Vous êtes sur la page 1sur 8

CONSUMER BEHAVIOR IN DIFFERENT AREAS

BARRY LENTNEK, STANLEY R. LIEBER, AND IRA SHESKIN ABSTRACT. Some facets of the evolution of consumer food shopping behavior are examined by a comparison of revealed preference studies in regions of differing economic development: Aguascalientes, Mexico ( 1968) ; Iowa ( 1934) ; Iowa (1960); and Michigan (1966-1968). The majority of rural Mexicans obey the Christallerian nearest neighbor axiom, but a few patronize the capital and regional centers regardless of distance. This behavioral variance is a function of household income. Comparison of the four study areas reveals that food shopping behavior may be universally subject to a dual assignment rule: households within a limited range of an opportunity exhibit a high probability of patronizing the closest place, whereas households at some distance from the nearest opportunity prefer shopping in larger places at greater distances. The absolute range within which the Christallerian axiom is applicable increases considerably (from two to thirteen miles) with level of economic development. KEY WORDS: Aguascalientes, Central place theory, Consumer behavior, Iowa, Mexico, Michigan, Multidimensional scaling, Multipurpose trips, Revealed preference analysis. PATIAL shopping behavior is a function both of the characteristics of the decision maker and of the decision-making environment. These characteristics are, in turn, functions of the level of economic development. The relationship between level of development and spatial shopping behavior is particularly pronounced for food shopping behavior for two reasons. First, the primary determinants of spatial demand (income and mobility) and the availability of alternative shopping opportunities tend to increase with level of development. Second, the proportion of food budgets allocated to nonindigenous goods, which may be

Accepted for publication 20 May 1975. Dr. Lentnek is Associate Professor of Geography ut the State University o f New York in Buffalo, N Y 14226; Dr. Lieber is Lecturer in Geography at the University o f Haifa in Israel; and Mr. Sheskin is a graduate student in the Department of Geography at Ohio State University in Columbus, OH 43210.
* W e gratefully acknowledge the support of the National Science Foundation (Grant GS-2018) for data collection and processing. Professor Gerard Rushton (University of Iowa) graciously made the data for the Iowa 1960 and Michigan 1966-1968 preference scale values available to us. Richard Buxbaum did a considerable amount of work both in Mexico and at the Ohio State University above and beyond the terms of his research assistantship during various stages of the project in data collection and processing.
@

available only at higher order centers, tends to increase with level of development. The purpose of this article is to explore some facets of the evolution of consumer spatial shopping behavior by a comparative analysis of research in areas differing significantly in level of development. The preferred investigatory procedure would be to obtain time series data for one region, but such data are not available. The best alternative is to contrast comparable studies in areas of varying development. Unfortunately, most shopping studies have employed differing and often noncomparable methodologies. Rushton used revealed preference analysis to study shopping behavior in Iowa in 1934, in Iowa in 1960, and in Michigan in 1966-1968.l We adopted this methodology for an examination of food shopping behavior in Aguascalientes, Mexico, in 1968. Given our thesis concerning the systematic covariation of consumer spatial behavior with !eve1 of development, we plan to: 1 ) analyze the Aguascalientes data for determinants of revealed spatial behavior;
1 Gerard Rushton, Analysis of Spatial Behavior by Revealed Space Preferences, Annals, Association of American Geographers, Vol. 59 (1969), pp. 391-400; and Gerard Rushton, Temporal Changes in Space Preference Structures, Proceedings, Association of American Geographers, Vol. 1 (1969), pp. 129-32.

ANNALS OF THE ASSOCIATION OF AMERICAN GEOGRAPHERS 1975 by the Association ot American Geographers. Printed in U.S.A.

Vol. 65, No. 4, December 1975

538

1975

CONSUMER BEHAVIOR

539

2 ) compare the spatial behavior patterns revealed by the four studies for evidence that spatial behavior varies systematically with level of development; and 3 ) discuss the relationship of the evolution in spatial behavior to the development of central places.
THE ACUASCALIENTES STUDY

Large cities dominate Latin Americas urban structure.2 Despite the proliferation of growth pole theories purporting to explain this primate city pattern, however, little is known empirically of the functional bases of primacy in terms of consumer behavior. We hope an empirical analysis of food shopping behavior in the State of Aguascalientes, Mexico, will shed some light on the behavioral processes leading to primacy at all levels of Latin Americas urban structure.

Data and Study Area


Aguascalientes is on the northern edge of the densely populated Mexican highlands (Fig. 1 ) . Its urban structure is typical of highland Mexico; i.e., the capital (Aguascalientes) contains almost half (160,000) of the states population, and dominates its trade area. The study area includes all of the rural State of Aguascalientes except the eastern and western peripheries, plus a small section of the Municipio of Luis Maya, Zacatecas, just to the north. Seven large towns (populations range from 550 to 2,600) and 122 villages (populations range from 30 to 500 inhabitants and consist almost exclusively of farmers and farm laborers) are within the study area. A field survey in 1968 used a spatially stratified random sample in the seven large towns and 122 villages. A special enumeration conducted by the Federal Board of Elections in May of 1968 provided the best data available, although base maps and enumeration lists contained considerable margins of error.3 The sampling ratio of heads of households was one-half
2 J. Miller and R. Gackenheimer, Latin Americnn Urban Policies arid the Social Sciences (Beverly Hills, California: Sage Publications, 1971), Chapter 1. 3 The basic document used was an estimate of population of individual villages and towns published by the State Electoral Commission, Commission Estatal Electoral del Padyon y Vigilancia Electoral (Aguascalientes: Delegacion Estatal del Registro Nacional de Electores, Division Sectional, 1968).

FIG. 1.

Principal market towns of Aguascalientes.

of one p e r ~ e n t .As household size averaged ~ six, the ratio of persons represented averaged three percent. We believe that those interviewed represent a broad and fairly accurate cross section of the population. The survey was conducted by two field teams, each consisting of an American graduate student and five Mexican campesinos who had an intimate knowledge of the substantive content of the questionnaires. The questionnaires contained a detailed review of the material aspects of rural life (including items such as time budgets and travel records). With few exceptions, quantity and price data were fully disaggregated. The schedules were checked for internal consistency and completeness, resulting in a loss of approximately 151 of 1,017 interviews, leaving 866 consistent and complete questionnaires.5 The variables employed in this analysis are:
4 The actual sampling ratio varied between 0.35 and 0.70 percent, with the bulk of the districts sampled falling within the 0.40 to 0.60 percent range. 5 A number of internal check questions were deliberately scattered through the questionnaire.

5 40

BARRY LENTNEK ET

AL.

December

1) the location of each interviewees residence; 2 ) the location of predominant food purchases; 3) the location of the place of work for interviewees receiving wages for off-farm labor; 4) the average level of household income for 1967-1968; 5 ) the average value of wage income for 1967-1 968 ; 6 ) the average value of food purchases for 1967-1 968; 7 ) the average value of subsistence food consumption for 1967-1968 (evaluated at Fall 1968 prices); 8 ) the average value of clothing purchased for 1967-1968; 9 ) the average value of bus fare expenditure for 1967-1968; and 10) the population of all settlements. Locational data are geocoded on a one-square kilometer grid. Revealed Preference Analysis As Rushton has pointed out:
A fundamental description of spatial behavior is one that describes the manner in which a population chooses between (that is, orders) a hypothetical set of spatial opportunities; this order is place independent, that is, it is unlikely to vary significantly from one area to another-unless major cultural boundaries are crossed.

26004

DISTANCE TO TOWN (KM)

MEXICO 1968
FIG.2. pers.
Indifference surface for Mexican food shop-

One method of protraying these spatial behavioral patterns is by indifference surfaces relating to attributes constituting spatial purchasing opportunities. The axes of such indifference surfaces are commonly town size ( a surrogate variable for attractiveness) and distances of consumers to towns.T Each town vis-8-vis the location of each consumer may be measured with reference to these two stimuli. The consumer decision-making process was considered to be a paired comparison of the nearby towns: i.e., a comparison of combinations of distances and town sizes. By patronizing a specific town, a consumer reveals his preference for a particular type of spatial opportunity relative to all other available spatial opportunities. Combin6 Gerard Rushton, Preference and Choice in Different Environments, Proceedings, Association of American Geographers, Vol. 3 (1Y71), p. 146. 7 Rushton, op. cit., footnote 1.

ing the decisions of a sample of respondents, a matrix was constructed depicting the relative similarity of all spatial opportunities (combinations of attributes). Using nonmetric multidimensional scaling techniques, an interval scale preference value for each combination of attributes was derived utilizing the rank order similarities of spatial opportunities.8 The scale values were then used to construct an indifference surface of revealed spatial behavior. The isolines of the indifference surfaces portray all combinations of attributes constituting spatial purchasing opportunities which were similar in attraction (Fig. 2).9 The largest town included in this analysis had a population of 2,600, be8 Reginald Golledge and Gerard Rushton, MultiDimensional Scaling: Review and Geographic Applications, Commissions on College Geography Technical Paper No. 10 (Washington, D.C.: Association of American Geographers, 1972). f) The statistical basis and interpretation of revealed preference surfaces have been discussed in R. Kern and G . Rushton, REVPREF: Paired Comparisons Analysis from Revealed Spatial Preference Data, Computer Institute for Social Science Research Technical Report No. 95 (East Lansing, Michigan: Michigan State University, May 1969); J. Kruskal, NonMetric Multi-dimensional Scaling: A Numerical Method, Psychometrika, Vol. 59 (1Y64), pp. 115-29; and R. N. Shepard et a/., eds., Multi-dimensional Scaling Theory and Applicaiions in the Behavioral Sciences (New York: Seminal Press, 1972), Vols. 1 and 2.

1975

CONSUMER BEHAVIOR

541

100-

90-

80-

4
7060-

g 3
u.

50-

D-w MICHIGAN A-A-4

IOWA
MEXICO

10

I5

20

25

MILES TO TOWN

FIG.3. Desire line map of Mexican food shoppers patronizing the capital city. The circles are three kilometer radii around the principal market towns.

FIG.4. Distance decay functions for food shoppers in three study areas.

cause the disproportionate size of the capital city precluded its use in the scaling algorithm. The indifference surface reveals a dual assignment rule for the 706 heads of households who were not patronizing the capital. The larger group patronized the nearest center regardless of town size, travelling less than three kilometers. This group obeyed the nearest neighbor postulate of classical central place theory. A much smaller group (69) travelled to regional centers regardless of distance. Most of these consumers lived more than three kilometers from the nearest town (Fig. 2 ) . Most of the 160 consumers patronizing the capital resided at distances greater than three kilometers both from the regional centers and from the city (Fig. 3 ) . A dual assignment rule thus may be induced both from statistical and from cartographic evidence. A consumer within three kilometers of the nearest urban place obeys the nearest neighbor assignment rule; a consumer at a greater distance travels either to a larger regional center or to the primate city. Although the duality in spatial behavior between short and long distances has been noted previously for shopping and migrational acts, the sharp break at approximately three kilometers warrants exp1anation.l Why do food shoppers more than three kilometers from the nearest urban place appear to disregard distance in destination choice? The answer ap10

pears to lie in the transport modes available: one must either walk or use the extensively developed bus system. The evidence suggests that most rural residents of Aguascalientes perceive the maximum walking range each way as three kilometers. This is the outer range of foodstuffs in the classical sense of outer range. Beyond three kilometers, most prefer to utilize bus service. Once on the bus, most shoppers prefer to travel to the large centers which have a greater variety of food shopping opportunities at lower prices because of increased competition and economies of scale.

Economic Influences Most respondents (73.5 percent) purchased food at the same level of place in which they resided (Groups A and B, Table 1). Almost
TABLE L-INCOMES OF GROUPS MEXICAN OF FOOD SHOPPERS
Mean family income ( z scores) Standard deviation Number of families

Village residents who shop in Villages (Group A ) -.125 .908 Towns (Group D) -.222 .608 City (Group E ) -.037 .768

414 58 139

Rushton (Proceedings), op. cit., footnote 1.

Town residents who shop in -.618 ,101 Villages (Group C) Towns (Group B) +.301 1.265 City (Group F) +.445 1.199

11 223 21

542

BARRY LENTNEK ET

AL.

December

all of these people lived in the place of pur- chase of food to or from work. The hypothesis chase. About one-quarter (26.5 percent) of the that place of wage employment affects place of food shoppers bypassed opportunities in their food shopping cannot be accepted (Table 2). home settlements and patronized places at Conclusions from the Mexican Study other levels (Groups C, D, E, and F, Table 1) . The standardized mean incomes (2 Three variates consistently accounted for the scores) and standard deviations of each group pattern of food shopping: distance, the hiersignificantly reduced variance in income (F = archical level of movement, and household in5.93 at 5 and 860 d.f., Table 1). In general, the come. Combining the spatial, hierarchical, and higher the paired order of origin and destina- income factors produced a dichotomous rule tion, the higher the level of family income, i.e., for the spatial behavior of food shoppers in y . TI,< < yF (where ?$ equals the Aguascalientes. The primary explanation for ,< mean income of Group C, etc.). Further- this behavior pattern is based upon the dual more, nonbypassing villagers are poorer than assignment rule. The secondary explanation is the nonbypassing townspeople. We can gener- that family income increases with hierarchical alize that the greater the hierarchical level of level of origin and destination. The underlying movement, the greater the income. assumption is that mobility is directly related to Four additional economic variables were ex- income, and both increase with the hierarchical amined to discern their relationships to hier- level of origin and destination. After much inarchical level of movement: vestigation, we remain unable to explain two 1) Average value of clothing purchases for anomalous departures from the two step rule. 1967-1968 was examined because multipur- First, eleven of the 866 respondents resided in pose shopping trips often include both food and towns, but chose to purchase food in Venadero, clothing purchases. Clothing may be available the village of their birth. Second, villagers who only in larger places, so clothing purchase shopped in towns had slightly lower incomes might explain some bypassing behavior for than villagers who shopped in villages. Two questions emerged from this analysis: food shopping. 1 ) how does the spatial behavior of rural Mexi2 ) Average value of subsistence food consumption for 1967-1968 was examined be- cans compare with the spatial behavior of rural cause place of purchase may be related to fre- Americans?; and 2) does spatial shopping bequency of travel, a variable which is related to havior develop conjointly with the development of central places? Attempts to answer these level of subsistence food consumption. 3 ) Average value of bus fare expenditure questions may lead to a fuller understanding of for 1967-1968 was examined because bus fare the development of central place systems. expenditure is related to trip frequency and trip COMPARATIVE ANALYSIS distance, and thus to hierarchical level of movement. Data on the food shopping preference be4) Average value of wage income or 1967- havior of rural people are available for 5,500 1968 was examined because farmers who are Iowa households in 1934, for 600 Iowa houseemployed off the farm should exhibit a greater holds in 1969, for 287 Michigan households in degree of urbanization than those deriving total support from their own farms. Wage 2.-NUMBER OF PERSONS BY PLACE OF WORK earners visit larger places more often, and TABLE AND SHOPPING LOCATION TYPE should be more cognizant of the advantages of purchasing in such places. Shopping location type

vE

These variables were examined in the same way as the income variate. The mean values for the various hierarchically defined movement groups had little pattern (through t tests). The possibility of a link between the location for food shopping and the location at which wages were earned was examined because bypassing behavior might result from the pur-

Place of nwrk

Village to ton n

Village to citv

Ton n to cit\

Village

Town
City Unemployed Not known Total

__

18 3 3 31 3

21 4 6 13 28
138

0 12 0 5

3 21

58

1975

CONSUMER BEHAVIOR
-1.5

543

-1.0

75

2 0

t
0

2600-

a 1300

.33

.32
.31

0 c

500 100

.30

O i i i i l b
DISTANCE TO TOWN ( K M ) DISTANCE TO TOWN (MI.)

MEXICO 1 9 6 8
-1.5 -1.0
8 000
2 0 c

75

0 c

6000

a d
0

a
P

4000

0 c

5
0

0 c

2000
1000
0

#,
IOWA 1934

l.0

-*

.s

0.0

10

15

20

25

10

15

20

25
(MI.)

D I S T A N C E TO TOWN (MI.)

DISTANCE TO TOWN

I O W A 1960

MICHIGAN 1 9 6 8

FIG.5. Indifference surfaces for four study areas. All non-Mexican graphs are from Rushton, op. cit., footnotes 1 and 6.

1966-1968, and for the 866 households in Aguascalientes in 1968. The four study areas differ in density of shopping opportunities, functional complexity of towns, mobility of households (Fig. 4 ) , and level of economic development. Only limited generalizations can be based on a comparison of the preference structures of these four areas. The four studies provide a rather small sample (Fig. 5 ) , and socioeconomic data for individual households were available only in Aguascalientes. Despite these

limitations, we believe it is useful to consider these areas as points along a continuum of increasing economic development, mobility, and income (i.e., Aguascalientes 1968, Iowa 1934, Iowa 1960, and Michigan 1966-1968). In Iowa Rushton concluded that the isolines for 1934 are almost vertical for choices made within eight miles of the households residence, but noted propensity to substitute larger town size for distance at greater distances:ll
11 Rushton

(Proceedings), op. cit., footnote I , p . 131.

5 44

BARRY LENTNEK ET

AL.

December

The preference structure for 1960 shows that the substitution was greatest for choices of towns smaller than 3,000 population and also greater for choices made among such towns where their distances from households were greater than ten miles. An interesting behavioral implication of this is the willingness to travel an extra mile, in order to pationize a larger town than would otherwise be available, increases as distance increases.

TABLE 3.-REGRESSION COEFFICIENTS FOR MODEL: PREFERENCE VALUE CONSTANT CITYSIZE = + DISTANCE

Standardized
Area (R) Constant City size

(RY) Distance (R?)

Mexico ( 3 4 ) Iowa 1960 (.80) Michigan (.65)

0.32 0.57 0.37

1.05 (.66) 0.58 ( . 3 5 ) 0.55 (.31)


City size (Error of estimate)

-0.44 (.17) -0.67 (.45) -0.58 (.33)


Distance (Error of estimate)

The inflection point was at about twelve to thirteen miles (20 km) in Michigan and at about two miles ( 3 km) in Mexico. The study areas may be ordered by the distance at which consumers convert from a stronger nearest neighbor bias (i.e., develop a strong tendency to shop at larger centers at greater distances). The order in which the indifference surfaces inflect in terms of distances to towns reflects the continuum of development: Aguascalientes (two miles), Iowa 1934 (eight miles), Iowa 1960 (ten miles) and Michigan (twelve to thirteen miles). The absolute range at which most consumers opt for the nearest place regardless of size increases with the level of mobility and family income. The ability to purchase larger unit volumes on any given trip and to store foodstuffs tends to increase with income levels and hence a greater willingness to travel. The indifference surfaces are not directly comparable (except for the shape of the isolines) because locational types (combinations of town size and distance) were not defined identically. Nevertheless, a standardized fixedeffects regression model was employed to describe each of these rather generalized preference patterns.12 The dependent variable was the scaled utility values of the different locational types, and the independent variables were the midpoints of the distance and town size categories for the locational types. The beta coefficients of this regression-like formulation represent the attribute trade-offs that consumers make when deciding among spatial purchasing opportunities. Slight changes in distance values have far more overall influence than changes in town size. The ratio of standardized coefficients of town size divided by distance showed a distinctly different trade-off pattern in Iowa in 1960 (0.58/0.76 = 0.89) and in Michigan (0.55/0.58 = 0.97) than in
13 The awkward statistical problems of using linear estimates for curvilinear functions are recognized, although a simple log-linear model resulted in a poorer fit ( R 2 ) than did the simple linear model.

Nonstandardized
Area

(N)

Constant

Mexico (18) Iowa 1960 (30) Michigan (30)

0.32 0.57 0.37

0.00002 (0.00000) 0.00017 (0.00003) 0.00016 (0.00003 )

-0.00232 (0.00065) -0.09477 (0.01224) -0.08120 (0.01609)

Mexico (1.05/0.44 = 2.38) ; rural Mexicans placed a higher premium on distance than did rural Americans (Table 3). Mexican consumers had lower incomes and placed greater value on distance than on town size (Fig. 6).
CENTRAL PLACE DEVELOPMENT

The trade areas for lower order central places in Mexico are much smaller than in the United States; the range is only three kilometers. Household income levels are low, and

01

OCMICHIGANI

I
91

8.

0 (IOWA.19601

I 0
0

,
05

.
1

C AGUASCALIENTESIO
I5 20

10

25

RATIO OF ATTRIBUTES (Town/Distance)

FIG.6. Number of distance units per unit of town size in relation to median family income.

1975

CONSUMER BEHAVIOR

545

household expenditures on food are much lower, so Mexican central places have a much smaller trade volume, thus precluding retail economies of scale and foodstuff variety. Higher income Mexicans bypass local centers for the primate city. The constant flow of income from rural areas to the regional center generates commercial and demographic primacy at the regional level, given the income and employment multiplier effects of retailing. Since commerce is the basic function of regional centers, the explanation for regional primacy should reflect the reasons for the concentration of central place activities. Our limited evidence suggests that consumer choice behavior in space is subject to lawful evolutionary processes. In particular, food shopping behavior may be universally subject to a dual assignment rule. Households within a limited range of an opportunity exhibit a very high probability of obeying the Christallerian nearest neighbor axiom, whereas households at some distance prefer the more varied shopping in relatively large centers. The absolute range within which the Christallerian axiom is applicable seems to increase considerably (from

two to thirteen miles) with the level of economic development. Although we have begun to explore the relationship of spatial behavior to change in urban hierarchical structure, we have presented very limited evidence, and results from other areas are necessary. Concrete and reliable data on threshold volumes of trade for small retailers in developing regions are not available, and it is difficult to specify, even stochastically, the rate of entry of intervening shopping opportunities. The most reliable source for determining whether an evolutionary process exists would be long term time series data for a single area, but such data are not available. Finally, our evidence relates only to food shopping. Despite these difficulties, we have attempted to shed some light on the development of central place hierarchies. A dual assignment rule explains spatial behavior patterns in areas with great social and economic differences. Additional studies using our methodological framework may provide the foundation for a behaviorally based model of a hierarchical evolution of central place.

Vous aimerez peut-être aussi