GiveDirectly, a charity that focuses on providing direct cash assistance to poor people in developing countries, decided to ask its team (or at least those members who were from or had spent time in the United States) if they do the same thing on the home front: do they stop to help homeless people soliciting money in their own cities?
2 people said they give most of the time, 5 said they did so sometimes, and 4 said they never did; you can see their full explanations here. Three of the four stating they never give cited the greater cost-effectiveness of giving to poor people abroad rather than the US, which is hard to argue with; aid money goes a lot further in developing countries than it does in rich ones like the US. The fourth argued that homeless people soliciting money might be more likely to misuse it than the rest of the homeless population. The two who said they did it usually noted that "the money is likely to be of more use to them than it is to me" and that the recipients as "likely to spend it as wisely as I could," respectively.
The five sometimes-givers gave a range of explanations , but I found this one particularly compelling:
I don't feel confident giving directly to the poor in the US without more evidence akin to what we have in most of the developing world, given the huge differences in culture and poverty. But, I give from time to time to make sure that saying "no" too many times doesn't harden my basic compassion for people who are suffering, regardless of why, or of what impact my money will have.
Besides helping sustain an altruistic spirit in the donor, there are two other factors that agitate in favor of giving to homeless people in the US. First, though it's rare, extreme, developing world-style poverty exists in the US. Harvard's Kathryn Edin and Michigan's Luke Shaefer found that more than 600,000 households lived on less than $2 a day per person in 2011, even after you take food stamps and other benefits into account.
Second, Penn State's Barrett Lee and the University of Alaska - Anchorage's Chad Farrell found in 2003 that panhandlers are worse off than the overall homeless population; they report lower incomes and less government aid, and are less likely to report working in the past month. They're also likelier to have substance abuse or mental health problems, and the data is a bit old, but this at least serves as suggestive evidence that money given to homeless people who are actively soliciting it might be well-targeted. Of course, it's impossible to say whether it's so well-targeted that the people being helped count as the "extreme poor" identified in Edin and Shaefer's study.
That said, a key difference between US poverty and homelessness and poverty abroad is that the government responsible for tackling the former is extremely wealthy and could easily make major strides if it chose to do so. For instance, it turns out that simply giving homeless people housing is a fairly cost-effective way of reducing it, and the US could absolutely afford to end or sharply reduce chronic homelessness that way. The government of Tanzania doesn't have those kinds of resources, which makes non-governmental intervention a significantly more crucial tool in poverty reduction efforts. That might be a reason to direct your donations toward GiveDirectly rather than to the US homeless population, while pushing for the US government to take the task of ending poverty more seriously.