Consider an espresso machine that can make only one drink at a time. Adding baristas without adding machines cannot increase the rate of espresso production.
This is not strictly true. Having multiple bar-trained counter staff allows baristas to rotate as needed, potentially minimizing espresso output rate loss due to declining barista productivity (and increasing error rate) over the course of the day, as well as interruptions in using the machine. Capital investment in “systems” would determine the extent to which this strategy can be employed.
I just made this idea up right now but I kind of like it: “Systems” capital is the subset of capital that determines how labor interacts with “physical” capital. Since systems are affected by both capitalists and labor, a small endogeneity problem arises between L and K. The effect of a change in K on MPL depends, in part, on a component of K that depends on QL (quality of labor), which would equal MPL with homogeneous workers. The systems capital contribution by a particular worker depends on that worker’s personal stock of human capital, and some systems are more dependent on workers’ own capital than others: car showrooms are useless without shrewd salesmen, while McDonald’s workers follow specific instructions that only the worst workers can screw up.
I don’t think this effect is big enough to be economically interesting on a macro level, but it might help clarify the language of “investing in human capital.” Horizontal differentiation in human capital might explain horizontal segregation among workers; horizontal differentiation in contribution of human capital to horizontally differentiated systems capital might explain a (little) bit more. Also, low-capital workers segregate into jobs in which systems capital does not depend much on workers’ human capital stocks.
What’s more interesting is the higher-order feedback between K, L, MPK, and MPL, the opportunity for cumulative advantage that results, and how that advantage can turn into disadvantage if any one of those components receives a large negative shock. Back to my example, if all the good baristas quit and are replaced by scrubs like me, systems fall apart and firm output eventually suffers.
But consider a labor market with incomplete information. Wouldn’t employers be very interested to know how job applicants will fit into existing systems? Is this what Thurow was driving at with his job competition model? What applicant signaling behavior would be consistent with this framework, and do we observe that behavior? If the model is “true,” would its effect be large enough to notice?