:235–260
Brock WA, Durlauf SN (2002) A multinomial choice model of neighborhood effects. Am Econ Rev 92(2):298–303
Brock WA, Durlauf SN (2006) Multinomial choice with social interactions. In: Blume LE, Durlauf SN (eds) The economy as an evolving complex system III. Oxford University Press, New York
Buehler R, Pucher J (2012) Demand for public transport in Germany and the USA: an analysis of rider characteristics. Transp Rev 32(5):541–567
Butts CT, Acton RM (2011) Spatial modeling of social networks. In: Nyerges T, Couclelis H, McMaster R (eds) The Sage handbook of GIS and society research. SAGE, Thousand Oaks
Butts CT, Acton RM, Hipp JR, Nagle NN (2012) Geographical variability and network structure. Soc Netw 34(1):82–100
Domencich T, McFadden D (1975) Urban travel demand. North Holland, Amsterdam
Dugundji ER (2006) Residential choice and the geography of family networks: some considerations. In: Proceedings of the 85th annual meeting of the transportation research board, Washington, DC
Dugundji ER (2012) Socio-dynamic discrete choice: equilibrium behavior of the nested logit model with social interactions. In: Interdisciplinary workshop on information and decision in social networks. Massachusetts Institute of Technology, Cambridge
Dugundji ER (2013) Socio-dynamic discrete choice: theory and application. Dissertation, Universiteit van Amsterdam, Amsterdam, Netherlands
Dugundji ER, Gulyás L (2003) An exploration of the role of global versus local and social versus spatial networks in transportation mode choice behavior in the Netherlands. In: Proceedings of AGENT 2003: challenges in social simulation. Argonne National Laboratory and University of Chicago, Chicago. http://www.dis.anl.gov/pubs/51409.pdf
Dugundji ER, Gulyás L (2006) Emergent opinion dynamics on endogenous networks. In: Trajkovski GP, Collins SG (eds) Interaction and emergent phenomena in societies of agents. AAAI Press, Menlo Park. Technical Report FS-06-05
Dugundji ER, Gulyás L (2008) Socio-dynamic discrete choice on networks: impacts of agent heterogeneity on emergent equilibrium outcomes. Environ Plan B, Plan Des 35:1028–1054
Dugundji ER, Gulyás L (2012a) Socio-dynamic discrete choice on networks in space: the role of utility parameters and connectivity. Proc Comput Sci 10:827–832
Dugundji ER, Gulyás L (2012b) Socio-dynamic discrete choice applied to intercity travel demand: issues in estimation. In: Proceedings of the federated conference on computer science and information systems (FedCSIS), Wroclaw, Poland. IEEE Press, New York
Dugundji ER, Walker JL (2005) Discrete choice with social and spatial network interdependencies: an empirical example using mixed generalized extreme value models with field and panel effects. Transp Res Rec 1921:70–78
Dugundji ER, Kapoen LL, le Clercq F, Arentze TA, Timmermans HJP, Veldhuisen KJ (2001) The long-term effects of multi-modal transportation networks: the residential choice behavior of households. In: Proceedings of the 9th world conference on transportation research, Seoul, Korea
Dugundji ER, Páez A, Arentze TA (2008) Social networks, choices, mobility and travel. Environ Plan B, Plan Des 35(6):956–960
Dugundji ER, Páez A, Arentze TA, Walker JL, Carrasco JA, Marchal F, Nakanishi H (2011) Transportation and social interactions. Transp Res, Part A, Policy Pract 45(4):239–247
Dugundji ER, Scott DM, Carrasco JA, Páez A (2012) Urban mobility and social-spatial contact—introduction. Environ Plan A 44(5):1011–1015
Fischel WA (2006) The Tiebout model at 50. Lincoln Institute of Land Policy, Cambridge
Fukuda D, Morichi S (2007) Incorporating aggregate behavior in an individual’s discrete choice: an application to analyzing illegal bicycle parking behavior. Transp Res, Part A, Policy Pract 41:313–325
Goetzke F (2008) Network effects in public transit use: evidence from a spatially autoregressive mode choice model. Urban Stud 45(2):407–417
Goetzke F, Andrade PM (2009) Walkability as a summary measure in a spatially autoregressive mode choice model: an instrumental variable approach. In: Páez A, Buliung RN, Le Gallo J, Dall’erba S (eds) Progress in spatial analysis: methods and applications, advances in spatial science. Springer, Berlin, Heidelberg
Goetzke F, Rave T (2011) Bicycle use in Germany: explaining differences between municipalities with social network effects. Urban Stud 48(2):427–437
Goetzke F, Weinberger R (2012) Separating contextual from endogenous effects in automobile ownership models. Environ Plan A 44(5):1032–1046
Hackney J, Kowald M (2011) Exponential random graph models of the Zurich snowball survey. In: Futurenet workshop: social network analysis in transport, Manchester, UK
Ioannides Y (2006) Topologies of social interactions. Econ Theory 28:559–584
Kenworthy J, Laube F (2005) An international comparative perspective on sustainable transport in European cities. Eur Spat Res Policy 12:11–50
Kollman K, Miller JH, Page SE (1997) Political institutions and sorting in a Tiebout model. Am Econ Rev 87(5):977–992
Kowald M, Axhausen KW (2012) Spatial distribution of connected leisure networks: selected results from a snowball sample. Environ Plan A 44(5):1085–1100
Louviere J, Train KE, Ben-Akiva M, Bhat C, Brownstone D, Cameron TA, Carson RT, Deshazo JR, Fiebig D, Greene W, Hensher D (2006) Recent progress on endogeneity in choice modeling. Mark Lett 16(3–4):255–265
Macal CM, North MJ (2010) Tutorial on agent-based modeling and simulation. J Simul 4:151–162
Manski C (1995) Identification problems in the social sciences. Harvard University Press, Cambridge
Maxwell DT, Carley KM (2009) Principles for effectively representing heterogeneous populations in multi-agent simulations. In: Tolk A, Jain LC (eds) Comp sys in knowledge-based environments. SCI, vol 168. Springer, Berlin, Heidelberg,
McFadden D (1978) Modelling the choice of residential location. In: Karlquist A et al (eds) Spatial interaction theory and residential location. North Holland Press, Amsterdam
Musgrave RA (1959) The theory of public finance: a study in public economy. McGraw-Hill, New York
Oates WE (1969) The effects of property taxes and local public spending on property values: an empirical study of tax capitalization and the Tiebout hypothesis. J Polit Econ 77:957–971
Páez A, Scott DM (2007) Social influence on travel behavior: a simulation example of the decision to telecommute. Environ Plan A 39:647–665
Páez A, Scott DM, Volz E (2008) A discrete-choice approach to modeling social influence on individual decision making. Environ Plan B, Plan Des 35(6):1055–1069
Rivers D, Vuong Q (1988) Limited information estimators and exogeneity tests for simultaneous probit models. J Econom 39:347–366
Samuelson PA (1954) The pure theory of public expenditures. Rev Econ Stat 36(4):387–389
Samuelson PA (1955) Diagrammatic exposition of a pure theory of public expenditures. Rev Econ Stat 37(4):350–356
Tiebout CM (1956) A pure theory of local expenditures. J Polit Econ 64:416–424
Timmermans HJP, Zhang J (2009) Modeling household activity travel behavior. Transp Res, Part B, Methodol 43:187–190
Train KE (2009) Discrete choice methods with simulation, 2nd edn. Cambridge University Press, Cambridge
Walker JL, Ehlers E, Banerjee I, Dugundji ER (2011) Correcting for endogeneity in behavioral choice models with social influence variables. Transp Res, Part A, Policy Pract 45:362–374
Acknowledgements
The authors would like to gratefully acknowledge discussion with Harry Timmermans, Theo Arentze, Cars Hommes, Frank le Clercq, Loek Kapoen, George Kampis, József Váncza and András Márkus, as well as the valuable and insightful comments from three anonymous reviewers which greatly improved the exposition in this paper. Very special thanks are also due to Guus Brohm and Nelly Kalfs at the Agency for Infrastructure, Traffic and Transport of the Municipality of Amsterdam, to Willem Vermin and the High Performance Computing support team at SARA Computing and Networking Services, Amsterdam and to David Sallach, Michael North, Charles Macal and the RePast development team. In addition we would like to express our appreciation more generally to a number of other committed scholars and out-of-the box thinkers that influenced our own thinking during formative years: Nigel Gilbert, keynote speaker at the first Lake Arrowhead Conference where the authors first met, and one of the team of visionary signatories of the European Social Simulation Association (ESSA); Axel Leijonhufvud, Robert Axtell and Masanao Aoki for eye-opening introduction to the world of adaptive economic processes; Kathleen Carley, a beacon for inspiration on the realm of possibilities of network analysis coupled with population scale social simulation; and Scott Page and John Miller, organizers of the Santa Fe Institute Graduate Workshop on Computational Economics, for pointing us to William Brock and Steven Durlauf’s work on multinomial choice with social interactions during an intensive two weeks of learning. Finally we would like to thank Clara Smith, Fred Amblard, Paul Chapron, Matthias Maillard and Samuel Thiriot for their wonderful initiative to bring together researchers in social network analysis and multi-agent systems at the lively SNAMAS workshop at ESSA 2011. The authors claim full responsibility for any errors.
Author information
Authors and Affiliations
Corresponding author
Appendix
Appendix
1.1 A.1 Individual choice probabilities for trinary nested logit
Recall the definition of the nested logit model presented in Sect. 3.1 and consider a scenario as in Fig. 1 where elemental alternative 0 is an isolate in its own “nest”, and elemental alternatives 1 and 2, assumed to be correlated, are nested together. The probability of choosing alternative 0 within nest0, conditional on having chosen nest0 is:
where the systematic utility for choosing elemental alternative 0 is a linear-in-parameter β function of the proportion p 0 of decision-making agents that have chosen alternative 0, and μ L is the scale parameter that characterizes the distribution of the Gumbel error terms within the nest. The inclusive value for nest0 is then:
The probabilities of choosing respectively alternative 1 and 2 within nest1, conditional on having chosen nest1, are likewise the binary choice probabilities:
where the systematic utilities for choosing elemental alternative 1 or 2 are respectively a linear-in-parameter β function of the proportions p 1 and p 2 of decision-making agents that have chosen alternatives 1 and 2, and μ L is again the scale parameter that characterizes the distribution of the Gumbel error terms within the nest. The inclusive value for nest1 is then:
The probabilities of choosing respectively nest0 and nest1 among the set of nests are:
where we have normalized the upper level scale parameter μ to unity. Thus the probabilities of choosing respectively alternatives 0, 1, 2 among all possible elemental alternatives in the choice set are:
1.2 A.2 Evaluation of eigenvalues of the Jacobian matrix
Let parameters β and μ L be real and finite with mode shares p 0,p 1 defined on [0,1]. Recall the converted system of equations from Sect. 3.1 which defines the steady-state equilibrium of the sociodynamic process for the trinary nested logit with social interactions:
This planar system of equations can be solved numerically for given empirical values of β and μ L . Depending on these values of β and μ L , the system may have more than one solution.
By evaluating the Jacobian J at a given solution p=(p 0,p 1), we can determine the stability of that solution.
Namely, if all the eigenvalues of the Jacobian matrix J have negative real parts, then the equilibrium point is asymptotically stable; if at least one of the eigenvalues of the Jacobian matrix has positive real part, then the equilibrium point is unstable. For the system given by (27), the four terms in the Jacobian matrix of g=(g 0,g 1) can be computed using the sum rule, the product rule and the chain rule as follows:
The characteristic equation to determine the eigenvalues of the Jacobian matrix is:
Solving (33) for eigenvalues using the quadratic formula we have:
For example, for the case study in Sect. 4.4 we can determine the stability of the solutions as in Table 9.
1.3 A.3 Analytical benchmark: a counter example
In this section we provide a hypothetical counter example to the results in Table 3 and Table 4 in Sect. 4.4 to demonstrate that: (1) the estimated parameter for the nest scale parameter need necessarily be statistically weak; and (2) the initial modal split in the data need not necessarily be a saddle point equilibrium of the minimal sociodynamic nested logit model. For convenience of comparison, suppose we again have 2913 decision-making agents, but instead the number of respondents in the data sample that chose to commute to work by bicycle or moped/motorcycle is 100, the number of respondents that chose to commute by external and/or internal system public transit is 500, and the number of respondents that chose to commute by car as driver or passenger is 2313. In such case, the modal split is roughly: 3,4 % mode share bicycle/moped/motorcycle; 17,2 % mode share public transit; and 79,4 % mode share car driver/passenger. Such a commuter modal split would be roughly similar to that in 2011 in the greater metropolitan areas of two of the top five most bicycle-friendly cities in the United States, Seattle and Minneapolis.
As in Sect. 4.4, we again estimate a minimal nested logit model, here on the basis of the hypothetical data. The only observed explanatory variable in the model is again the network interaction variable. Unobserved heterogeneity across the transportation mode choice alternatives is again captured by nesting the alternatives that are assumed to be correlated. Since we again have only three elemental choices, there are only three possible nesting structures, namely public transit nested with bicycle, public transit nested with car, and bicycle nested with car. Estimation of the three successive nested logit models shows both the first and the third nesting structure to be indicated, namely public transit nested with bicycle, and bicycle nested with car. The unobserved heterogeneity might represent here for example individual preference for freedom from parking requirements, and freedom from schedules, respectively. The estimation results for both of these models are given Table 10; in both cases the nest scale parameter is highly significant as evidenced by the t-statistics against unity, 10.1 and 9.14 respectively. The nesting of public transit with car is not indicated in this example.
Using the approach outlined in Sect. 3.1 and the previous section of this Appendix, we determine the equilibrium solutions for the long-run behavior of the minimal nested logit model with sociodynamic feedback with global interactions and characterize their stability, for the particular estimated parameter values in Table 10. We find that there are seven equilibrium solutions for the hypothetical model with the transit-bicycle nest and five equilibrium solutions for the hypothetical model with the bicycle-car nest. See Tables 11 and 12, respectively. These estimated parameters show the feedback in the system without any other variables is again “dominant enough” to cause runaway flocking as evidenced by the multiple stable steady-state equilibria. However, here contrary to the case in Table 4, the initial modal split is indeed one of the stable equilibria (solution nr. 3 in both Tables 11 and 12).
Rights and permissions
About this article
Cite this article
Dugundji, E.R., Gulyás, L. Structure and emergence in a nested logit model with social and spatial interactions. Comput Math Organ Theory 19, 151–203 (2013). https://doi.org/10.1007/s10588-013-9157-y
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10588-013-9157-y