European Advances in Consumer Research Volume 1, 1993 Pages 191-197
THE EFFECTS OF INTRINSIC, COUNTRY-OF-ORIGIN AND PRICE CUES ON PRODUCT EVALUATION AND CHOICE
John Liefeld, University of Guelph, Ontario, Canada
Marjorie Wall, University of Guelph, Ontario, Canada
[The data reported herein were supported in part by strategic grants from the Social Sciences and Humanities Research Council of Canada and by manufacturers of some of the stimulus products.]
Product choice and ratings of quality and value were employed as dependent variables in a multi-cue experiment. Product stimuli, varying in intrinsic properties, were presented to New Zealand consumers for choice and rating, along with extrinsic information cues accessed by a computerized information display board. Chi-square, Phi coefficients and analysis of variance revealed that all models yielded low magnitudes of explanation, although the combination of intrinsic and extrinsic cues improved the explanation levels, particularly when product choice was the dependent variable.
Prior to the 1960's, research on information cue use in product evaluation and choice was concentrated on the extrinsic cues of price and brand, using survey methods. In the 1960's the research was broadened to include other extrinsic cues such as packaging, seller, and country-of-origin (Wall et al, 1991) and the first experiments were reported (Schleiffer & Dunn, 1968; Schooler, 1965). In the 1970's and 80's many surveys and experiments focused on the country-of-origin (CO) cue in consumer ratings of product quality (Baughn & Yaprak, 1992; Liefeld, 1992).
Although the CO cue has consistently been found to be statistically significant, the strength of effect for this and other extrinsic cues employed in survey and experimental research is disappointing. Seldom have R Squares greater that 0.10 been reported, even when many variables were included in the models. One explanation for these weak effects suggests that the intrinsic properties of products may play more important roles in product evaluation than extrinsic cues such as brand, price or CO (Szbillo & Jacoby, 1974; Wall et al, 1991; Zeithaml, 1988). Research to date has been constrained to evaluating one product at a time and asking respondents to rate its quality and/or other attributes. In the marketplace, however, consumers make choices from a variety of competing brands, models, styles, options etc., simultaneously. Thus the external validity of past research is questionable.
Survey research is constrained to single cue measurement, ie., respondents are limited to making judgments about one product stimulus at a time for one cue at a time. Several experimental studies of CO cue effects have employed multi-cue designs in which the respondents could take into account product stimuli with several extrinsic cues, such as price and brand, presented simultaneously. Multi-cue designs brought these investigations closer to the realism of the marketplace yet did not match its multiple product stimuli nature. Rating scales for perceived quality, risk, value, etc., and semantic differential image scales have constituted the dependent variables in both survey and experiment based research (Nagashima, 1970; Nes, 1982). Only one experiment reported product choice as a dependent variable (Schellinck, 1989); yet, in the marketplace, choice and purchase are the conclusive measures of consumer behaviour.
When making choices, consumers may or may not make explicit judgments about separate attributes of a product such as quality, risk in purchase, value, etc. If they do make separate judgments, the model by which such judgments are combined to result in the choice of a specific product alternative is not well understood by consumer researchers. This is evidenced by the plethora of theoretical models and contradictory findings about consumer decision processes reported in the literature.
This paper reports one of a series of studies designed to improve upon past research by: using product choice, in addition to ratings of value and quality, as dependent variables in a multi-cue experimental design; using multiple product stimuli in order to include manipulation of the intrinsic attributes of product alternatives; and employing an information use tracing methodology to attempt to capture measures of the extrinsic cues consumers attend to while engaged in the choice process.
The research described herein was replicated in three countries, Australia, New Zealand and Canada. This paper reports results for the first two goals for New Zealand only. The specific objectives were to answer the following research questions.
1. Are CO and price (Extrinsic Factors) related to product choice?
2. Is sample appearance (Intrinsic Factor) related to product choice?
3. Does the model 'size of effect' (explained variance), improve when the intrinsic cue of sample appearance is combined with the extrinsic cues of CO and price, with product choice as the dependent variable?
4. What are the combined effects and relative strengths of CO, price (Extrinsic), sample appearance (Intrinsic), age and gender, with self ratings of perceived ability to judge product quality held as a covariate, on a) ratings of product quality, and b) ratings of product value?
5. Do models using product choice, ratings of quality or ratings of product value as the dependent variables, yield greater magnitudes of explanation (size of effect)?
6. Do the relative strengths of the cues (Extrinsic and Intrinsic) differ between models using product choice as the dependent variable and those using ratings of quality or value?
RESEARCH DESIGN AND METHODS
Consumers were asked to pretend they were shopping for a product for which four alternative items which were substitutable but varying in their intrinsic properties were provided. The extrinsic information of brand, price, CO, etc., was removed from each sample and made available in an Information Display Board (IDB) matrix on a computer screen (Mouselab; Johnson et al, 1988). The subject moved a pointer on the screen with a mouse to open boxes which revealed the information of interest. The first task was to decide "which one would you buy?" The respondent could examine both the intrinsic physical properties of the four alternatives and the extrinsic information on the IDB screen. After choosing the one they would buy, the second task was to rate the quality and value of each item. The information matrix was again available on the computer screen. This procedure was repeated separately for three product categoriesBmen's business shirts, smoke detectors, and jars of pickles.
Product StimuliBExperimental Intrinsic Factors
The samples for each product category were chosen to provide a realistic dispersion across the intrinsic properties of each product. For men's business shirts the colours and patterns were varied. For smoke detectors appearance and shape of the cases (buttons etc.), were varied. For pickles the colour, size and the presence of dill seeds and onion bits were varied. These variations in intrinsic properties between the product samples formed a 'within-subject' factor in the experiment.
Product StimuliBExperimental Extrinsic Factors
For each product category, two extrinsic variables were manipulated in the experiments,B CO (four levels) and Price (high/low), to produce a 'between-subjects' design with eight treatment conditions as follows: men's business shirts could be made in New Zealand, Canada, Taiwan, or China, and have the price of $34.00 or $59.00; smoke detectors could be made in New Zealand, Canada, Ireland, or Mexico, and have the price of $22.95 or $39.95; and pickles could be made in New Zealand, Canada, Hungary or Czechoslovakia, and have the price of $.33/100gms or $.44/100gms.
Product StimuliBOther Factors
To provide greater mundane realism, several other information cues were provided on the IDB screen for each product category but were not experimentally manipulated. These included brand name and information about other attributes and were held constant across the experiment treatments as follows:
Sample ABBrand = Ives St. Laurent, 100% cotton.
Sample BBBrand = Van Heusen, 50/50 cotton/poly blend.
Sample CBBrand = Summit, 50/50 cotton/poly blend.
Sample DBBrand = Lichfield, 100% Polyester.
Sample ABBrand = First Alert, escape light,
Sample BBBrand = Dicon, alarm pause.
Sample CBBrand = Goldair, test button.
Sample DBBrand = Click, low battery warning.
Sample ABPolskie Ogorki.
Sample BBBaby Dill.
Sample DBDill (regular).
The dependent variable of product choice was recorded by moving a pointer to a box labelled A, B, C or D at the bottom of the IDB screen and pressing the mouse button. The dependent variables of product quality and value were obtained using paper and pencil rating scales. Eleven point interval scales ranging from 0B10 were labelled 'low qualityBhigh quality', or 'low valueBhigh value' at their polar ends.
Respondents were recruited in two urban shopping malls, one in Auckland and one in Dunedin in New Zealand. Shoppers were screened to ensure they were between 21 and 55 years of age and had eaten pickles in the past year. Qualified consenting subjects were randomly assigned, using quotas for age and sex, to one of the eight treatment conditions for each product. Respondents first completed a short questionnaire regarding brand awareness for pickles and smoke detectors, then were seated in front of a computer screen which welcomed them and asked them to move the mouse around to acclimatize positioning the pointer on the screen. The second screen asked them to pretend they were shopping for a men's business shirt and that there were four styles available in the store. Four shirt samples were placed on the table in clear plastic bags labelled A, B, C, and D and respondents were informed that the next (third) screen would provide the information for each shirt which normally could be found on the labels. They were asked to "look at whatever you look at when you buy a shirt and tell us which shirt would you buy?" The information provided in the computer screen boxes depended on the treatment condition to which the respondent was assigned. The respondents could examine the shirts, and obtain any desired information from the screen about the brand, price, country-of-origin or material. When they had decided which shirt they would buy, they picked the appropriate box with the mouse. Then the fourth screen explained that the second task was to rate the quality and value of all four shirts using paper and pencil rating scales, and in case they wanted to review any of the label information, the next screen (#5) would provide the same information matrix they had used earlier (screen #3). Following completion of the rating scales for the men's shirts the same series of tasks was repeated for the smoke detectors and pickles.
After the computer phase, respondents completed a questionnaire which asked information about prior experience in purchasing these types of products, their ratings of how good a judge they thought they were of the quality of the three product categories, ratings of the quality of products in general from a list of 14 countries and demographics. In Auckland 669 contacts were made in the shopping mall of which 96 (14.3%) completed the experiment. In Dunedin 473 contacts were made of which 97 (20.5%) completed the experiment. Demographic characteristics of the total sample of 193 are shown in Table 1.
To examine the effects of the extrinsic and intrinsic cues on product choice two forms of analysis were employed. Because the dependent variable (product choice) and the independent variables (treatment conditions) were nominal data, crosstabulations were used to determine the strengths of effects of the extrinsic and intrinsic cue treatments on choice. Chi Square was used to test significance and the Phi Coefficient, which is equivalent to R2 and is an estimate of the proportion of variance explained, was used to test the 'size of effect' (Rosenthal and Rosnow, 1984). To examine the effects of the intrinsic and extrinsic cues on ratings of product quality and value, repeated measures analyses of covariance (ANCOVA) were executed with age and gender as additional independent variables and self ratings of ability to judge products as a covariate.
Extrinsic factors: Crosstabulation and Chi Square analysis yielded very weak and non-significant price effects. Consequently, for the purposes of examining country effects, the price conditions could be collapsed to reduce the analysis to four treatment conditions rather than eight, thus, leaving CO as the primary experimentally manipulated extrinsic cue. The actual and expected frequencies under the null hypotheses of no association are shown in Table 2, for each product sample, across the four collapsed treatment conditions. Country was significantly related to choice for all product categories, with products made in the home countryBNew Zealand, chosen most often, followed by products from Canada, while products from the remaining countries, which represented less developed nations, were chosen least often.
DEMOGRAPHIC CHARACTERISTICS OF THE SAMPLE
FREQUENCIES OF CHOICE BY TREATMENTS
Intrinsic factors: To determine the aggregate effect of the intrinsic factors, choice of product samples across all treatment conditions (the total frequency of choice for each sample) was crosstabulated with the null hypothesis of equal probability of choosing any of the four samples. For shirts, the Chi Square value was 24.58 (3 df, p < .005, phi coefficient = 0.356). For smoke detectors, the Chi Square value was 8.07 (3 df, p < .05, phi coefficient = 0.203). For pickles, the Chi Square value was 16.4 (3 df, p < .005, phi coefficient = 0.292). Shirt A, Smoke Detector A and Pickle B were preferred by most, regardless of treatment condition while Shirt C, Smoke Detector B, and Pickle A were least preferred. Thus the variation in intrinsic factors significantly influenced product choice.
The Chi Square values for each product category in Table 2, indicate that when the CO treatment conditions were combined with the intrinsic effects, the probability of accepting the null hypothesis was even lower yet (p < 0.002, 0.000 and 0.000 respectively), and the Phi Coefficients were larger (0.364, 0.457, & 0.430 respectively). Further, when the crosstabulations were executed including the extrinsic factor of price (even though the individual price effects were insignificant), the Phi Coefficients were higher again ( 0.419, 0.482 and 0.465 respectively).
By arraying the Phi Coefficients for these three levels of analysis (Table 3), the increased explanatory power achieved by combining intrinsic and extrinsic factors is demonstrated. The percentage gains in Phi2 for men's shirts by adding CO and Price to the model are modest, 4.0% and 37.8% respectively, and the Phi2 for the intrinsic factor alone is quite large. For smoke detectors the intrinsic effects alone do not explain much variance (Phi2 only 0.04), but the percentage gain from adding CO to the model is very large (400%) and the price factor increases the cumulative gain in explanation over the intrinsic effect alone, to 455%. The Phi2 of the intrinsic effect for pickles is a modest .085. By adding CO the gain in explanatory power is 117%. Adding the price effect increases this gain to 154%.
SIZE OF EFFECT - PHI2
These results provide answers to research questions 1, 2, and 3. CO effects were much stronger than price effects (Q1). The intrinsic factor was much stronger than the CO effect for men's shirts, while for smoke detectors, the extrinsic factors were much stronger than the intrinsic, and for pickles, the explanatory power of the intrinsic and extrinsic factors were about equal (Q2 & Q3). Finally, the 'size of effect' of the model improved when intrinsic and extrinsic factors were combined (Q3).
These findings have intuitive appeal. Men's shirts are a "style" good and most respondents were experienced as purchasers. Smoke detectors and pickles had low sales in New Zealand at the time of the study, hence respondents were not experienced in purchasing these products. Additionally, smoke detectors are likely evaluated more on function than appearance. Pickles have an important component of taste and the intrinsic properties of size, colour and appearance in the jar are not direct indicators of flavour. This evidence supports a conclusion that the effects of intrinsic and extrinsic cues are product dependent.
Ratings of Product Quality and Value
Since the dependent variablesBratings of quality and value, were intervally scaled and each respondent rated all four product samples, analysis of variance using a repeated measures model was employed to answer research question four. To improve the model both age and gender were included as factors along with a covariate of respondents' self rating of ability to judge product quality. Six separate Covariance models were run under SAS-GLM. The general form of the model was as follows.
Q/V (product) = f Sample Appearance, Country, Price, Gender, Age, PA
Q/V = rating of quality or value
PA (covariate) = perceived ability to judge product quality
The output from the six repeated measure analyses of covariance is summarized in Table 4. All six models were statistically significant but the R2 values, compared to the Phi2 values in the crosstabulation analyses with product choice as the dependent variable, are much lower. Except for Shirt Quality, the R2 values are very similar to those reported in prior experiments (Ettenson et al, 1988; Hong & Wyer, 1989; Schellinck, 1989; Nes, 1982; Wall et al, 1991). In fact, multi-cue experiments using ratings of product quality and value did not achieve improved R2 values compared to the earlier single cue experiments using rating scales. This finding is similar to findings reported by Liefeld (1992), using a meta-analytic approach.
The covariate, perceived ability to judge quality for the product class, was strong for the detector and pickle analyses but not for shirts. Gender was not statistically significant for any dependent variable but approached significance for value ratings of smoke detectors. Age was statistically significant for shirt and smoke detector quality. For both, older persons gave higher ratings of quality. The intrinsic factor, sample appearance, was twice as strong as CO or price for shirt quality, and even stronger than the extrinsic cues for shirt value. Also the age factor was stronger than CO or price for shirt quality. The shirt quality model achieved the highest R2 (0.14), and all factors were significant except gender and perceived ability to judge shirt quality. For detector quality and value the CO factor was almost four times stronger than the intrinsic factor of sample appearance. For pickle quality the CO factor was marginally stronger than the sample appearance factor, but for pickle value, their relative strengths were reversed. The price cue showed some relative strength for the detector and pickle value rating. These findings mirror the results of the analysis with product choice as the dependent variable, except ratings of quality and value gave lower sizes of effects.
The directional nature of significant factors in these models is revealed in Table 5. For the intrinsic factors, the direction of effects is identical to those found with the analysis with product choice as the dependent variable. Shirt A, Detector A, and Pickle B, all had significantly higher scale means. The home country, New Zealand, had the highest means, and the other developed country, Canada, was also very high. Older respondents gave higher ratings for quality than did younger respondents. A higher price was related to higher ratings of quality and lower ratings of value.
The magnitudes of the differences between the means of the factor levels are much smaller than the differences found in the choice frequency analyses with product choice as the dependent variable. Clearly, rating scales did not uncover the degree of difference captured when the dependent variable was product choice. The validity and reliability of rating scales should be questioned. In a series of experiments measuring the response function associated with rating scales, Saris and his colleagues concluded that "variations in response functions (ie, variations in the way respondents answer questions) can disturb the correlations between variables so much that we do not see how one can rely on analyses across respondents" (Saris, 1988). Others have also reported problems associated with rating scale type measures (Lodge, 1981; Rule & Curtis, 1982). For our purposes here it is sufficient to note that the magnitude of effects were much larger when the dependent variable was product choice and lower when the dependent variable was a rating scale.
In summary, when rating scales were employed as the dependent variable, the CO effect was stronger than the price effect; the intrinsic factors were stronger than the extrinsic factors for men's shirt quality and value, about the same strength for pickle quality and value, while the CO effect was much stronger than the intrinsic effect for smoke detector quality and value. These findings mirror those found for the analysis with product choices as the dependent variable but the R-Squares of these models were much lower than were the Phi Coefficients for the choice analyses.
EFFECTS OF INTRINSIC AND EXTRINSIC CUES, AND AGE, GENDER AND PERCEIVED ABILITY TO JUDGE PRODUCT QUALITY ON RESPONDENT RATINGS OF PRODUCT QUALITY AND VALUE
The effects of CO and price on New Zealander consumer ratings of product quality and value confirm an earlier small experiment conducted in New Zealand (Driscoll, 1989). Also, New Zealanders responded to CO and price cues (i.e. favouring home country and lower priced products) in a similar fashion to Canadian and American consumers as reported in many North American experiments (Liefeld, 1992).
This study reveals that when intrinsic cues are included in the experiment, and when product choice is employed as the dependent variable, the explanatory power of the models, as revealed by Phi Coefficients, is increased more than twofold in most cases compared to R-Squares attained in ANOVA's of product quality and value. Using product choice as the dependent variable provides greater magnitude of explanation, but it is important to note that this greater magnitude is not very high compared to rules of thumb for interpreting R-Squares or Phi Coefficients in which values below 0.30 are considered weak (Kinnear & Taylor, 1991). Even more interesting is the fact that the R-Squares found in these models for ratings of product quality and value were no higher than those reported in earlier experiments which did not include intrinsic factors, or other explanatory variables such as perceived ability to judge quality (Liefeld, 1992).
These findings reveal that, within a product category, the relative strengths of the various cues are about the same in both types of models,Bthose with product choice as the dependent variable and those with quality and value ratings as the dependent variables. These findings do not hold across product categories. For the shirts, in both types of models, the intrinsic factors were strongest, CO cues were relatively weaker, and price not important at all. For the smoke detectors the CO cues were strongest and the intrinsic and price cues were minor in effect. For pickles the intrinsic and CO factors were relatively equal in effect and again price was much weaker. Thus while the absolute level of explanation differs between models with product choice versus ratings of quality and value, the relative effects of the factors within the models are about the same, and appear to be conditional upon the product category.
The implications of the relatively low observed coefficients of determination for either type of model suggest that theoretical and methodological improvements are needed. Continuing to use measurement tools and perspectives which appear to be insufficient to the task at hand will continue to yield limited insights. Anecdotal evidence from the data collection sites in New Zealand suggests that alternative models of consumer behaviour may be at work. Some subjects were quite willing to choose a product without accessing extrinsic cues from the computer screen, suggesting that others may have accessed it because they thought they were expected to, rather than this being their typical shopping pattern. This suggests demand effects in the choice situation employed in this research design and challenges the presumption that consumers always pay attention to and consider extrinsic information cues in their decision processes. Other subjects commented that the separate ratings of quality and value were strange procedures to them as they did not think this way when shopping. It could be that New Zealanders are less familiar with rating scales than are North Americans. But the main issue this raises is the external validity of rating scale tasks. If consumers use a gestalt process for making choices in the marketplace and do not make individual judgments on each attribute for all choice alternatives, then rating scales may be measuring mostly how consumers answer rating scales, and not what the scales purport to measure. Measurement tool development is needed to better adapt the measurement processes employed to be consistent with the decision processes consumers undertake when making decisions. A more realistic experimental situation, such as placing respondents in a simulated shopping situation and manipulating the intrinsic and extrinsic information cues on the labels and shelf stickers, could provide higher mundane realism, but at the same time would make process tracing measures extremely difficult.
FACTOR LEVEL MEANS - (SCALE 0-10) SIGNIFICANT FACTORS
Baughn, C.C., & Yaprak, A. (1992 Forthcoming). Mapping the country-of-origin literature: recent developments and emerging research avenues. In Nicolas Papadopoulos & Louise Heslop (Eds.), Product and country images: research and strategy. Haworth Press, New York.
Driscoll A. (1989). Title to be added. Unpublished undergraduate thesis, Department of Marketing, University of Otago, Dunedin, NZ.
Ettenson, R., Wagner, J., & Gaeth, G. (1988). Country of origin and the "Made in the USA" television campaign: a rating and choice analysis of consumer decision making. Working Paper, University of Maryland.
Hong, S., & Wyer, R.S. (1989). Effects of country-of-origin and product-attribute information on product evaluation: an information processing perspective. Journal of Consumer Research, 16, 175-187.
Johnson, E., Payne, J., & Bettman, J. (1988). Monitoring information acquisitions and decisions: a process tracing methodology for decision research. ONR Tech Report.
Kinnear, & Taylor . (1991). Marketing research: an applied approach (4th ed.). New York: McGraw Hill.
Liefeld, J.P. (1992 Forthcoming). Consumer use of country-of-origin information in product evaluations: evidence from experiments. In Nicolas Papadopoulos & Louise Heslop (Eds.), Product and country images: strategy and practice. Haworth Press, New York.
Lodge, M. (1981). Magnitude scaling: quantitative measurement of opinions. Beverly Hills: Sage University Paper Series on Quantitative Application in the Social Sciences.
Nagashima, A. (1970). A comparison of Japanese and U.S. attitudes towards foreign products. Journal of Marketing, 34, 68-74.
Nes, E.B. (1982). Consumer perceptions of product risk and quality for goods manufactured in developing versus industrialized nations. Unpublished doctoral dissertation, University of Wisconsin, Madison, Wisconsin.
Rosenthal, R., & Rosnow, R.L. (1984). Essentials of behavioral research: methods and data analysis. New York: McGraw Hill.
Rule, S.J., & Curtis, D.W. (1982). Levels of sensory and judgmental processing: strategies for the evaluation of a model. In B. Wegener ( Ed.), Social attitudes and psychophysical measurement, Hillsdale: Erlbaum, pp 107-123.
Saris, W.E. (1988). Variations in response functions: a source of measurement error in attitude research. Amsterdam: Sociometric Foundation.
Schleiffer, R.D., & Dunn, W. (1968). Relative effectiveness of advertisements of foreign and domestic origin. Journal of Marketing Research, 5, 296-299.
Schooler, R. (1965). Product bias in the Central American Common Market. Journal of Marketing Research, 2, 394-397.
Schellinck, A. (1989). Determinants of country of origin cue usage. In Alain d'Astous (Ed.), Marketing (pp. 268-275). Montreal: Administrative Science Association of Canada.
Szbillo, G., & Jacoby, J (1974). Intrinsic versus extrinsic cues as determinants of perceived quality. Journal of Applied Psychology, 59, 74-78.
Wall, M., Liefeld, J.P., & Heslop, L.A. (1991). Impact of country-of-origin cues on consumer judgements in multi-cue situations: a covariance analysis. Journal of the Academy of Marketing Science, 19, 105-113.
Zeithaml, V.A. (1988). Consumer perceptions of price, quality and value: a means-end model and synthesis of evidence. Journal of Marketing, 52, 2-22.