A study by Miranda and Jegasothy (2009) aimed to explain consumer expectations (in Australia since that is the study setting) about what retailers should do in respect of four types of goods that are commonly returned (the designated dependent variable, DV) in respect of five essentially unrelated (among themselves) independent variables:
- Why shoppers became dissatisfied and returned the goods in question;
- Socio-demographic variables like age of shopper, occupation, and family size;
- Experience with the store where the purchase was made or which was designated by the manufacturer’s Help Line as the official returns, repair, and replacement center for the area.
- “Inclination to repurchase the returned brands” (369) or product class; and,
- How satisfactory the return process was deemed to be.
The relevance of studying this designated DV, the authors affirm, is heightened by a combination of factors: the sizeable value of goods that are returned each year (about $35 billion but this figure is for the American market) and which manufacturers who wish to maintain high rates of customer satisfaction feel they must replace no questions asked; the growing concern about environmental issues that could impel questions about how manufacturers dispose of returned merchandise; friction in the value chain when manufacturers do not reimburse retailers for the tasks of accepting and refurbishing returned goods; and the damage is done to brand equity by-products so defective consumers trouble themselves to complain and return them.
Based on Compliance Strategies Branch, Australian Consumer and Competition Commission rankings, the authors confine the DV understudy to the top four categories of returned products: packaged food, toiletries, apparel, and appliance product categories. In turn, the IV’s are operationalized as a combination of qualitative/categorical and ordinal variables:
CATEGORICAL: Gender, age, occupation, the incidence of returning defective products, whether single or multiple items, whether used items before returning, whether returned for cash, same model or another model, the reason for return, type of refund sought, whether an external stimulus was needed to impel return, and whether agreeable to purchasing reconstituted goods in general and the returned brand in particular.
ORDINAL: Recency of returning defective products, how long returned item was in use, how obligated a manufacturer is to accept returned goods, satisfaction with store handling of a return procedure, the price class of the returned good,
The study employed convenience sampling in the form of mall intercept interviews. Over two weeks, a total of 426 “randomly intercepted adult shoppers” (Miranda and Jegasothy, 2009, p. 374) outside the Chadstone and High Point shopping malls in Melbourne agreed to be interviewed about any of the four product classes of concern that they had returned at least once in the prior three years. The overall sample size was a function of attaining the degrees of freedom required for statistical analyses later on. After all, the set of four product classes and minimum cross-breaks (by gender) effectively created a 4 X 2 research design in analysis.
The authors would say only that the sampling method was random intercept to permit projecting to a target population. Since true random selection cannot be left up to contractual project hires who undertake fieldwork because of the temptation to self-select amenable-looking shoppers, it is likely that the authors enforced systematic interval sampling for a more truly random approach. However, justifying random-intercept convenience sampling as representative of the population of interest is flawed by two considerations:
A. The research team does not state whether they assume the results to be projectible to Melbourne shoppers only or Australians across the nation. The research report concludes with sweeping statements “consumers”, presumably to convince the gullible reader that results are applicable in other countries and cultural contexts.
B. Relying solely on the contention of Malhotra, Hall, Shaw, and Openheim (2002) that such a sampling approach is representative. No other research methods source agrees that convenience sampling is as reliable and rigorous as systematic random or stratified sampling. For one, mall intercepts do not account for those who patronize off-mall specialty boutiques, by mail order, through the Internet or work-from-home direct sellers such as the ubiquitous Avon ladies.
The researchers can at least be credited for attempting to comprehensively account for a multitude of variables around the return of spoiled food or defective clothing, toiletries, and home appliances.
It is rather difficult, however, to accept the findings as breakthrough research. For one, some findings are self-evident and no longer bear investigating. For instance, there is only ever one valid reason for returning food and drink. And one doubts whether regulatory authorities anywhere will accept any method of disposal except destruction. Secondly, internal validity as an “experiment” is flawed, and external validity is well and truly diminished by overreaching when it comes to data analysis.
It is like an experiment that one manipulates an IV, sets up at least one experimental and one control group, and implements measures to hold intervening or confounding variables constant. But this is a natural experiment at best and can even be downgraded to an observational study. If the latter is true, then validity is degraded by selection bias, and the entire study design is affected by low or non-existent between-groups equivalency in probability terms. As a natural experiment, the Miranda and Jegasothy effort can perhaps be lauded for including all possible variables around the purchase return phenomenon. But the need to hold some constant while accounting for cause-and-effect hypotheses is achieved only by assigning dummy variables in logit regression analysis.
There are at least two technical flaws. One is that unrelated IVs are required in natural experiments. Had autocorrelations indices been run, one would likely have found a high degree of correlation between age and occupation on one hand, and the third to the fifth IVs on the other hand. The second is the acceptability of running a regression method on DVs and IVs that are categorical, notwithstanding the recourse taken of setting base categories to “1” in data analysis. Considering the nature of the variables, a series of chi-square and Kolmogorov-Smirnov tests would have been more rigorous. Here is another case where simpler methods would have been best.
In attempting to construct a comprehensive model around DVs, lastly, the researchers blur the lines between correlation and causality. It is difficult to explain how IVs like prior experience with the store or satisfaction with the return process cause consumer expectations about the disposal of returned goods to change. The IV about purchasing refurbished merchandise is not even an antecedent, much less a cause, of the stated DV on expected disposal or recycling of returned goods. Rather, the latter is more likely the “cause” or the compelling factor that explains purchase propensity for recycled goods. One may well conclude, therefore, that the researchers have expended a great deal of time, statistics, and journal pages on “proving” such hoary truths as consumers expect spoiled food products to be destroyed and think less of a defective brand.
References
Malhotra, N. K., Hall, J., Shaw, M. & Openheim, P. (2002). Marketing research (2d ed.), New South Wales, Australia: Pearson.
Miranda, M. J. & Jegasothy, K. (2009). Does consumers’ future buying behaviour regarding products that are returned influence the way consumers want returned goods to be disposed? Journal of Marketing Theory and Practice, 17 (4); 369-392.