Volterra equations with affine drift: looking for stationarity

Abstract

We investigate the properties of the solutions of scaled Volterra equations (i.e. with an affine mean-reverting drift) in terms of stationarity at both a finite horizon and on the long run. In particular we prove that such an equation never has a stationary regime, except if the kernel is constant (i.e. the equation is a standard Brownian diffusion) or in some fully degenerate pathological settings. We introduce a deterministic stabilizer $ \varsigma$ associated to the kernel which produces a {\em fake stationary regime} in the sense that all the marginals share the same expectation and variance. We also show that the marginals of such a process starting from when starting various initial values are confluent in $L^2$ as time goes to infinity. We establish that for some classes of diffusion coefficients (square root of positive quadratic polynomials) the time shifted solutions of such Volterra equations weakly functionally converges toward a family of $L^2$-stationary processes sharing the same covariance function. We apply these results to (stabilized) rough volatility models (when the kernel $K(t)= t^{H-\frac 12}$, $0<H<\frac 12$) which leads to produce a fake stationary quadratic rough Heston model.