Babson College - The Stability of Retail Shopping Choices over Time and Across Countries

Because retailing originated as a geographically local market activity, it was natural for retailing theory and research to begin from the same local market perspective. Research in substantive retailing decision areas, such as store location, retail positioning, and patronage behavior, began and has almost always remained focused at this level. However, retailing practice has changed over the past 50 years, due to the spatial diffusion of retail innovations, the aggressive geographic market expansion of successful retail organizations, and the development of a global consumer market. This expansion of retailing practice now raises urgent questions about whether retailing research findings and normative conclusions developed from the local market paradigm can be generalized more broadly. The potential for savings in expense and time would be enormous if this were possible.

For example, can one assume that the factors that drive store choices in a market where a research study was conducted, typically for convenience reasons also will drive store choices in other markets? Can one assume that any new retail format that is successful in the USA, such as a new category killer, could also be successful if introduced into other markets? The sheer variety of social political, cultural, legal and historical forces at work in individual countries would continue to suggest that the answer is negative because institutional diversity rather than uniformity has been the hallmark of retail innovation.

Only one study, Arnold, Oum & Tigert (1983), has systematically examined these issues, testing for the consistency of the effects of perceived retail attributes on patronage choice, using data collected in different seasons, time, cities and countries. Their study of supermarket patronage retained the null hypothesis that their multinomial logit (MNL) parameter vectors were equal across four quarterly sets of seasonal data from the Toronto market in 1980. However, they rejected the null hypotheses that the multinomial logit model parameter vectors for supermarkets were identical (i) over time (1974, 1977 and 1980 Toronto samples), (ii) over US cities (Cleveland, St. Louis, and Tampa), and (iii) over cities in different countries (Toronto, Cleveland, Birmingham, UK and Amsterdam).

But, since their study a new paradigm has emerged that relies on random utility theory as a basis for developing choice models and comparing sources of choice data. This new paradigm can isolate whether apparent differences in choice model parameters across data sets are due to real differences in preference structure or to differences in random error component variance. We employed this new paradigm in a fresh look at survey data taken across a number of cities in space and time.

Surprisingly, the application of this new model to cities in different countries supports a hypothesis of spatial consistency of consumer preferences. Similarly, tests of stability over two years for supermarkets and four years for shopping centers also show no differences in preferences as both the random component variances and the coefficients were unchanged over period studied.

Managerially, our results suggest that similar retailing strategies will yield roughly the same results over time and many geographical locales. This is consistent with consumer shopping patterns that can be observed in many cities throughout the world. What may differ, however, and further research is needed to determine its extent, is the magnitudes of changes in choice behavior in response to different retailing strategies and tactics. That is, common utility parameters lead to similar strategic conclusions, but differences in scale ratios lead to different expectations about the magnitudes of the impacts of those strategies from place to place or time to time. Put another way, different scale ratios imply different response elasticities, such that smaller random components lead to larger elasticities and larger random components leads to smaller effects. For example, our results suggest that if one applied the same retailing strategy to shopping centers in Edmonton and Oslo, the strategy should have
larger effects on choices in Oslo, all else equal.

The latter point is non-trivial because managers learn from experiences, and they try a particular strategy that fails to meet expectations in place A, our results suggest that this does not necessarily mean that it would not meet expectations in place B. Similarly, if a strategy is tried in place A, and meets expectations, it does not mean that it would do likewise in place B. In turn, this suggests that experiential learning may be much riskier than relying on science, especially in light of the fact that RUT provides researchers and managers with powerful tools with which to analyze and predict behavior. Thus, while it is common for retailers to "try" something to see what happens, but this may be sub-optimal in the absence of a systematic strategy that permits comparisons and generalizations.

Print friendly version

E-mail this page to a friend