The Effects of Missing Information on Consumer Product Evaluations

Citation:

Deepak Sirdeshmukh and H. Rao Unnava (1992) ,"The Effects of Missing Information on Consumer Product Evaluations", in NA - Advances in Consumer Research Volume 19, eds. John F. Sherry, Jr. and Brian Sternthal, Provo, UT : Association for Consumer Research, Pages: 284-289.

Advances in Consumer Research Volume 19, 1992      Pages 284-289

THE EFFECTS OF MISSING INFORMATION ON CONSUMER PRODUCT EVALUATIONS

Deepak Sirdeshmukh, Ohio State University

H. Rao Unnava, Ohio State University

Previous research on the effects of missing information on consumer evaluation tended to present attribute dimensions to consumers and then inform them that the information on those dimensions is missing. We argue that the mere presentation of an attribute dimension is sufficient to induce inferences that might otherwise not have been made by subjects. The results of an experiment where subjects were presented only those attributes on which information was available show that missing information has little effect on consumer evaluations. Implications for research on missing information are discussed.

INTRODUCTION

Early models of consumer choice assumed that consumers use only explicitly available information in making choices (e.g., multi-attribute models). However, recent studies have shown that the inferences consumers make upon exposure to advertising information play an important role in consumer decision making (Kardes, 1988). Because consumers are more likely to make inferences when product attribute information is missing, much attention has been directed toward understanding how consumers' product evaluations are affected by missing attribute information (Meyer 1981, 1982; Yamagishi and Hill 1983; Dick, Chakravarti, and Biehal 1990; Huber and McCann 1982). Several recent studies have shown that consumers use different strategies to impute values for attributes about which information is not provided, when arriving at an overall evaluation for a product (Jaccard and Wood 1988; Johnson and Levin 1985).

Some recent research, however, has questioned the validity of the findings reported in the missing attribute literature. For example, Lim, Olshavsky and Kim (1988) and Ford and Smith (1987) suggested that the methodology used in most of the research on missing attributes may have produced effects which overstate the presence and amount of inference making. This criticism rests on the common practice in missing information research of asking subjects to first impute a value for the attribute on which information is missing prior to making an overall evaluation. Such procedures would force subjects into making inferences that they might not have made otherwise.

The present study represents a further attempt to examine the effects of stimulus presentation format and instructions on an individual's tendency to infer. Specifically, past studies, including those by Ford and Smith (1987) and Lim, et al. (1988), have examined the effect of missing attributes on product evaluation by presenting subjects with the description of a brand along a few attributes. The information on one or two attributes is left missing. Product evaluations in the missing attribute condition are then compared with the evaluations in the control condition where subjects have access to all information about the product.

We believe that the very presentation of an attribute or evaluative dimension, regardless of instructions to infer the value, may result in inference making. That is, the mere presence of an attribute name, even without any accompanying information, is sufficient to induce inferences. Therefore, even in studies that have attempted to measure inference making 'unobtrusively', the presentation of an attribute might have been sufficient to cause subjects to make inferences that they would normally not have made. Further, from a practical point of view, an advertiser is not likely to mention an attribute and then not provide any information about the product on that attribute. The objective of this research, therefore, is to examine the effect of missing information on consumer evaluations when consumers are not explicitly made aware of the attributes on which product information is missing.

A second issue concerns the number of dimensions on which information is missing. Most studies that have examined the effects of missing information have typically provided subjects with information on two or three attributes at most. In the missing information condition, subjects are deprived of information on one or two of those attributes so that subjects have information on only one product attribute (e.g., Johnson and Levin 1985; Huber and McCann 1982). Subjects' evaluations are then collected in these relatively information impoverished conditions. The likelihood of inference-making should be artificially enhanced under these conditions because information available for making an informed judgment is limited. This research attempts to examine the effect of missing information on consumer evaluations as a function of the amount of available information.

LITERATURE REVIEW AND HYPOTHESES

Several studies that have examined the effect of missing information on consumer evaluations have concluded that consumers rate products less favorably when confronted with partially described alternatives (Huber and McCann 1982; Jaccard and Wood 1988; Johnson and Levin 1985; Lim, et al. 1988). This negative reaction is attributed to consumers' tendency to discount the imputed value of the missing attribute. For example, Huber and McCann (1982) examined the extent and impact of inference making, and the implications of inferences for cognitive algebra. Subjects were asked to evaluate products on several dimensions, and information on some dimensions was missing. Inference making was either explicitly prompted or was implicit. Subjects in the explicit condition were instructed to generate a value for the missing information and then come up with overall evaluations. In the implicit condition, subjects were presented with identical stimuli and only asked to provide overall evaluations. The results showed that missing information affected purchase likelihood. When information was missing, subjects expressed lower likelihood of purchasing the target brand. Further, subjects' inferences had a significant impact on purchase likelihood with or without explicit instructions to infer a value for the missing attribute.

More recently, Ford and Smith (1987) examined the effect of explicitly prompting inferences. The authors suggested that studies examining inference making by asking subjects to provide a value for the dimension on which information is missing may be subject to demand artifacts. In their study, some subjects were directly prompted to infer the missing value (explicit condition) while others were given no such instructions (implicit condition). However, both groups of subjects were given identical stimuli where the value of the missing attribute was indicated with a question-mark. Results showed that the two tasks provided conflicting results. Prompting an inference produced results that differed in magnitude compared to those produced without such prompting. The authors discussed the inflationary effects of prompting inferences which call into question the effect sizes obtained in previous literature. Consistent with Ford and Smith (1987), another recent study by Lim, et al. (1988) reported the absence of any effects of missing information when subjects processed product attribute information under conditions that mimicked the implicit condition of Ford and Smith (1987) and Huber and McCann (1982).

In sum, while prior studies have reported significant effects of missing information on consumers' evaluations of products, recent research suggests that most such effects are exaggerated by the methods employed. The most serious criticism that has been leveled against prior research is focused on the practice of explicitly having subjects infer the value of a missing attribute before providing their evaluations. Under such conditions, it is argued, the obtained effects may be more due to the experimental task that requires subjects to make inferences rather than to the tendency in subjects to make spontaneous inferences before evaluating a product (Ford and Smith 1987).

Interestingly, even studies that have attempted to examine inference making without first asking subjects to impute a value to the missing attribute may have generated inferences due to the methods employed. This is because in all those studies (e.g., Lim et al. 1988) subjects are presented with an alternative where the dimension on which information is missing is also listed. Therefore, even in cases where subjects are not explicitly prompted to infer the value of the missing attribute, the very presence of the dimension, could generate some amount of inference making. A more unobtrusive strategy to examine the issue of inference making, therefore, would be to provide subjects with incompletely described alternatives where the attributes on which information is missing will not even be mentioned.

The procedure of omitting the attributes themselves, and not just the information about the attributes, should provide a truly unobtrusive examination of consumer inferences. If consumers are prone to infer the value of missing attributes, they should do so even when the attributes are not listed. On the other hand, differences in the pattern of evaluations between this condition and a condition where the attributes are listed without providing values would suggest that the methodology that has been adopted in more recent research (e.g., Lim et al. 1988) is inducing inferences that would not normally occur or inflating effect sizes.

A second issue concerns the number of dimensions on which information is available when subjects make judgments. As discussed earlier, past studies have measured the effects of missing information under conditions where the available information was unrealistically scarce. For example, in the Lim, et al. (1988) and Huber and McCann (1982) studies, subjects were presented information on only one attribute, with information on the other attribute missing. We believe that having subjects judge a product on the basis of only one attribute would greatly exaggerate the effect of missing information. Subjects should be more motivated to make inferences because they have very little information to make judgments. Thus, by providing little information on the experimental products, past studies might have inadvertently inflated the effects of missing information.

The research reported in this paper addresses the two aforementioned concerns with the existing research on missing information. First, we provide information to subjects on only those attributes for which attribute values are available. Subjects are not made aware of the other attributes on which information is missing (implicit condition). This condition is contrasted with the condition where subjects are made aware of the attributes on which information is missing (explicit condition). Subjects, however, are not asked to impute values to the missing attributes before providing product evaluations. Thus, the explicit condition in this study is similar to the implicit condition in some of the past research (e.g., Ford and Smith 1987). Second, the amount of information available to subjects is systematically varied so that subjects have information on a minimum of four attributes and a maximum of ten attributes before a judgment is rendered. It is expected that missing information affects product evaluations more strongly in the explicit condition than in the implicit condition. Further, it is expected that the effect of missing information on evaluations would decrease as the amount of available information increases.

METHOD

The effects of information presentation format and the amount of available information on subjects' evaluations were investigated in a 2 (explicit vs. implicit presentation) X 4 (information available on ten, eight, six or four attributes) between-subjects design. 192 subjects recruited from an undergraduate marketing class participated in the study for bonus points in their coursework. All subjects were randomly assigned to the treatment conditions.

STIMULUS MATERIALS

Color television sets were used as the target product in this study. This product was chosen because it was relevant to the student population and the task of evaluating televisions was not novel for them. Further, a product like television is normally evaluated on several attributes before a choice is made. Therefore, the presentation of information on several attributes is neither unusual nor unexpected for the student subjects.

The attributes on which information was presented in the final study were derived from a pretest. Forty six students from the same subject population were asked to list the most important attributes that they would be seeking information on while choosing a television set for themselves. The ten most frequently listed attributes by these subjects were selected for the final study. The attributes generated from the pretest are listed below in order of their importance as defined by the number of times an attribute was mentioned.

1.Screen Sizes Available

2.Sound Quality

3.Remote Control Quality

4.Color Quality

5.Picture Quality

6.Warranty

7.Cable Ready

8.Style and Appearance

9.Service Facilities

10.Hook-up to VCR, etc.

Although price was one of the most frequently mentioned attributes, it was not included in the study because price could have either an allocative role or could be a proxy indicator for quality (e.g., Lim et al. 1988). It is also possible that price could overwhelm the other attributes in affecting quality ratings which would affect the results of this study.

PROCEDURE

Subjects were given a printed questionnaire which contained instructions, manipulations and dependent measures (see appendix).The amount of information available and the explicitness of missing information was manipulated through the presentation of the consumer agency ratings. For example, the appendix displays the questionnaire used in the explicit presentation condition, with information available on 6 attributes. In the implicit presentation condition, with information available on 6 attributes, the names of the 4 missing attributes were not included in the questionnaire. Other conditions were manipulated accordingly. Subjects' overall evaluation of the television brand was measured by two seven-point scales (7= very good, very desirable; 1= very bad, very undesirable). Subjects' perception of the adequacy and sufficiency of the presented information was also collected. Two seven-point scales (very adequate, very sufficient vs. very inadequate, very insufficient) were used for this purpose. This measure was incorporated to check the missing information manipulation. If missing attributes were stated explicitly, thus drawing subjects' attention to those attributes, then subjects in the explicit condition should perceive greater inadequacy of information in the explicit than in the implicit condition. After completing this measure, subjects were debriefed and dismissed.

RESULTS

Attitudes

The two items constituting the attitude scale were averaged to form the mean attitude score for each subject. The cell means are reported in the Table. An overall ANOVA on the attitude score with presentation explicitness and information availability as the independent variables revealed a significant main effect of presentation explicitness (F=31.3, p < 0.0001). Missing information negatively affected subjects' evaluations more under the explicit condition (M= 4.49) than under the implicit condition (M= 5.29). This finding supported our expectation that the very presence of an attribute name enhances the effect of missing information on subjects' evaluations.

The effect of information availability was not significant (p < 0.15). However, the interaction between presentation explicitness and information availability was significant (p < 0.001). Subjects evaluations were affected more by presentation explicitness when the amount of available information was only four attributes (Ms=5.5 vs. 3.7) than when it was more than four attributes (Ms = 5.21 vs 4.73). This finding offers preliminary support to our expectation that the effect of missing information is exaggerated when the available information is limited.

A series of a priori contrasts were then conducted to examine the attitude effects in more detail. The contrasts revealed that missing information had no effect on subjects' evaluations (all p-values > 0.25) when they were not specifically told that the information on several attributes was missing (implicit condition). That is, subjects' evaluations of the target product remained unchanged whether they had information on four, six, eight, or ten attributes.

The results, however, were different when subjects were exposed to the names of the attribute dimensions on which information was missing. Subjects' evaluation of the target product was significantly less positive when information was available on only four attributes versus when information was available on six, eight, or ten attributes (all p-values < 0.05). Interestingly, even under explicit presentation conditions, subjects' evaluations remained unaffected by missing information when the amount of available information was at least six attributes (all p-values > 0.1). The absence of effects of missing information on product evaluations even under explicit presentation conditions, when the amount of available information is sufficiently large, suggests that previous studies might have reported inflated effects of missing information when they provided information on only one or two attributes.

TABLE

CELL MEANS AND RESULTS OF CONTRASTS

Information sufficiency

Subjects' perceived sufficiency of the available information in making their evaluations was measured using two seven-point scales. These two measures were averaged to form a mean perceived sufficiency score for each subject. It was expected that subjects' perceptions of information insufficiency would be greater when the stimulus material explicitly states that attribute information is missing than when the stimulus material does not.

As expected, the overall ANOVA on the perceived sufficiency score with presentation explicitness and information availability as the independent variables showed a significant main effect for presentation explicitness (F=26.33, p<0.0001). Missing information negatively affected subjects' perceived sufficiency of information to a greater extent in the explicit condition (M=3.29) than in the implicit condition (M=4.41). Further, a significant main effect of information availability was observed (F=22.12, p<.0001) suggesting that subjects' perceived sufficiency of information systematically decreased as the amount of available information was reduced (Ms = 4.85, 4.36, 3.60, 2.50 for ten, eight, six and four attributes respectively). Thus, while all subjects were sensitive to the variation in the amount of information available to them in making their evaluations, the amount of information manipulation exerted stronger effects in the explicit condition than in the implicit condition.

DISCUSSION

Much of the past research on missing information and inference-making has adopted the common procedure of presenting subjects with the product attributes on which information is missing, along with the attributes on which information is available, and then measuring subjects' evaluations of the product. The results of this past research suggest that consumers have less favorable evaluations of products when attribute information is missing than when it is available. Theories of how consumers treat missing information when arriving at an overall evaluation have been developed based on these findings.

In our research, we contended that the very presence of a missing attribute name in the information set presented to a subject is sufficient to induce inferences about that attribute. It was our belief that when subjects are not cued to a particular attribute, the probability of inference-making about that attribute is reduced greatly. The reason for the curtailment of inference-making is the reduced probability of a consumer dredging the attribute from memory, discovering that information on that attribute is missing, and then imputing a value for that attribute.

The results of our experiment were supportive of our expectations. Consumer evaluations of a television brand were virtually unaffected when consumers were presented missing attribute information implicitly (missing attribute names not presented along with the attributes on which information is available). On the other hand, explicit presentation of missing attributes (missing attribute names presented along with the attributes on which information is available) caused subjects' evaluations to be more negative when compared to the full information condition. This seems to suggest that individuals either simply chose to ignore the missing information, or, even if an inference procedure was at work, it differed markedly from the one used when the dimensions were listed and values were missing.

The effect of missing information was also found to depend on the amount of information that was available to the subjects when they made their evaluations. When subjects had information on only four attributes, with information on six attributes missing, their evaluations were significantly less positive than when subjects had information on six, eight or ten attributes. This effect, however, was limited to the explicit presentation condition.

The finding that missing information did not affect subjects' evaluations when subjects were not explicitly told that the information was missing is important because it casts doubts on the robustness of the findings reported in previous research.

APPENDIX

Specifically, it suggests that unless subjects are actually told that information is missing on an attribute, they do not seem to be affected by it.

One possible alternative explanation for this insensitivity of subjects to missing information might be their lower level of task involvement. Stated simply, given the task environment, subjects may not have bothered to think of the information that was required to make an evaluation and therefore exhibited no sensitivity to missing information (e.g., Kardes 1988). However, this argument is not supported by the finding that both in the implicit and explicit conditions, subjects' ratings of the adequacy and sufficiency of the information presented to them was affected by the experimental manipulation of missing information. It thus appears that subjects were aware of the fact that the information given them was not complete. Their evaluations, however, were affected only when they were cued with the attribute names on which information was missing.

Another possible explanation for the results may be the "recall inhibition" effect (Alba and Chattopadhyay 1986). The recall inhibition explanation would suggest that the presence of the available attributes may have interfered with the retrieval of other attributes. Consequently, given this inability to recall, subjects may not have been sensitive to the missing attributes. Such a possibility may bear further examination.

We would also like to note here that previous research on missing information has employed experimental manipulations and subject pools (students) similar to those used in our study. Hence we are comfortable comparing results of our study with those found in the studies reviewed earlier.

Our findings also question the practice of employing stimuli that typically provide information on only one attribute with information on the other attribute missing. As argued earlier, the potential for artificially inducing inferences is very high under these conditions. The finding, in this research, that subjects are relatively sensitive to missing information when they are provided information on only four attributes, but not when they have information on six or more attributes, demonstrates that the effects of missing information are conditioned on the amount of available information. The findings reported in several earlier studies on missing information, then, have to be interpreted with caution in light of our findings.

Much of the past research on the effects of missing information on consumer evaluations has concluded that consumers evaluate products less positively when information is missing. Some recent research has questioned the methodology adopted in this literature and has cast doubts on the robustness of the findings reported in this literature. Our research extends this criticism by showing the relative insensitivity that consumers seem to exhibit toward missing information, in arriving at product evaluations, when the missing attribute information is not explicitly stated. Future research should examine the conditions under which consumer evaluations exhibit sensitivity to missing information.

REFERENCES

Alba, Joseph W., and Amitava Chattopadhyay (1986), "Salience Effects in Brand Recall," Journal of Marketing Research, 23 (November), 363-369.

Dick, Alan, Dipankar Chakravarti,and Gabriel Biehal (1990), "Memory-Based Inferences During Consumer Choice," Journal of Consumer Research, 17 (June), 82-93.

Ford, Gary T. and Ruth A. Smith (1987), "Inferential Beliefs in Consumer Evaluations: An Assessment of Alternative Processing Strategies," Journal of Consumer Research, 14 (December), 363-371.

Huber, Joel and John McCann (1982), "The Impact of Inferential Beliefs on Product Evaluations," Journal of Marketing Research, 19 (August), 324-33.

Jaccard, James and Gregory Wood(1988), "The Effects of Incomplete Information on the Formation of Attitudes Toward Behavioral Alternatives," Journal of Personality and Social Psychology, 54 (4), 580-91.

Johnson, Richard D. and Irwin P. Levin (1985), "More Than Meets the Eye: The Effect of Missing Information on Purchase Intentions," Journal of Consumer Research, 12 (September), 169-77.

Kardes, Frank R. (1988), "Spontaneous Inference Processes in advertising: The Effects of Conclusion Omission and Involvement on Persuasion," Journal of Consumer Research, 15 (September), 225-233.

Lim, J., Richard W. Olshavsky, and John Kim (1988), "The Impact of Inferences on Product Evaluations," Journal of Marketing Research, 25 (August), 308-316.

Meyer, Robert J. (1981), "A Model of Multiattribute Judgments Under Attribute Uncertainty and Informational Constraints," Journal of Marketing Research, 18 ( November), 428-441.

Meyer, Robert J. (1982), "A Descriptive Model of Consumer Information Search Behavior," Marketing Science, 1 (Winter), 93-121.

Yamagishi, Toshio and Charles T. Hill (1983), "Initial Impression Versus Missing Information as Explanations of the Set-Size Effect," Journal of Personality and Social Psychology, 44(5), 942-951.

----------------------------------------