households’ choice of residential location: empirical evidence and methodological issues

Lucia ROTARIS*, Edoardo MARCUCCI+ and Romeo DANIELIS*

* Università di Trieste, piazza Europa 1, 34127, Trieste

+ Università di Roma Tre, via G. Chiabrera 199 - 00145 Roma

ABSTRACT

The regulation policies managing the urban traffic activities and aimed at preserving a good accessibility level but also an acceptable environmental protection are continuously discussed and criticized, the question being if it is more important to guarantee the first or the second goal. Knowing the preferences of the population about this issue would help the public administrators deciding the priority order of their actions. The aim of the research reported in this paper is to investigate this issue. In order to do so a sample of households living in the Friuli Venezia Giulia or in the Marche region (Italy) had been interviewed using as data collection methodology the conjoint analysis technique, and as analytical approach the group discrete choice models. The research is innovative both in terms of the attribute used to perform the stated preferences test, and for the methodology used to analyze the data, as we tried to verify (1) if the representative member theory accordingly to which the household preferences can be investigated using the opinions of one member of the family is correct or not; (2) if it is possible to measure the power that each member of the family has leading the households decisions about the residential location. The results demonstrate that for the sample that we have interviewed the most important attribute of the residential location is the absence of air pollution, followed by a good accessibility level and finally by a low noise level. Under the methodological point of view our results show that the interaction taking place among the family members for decisions related to long lasting goods is extremely important and can not be ignored by the annalists in the name of the representative member theory.

1  INTRODUCTION

The increasing mobility of passengers and freight stimulates administrators to develop and extend the transport system, although the negative externalities produced by the transport activities (e.g. congestion, air pollution, and noise) cause a growing deterioration of the environmental context characterizing residential locations. It is quite difficult to predict which factor counts the most for the citizens: accessibility or environmental protection? This study is aimed at answering this question.

In order to do so, a stated preference (SP) exercise was administered to 38 households. The interviewees were asked to choose among several hypothetical residential locations characterized by different accessibility, pollution and noise levels.

Indeed, the choice of residential location is a typical group decision, involving all the family members’ preferences and constrains. Sometimes the bargaining power of the members of a group is equal or evenly distributed among the group members, more often one member leads the group decision process, driving the final group choice as much as possible toward his needs. As a matter of fact some empirical experiments (Molin, et al. 1997; Arora and Allenby, 1999; Aribarg, et al., 2002; Gliebe and Koppelman, 2002; Dosman and Adamowicz, 2003; Puckett and Hensher, 2006; Mok, 2007; Paglione et al., 2007) have demonstrated that under some circumstances the group choice is quite different from the average choice of its members. This implies that the representative agent theory identifying the group behavior with the representative member behavior is not always a reliable representation of the actual group decision making process. For this reason we tried to measure the bargaining power of each family member relatively to each characteristic used to describe the hypothetical residential locations, that is air pollution, noise, accessibility, and rent.

The paper is structured as follows: in section 2 the literature dedicated to the residential location choice and to the group decision theory is reviewed, in section 3 the characteristics of the SP experiment performed within the research program are described, in section 4 the econometric results are depicted, and in section 5 some conclusions about the results obtained and the methodology used are reported.

2  LITERATURE REVIEW

Traditionally, the monetary value of the factors characterizing a residential location have been studied using hedonic price methods (Rosen, 1974). This is based on random utility theory according to which an individual, when choosing a house, is driven by the willingness to maximize his utility, and on the hypothesis that the price of a house depends both on its intrinsic technological characteristics (number of rooms, age of the building, …), and on its extrinsic characteristics, that is the accessibility level (distance from downtown, supermarkets, schools, etc.) and the environmental quality of the context where the house is located (air pollution, noise, etc.). It is possible to quantify the premium that residents are willing to pay for living in a house located in a comparatively better context, or in an area with some particular urban features via the estimation of a regression model having as dependent variable the market price of the sampled houses and as regressors their intrinsic and extrinsic features, (Tyrvainen, 1997; Luttik, 2000; Deodhar, 2004). This technique is based on aggregate data and does not allow to investigate the mechanisms driving the decision making process, especially if the choice involves a household, that is a group of at least two people interacting and exchanging opinions in order to reach a joint decision.

More recently, both to overcome these limits and to estimate the willingness to pay for context features that are not traded in the real market, hedonic price methods have been substituted by the use of stated preference data collected via conjoint analysis experiments and analyzed by discrete choice models[1]. The main features and results of some of these studies (Molin et al., 2001; Hunt, 2001; De Dios Ortuzar and Rodrigue 2002; Wardman and Bristow, 2004; Blijie, 2005) are summarized in table 1. They all make use of conjoint analysis as it allows the estimation of decision-makers’ preferences on the basis of their choices among hypothetical residential locations differing in terms of house and environment characteristics.

Authors; Place / Aim of the research / Sample and data collection methodology / Results
Blijie (2005); the Netherlands / To estimate the importance of accessibility on residential choice behavior. / 11.000 Dutch households; Revealed Preferences (National Housing Survey): the choice set is a random sample of 60 dwelling taken form the full set of available alternatives within an area with a max distance varying accordingly to the socio-economic characteristics of the household / The overall influence of accessibility is very small, whether or not interacted for household specific characteristics, probably because of the relative high quality of the Dutch transport system and the spatial distribution of services. The authors did not find a strong preference for the residential living environment; it seems that the dwelling type is more importance.
Wardman and Bristow (2004); Edinburgh
(UK) / To estimate the monetary value of changes in traffic related noise levels and air quality. / 398 interviews; Stated Preferences: both Conjoint Analysis and Contingent Valuation / Conjoint Analysis Results: WTP per month to avoid a 50% increase of traffic noise = $33,55; WTP per month for a 50% decrease of traffic noise = $22,75; WTP per month for a 50% decrease of air pollution = $39,91
Contingent Valuation Results: WTP per month for a 50% decrease of traffic noise = $10,63 – 18,33; WTP per month for a 50% decrease of air pollution = $11,49 – 19,30
de Dios Ortuzar and Rodrıguez (2002); Santiago (Chile) / To estimate the WTP for reducing the amount of atmospheric pollution in residential location. / 107 selected households; Stated Preferences: the whole families were asked to rank 10 options arising from variations in travel time to work, travel time to school, rent of the house and number of Days of Alert in terms of concentration of PM10 / The WTP is at about 1% of the family income for reducing one Day of Alert in terms of concentration of PM10 per year and it is 26 – 36 Ch$ for one minute of commuting time.
Molin et al (2001); Eindhoven and Veldhoven (the Netherlands) / To verify whether residential preferences are strongly associated with socio-demographic variables or whether they are idiosyncratic. / 147 selected households; Stated Preferences: the group members were required to express their overall preference for each of the 16 profiles varying in terms of dwelling type, number and size of bedrooms, monthly costs and tenure, type of buildings in the neighborhood, frequency of public transport, travel time to work or to study. / Families prefer higher to lower frequencies of public transport, although the importance perceived for this attribute is rather low. On the other hand utility decreases with increasing travel time, which applies to all family members. Families on average attach most importance to the child’s travel time, followed by the mother’s travel time, while they attach least importance to father’s travel time. The results of this study show that family’s residential preferences are highly idiosyncratic, or at least not systematically related to the selected socio-demographics.
Hunt (2001); Edmonton (Canada) / To estimate the relative importance placed by the population on mobility, air quality, traffic noise, neighborhood streets, development densities, and funding sources such as taxes. / 1277 randomly selected households; Stated Preferences: the respondent was asked to participate in four separate stated preference ‘games’, with four different hypothetical home location alternatives and to rank them from the most to the less preferred. / Housing type is the most important element, followed by traffic noise, air quality, municipal taxes, and accessibility.

Table 1 Empirical evidence of households’ preferences for residential location

Although numerous studies have demonstrated that generally family members take joint decisions on important issues such as housing or durable goods (e.g. Davis and Rigeaux, 1974; Munsinger et al., 1975), the conventional approach has been to collect the hypothetical choices of a representative member for each household (typically the family head), assuming that his responses were the expression of the preferences of the family as a whole. However this technique has two limits. The preferences of the representative agent may differ from those of other family members, and he may not be fully aware nor of these differences, nor of the influence that each fellow member has on the family final decision. To overcome these limits the responses of all family members can be collected and aggregated in order to estimate the “family” model, but, (1) as the members do not have the same influence over family decisions, the power that each member has to shape the family decisions should be estimated too, and (2) as members’ preferences may change during the decision making process involving all the family, the estimates of each member preferences could be inconsistent if they are measured before the decision process involving all the group had taken place. A solution adopted by Molin et al. (1997) is to convey family members together and ask them to jointly choose the preferred hypothetical residential location. This procedure allows family members to learn about mutual preferences, and, if relevant, to modify their individual choices, simulating the interaction processes that actually occurs in real-world decision-making processes (Molin et al., 2001).

2.1  Group choice models

There is a growing and diversified literature specializing on the analysis of group decision making dynamics that could be ideally divided in the following methodological approaches: game theory, decision making theory and discrete choice theory. The last approach is characterized by two specific research interests: individual choice models with social interactions, studying the influence that the decisions made by some agents, like friends, peers, neighbors, have on one agent’s choice process, and group choice models, analyzing the interaction among group members when they have to reach cooperative choice (Paglione et al., 2007). The research described in this paper pertains to this second approach[2].

The theoretical foundation of group choice models is represented by the individual Random Utility Maximisation (RUM) model which assumes that each member n of a group g chooses the most preferred alternative i within a choice set Cn faced by all members of the group as if he is a utility maximizer, so that the probability that n chooses i is equal to the probability that the utility that he gets from i , that is , is larger than the utility he gets from all the other available alternatives belonging to the choice set[3]:

(1)

The theory can be extended to groups hypothesizing that the aim of the choice makers is to maximize the group utility function (GUF) that can be specified by different mathematical functions (additive, multiplicative, multi-linear, etc.) (Timmermans, 2006). The multi-linear specification is the following

(2)

It is characterized by an additive component, explicating the utility perceived by each member n (), and by a multiplicative component, representing the interaction effects among the members of the group. is the parameter representing the leading power that individual n has over the group choice and are the parameters representing the interaction effects among the members of the group (meaning that as the value of these parameters increases, the importance of reaching a choice satisfying all the members increases).

If the group is made of two people, A and B, and the GUF has an additive form, the utility of the group can be represented with the following expression:

(3)

where and represent the utility perceived by individual A and B if they choose respectively alternative iA and iB, while and are the parameters measuring the power that each member has over the group choice. Similarly to the “standard” individual RUM, it is assumed that the group will choose the alternative maximizing the utility of the group as a whole, that is the alternative yielding the highest .