• No results found

Behavioural models and spatial planning : some methodological considerations and empirical tests

N/A
N/A
Protected

Academic year: 2021

Share "Behavioural models and spatial planning : some methodological considerations and empirical tests"

Copied!
15
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

Behavioural models and spatial planning : some

methodological considerations and empirical tests

Citation for published version (APA):

Timmermans, H. J. P., & Veldhuisen, K. J. (1981). Behavioural models and spatial planning : some methodological considerations and empirical tests. Environment and Planning A, 13(12), 1485-1498. https://doi.org/10.1068/a131485

DOI:

10.1068/a131485

Document status and date: Published: 01/01/1981 Document Version:

Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers) Please check the document version of this publication:

• A submitted manuscript is the version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website.

• The final author version and the galley proof are versions of the publication after peer review.

• The final published version features the final layout of the paper including the volume, issue and page numbers.

Link to publication

General rights

Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain

• You may freely distribute the URL identifying the publication in the public portal.

If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please follow below link for the End User Agreement:

www.tue.nl/taverne Take down policy

If you believe that this document breaches copyright please contact us at: openaccess@tue.nl

providing details and we will investigate your claim.

(2)

Environment and Planning A,1981, volume 13, pages 1485-1498

Behavioural models and spatial planning: some

methodological considerations and empirical tests

H J P Timmermans, KJ Veldhuisen

Department of Arehiteeture. Building and Planning, University of Teehnology, 5600 MB Eindhoven, The Netherlands

Presented at the 2nd European Colloqu ium on Quantitative and Theoretical Geography, Cambridge, 11 -14 September 1980

Reeeived 16 February 1981

Ab traet. This paper is concemed with the relevance of various geographie models of spatial shopping and residential cholce behavi.our to physical planning. It is argued that models relying on areal aggregation and overt patiat interaction patterns gene rally do not satisfy a number of meth dological r quirements considered extremely relevant inan applied context. In addition, itis

rgued that behaviour I models provide a potentially more valuable approach for predieting consurner response to policy decisions with regard to spatial structure. Empirical evidence substantiating the claim that consumer ev lu tions bear some systematic relationship withobjective attributes of spatial alternatives and overl ehoice behaviour is provided in lhe context of spatial shopping behaviour and residential choice behaviour.

lnl r duction

Since 1965 ub tantial number of article have emerged in geography and regional ci nce which are concerned with developing and testing models to explain aggregate p tial-int raction patt rn. Although clear difference in terms of pecification and calibr tion betw n th models prevail, they may be clas ·fied into two genera1

ppr h . irst, th r ar th gravity spatiaI-int raction and entropy-ma im° ing

mod I. B i Uy th mod Is urn that aggregate patia1 choice ptteros caD he

p dict d from th ttr tiviti of the 11 mativ and distanee decay effec 0 [n

ition, th y ssum th t ob IV d int ra tion pttems m y be u d s in ut data to calibrate the model and make conditional forecasts of the likely effects of changes in attractivities and distance decay to consumer choice behaviour. Second, there are the lessGT developed behavioural modeIs. These share with thè rirst class of models the assumption that consumer choice behaviour is contingent on the attraction of a number of alternative choices and distance separations. However, unlike the first c1ass of modeis, the behavioural models predict choice behaviour by reference to individuaJ's preferences. valuations, attitudes, perceptions, and/or judgments, which are generally measured with questionnaires or in laboratory experiments.

The present paper examines the relevance of bath approaches to physical planning. Specifically. it discusses some methodological issues involved in applying these

approaches to predict consumer reactions to changes in spatial structure in the context of residential choice behaviour and spatial shopping behaviour. The central thesis is that behavioural models offer a potentially more valwble approach to predict the likely effGcts of physical planning schemes as compared with the gravity-type approaches. The second part of this paper reviews some of the authors' recent research findings to

substantiate the claim that consumers' evaluations bear systematic relationships with owrt behaviour and with the physical attributes of spatial alternatives.

The appropriateness of the approaches to physical planning

To assess the appropriateness of the approaches to physical planning it is first necessary to define the essential characteristics of physical planning on the basis of which the approaches may be evaluated. Physical planning is basicalJy con erned with the

(3)

1486 HJP Timmermans, KJ Veldhuisen

10 tion, int nsi and arnount of land development for variou space-demanding tivities. This ta k pre nt not a goal in itself; land-use plans are the result of a more n ral planning proc ss aim d at fulfilling objective of economie and social \ ll-being. Ph sical planning i also ssentially future-ori nted. It is concerned with tb n ds of pa -using functions in th for seeable future. Bath characteristics n 'tat the u of res ar h. That is, re earch is ne ded to evaluate alternative plannin programs in terms of fulfilling their underlying objectives; it is also r quired to ~ r t th futur stat of the system under investigation. If model building is to contribut t the olving of physi 1 planning problems, at least two onditions should th ref r b satisfied; gen ralisability of the results of a model and the in I ion of poli -sensiti independent varia bles. Generalisability in general refers to th fact that behaviour in on area n be predicted from observed patterns in another ar . Gen rall bility thus refer to the condition that results pertaining to on group of respondent in on parti ular study area can be used validly to predict tbe behaviour of anoth r group of r pond nt in a fundamentally different study

a, and 'bly in a totall different time period. The term refers to the condition that tb structure and arameters of th model may he used to predict future

heh viOUT in a different spatiotemporal Hing. Parameter estimates should be independent of initial conditions.

Generalisability therefore involves three interrelated and overlapping issues: fiTst, Jalisation of the results pertaining to one group of respondents to another group of respondents· second, generalisation of the results to a future point in time; and t.hird, n ralisation of tbe r uIts to a totally different region, as will aften be the

in tb conte t of physica1 planning where there are major chan within the dy it lf. The flI'St issu lar Iy ms to be a matter of uitabl sampling d therefore refers to a technicaJ matter. The ond i ue however, ha ron conceptuaJ irnpli tion and ntiaUy refers to tbe tempor I tabiJjty of

b name bein di d. Th third' e bath me con ptual nd t hni I

p' tio and primarily' concern d with th pro of mod I pe ifi tion and parameter estimation. Two ma el ing approaches previou Iy mentioned ar now discussed on these dimensions.

Generalisation of the results of a model to another group of respondents, but within the same study area and at the same point in time, only presupposes a suitable sampling design. Within a particular study area a number of consumer types, which may be defmed on the basis of their decisionmaking characteristics, may be iden tified. Irgeneralisation is needed, it seems therefore necessary that the sample gives a good representation of the consumer types which occur in the population. Thus, a random and sufficiently large sample design is needed to generaLise the results of a model which pertain to one group of respondents to another group of respondents, providing bath these groups have sirnilar decisionmaking profiles. Because the decisionmaking process of individuals may be governed to some degree by spatial factors, the sample should also be random in a spatial sense.

TemporaI stability of the phenomena being studied relates to the fact that model calibration is dependent upon observations at one point in time. Evidently, if models are used for conditional forecasting a necessary condition to yield meaningful results is that the observed relationships between environmental factors and behaviour can be taken to apply in future time periods. That is, stability of behaviour patterns

presupposes stability of antecedent conditions. Models of residential choice behaviour relying upon observed spatial-interaction patterns for calibration such as maximum-entropy formulations of the Lowry model (for example, Senior and Wilson, 1973; Putman, 1977a; L977b; 1978; Putman and Ducca, 1978a; 1978b), Iinear programming models (for example, Härsman and Snickars, 1975; Wheaton, 1974),

(4)

8ehavioural models and spatial planning 1487

syntheses of these approaches (Anas, 1973; 1975; Sl.:nior and Wilson, 1974; Williams and Senior, 1978), semi-Markov models (for example, Ginsberg, 1971; 1972a;

1972b; 1973; 1978; 1979), and deve10pments at UCLA (Clark and Huff, 1977; Huff and Clark, 1978; Clark et al, 1979; Huff, 1979; Smith et al, 1979; Smith and Mertz, 1980) appear not to satisfy this condition. Observabie intraurban migration patterns are not only, nor mainly, the result of household's preferences, housing demand, and housing supply but are also the result of constraints deriving from personal, environmental, and social factors. In addition, the effect of each of these factors on ob rvabie re identia1 choice patterns cannot r adily be determined. Since it is well-known that these antecedent conditions may change dramatically even within the period of a few years, it is hazardous to use observabie residential choice patterns for prediction. The effects of imperfections in the urban housing market on market choice cannot readily be assessed, and Menchik (1972) therefore also opts for using questionnaires to elicit residential preferences. The underlying contention is that household's preferences and evaluations are relatively stabIe and weU-deîll1ed for various life-eycles and income groups un1ess major ultural boundaries are crossed. At 1ea t th rate of change in preferences and evaluations will be much slower as compared with the rate of change in environmental and social constraints. Hence, one should attempt to measure sep rately housing demand and supply. Itissuppased th t these pre erences and evaJuations can be measured properly. Recent research findings tend to support these assumptions (Veldhuisen, 1979b; Ve1dhuisen and

immerman , 1981 ; 1981b).

An implication of this line of reasoning is that only discretionary behaviour may be modeU d adequatelyon the basis of overt spatial interaction patterns (see also Pirie, 1976. Unlike residential choice behaviour. spatial shopping behaviour may be con ider d a form of discretionary behaviour. AJthough it isaccepted that spatial hopping b haviour m y be innuenced by earch and learning, imperfect infonnation and unc rtainty eL nnan nd William , 1979), it· argued neverth Ie that the f eto a rel ti ly unimportant in inf1u ncing consumer hopping behaviour (seealso

imm rm n and Ru hton, 1979). onsequ ntly, mod Isrelying upan reveal d choice behaviour for calibration (for example, Pipkin, 1977; Smith et al, 1977; Pankhurst and ROè, 1978; Recker and Kostyniuk, 1978; Hay and 10hnston, 1979; Hubbard, 1979) reasonab1y satisfy the condition of temporal stability. It may even be argued that these models have the potential advantage that they model revealed behaviour whereas they do not share the difficulties of measurement, qu~stionnairedesign, and administration sa typica1 for models relying on expressed preferences. Revealed preference models do not have to demonstrate that expressed preferenee bears some systematic relationship with overt b('haviour.

As has been noted, generalisability refers to the condition that results pertaining to one group ofrespond~ntsin one particu1ar study area can be used va.lidly to prediet the behaviour of anather group of respondents in a fundamentally different study area, possihly in a totally difkrent time period. The term r('fers to the condition that the structure and parameters of th(' model may be used validly to predict future behaviour in a different spatiotempora1 setting. Thus generalisability involves the question of whether results obtained by modeUing one particular set of spatial aIternatives with specific characteristics may be transferred to different sets of spatial altematives.

Models relying on n~veaJedbehaviour clearly have the disadvantage of restricting themselves to the domain of experience. Revealed behaviour, by definition, concerns the choice of actual spatial altematives. Since this et of altematives is only ane subset from among al1 po sible sets of spatial alternatives (and h nce the paramete of the model are not based on nove1 alternatives), these models involve the problem

(5)

1488 H J P Timmermans, KJ Veldhuisen

of trapolating the resllltS of the model beyond the actual types of spatial alternatives ob erv d. On th ontrary, laboratory experiments and questionnaires permit the

ying of th levels of tbe stimuli in every possible way even to the point of

ifying spatial aItemativ s b yond the domain of experience. Therefore, in theory at 1 t, tb r ults of laboratory experiments and questionnaires are transferable to real-w rld situations which previously did not exist. The problem, however, is that on will h to d monstrate that subjects view the hypothetical choice of alternatives

real-world ph nomena and that these alternatives can provide reliable infonnation. Th t is, if tbe results of laboratory experiments and questionnaires are to be used to predi t real- orld eh i behaviour it has to be demonstrated that these results bear som ystemati relationship with overt choice behaviour. Recently, a large amount of mpirical e .den e upporting this assumption has been accumulated in a variety of spati onte ts (Lieber, 1978' 1979; Louviere and Wilson, 1978; Louvier et al, 19 0).

Generalisability implies tb t the paramet TS of til model be independent of any p tiotemporal stro tur. During the last f w ars s veral authors have argu d or demo trated that the parameters of gravit and entropy-maximising models ar dep ndent upon the spatial structur of tbe study area (Curry, 1972; Johnston,

19 3· 19 5; ]9 6- liff tal ]9 4' 1975; 1976; Ol son, ]975; urry et al, 19 - Sh ppard et al, 19 6; Sheppard, 1979; Ve]dhuisen and immennans, 1979; Fotheringham and ebber, 1980' Griffith and Jones, 1980). Distance decay effects ren t bath the influence of map pattern and the actual wiUingness to travel of individu . For the moment there are no gen rally accepted ways of parating the effe t of t t 0 ource from ob rved interaction patterns e cept perhaps for

Rushton' preference aling model (Rushton 1969; Girt, 1976; Timmennan 19 9. The preference sealing approach may yield accura te preferenee e tima t

ro Odin th study' efuUy designed. Espe iaUy it is important to d sign a pwpo~seful p] hi h enabl tiro tion of tbe preferenee funclion throu hout it

. If tb conditions are re nably tisfied, me evidence exi t that preferenee structures derived from the spatial choice behaviour0 one group 0

respondents in one particular study area can be used successfully to predict the spatial choice hehaviour of a totally different group of respondents in a completely different environmental setting (Timmermans, 1981 b). Ifone uses behavioural models relying on laboratory experiments or questionnaires in which distance is considered as a negative stimulus entering an individual's decisionmaking process rather than as a variabIe increasing or decreasing the Iikelihood of an interaction, paramt:>ter estimates will he independent of spatia1 structure because the hypothetical spatial alternatives may be anywhere. The advantage of behavioural models is that preferences, eva]uations, or judgments are distinguished from opportunities. In other words, behavioural models are relatively indept:>ndent of any spatial structure compared to the mode1s relying on areal aggregation and aggregate spatial-inkraction patterns. Consequently, the level of generalis:.1bility of behavioural models is potentially high whereas, strictly speaking, gravity-typt: models can only bl:l used vali.dly to make conditional forecasts if the spatiaI structure of the study area will remain unchanged. However, environrnental planning programmes are generally directed at changing the spatial structure of the study area and, hence, it is methodologically unjustified to use the parameters of gravity-type models for forecasting.

A further problem militating against the generalisability of the resuIts of gravity-type models concerns the way in which these models are calibrated. As Louviere (1979) basnoted, a necessary condition for generalisability seems to be that the set of observations are balanced; that is, that one has an. approximately equal number ot observations at every ]eveI of the attributes of the spatial alternatives and similarly

(6)

8ehavioural models and spatial planning 1489

even joint distributions. However, since gravity-type medels are calibrated on the basis of real-,:,orld data this will amost certainly not he the case. Consequently, the parameters ot the model will be weighted more according to certain ob ervations. Again, because environmental planning programmes are generally directed at changing ~h~se ~e~l-woriddistributions and these changes will affect the estimated parameters, lt IS dlfflCUlt to see how gravity-type models can yield parameter estimates which can be used validly to predict the likely effects of policy decisions on overt behaviour.

The second condition-the availabiUty of poIlcy-sensitive independent variables-will be self-explanatory. This condition po es a number of problems both for the gravity-type models and the behavioural models. In the case of shopping models, consumer shopping patterns are usually explained by some measure of the attractiveness of the shopping zones (for example, floor-space) and same measure of distance decay. Similarly. many gravity-typ models explain residential choice behaviour by reference to the distribution of mployment and some measure of distance decay. Hence, thc e model use surrogate variables to explain overt spatial choke patterns. The problem is, however, that one does not know the form of the relationship between the e urrogate varia bles and the variables actually influencing spatial choice hehaviour. lf these relationships are nonlinear, as might be expected, how can one assume that manipulating the surrogate variables will have the desired policy effect on consumer b haviour? xpressed in different terms, this means that, if the relationships are nonlinear, changes in the surrogate variables will not resuIt in one-t~nechanges in terms of consumer responses, and hence the predictions of the model may be infenor compar d with models incorporating nonlinear relationships.

part from the problem of using surrogate variables, there is also the problem of the number of variables. When using models for description, one may wish to u as

m II a numb r of independent variabl as po ible to account for the variability of th d pndent variabie. 1fone u a model in the context of physical planning, however, the prime objective wiU be to predict the lilcely effects ofall variabie which are consid r d rel v nt to a particular pI nning t . Usually, this wW involve more than two variabie , which in turn may lead to vere data~atheri.ngproblem and also to estimation problems due to possible near-multicollinearity among the independent variables (Timmermans, 1981a). In theory at least, behavioural models do not share these problems. A fairly large number of variables may be included in questionnaires. In addition, .independent parameter estimates may be obtained by assessing the relationship between consumer evaluations and manipulabk attribute.f; ot spatia1 alternatives. Ilowever, behavioural models share with the gravity-type models the problem whether equations can be developed solely on the basis of policy-sensitive factors, or whether other intluentia l factors should be included. For example, a number of empmcal studies (Hudson, 1976; Potter, 1979; SchuIer, 1979;

Timmermans, 1980) have suggested that price and product quality are very influential variabks in consumer shopping behaviour. These factors, however, cannot be

manipu!ated within environmental planning programmes. Consequently, a behavioural modd which does not inc1ude these variables may be inferior both from a theoretical and from a practical point of view. On the other hand, it is not readily evident how such factors may be included anyway when using the model for prediction, unless

very sirnple assumptions are made. .

In conclusion, mode1s relying on overt aggregate spatial-interaction pattems, Wlth tht' possible exception of the preference-scaling model in the context of shopping

bl~haviour,

suffer from a number of methodological issues which restrict their potential contribution to the solving of physical planning problems. The reason is that these mod~lsprimarily give a description of interaction pattems; they cannot heinterpreted in t(;rms of causa! relationships. Of course, several authors have provided e post

(7)

1490 HJP Timmermans, KJVeldhuisen

rationalisations of the gravity model, linking the structure of the model to a set of umptions regarding the decisionmaking of individu als. Examples include the derivati n f gravity-type mod Is from deterministic utility theory, random utility th ory, and ps cho10gicaI choice th ories (for example, Choukron, 1975; Golob and B kmann, 1971; Golob t aI 1973; Hyman,1977; Niedercorn and Bechdolt, 1 69' ükamp 1975' Smith,1975; 1976; White, 1976). However, these d elopm nt \Vere criticised by Sheppard (1978) and Williams (1977); utility fun tions may ary consid rably among individuaIs, with the assumptions involved

beingrath r unre li tic and often not substantiated by empirical research findings. nrationaI behaviour imperfect information, constraints, and perception are often not in luded in the specifi ation of gravity-type modeis.

The failure of these models is that th y do not capture the mechanisms that give ri to ob ervable aggregate interaction patterns. On this point we agree with the

ning of ay r (1976; 1979a; 1979b). However, unlike Sayer we believe that om oftb me hanisms may b captured by behavioural modeIs, at least for less mple type of spatiaI behaviour and decisionmaking. Gravity-type models suffer from tbe fa t that they are not e plicitly developed for application in the context of ph 'caI planning. Consequently, methodologically relevant issues such a the

enera.Iisability of their results. the stability of th ir result , and mu1tiattribute etensions have received only minor attention. t present, gravity-type models may only he usedvalidly to predict consumer response to policy decisions of the

ant edent conditions of the behaviour and decisionmaking under investigation remainUIl hang d. However, in practice tbis wi1I rarely be th case, peciaIly because poli y decision generally aim at changing the e antecedent conditions. We

been arguing tb t behavioural models offer a potentially valuable approach

in tb< ory at least, avoids mo t of these methodological prob1em. Thi i not to y tha behavioural models will be superior in terms of accounting for the

. biliy of ob rved interaction patterns. or can it he id th t haviOUI I

free from problems. On the contrary, although behavioural m cL I void some problems of the traditional modeis, some new types of problems are introduced. We therefore do not suggest that behaviouraI mode1s are better than the traditional models in every possible way, but we merely argue that behavioural mode1s represent an aIternative approach wbich deserves further attention. In theory at least, it satisfies same methodological conditions wbich are considered extremely important when using models in the applied context of physica1 planning. To illustrate ome of the advantages and difficulties of behavioural models, the second part of this paper discusses some theoretical developments and empiricaI research findings in the application of behavioural models to the study of shopping and residential choice behaviour.

Theoretica1 considerations

Consider an environm~nt or physical space wmch may be described by a matrix [Xij],

denoting the level or value of location i on attribute j. Each destination is conceived of as a bundie of objective attributes. Some of these attributes are of a quantitative nature, others are tru1y qualitative. These attributes may include aspects of site, neighbourhood, and relative location. Spatial choice behaviour is concerned with the way in which individuals choose a particular destination from among the set of all possible destinations. How may tms decisionmaking process be modelled?

FoUowing a long-standing tradition in behavioural geography (for example, Louviere et al, 1980), it is assumed that spatial choice behaviour is contingent upon the

attribute levels of the destinations. We assume that individuaIs have established an evoked set of alternatives. Thisset is a subset from the set of aU possible destinations.

(8)

Behavioural models and spatial planning 1491

The evoked set is bounded by locational, income, family size, and other environmental and personal fa.ctors. Given this evoked set of alternatives, it isassumed that

inJividuals evaluate these alternatives on the basis of the attributes they consider relevant to their decisionmaking task. This evaluation process isguided by the individual's cognitive representation of the physical space, (x~), his value system, his motivations, and possibJy by other personal characteristics. Individuals are assumed to combine their evaluations of the values or levels of the relevant attributes according to some combination rule into an overall evaluation. The result of this process is a subjective preferenee scale(E;,i

=

1, ... ,n) which gives an ordering of then alternatives on the basis of their utility. Ttis assumed that an individual will select that alternative which received the highest overall evaluation or utility.

The theoretical considerations underlying the applications may be formalised as fol1ows in terms of a number of functional relationships:

Xij =f(Xlj ) , (I)

EI

=

g(Xlj) , (2)

Bi

=

h(EI) , (3)

BI = II{g(f(Xij )]} , (4)

where BI defines the probability that alternative i will be chosen. This recursive equation system shows that consumer behaviour is considered to he a response to a set of objective attribu tes of spatial alternatives through a decisionmaking process by which consumers evaluate these alternatives. VaLidation of these equations requires the identification of the relevant attributes of the altematives, the establishing of the relation hips betwen the levels of these attributes and the corresponding evaluations of con umers, the mferring of the nature of the combination mie involved in the decisionmaking proce ,and, rll1ally, the obtaining of the relatianship between overall ev luatian and choice behaviour. Elements of this approach have been appLied

ucce fully by Louviere and his coUeagues in a variety of contexts (far example. Louviere,

r

976; Louviere and Wilson, 1978. We will now illu tr t som ope tional aspects of thc approach by two examples.

Fxample j : spatial shopping behaviour

The conceptual model previously outlined was employed in a study of spatiaJ shopping behaviour in part of the region of Southeast Brabant in. the South Netherlands. 771 households participated in the survey, and respondents were randomly sclect~d on thc basis of their willingness to participate in the survey. They were asked to mention th~ shopping area they usual1y patronise to buy clothing, and the total number of visits to each shopping area was taken as a measure of choiet: behaviour.

To calibrate the model, it is first necessary to identify the relevant attributes of thc shopping areas. For tht' present analysis three attributes were identified: seJection, distance, and availability of parking space. All three attributes are policy s~nsjtive. Their choice was dictated by alternative retailing pbns, which differed on these dimensions. Respondents were asked to provide a numerical evaluation of each shopping area for each attribute on a 1-9 rating scale, the ends of which were defined as 'worst possible place to shop for clothes' (1) and 'best possible place to shop for clothes' (9). Respondents were asked only to evaluate the shopping areas they felt familiar with. Consequently, sample sizes for shopping areas were unequaI. Since the analysis was conducted at the aggregate level individual responses were averaged over respondents to yield mean ratings, and these mean ratings were assum~d to oorresp?nd to the x- of equation (1) for the aggregate sample. Respondents proVlded a numencal

'I . .

(9)

1492 H J P Timmermans, KJ Veldhuisen

The OIT ponding physical attributes of the shopping alternatives were measured

a follows: s Ie tion was measured by totalling the number of shopsineach shopping a' di tan as measured in terms of travel time to the shopping area, and

availabilit of parking facilities was measured as the average percentage of occupied parking pIac on an av rage Saturday aftemoon. In order to calibrate equation (I) and a e th strength of this relationship Spearman rank order correlations between th ph sicaJ measures f the hopping aItematives and the average subjective

valuation of these measures were comput d. Values obtained were +0·96 +0·99 and

+

1·00 re pe ti el indi ating almost perfect monotonic relations. Ite:ative ' I t-squ procedures vere u ed to obtain the functionaI relationships between the pairs of variabJ s. The b st fitting functions \Vere:

Xi ( Ie tion)

=

+

0·94 i ·27 i (5)

Xj (parking)

=

7·6_-0·09

70Pi (6) Xl(di tan e)

=

7·94 - 0 . 58 i '3

T?

(7)

'h re ~i i th number of hops,Pi th percentage of occupied parking places, and

Tl

the travel tiro for 10;ation i. The e re uJts show that the relationships are nonlinear. Hence, hanges in physical attributes of spatial tructure will not result in one-to-<>ne hanges in evaJuations and consumer behaviour, providing these empirical relationship are aIso theoretically true.

In order to validate equation ( ) a laboratory experiment was conducted. A

f toriaJ de'go s used to infer the algebraic composition rule subjects use to arrive

t an 0 erall evaIuation. Eighteen subjects participated in the experiment.

Twenty-n hypotheticaJ shoppiTwenty-ng aItemative , each describiTwenty-ng a three-attribute profile were d stimulus combinations. Each attribute was varied over three I vels: number of shops ( ) varied over 10,40, and 70; travel time(r) varied over 15, 30, and 45 minut . and time to rmd a parking place (TP) varied over 3, 6, and 9 minutes. S

bi

e u ted e ch of the twenty-seven hypotheticaI stimulu combination . The subjects also evaluated two additional stimulus l:ombinations \ bi h re more

extreme than the design combinations; these they were asked to consider 'best' and 'worst', on a 1-100 mm scale. The resulting data were subjected to an analysis of varianee. All main effects and the NS x Tl effect were statistically significant whereas

theNS xTP and Tt xTI' effects were not statistically different from zero beyond the

o

·05 level. Hence, one cannot conclude [irmly the presence either of a multiplicative or of an additive composition mIe.

Equation (3) suggests that consumer choice behaviour is related to consurner's overall evaluations. Since- the laboratory experiment did not yie-ld condusive findings, both the additive and the multiplicative composition rules were tested, assuming that consumer choice beha"iour is linearly related to overall evaIuation. The multiplicative mIe yielded the best results. Thc relationship between spatial choice behavinur and overall evaluation based upon a multiplicative rule is reason.a bly linear [r = O' 85;

F(I, 14)= 36·45; p

<

O·OI]. The a<lditive compositive rule' performs less welI in terrns of its F-value [r

=

0,82; F(I, 14)

=

28 ·57; p

<

0 ·Ol]. These results tentatively suggest that a multiplicative composition rule is superior in deseribing eonsumer shopping choice behaviour. In terms of retail planning this result would suggest that a new retailing development will probably only eome up to expectations if every important attribute of the new development is above some minimum .level.!

Example 2: residentwl choice behaviour

Ithasbeen argued that the environmental, individual, and sociaI eonstraints may change dramatieally through time, and, henee, that predietion from observed interaetion

(10)

Behavioural models and spatial planning 1493

patterns may yield inferior results. Instead preferences and utilities should be measured directly at the personal level by use of questionnaires; the results should be used in models to calculatc the dcmand so that consequently the residential choice behaviour patterns can be predicted from the confrontation of demand and supply. lhis applies espccially to residentiallocation as dwellings can normally only be occupied by one houschold at the time.

In the preceding example it has been shown that utilities of shopping centre can be measured and used successfully for the prediction of shopping behaviour. In the second example we will show that the same type of measurement can be used for establishing housing utilities; ho wever , assessment of the relationship between housing utilities and consequent observed behaviour is not possible as the constraints are very i:mposing. What may be done is the fonnulation of a model in which the utilities are, amongst other things, used to predict location and relocation pattems of households and afterward to make a comparison with reality. In this case a possible model is being discussed.

A housing situation i essentially a multiattribute situation (for example, Knight and Menchick, 1976' Louviere, 1976; and others), which contains at least six or seven important attributes. Because of this number of attributes, overall utility cannot be measured directly (for example Slovic and Lichtenstein, 1971; Fischer,

1975; Veldhuisen and Kapoen, 1979). The measurement of the utilities of the separate attributes can be performed either directly (through sealing) or through

factorial designs. lthas been shown (Green et al, 1972; Veldhui.sen and Timmermans, 1981 a; 1981 b) that results of factorial designs can be approximated by the

comparatively easy direct measurement methods. Consequently the utility of a number of attributes were measured directly in a survey of 560 households in the area of Southeast Brabant in the South etherlands. Valuations were gathered (amongst others) for the following list of physical and spatial attributes:

1 the noor p ce of the dwelling; 5 the distance to (desigJuted) playing facilities; 2 the area of private land around the dwelllng; 6the situation of the dwe1lingwith regard to traffic; 3 the area0 pubüc space (green ; 7 the distance to shops for con enience goods;

4 the distance lo playgrounds; 8 the distanees to the public transportation system.

The relationships between spatial-physical propertjes and their valuations [equation (1)] have been worked out for households differing in their stage of the family cycle (the number of household categories being 7). In all of the cases, the form of the relationships was nonlinear and could he described by a function of the form:

m [(x)

=

1 +axb ,

excellent

extJemely paor 0 100 150 200

o

5 10 15 20 5

time taken towa11c to park (mins) floorspace (m')

Figure 1. The dweUers' evaluations of 'Ooorspace' and 'distance to park' as a fWlctïon of these attributes.

(11)

H J P Timmermans, KJ Veldhuisen h i f th valuation. si al at ribut • and minimis Z =

c x

bj t to =

c

= -h Uin

J

in inu t n nd r th vil&C:Ulcies (suppty m I By maIllpu. h ulm 0 t n d

introduced into the allocation model: (1) the home-work dlslance " l br'

winner(s). (2) the social integration of the household, (3) the renter or ownership of the dwelling. (4) the type of dwelling (single-family unit versus multiple-family unit), and (5) the price. Preliminary results of using this model are promising.

Conclusions

ft has been arguèd that geographic models relying upon area! aggregation anJ observeJ interaction pattems suffer from a number of methoJologicai shortcomings which make them less suited for predicting the effl:cts of policy decisions on consumer spatial behaviour. Tt has also been argued that, in theory at least, behavioural models avoid most of the problems related to such issues as generalisability and policy-sensitive independent variables. Ho wever, to apply behavioural models it has Lu be demonstrated that con urner evaluations bear same systematic relationships both with rnanipulable attributes of spatial alternatives and with overt choice patterns. This paper has provided some additional evidence that these relationships prevail. In view of the quality of physical planning it is hoped that these fmdings will stimulate the further development and application of behavioural model in geography and regional science or at least point to the direction in which fruitful research can be pur u d.

Rderencea

An.aJ A. 1973 "A dynamic disequilibrium model of residenliallo lion"EnvironmentandPlanning. 5633-647

An.aJ A. 1975 ''Theempirical calibralion and te ting of a simulation model of residentiall cation" EmtironmentandPlanning A 7899-920

(12)

Behavioural models and spatial planning

1495

Brown L A, Horton F E, Wittick R I ] 970 "0 ' ,

urban migrants"Demography 7 175 -183 n place utllity and the norrnative allocation of intra-Chuukron J H, 1975 "Ageneral framework for th '

mode Is" Regional Scïence and Urban

Economic~ ~e~~I~:ment

of gravlty-type trip distribution Clark WA V, Huffl 0, 1977 "Some em ''caJ 2~2

migration " Environment and Planning~19 I;~~~ 10:7~uratlOn-of-stayeffects in intraurbao CI"rk W A V, HuffJ 0 Burt l E 1979

"c

I"b t'

and Planning A ]1689-705 ' a1 ra mg a model of the decision to move"Environment GiffA D, Martin R OrdJ K 1974 "E I ' , ,

modeis"RegionaÎ Studies

8

281 _ 28;a uatmg the fnctlOn of distance parameter in gravity CliffA D, Martin R, Ord JK 1975 "Ma t ' ,

comments by Rl Johnsto~ and b

r!'

Ja tem andf~ctlonof distance parameters: reply to 9285-288 ' Y urry, D A Griffith, and S Sheppard"Regional Studies

liff D,

Mart~~

R.

O~d

J K, I

~76

"A reply to the fmal Comment"Regiona/ Studies 10341-342 Curry L. 19:72 Aspahal analysls of gravity flows"Regional Studies6 131-147

CU~19:i~~lffith

D A, Sheppard E S, 1975 "Those gravity parameters again"Regional Studies9

Fischer G,~, 1975 "ExperimentaJ applications of multi-attribute utility modeIs" in Utility Fo~~bllltyand Human Decision Making Eds DWendt, C Vlek (Reidel, Dordrecht) pp

1-46

rm~ S, W,ebberM J. 1980 "Spatial structure aod the parameters of spatial interaction models Geographlcal Ana/ysis ]2 33 -47

insberg R B, 1971 "Serni·Markov processes and mobility"Journa/ of Mathematica! Sociology ]

233-262

Ginsberg R B. ] 9713 "Critique of probabilistic modeIs: application of the semi.Markov model to migration"Jounla/ of MathematiCtll Sociology2 63-82

Gin berg RB" I972b ,"Incorporating causaJ strueture and exogenous information withprobabilistic modeIs: Wlth SpeCial reference to choice. gravity, migration and Markov chains"Journal of

arhemariCtll Sociology 2 83-103

Gin berg R B, 1973 "Stoch tic m deis of residential and geographic mobility for beterogeneow popuJations" Envir mentandPlanning S 113 -124

Ginsberg R B, I 78 "The relationship between timing of moves and choice of destination in

t ha tic m d Is of migr tion" Env' nmentandPlanningA 10667-681

Gin ber R B. 1979 "Te ts of toch tic models of timing in mobility histories: comparison of int: rrn ti n derived from different bservation plans"Environ~ntandPlanning A ]I

1387-1

in J L, 1 7 "om xlen ion of Rushton's sp tia! preference sealing model" GeographiCtJ/ Analysis8 137-157

Golob T F, Beckrnann M T, 1971 "Autility model for travel forecasting" Transportation Science

579-90

Golob T F, Gustafson R L, Beckmarm M T, 1973 "An economie theory approach to spatial interaction " Papers and Proceedings of the Regional Science Association 30 159 -182

Green PE, Carmone F l, Wind Y, 1972 "Subjecrive evaluation models and conjoint measurement"

Behavioral Science 17288-299

Griffith D A,lones KG, 1980 "Explorations inta the relationship between spatial structure and spatial interaction " Environment and Planning A 12 187 -203

lllrsman B, Snickars F, 1975 "Disaggregated housing demand models: some theoretical approaches"

Papers and Proceedings of the Regiullal Sciellce Association34 121 -145

Hay A M, lollOston Rl, 1979 "Search and the choice of shopping centre: two models of variabitity in destination selection"Environment alld PlanningA 11 791-804

Hubhard R, 1979 "Parameter stability in cross-sectional moJels of ethnic shopping behaviour"

Environment and Planning A 11 977 -993

HuJsun R, 1976 "ünking studies of the individual willt m(\dels of aggregate behaviour: an empirical exampJe" Transaetions of the Institute ofBrit~sh Geo~GJ?hersnew series I 1,59~174 HuffJ 0, 1979 "Residential mobility patterns and populatlOn redistnbutlOn Wlthin the CIty

Geographical Analysis 11 133 -148 , , ,

Huffl 0, Clark W A V, 1978 "CumuJative stress and cumulahve mertla: a behavioral model of the decision to move"Environment and Planning AIO I 101 -11 ] 9

Hyman G M, 1977A Local Value Theory of Spatial OzoiceOccasiooal Paper 1 (Centre for Environmental Studies, London) " "

(13)

6 H J P Timmermans. KJ Veldhuisen

effici nts in comparative studies of the frictions of distance"

ud nary note"nïdeMIr e mpiri aUy valid Lin

vity I w' of p t IInter·

Y.

12

J • fi nt

urban travel"GeograFlSka Annaier B 59 82-94

PirieG H, 1976 "Thoughts on revealed preference and spatial behaviour" Environment and Planning A 8947-955

Potter R B, 1979 "Factors influencing consumer decision making"Psych%gical Reports 44 674 PutmanS H, 1977a "Calibrating a disaggregated residential allocation model-DRAM" in Londoll

P(f{Jers in Regional Science 7: Altemative Frameworks for Analysis Eds D B Massey, P WJ Batey (pion, London) pp 108-124

PutmanS H, 1977b "Calibrating a residential location model for nineteenth century Philadelphia" Environment and Planning A 9 449 -460

PutmanS H, 1978Studies in Applied Regional Science 13. Urban Residenti.al Location Mode/s (Martinus ijhoff, Leiden)

Putman S H, Ducca F W, 1978a "Calibrating urban residential models 1: Procedures and strategies" Environment and Planning A Ir633-650

Putman S H, Ducca F W, 1978b "Calibrating urban residential models 2: Empirical results" Environment and Planning A 10 1001-1014

Recker W, Kostyniuk L, 1978 "Factors influenctng destination choice for the urban grocery shopping trip" Transportation 7 19 -33

Rwhton G, 1969 '''The sealing of locational preferences" inBehavioral Probiems in Geography, A

Symposium Eds K R Cox,R G Golledge (Northwestem University Press, Evanston, (IJ.) pp 197-227

Sayer R A, 1976 uA critique of urban modelling: from regional science to urban and regional political economy"Progress in Planning 6 187-254

Sayer R A, 1'179a "Understanding urban models versus understanding cities"Environment and Planning A 11 853-862

Sayer R A, 1979b "Philosophical bases of the critique of urban modelling: a reply to Wilson" ErwvonmentantiPlanning A 11 1055-1067

(14)

Behavioural models and spatial planninq 1497

Schuier H, 1979 "A disaggregate store-choice model of spatial decision-maJdng"Professional Geographer 31 146-156

Senior ML, Wilson AG, 1973 "Disaggregated residentia1 location modeis: same tests and furtner theoretical developments" inLondon Papers in Regional Science 4: Space- Time Concepts in Urban and Regional Models Ed. EL Cripps (Pion, London) pp 14l-172

Senior ML,Wilson A G, 1974 "Explorations and syntheses of Iinear programming and spatial interaction models of re idential location" Geogmphical Analysis 6 209-238

SheppardES, 1978 "Theoretical underpinnings of tht> gravity hypothesis"Geographical Analysis 10386-402

Sheppard E S, 1979 "Gravity parameter estirnation" Geographical A nalysis 11 120 -133 Sheppard ES, Griffith D A, urry L, 1976 "A linal comment on mis-specification and

auto-correlati n in those gravity parameters"Regional tudies 10337 -339

Slovic P, uchtenstein S, 1971 "A comparison of Bayesian and regression approaches to the study of information processing in judgment"Organizational Beht1llior and Human Performance 6 649-744

mith AP, WhitebreadPJ, Macketl R L,1977 "The utilization of services" in Models of Ciries and Regions Ed AG Wilson,P H Rees, M Leigh (Joho Wiley, Chichester, Sussex) pp 323-403

mith T E, 1975 "A ehoice theory of spatial interaction"Regional Science and UrOOn Economics S 137-176

Smith T E, 1976 "Spatial discounting and lhe gravity hypothesis" Regional Science and Urban Economics 6 331 -356

Smith T R, Qark W V, Huff JO, Shapiro P, 1979 "A decision-making and seareh model for intraurban migration" Geographical Analysis 11 1-23

Smith T R, Mertz ,1980 "An analysis of the effeet of information revision on the outcome of housing-market arch, with special reference to the influence of realty agents"Environment anti

PJonningÁ 12 155 -174

Timmermans HJP, I 79" spatial preference model of shopping behaviour" Tijdschrift voor

Economjsch~ ~n SociJzl~Geografie 70 45-48

Timmermans HJ P, 1980 "Con umer spatial choice strategies: a comparative study of $Ome altem tive behaviour I hopping models" Geoforum 11 123 -131

Timmerm ns H J P, 19 la "Multiattribute shopping models nd ridge regre ion analyse " Environment and PIonningÁ 13 43 -56

Timmerman I1 J P, 19 Ib " p lial choice behaviour in different environmental settings· In

appli tJ n f the revealed preferenee ppro eh"Geografiska Anna/er 63 57-67

Timmermans H J P, Rushton G, J979 "Revealed space preference theory: a rejoinder" Tijdschrifr 1J()()r Economische en Sociale Geografie 70309-312

Veldhuisen KJ, 1979a "A utility model for the aUoeation of households" paper presented at the Regional Science Conference, 1979, London, available from the author

Veldhuisen KJ, 1979b "De meting en verklaring van woonwaardering"· (The measurement and explanation of residential evaluations)Mens en Maatschappij 54 144-169

VelclilUisen KJ, Kapoen L L, 1979UAregionallocation model" inNew DeIJelopments ill Modelling TralJe{ Demand and Urban Systems: Some Results of Recent Dutch Research Eds G R M Jansen, P H L Bovy, J P J M van Est, F Ie Clerq (Teakfield, Farnborough, Hants)

Veldhuisen K J, Timmermans HJ P, 1979 Spatial Structure. Spatial BehalJiour and Spatial Interaction Modeis: Some Critical Commellts on tlle Use

vt

GralJity and Entropy-Maximizing Models for Forecasting Purposes WP-2, Department of Architecture, Building and Planning,

University of Technology, Eindhoven, The Netherlands

Veldhuisen KJ, Timmermans H J P, 1981 a "Voorkeuren en waarderingen: Een drietal meet-systemen" (preferellces and evaJuations: three measurement approaches)Mens en Maatschappij S6 154-172

Veldhuisen KJ, Timmermans H J P, 1981b "Voorkeuren en waarderingen: Toetsing van een drietal meetsystemen " (Preferences and evaluations: tests of three measurement approaches Mens en Maatschappij S6 275-293

Wheaton W, 1974 "linear programming and locational equilibrium; the Herbert-Stevens model revisited"Joumal of UrOOn Ecollomics I. 278 - 287

White RW, 1976 "A general.isation of the utility theory approach to the problem of spatial interaction"Geographical Analysis 8 39-46

WiJliams He W L, 1977"On the formation of travel de mand models and economie evaluation measures of user benefit"Environment and PJonning A 9285 -344

(15)

1498 H J P Timmermans, KJ Veldhuisen

illiams H L. nior ML,1978 "Accessibility, spatial interaction and the spatial benefit an 1 ·s of land use transportation plans"inSpatiallnteraetion 17leory and Planning Models Eds Karl vi t. L Lundqvist, F Snickars JW Weibull (North-HoUand, Amsterdam) pp 253-287 lpertJ I 5 "Behavioral ape t of the decision to migrate"Papers and Proeeedil1gs of the R . n 1 Scinee ssociotion 15 159-169

Referenties

GERELATEERDE DOCUMENTEN

In Proef C het •n ondersoek na die invloed van sommige vitamines, aminosure en ander stowwe op kallusvormingsverm~ van apikale snyvlakke van winterlote van Salt

Daarbij wordt onderscheid gemaakt in de van SDRC afkomstige software, het Marc-pakket, eenvoudige programma's voor gebruik in het standaard onderwijs en

Veelzeggend is bovendien dat de beide hoofdpersonen de actieve politiek niet eens bereiken: de een gooit tijdens een talkshow zijn eigen glazen in nog vóór men hem in de Kamer

Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of

Furthermore, measuring methods are used in non-evaluative ways; for example, in exploring learning strategies that are executed during studying expository texts (e.g., Schellings et

This section will highlight some of the key challenges that face the informal public transport industry in Sub-Saharan Africa..

Bij de tweede rooidatum nam het aantal gezonde bollen na een wwb bij 43,5°C af naarmate meer voorwarmte was gegeven waardoor er meer zieke bollen werden geoogst en minder

The central goals of sustainable community planning can be achieved by applying planning principles to development projects that promote and reflect the different