Binding vs. non-binding treaties

David Victor (Global Warming Gridlock, 2011) suggests that non-binding treaties may, in at least some circumstances, be better than binding treaties. Looking at a set of treaties that have included both, he states,

The experiences with international cooperation to regulate pollution in the North Sea, the Baltic Sea and to regulate the emissions that cause acid rain in Europe are all examples where binding and nonbinding instruments were deployed side-by-side. In each, the nonbinding rules proved much more effective. (226)

Granted, binding rules are more likely to be complied with than non-binding rules:

...the few studies that have tried to measure compliance with binding international pollution agreements generally find that compliance is just about perfect. By contrast compliance with nonbinding accords is generally much lower. (228)

But nonbinding accords still net a better outcome:

Good news about compliance is often bad news about effect cooperation. When dealing with problems marked by lots of uncertainty, a pristine record on compliance is an indicator that governments are setting commitments too modestly. Looking just at compliance, binding accords almost always perform better than nonbinding ones. But when correcting for the content of the accords - which is nearly always more ambitious in comparable nonbinding accords precisely because there are fewer penalties for falling short - often the nonbinding accords perform better. (228)

In other words, half of an ambitious loaf is better than all of a conservative loaf.

I can't help suspecting, though, that this result is more reliable when most or all of the parties actually want to cooperate (ie, if they'd all actually like to promise more and are restrained from doing so by the potential for penalties if they overpromise and fail in a binding accord) than it is in a case where some or many of the parties may actually prefer to do as little as possible and free-ride on the efforts of others.