Union membership and support for Bernie Sanders in 2016 election

The 2016 Election
The Research Question
My question regarding the 2016 election is this: why did Americans vote for Donald Trump? This is a decidedly open-ended question, which means it would involve exploratory research rather than the testing of a hypothesis. The research could consist of a mixed-methods design with both qualitative and quantitative research conducted. Individuals could be asked both open-ended questions and closed-questions (with Likert-scale measures in place).
How I Would Answer This Question
A good means of collecting data for exploratory research is the interview method or the survey method (Leonard, Noelle et al.). First, I would break up the voting sample into demographics: (a) gender, (b) age, (c) voting history (Party affiliation), (d) region of the country, (e) religion, (f) ethnicity/race, and (g) ideology (ex., conservative, liberal, libertarian, reform, etc.). Then I would seek to obtain a sample of participants that could provide me with a significant set of data that I could draw reasonable conclusions as to why different voters voted for Trump. I would like to post a survey online using social media—Facebook and Twitter—and also obtain individuals for interviews using social media as well. The interviews would provide me with a qualitative data set and the surveys would provide me with a quantitative data set.
Evidence I Would Need to Obtain but Cannot Obtain
Self-reported data is not always the most objective and does not always offer the most empirical of proofs. It is not like observing the results of test and obtaining clear evidence. In other words, the objective data that I would like to obtain would primarily still be highly subjective because it would rely on individuals stating their reasons for voting for Trump. Sometimes people fail to articulate their reasons sufficiently or might hide their true reasons for fear of being judged. The type of objective, clear-cut evidence I would like for this study would thus be impossible to obtain by conducting interviews and surveys. The data would still be considered subjective because of the self-reported nature of the data set. The perfect evidence to support my argument would be an empirically-provable data set where the data is completely objective, valid and reliable—but there is no clear way to discern this data set, as human motivation is a highly subjective subject that can only be explored. It can only be described in qualitative terms, even if quantitative data (as from a Likert-scale survey) were used to rank responses.
Whether There is a Difference between the “Perfect” Evidence and What I Can Actually Obtain
The difference between what I could actually obtain and the “perfect” evidence is that the former is basically subjective and dependent upon self-reported data while the latter would remove the subject from the reporting process and depend solely on a method of data collection that is objective. The difference is such that it could affect my argument by challenging the validity and reliability of my research. A study that is based solely on the subjective responses of individuals will lack quantitative merit. However, as I would seek to combine qualitative with quantitative data, I could at least provide a bit of quantitative merit, even if the data is still relying on self-reporting. As there is no other way to understand the motives of voters, this mixed-methods approach is at least a way to develop a theory.
References
Leonard, Noelle R., et al. “A multi-method exploratory study of stress, coping, and
substance use among high school youth in private schools.” Frontiers in Psychology 6 (2015): 1028.