A COMPARISON OF RATING AND CONJOINT ANALYSIS METHODS TO MEASURE CONSUMER PREFERENCES FOR HEARING AID ATTRIBUTES

Purpose: Low utilization of hearing aids among those who would receive clinical and quality of live benefits from using them has drawn researchers to the study of consumer preferences using both ratings and conjoint analysis methods. We compare valuations of hearing aids attributes from Likert-type rating and conjoint analysis methods.

Methods: Following a literature review, qualitative analysis (n=15) and pilot (n=25), 75 outpatient with hearing loss were presented with seven attributes of hearing aids: performance in quiet settings, comfort, feedback, frequency of battery replacement, purchase price, water and sweat resistance, and performance in noisy settings. Preferences were assessed using a five point Likert-type scale and via eight paired-comparison conjoint tasks designed from a main-effects minimum orthogonal design. To make the two methods more directly comparable, attributes were described with explicit levels (rather than left unbounded) before the rating tasks. Conjoint analysis results were estimated via logistic and linear probability models (LPM). Willingness-to-pay estimates were compared using Pearson’s correlation. 

Results: Of the 75 respondents, 56 (75%) provided complete responses. Respondents tended to be male (66%), aged 51-70y (46%), well educated (43% has some college), wealthy (48% had income over $100,000), and 44% of respondents had never had a hearing aid. All methods identified performance in a noisy setting as the most preferred attribute (with a WTP ranging from $2,674 to $9,000), with all results presented in figure one. Overall the correlation between the ranking and conjoint methods were high, but were higher for the logistic estimates (P=0.03) than the LPM (P=0.06), yet there was little difference between the logistic and LPM estimate (P=0.002).

Conclusions: We find a high level of concordance between the valuations estimated using rating and conjoint analysis valuation techniques – a result that is in stark contrast with the previous literature.  We conclude that this result stems from our constraining the rating of attributes to between defined levels.  While conjoint analysis has many benefits, including the absence of ceiling and floor effects common in rating-based elicitation, previous studies may have exaggerated differences between the two methods when leaving levels unbound. Future studies need to differentiate between “attribute importance” and “level importance” and aim to estimate valuations over policy relevant ranges. Figure 1: Willingness-to-pay estimates for hearing aid attributes