Related simulation studies
A related but more complicated model in Aoki (2002, s. 7.1) was examined by Ono (2007), who simulated several two-sector models using both the cumulants and the Monte Carlo method for the case of one technically advanced sector and another less technically advanced sector. The model has many more parameters than the one described in the section above.
The broad conclusion of this analysis is that the technically advanced sector behaves in a non-self-averaging manner and the other sector in a self-averaging manner.
Sector 1, which is initially of size n1 = 250, and technically advanced, behaves in non-self-averaging way, while sector 2, which is larger with initial size n2 = 500,000 than sector 1 but less advanced technically, behaves in a self-averaging manner. The size of the CVs in Ono's simulation a re approximately 0.07 for sector 1 and about 0.014 in sector 2. In other words, the standard deviation of n1 , that is, the square root of ки is about 7 percent of the mean of n1 in sector 1, and that in sector 2 is about 1.5 percent. This conclusion holds both with the cumulant analysis and with Monte Carlo simulations. The correlation between the n1 and n2 sectors is completely negligible in both the cumu- lant analysis and the Monte Carlo runs.
Interestingly, self-averaging behavior was not found to be a robust outcome for one of the two sectors. Altering the rate at which firms change their types in Monte Carlo simulations could induce nonself-averaging behavior in both sectors: a result of particular importance for economic policy designs, as mentioned above.
The overall conclusion of the cumulant analysis is the realization on the part of policymakers that in real economies with several sectors CVs are likely not to be the same across sectors but to vary from sector to sector.