Mark Kleiman has an interesting post on America's lousy record in training for high skills jobs. The problem is straightforward - employers have little incentive to invest in training low skill workers because these workers tend to move from job to job, taking their (expensively acquired) skills with them. Thus, there's a collective action problem. Even if a skilled workforce is in the interests of all employers, each individual employer has an incentive to underinvest in skills training, because she doesn't capture all the benefits of improving workers' skills. Kleiman argues that in a world where we don't have guild apprenticeships, indentured servitude or slavery, the obvious way to fix this problem is through financial incentives; give employers a chunk of low-skilled workers' future social security contributions, which would allow these employers to recapture some of the investment that they had made in training. It's an interesting proposal, but there are other ways to tackle this problem, which have a stronger track record - they've been shown to work in practice.
First - you can oblige employers to contribute collectively to skills and training, through industry level associations that work in cahoots with the state. This works pretty well in Germany, as Wolfgang Streeck has shown in a series of publications. Firms sometimes need to be compelled to do what is in their collective interest, precisely because it is their collective, rather than individual interest which is at stake (this is the basic point of the collective action literature). Second - you can institute a strong welfare state. This doesn't affect the incentives of firms so much as the incentives of individuals. Individuals are more prepared to invest heavily in job and industry specific skills, because they are less worried about what is likely to happen if their firm fires them, or their industry goes into recession. As Torben Iversen and his colleagues show, there is strong empirical evidence that specific skill training goes hand in hand with strong welfare systems. The rub, however, as Iversen et al. point out, is that economies which favour generalist training (such as broadly focused university degrees) do some things better than economies with strong vocational training. The former tend to adapt better to rapidly changing economic circumstances, and to produce more in the way of radical innovation, while the latter take better advantage of stable economic conditions, and provide better incremental innovation.
Update William Sjostrom comments on Kleiman's post (as an economist, he's naturally sympathetic to the idea of market incentives, if uncertain about whether the incentives will actually do what they're supposed to).