Title:SDEs with uniform distributions: Peacocks, Conic martingales and ergodic uniform diffusions

Abstract:It is known since Kellerer (1972) that for any process that is increasing for the convex order, or "peacock" as in Hirsch et al. 2011, there exist martingales with the same marginals laws. Nevertheless, there is no general constructive method for finding such martingales that yields diffusions. We consider the uniform peacock, namely the peacock with uniform law at all times on a generic time-varying support [a(t); b(t)]. We derive explicitly the corresponding SDEs and prove that, under certain "conic" conditions on a(t) and b(t), they admit a unique strong diffusive solution. To guess the candidate SDE we resort to the approach of inverting the Fokker Planck equation. Dupire (1994) did this for volatility modeling. Here we tackle the inversion with the caveats needed when dealing with uniform margins with conic boundaries. This was done originally in the unpublished preprint by Brigo (1999). Independently, Madan and Yor (2002) obtained the result as a simple application of Dupire. Once the SDE is guessed, we analyze it rigorously, discussing the cases where our approach adds strong uniqueness of the solution of the SDE and cases where only a weak solution is obtained. We further study the local time and activity of the solution. We then study the peacock with uniform law at all times on a constant support [-1; 1] and derive the SDE of an associated mean-reverting diffusion process with uniform margins that is not a martingale. For the related SDE we prove existence of a solution. We derive the exact transition densities for both the mean reverting and the original conic martingale cases. We prove limit-laws and ergodic results: the SDE solution transition law tends to be uniform after a long time. Finally, we provide a numerical study confirming the desired uniform behaviour. These results may be used to model random probabilities, recovery rates or correlations.