Monte Carlo simulation of macroeconomic risk with a continuum of agents: the symmetric case (Q1404151)

The authors aim at extending the Monte Carlo method of evaluating a multiple integral in order to simulate macroeconomic uncertainty when many agents face individual risk which is modelled by a process with a continuum of random variables. The following notions and results are basic for the paper: probability spaces, a Polish space, \(\pi\)- and \(\lambda\)-systems, Dynkin's theorem, essentially countably generated \(\sigma\)-algebras, conditional expectations and regular conditional distributions, a continuum of random variables, pairwise measurable probabilities, Monte Carlo convergence and Monte Carlo \(\sigma\)-algebras, i.i.d. case, essential: pairwise independence, symmetry of processes, pairwise exchangeability. The main theorem deals with a measurable mapping from a sample probability space to one of measures on a Polish space equipped with the Borel \(\sigma\)-algebra corresponding to the topology of weak convergence of measures and establishes the equivalence, in particular, between the properties: ``a process is essentially i.i.d conditioned on the \(\sigma\)-algebra (countably) generated by this mapping'' and ``the process is Monte Carlo convergent, in a suitable product probability measure''. The paper is concluded with the following assessments: \textit{``... as a by-product of our work, we have shown that the fundamental probabilistic concept of (essential) independence constitutes a necessary condition for the classical sequential law of large numbers to hold. This converse result is entirely new in the extensive mathematical literature of the subject.''} (p. 765).