This is an important problem in jurisprudence. As Richard Epstein says in Skepticism and Freedom, "Classical liberalism requires us to maintain the distinction between liberty and coercion, to advance the former while constraining the latter."
The paradox of liberty and coercion was posed in its most salient form by Robert Lee Hale. Hale's key philosophical move was to assume radical skepticism about the merits of competing claims to scarce resources. Hale thereby demonstrates how any distinction between liberty and coercion must be one of degree. For Hale, in a world of scarcity, we are "free" only to choose among different forms of coercion, if that.
Epstein suggests that the weakness in Hale's argument originates in the lack of distinction made between individuals and groups under Hale's definition of "coercion." "When a manufacturer is unable to sell his wares," asks Epstein, "which of the millions of its potential customers is guilty of coercion and why?" On my view, Epstein is correct in his judgment that any coherent theory of liberty and coercion must stay consistent as it moves from the scale of individuals to the scale of large groups.
But Epstein appears to side-step the problem of defining liberty or coercion by taking the pragmatic route of merely observing that most refusals to deal will not, in fact, be the result of coercion. In effect, Epstein's answer is that Hale has a good philosophical point, but we shouldn't get too cute with our philosophy when we're talking about ordinary, everyday commerce. Our natural moral intuitions are often correct, regardless of whether they can be analyzed through any coherent framework of moral philosophy.
I like the pragmatic, instrumentalist strategy that Epstein follows. I can't see how we'd make it through life without it. But I want here to try to push out the edge of the philosophical theory of liberty a bit further.
The best jurisprudential definition of liberty that I know is found in Judge Richard Posner's Problems of Jurisprudence. In this book, Posner offers the theory (which he attributes to Quine) that a person's choice is "free" so long as her preferences are a link in the chain of causation that leads to an act. According to Posner then, coercion arises when a person is forced to act either before their preference can be formed or without their preference being part of causal chain that leads to the act.
Posner's definition is very useful because it suggests a judge-invariant method whereby coercion may be identified. Judges need not hesitate in condemning coercion when defendants can show that their act did not or could not reflect their rational preferences. Perhaps most importantly, there is no need to peer into the hearts or minds of plaintiff or defendant in making this judgment.
But as Posner points out this definition leads to several paradoxes, such as that of the well-supplied drug addict, the overgrown child, or animals. In each case, the preferences of these actors would seem to be a link in the causal chain that leads to their action. Yet in each case, at least some people will suspect that these are not truly free actors.
My humble suggestion is that some of these paradoxes may be resolved by considering, in addition to rationality, the periodicity of the preferences of these actors. Rationality and periodicity are orthogonal in this analysis. An act can be rational, but ill-timed; conversely, an act can be well-timed, but irrational. An actor becomes more "free" when she optimizes the fit of an act to both her self-interest and her rhythms.
For example, the drug addict indeed prefers to use drugs, and with an increasing frequency as the addiction becomes stronger. But so long as we know that the rational preference of the drug user to use more drugs is not sustainable at increasing frequencies of consumption, then we may question -- again without peering into the mind or heart of the drug user -- whether, in fact, their decision is "free." Similarly, we may ask whether an overgrown child is free who does not begin to engage in work and social activities with an increasing frequency beyond some point in time.
On the other hand, after adding considerations of periodicity to the definition of liberty, animals remain "free," because like humans, their preferences display patterns of periodicity. This observation has applicability to infants and the mentally ill as well.
Judge Posner implicitly acknowledged the importance of periodicity in his discussion of coerced confessions and the defense of duress. In each case, the rational preference of the actor is part of the causal chain that leads to the act; but the act is "forced" because it occurs before the actor would act independent of coercion. The temporal dependence of the argument hints at the role of periodicity.