Something
Brad DeLong explained to me which is not totally obvious
In standard general equilibrium theory, the market outcome is Pareto efficient. One of the standard assumptions is that people are totally selfish and care only about their own pleasure minus pain due to consumption. One might wonder if the result can be generalised to the case of people who care twice as much about their own pleasure minus pain due to consumption as that of others. I think that, decades ago, I guessed that assuming that people could give to others would imply the result that free trade and free giving would lead to a Pareto efficient outcome. If I did, I was wrong. The assumption of total selfishness is required for the conclusion that Laissez Faire is Pareto efficient.
If people are altruistic but not 100% selfless then coercion in helping the poor is Pareto improving (Brad explained this to me by the way). Let's say we care about strangers one half as much as I care about people in my family. This is an immense degree of atlruism (my poor kids would be living at lower third world levels if I did). However, at the point when I am indifferent about giving more, you would want me to give more to the poor. The reason is that you don't care any more about me than a poor person.
This is true even if one of the reasons I want to give to the poor is that I know it will make you feel better (and I care about you too).
This means that we would both be happier if we were both coerced to give more than we want. The poor, needless to say, would be much happier.
If people are altruistic, then there is an externality from giving (you give to the poor so I feel better because the poor are less poor). This means that the amount of giving optimal for the rich is more than we would choose acting as individuals.
Charity with no is Pareto efficient only if people are totally selfish or if they are totally selfless caring no more about their flesh and blood than total strangers.
