Doug is right. I've spent a lot of time on experimental design (just landed a PhD) and what you're suggesting is occasionally called a "convenience sample."

The big problem with it is this: not everyone drops receipts. So you won't be learning about people in general, just receipt-droppers in general.

However, if you can accept that, then you're in a good spot. Since you're measuring receipt-droppers and only receipt droppers at every ATM machine, you can compare what happens between them.

So, what you'd be asking is, what can you learn from receipt droppers in a poor neighborhood vs. a rich one? And do you have a good reason to believe that what you learn is true of the larger population of humans that also includes non-receipt droppers? You'd have to make a case for the second but I think the answer is yes.

You also have their privacy to consider, which I think is the more important question by a very wide margin. I suspect the dropped ATM slips you're referring to don't have personally identifiable information (if they do, probably don't look at them, period) but at the same time they are still personal, and for people to feel comfortable in society they need to have some faith that their small mistakes and irresponsibilities are carried away by the wind.

Many ethics review boards also care a LOT about getting explicit permission from participants to use their data. I think your anonymous dropped ATM receipt approach would be a hard sell.

Also you're not weird. You've just got an eye for data, like a good social scientist. Lots of people won't understand it or respond because they're not trained and it's an area of life they're unfamiliar with. If you want genuine feedback on your idea try this; pick a good college and find one of their social sciences departments (Economics and Social Psychology are good choices). Find a professor who looks approachable and ask them about your idea, and if you want a wide round of feedback ask if you can also talk to the graduate students in the department. Your idea sounds exactly like the creative approach one of them would use for a dissertation or a publication (with the caveat that it may be too intrusive to be feasible in academia).

For comparison I've been tossing about an idea that you can learn something meaningful about the distribution of commerce in a city by taking restaurant menus from a random sample of restaurants and collapsing each one to an "average menu price" that could be compared across neighborhoods. I'll probably never go anywhere with it, but if you speak with data analysts and social scientists you will find that quite a few of them think about data collection in that slightly offbeat way.

And if you're looking for a book that will resonate with you I highly recommend looking up Sam Gosling. His book "Snoop" is about all the amazing things we can learn about people using what he calls "behavioral residue"--which is the category your dropped ATM slips fall into.

For example, you can tell which display in a long-standing museum exhibit is the most popular just by glancing at the wear-and-tear on the floor.