Comments of Evolving Country of Origin Research

Carl Obermiller, Seattle University
[ to cite ]:
Carl Obermiller (1993) ,"Comments of Evolving Country of Origin Research", in NA - Advances in Consumer Research Volume 20, eds. Leigh McAlister and Michael L. Rothschild, Provo, UT : Association for Consumer Research, Pages: 690-691.

Advances in Consumer Research Volume 20, 1993      Pages 690-691


Carl Obermiller, Seattle University



(Li, Leung, and Wyer)

By the time I reached the results section of this paper, I had several critical comments. First, why examine two processes with two manipulations and consider only the main effects of one manipulation on one process and the other manipulation on the second process, as if this were two studies conducted on the same subjects. Both information load and motivation are apt to influence both the roles of CO as signal and as attribute; yet, no hypotheses are presented for half the outcomes of the design. Second, the logic of the second, and more interesting, hypothesis is not compelling. The hypothesis assumes that the attribute value of CO is less salient than other attribute information. No justification for such an assumption is given, and several counter-examples come quickly to mindCFrench champagne, Russian caviar, German automobiles, American politicians. For American consumers of any of these products, CO labels confer authenticity, social standing, or patriotism directly and immediately; for some products, at least, CO is a highly salient cue. An alternative interpretation of H2 logic is that CO is but one of many cues that might be processed; and, higher motivation leads to more cues' being processed. This interpretation assumes that increasing motivation has no effect on the relative salience of the cues. One might argue that increasing motivation shifts salience away from such peripheral cues as CO labels. In which case, higher motivation might lead to a reduced attribute effect. Thus, although H2 was an interesting conjecture, I felt it lacked sufficient a priori justification for an hypothesis.

My third criticism of Li, et al, (up to the point at which they presented results) was the confounding of information level with time of presentation. The high information condition was about 300% longer duration. Although in both conditions the important eight attributes immediately preceded the dependent measures in both conditions, one could expect that attention to those eight attributes waned in the high information condition. Thus, any lack of effect of signalling might be due to a design artifact. An alternative procedure might have been presenting the information simultaneously, as in a print format; which would allow the viewer to scan and focus attention on important bits of information.

My more serious questions of Li, et al arise with the presentation and interpretation of the results. At the heart of the problem is the lack of a measure of the effect of attribute processing. Recall that the dependent measures were 32 attribute scales and several scales of global quality. It is surprising that, having done prior research demonstrating the attribute effect, no measures specific to it were included. In fact, the only evidence presented for the hypothesized attribute process effect is the result of step-wise regressions in which the attribute ratings are entered first, followed by the CO labels. The significant change in adjusted R2 is attributed to the attribute process. While it is clear from such an analysis that the attribute ratings do not explain all the variance captured by the CO labels; it is also true that we cannot simply assume that the additional explained variance is due to the CO-as-attribute processing. Something else is contained in the CO labels, but this study cannot claim to have ruled out all but two possible effects of country-of-origin labels. The lack of a direct measure of the attribute effect is a serious flaw.

Equally distressing and even more confusing is the introduction of another variable in the results section. The 32 attributes are factor-analyzed into two categories, interpreted as style and function. Most of the ensuing discussion of the results focuses on the different outcomes for style and function evaluations, with considerable effort to present these differences as support for the original hypotheses. The discussion is so dedicated to style and function that one might wonder if the second half of this paper were not mismatched with the first. It is unclear if the style/function dichotomy ought to have been anticipated in the hypotheses. I think the problem reflects the failure to design the research with adequate dependent measures. The hypotheses of the study do not warrant measures of 32 specific attributes (but do warrant measures of the attribute effect). The style/function split is sensible; but the different effects of these two dependent measures, combined with the indirect assessment of attribute effects, results in a confusing presentation of the studies results and a tenuous interpretation.

Included in the interpretation are three points that raised questions. First, the limited effect of information on signalling (for style but not function) was explained in terms of information economics. Why is this theoretical base introduced in the discussion section rather than as justification for the hypothesis? Second, the dominance of the signalling role over the attribute role (an arguable conclusion) is suggested as explainable if we assume that consumers do not use any piece of information more than once. This assumption is supported by the argument that CO would be used as a signalling cue rather than an attribute. The justification is that a signalling cue can be used to infer "several different attributes." Why should we assume that consumers can use a cue to infer multiple attributes but not to infer an attribute and a value? Moreover, we have substantial evidence to suggest that consumers show no such limitation in their interpretation of price cues, which are used both to infer other attribute values and to imply economic sacrifice and to impute social standing and to estimate future prices, etc, etc. Finally, the stress of inadequate dependent variables leads to the speculation that CO labels have stronger effects on "affective" products than on "functional" products. While some marginal difference may exist, and the question may lead to fruitful investigations, both the premise and the implication appear erroneous. The premise is that product type differences dominate other factors. I think it unlikely that a such a gross dichotomy as affective/functional product type will be as fruitful as psychological variables. Not only are few products likely to be purely one or the other, there may be no products for whom there is consensus of perceptionCwhat is affective for one person is functional for another. The implication, that one should expect a main effect across products, runs contrary to most of the CO research that has been done, which has examined a long list of primarily functional productsCincluding, automobiles, computers, cameras, camping stoves, and stereo systems.



(Alden, Hoyer, and Crowley)

I found the Alden, et al study a sophisticated research design; my two questions address its flanks, the ramparts of its middle being well fortified. First, the question of applying the results: The authors cite the impending free trade agreement as a reason to study the Mexico label and as a reason to expect CO to become more relevant as product information. No evidence is offered, however, to indicate the extent to which consumers outside the lab attend to CO labels. In fact, I would use the study's own pretest results to argue that consumers may gloss over the CO label. I can offer no other explanation for the finding that "Made in Mexico" as part of a description of toothpaste did not differ in typicality from "Made in the U.S." I should require a great deal of persuasion to be convinced that consumers equally expect their toothpastes to be produced in these two countries. On the other hand, I can easily accept that consumers pay little heed to where their toothpastes are made. Low involvement products are low involvement products, after all; and, before we extend our findings too far from the lab, we should determine the extent to which CO cues are processed under low involvement. ("Low" involvement outside the lab is likely to be considerably lower than inside.)

The second question is both more and less important. Less important because the results stand regardless of the naming of the constructs; more important because the results may be misunderstood as conceptualized. This question involves the three constructsCinvolvement, perceived risk, and typicality. Alden, et al argue that higher perceived risk, as indicated by CO, triggers a shift in involvement; independent of the typicality of the CO cue. My question is, how can the risk associated with the CO cue be interpreted unless involvement is high. As I reason above, for low involvement products, consumers are not apt to examine peripheral cues in order to make evaluative inferences. In short, consumers will not recognize any implications of the CO cue unless the country is atypical. An unexpected bit of information (atypical CO) would increase motivation to process, shifting evaluation from category- to attribute-based, and lead to a negative effect of higher perceived risk. Given the study that was done, the confusion of constructs is difficult to sort out, since Mexico is arguably both less typical and higher in perceived risk. A test of the differences might be conducted with CO's of equal perceived risk but different levels of typicality and different implications for evaluation. For example, Italy and U.S. may be equally risky CO's for dress shoes; Mexico, again, more risky. Italy may signal the highest quality, U.S. middle quality, and Mexico low quality. If this is so, the Alden, et al logic would suggest equal evaluations for U.S. and Italy, lower for Mexico, based on the perceived risks. I would suggest both the Italy and Mexico labels would seem atypical and would shift consumers to further analysis, with the results positive for Italy and negative for Mexico, relative to the U.S.




This paper does not deal with CO, although I see its significance to that research stream. Like the other authors, Baumgartner examines alternative processing modes. He discusses holistic and analytic perception, presents a methodology for examining them, and tests the methodology in a product perception study. The dichotomy Baumgartner addresses is quite similar to the category/attribute distinction considered by Alden, et al. As is typical in consumer behavior's borrowing of theory from psychology, some of the limitations have been loosened. Consumer behaviorists who have examined category-based processing do not limit processing to integral attributes as holistic processing is characterized in psychology. (Integral attributes are characteristics of an object that cannot be considered separately from the object, e.g. brightness.) In general, Baumgartner applies holistic processing to product perception much as Alden, et al do category-based processingCperception based on little motivation, little elaboration. Baumgartner focuses on perception; but, his logic is consistent with the subsequent differences in evaluation that are the primary concern of Alden, et al.

Baumgartner's study is an adequate introduction to a new methodology and a potentially fruitful theory; but, several aspects are troublesome. The first is his appeal to low statistical power in defense of marginal significance levels. On the one hand, he suggests power may be low; he should compute the power and answer his own conjecture. On the other hand, he ought to have considered power beforehand and designed the study with adequate statistical conclusion validity. (Too few of us consider power until it's too late; Mea culpa.) A second problem is the reversal of endpoints for the classification index (in the revised draft; I do not know if this error persisted to the final copy). Holistic processors may skip over the mistake; analytic processors are apt to be confused and dubious. My final disagreement is with Baumgartner's conclusion that the finding that decision time was related to classification process implies that consumers' time budgets will influence their decision processes. Although I agree with him, I think the conclusion should be tempered; the results are consistent with the conjecture, but not suggestive. The manipulation of decision process caused a change in decision time. One cannot logically conclude that a change in decision time would, therefore, cause a change in decision process.


None of the papers focused on the description or meaning of specific cues. Early research on country-of-origin accomplished little beyond identifying which countries supplied positive, which negative effects on product evaluations. Much of the motivation of the research stream came from disagreements about whether CO cues did or did not have significant effects; but, of course the discrepancies came, in part, from using different countries, different subjects, different products, and different processing situations. The current papers show a marked advance in our level of analysis. They share a strong grounding in theory. The two papers that address CO could as easily have dealt with other bits of product informationCprice, brand name, store name, or product category. Specific country effects on product evaluation can, as a result, be expected to differ as conditions change their typicality, perceived risk, or their value as attributes. Moreover, the nature by which consumers process CO cues depends on basic forces such as motivation, time, ability, and expectations. Different findings for specific country effects may be explainable as differences in the underlying theoretical framework for product evaluations.

An important next step, if CO research is to remain vibrant, is to address the question of relevance. More work should be done on the way consumers use CO information in mundane situations. How salient is CO? Does CO salience differ by product involvement? Outside the laboratory, what determines the extent to which consumers process categorically/holistically versus by attribute/analytically? And, how does CO influence the process of evaluation, and how is it treated in each? To answer these questions, we need to do more field research, surveys, observations, intercept questionnaires, etc. And, we need to design laboratory studies that simulate the relevant dimensions of natural environmentsClack of focused attention, realistic levels of involvement, perhaps even controlled levels of distraction such as might be encountered in stores or in advertising viewing environments. Although I have raised questions about many of the details of the research in these papers, I commend the authors for their grounding of the work in theoretical frameworks that can be generalized. Now, I think we can turn some attention to expanding those theories, making them robust enough to survive outside the lab.