Confession: I have on occasion contracted the services of someone to help clean the RV. Earlier, I had some qualms about doing this, but those subsided when I realized how many colleagues I know (in sociology and elsewhere) hire cleaning help, including colleagues who have fewer square feet of living space than I do. She is, like many people in her line of work, a woman of color and low socioeconomic status. While there may be wonderful souls out there who would clean my house for free, this woman is not one of them. Instead, when I asked how much it would cost, she said $40. However, it is possible, had I played hardball, I could have told her that I wasn't paying more than $25 and she still would have done it, as she does need the money. As turns out, I actually pay her $50 because, for whatever reason, that feels more like the right compensation to me (the job takes her a little more than 2 hours to do).
Meanwhile: Some fellow scholars are currently involved in a project that will involve surveying many persons of color, a large portion of whom will be of low socioeconomic status, like the woman who cleans my RV. The interview is long; for complicated reasons, it may take some people 2 hours to complete. My colleagues would like to pay respondents $50 to do this survey. However, it is very possible that this plan will run into trouble from their IRB* by virtue of being judged as potentially "coercing" poorer respondents to do the survey. The idea being that if respondents do the survey because they feel like they can't pass up the money, then they are effectively being coerced into participating and so they are not really consenting research participants.
My colleagues will be much less likely to have these problems with the IRB, they've been told, if instead they only ask to pay respondents $25, since $25 is less money and so less coercive. Consider the following scenario: the interviewer tells a low-income respondent that she will receive $25 if she does the survey; she says she will do it for $50; the interviewer tells her that $25 is the most she can get; she agrees to do it for $25 since, in her situation, any money is better than nothing. In the IRB best-case scenario, my colleagues would only use respondents who were willing to do the survey for nothing, but IRBs do recognize that people will often not part with two hours of their time if some compensation is not involved.
If you asked the woman who cleans my RV whether she would rather get $50 by cleaning my RV or by sitting and doing a survey, I would not be surprised if she chose to do the survey. However, in the view of many IRBs, it is unethical for her to be tacitly given this choice.
I should note that when the woman cleans my RV, I'm not really fussy about whether she does every little thing that someone might expect as part of her job. However, I'm sure there are people who are far more demanding and insist on everything being done properly before the person is paid. There could be tasks that she really dislikes doing, but she has to do them if she is going to get her $50. On the survey, meanwhile, compensation does not depend on the person answering every question. Instead, the person can refuse to answer any question that makes her feel uncomfortable or which, for whatever reason, she'd rather not answer, and she will still get her $50.
Curiously, IRBs would likely have less problem with offering respondents $50 to do the survey if the respondents had higher incomes. The common view is that the less people can use the compensation that researchers might be able to provide them, the more compensation you can ethically give them.
(I might add, incidentally, that I would understand the logic of the "coercive" argument better--which is still not to say I would necessarily be convinced of it--if we are talking about research that places the subject at some kind of risk, as some medical research does. Offering someone money they desperately need to assent to research that implies a health risk is a different ethical matter. Doing a survey does not add any risk to health or well-being. It is true that surveys sometimes ask people questions they don't want to answer, but, again, respondents can be told and re-told during the interview that they can refuse to answer any questions they want and they will still receive their compensation.)
* [I]nstitutional [R]eview [B]oard