Marx's distinction between productive/unproductive labor does not map out in terms the division between mfg and service employment.
However, there seems to be a big wage gap between those who work in mfg and those in pure services--secretarial work, catering, janitors, landscaping, house cleaning, etc.
How are we to explain this gap?
According to James Galbraith, pure service employment seems to be the source of net job growth in the US (workers who deliver, assemble, and repair factory output could be considered mfg workers), and temping is concentrated in this sector as well. Now it would seem that only when capital has made a big investment in physical capital which it must amortize are employers willing to pay some wage premium and provide some extra benefits in order to ensure that the amortization will not be interrupted by excessive turnover or irresponsible work. But then it will pay capital to have such (relatively) well paid workers do overtime and thus conserve on benefits than hire new workers in a seasonal or even more long term expansion (the use of overtime in the better paying mfg sector seems a plausible candidate to explain increased wage inequality,e.g., the stories of $100,000 year autoworkers putting in 60 hour weeks--any studies on this?) This use of overtime will have the effect of concentrating new workers in the pure service industry in which no benefit temping is rampant and unions non-existent (as there is no expensive physical capital which would be idled by pure service workers' strikes, their power is seriously limited; ultimately I do think that only productive mfg labor has the latent social power with which to bring society to its knees).
Another question on the unemployment angle: so many of these pure service jobs seem to be pseudo jobs. Real practical unemployment seems quite high in the US. Some have argued that our unemployment indicators are biased downward (e.g., prisoners and uninsured drop outs, like me, are not included--my dept at UC Berkeley gives us 12 semesters of teaching which I used up in rapid succession; there is a lot of grad student unemployment). Doug has argued that while this is true, this has always been true, so there can be no question that there indeed has been an employment boom of late. But then are these real jobs? And second, the kinds of unemployment our indicators miss may indeed have been increasing of late (prisoners, underemployment of 'illegal aliens'). That is, there may not have been much of that unemployment our indicators are biased to disguise 25 years ago but now there is a lot of that kind of unemployment. So the indicators are really only doing their most effective propanganda work now. The collapse of Keynesianism may have had more horrific impact that commonly acknowledged. A possibility?
best, rakesh