I have a son-in-law who loves steak but abhors the consistency of burgers. He would be a 12-none guy.

Given that only 16/800=2% of your sample did this, it would indicate to me that the "didn't understand instructions" possibility is not a realistic one. I would put the weight of evidence on this being "real" behavior, not something task-induced. I'd say you should include them, and better yet, think about what this phenomenon could be representing in terms of decision process. Maybe exploring this could be your next project?

Please, refer to my article

Opt-out Forced Choice Effect in Combined Revealed and Stated Preferences Discrete Choice Models: Testing Gender-specific Convergent Validity.

All the best

I agree with Joffre Swait's suggestion not to exclude observations. Your study is telling you there is something to be explored and explained in your sample population. Their response could be a genuine reaction either induced by the study design or actual preferences that need to be resolved.

1. Yes, exclude those 9 respondents, who always selected 'None of These', simply because they don't have utility and resultantly 'protest'.

2. Besides, estimate models excluding and including them as the number of protesting respondents as you mentioned is only 9, which quite a low number if we review the previous literature.

3. Why didn't you apply genuine/non-genuine protesting respondents as explained in one of Jacob Ladenburg's (a Danish Economist) to reduce the protests, which are already low in number. For instance, Discrete Choice Analysis makes researchers worry when protesting respondents make adequate enough proportion of the total sample, say 20%, although there is no hard and fast rule !

I hope this helps. Regards and best wishes.

Why are we all arguing out of the blue, arguably because the researcher didn't ask the respondents about the reasons behind their 'opt-out' behaviour. Instead, s/he asked this question very straightforwardly that whether these respondents may be included or excluded in the analysis! ! . . . Because of this research mistake as it seems to be, all the researchers answering this question as it may look like are going into the different directions!! :-)

Dear Julian Sand,

Serial non-demanding behavior is neither an indicator for 'irrational' response, neither an indicator for straight-lining. Excluding them would bias your predictions (e.g., market shares).

Straight-lining behavior could be inferred from participants who always choose Alt A, B or C in your (first) forced question.

In any case, this should be confirmed by looking at other indicators such as response time, or variability in post-choice (un)certainty, or any other additional information you collected regarding survey understanding or commitment.

I agree with Joffre Swait , investigating determinants (or decision process) of serial non-demanding is always useful. Might be something to think about during pre-testing, e.g. using think aloud methods.

Best wishes,

Jonathan

In one of my choice experiment study on beach recreation, only 3 respondents selected 'none of beaches'. Those all 8 alternative beaches are free public goods with no entrance fee. So, monetary attribute I selected was an Entrance Fee, even with no entrance initial level. When I afterwards asked them about the reason for not visiting any of the alternative beaches, they said we are 'not WTP'. When I further asked them, they replied it's government's responsibility to arrange financial resources from various types tax revenues. And, I forced they strongly objected and said they are still not WTP. Hence, I excluded those 3 respondents. Very simple. We must ask about the reasons behind protest or zero WTP behaviour. Very simple.

There exists different techniques to assess 'none these' behaviour while conducting Discrete Choice Experiments;

1. Reasons behind choosing 'None of these', e.g. probably because of no-monetary attribute or payment vehicle (i.e. a protest),

2. Non-attendance of (or ignoring) one or some attributes,

3. Choice certainty scale (1 - 10), and

4. Entreaty explained to the respondent to re-consider his/her choice.

The above all four techniques are applied by the different researchers, according to the Lietrature Review.

However, without applying those recent techniques in case of diagnosing 'none of these' issue and rather only support it with a bunch of assumptions will bias the results. For example, no matters even its about calculating market share ! !