Flexible Mixture Priors for Large Time-varying Parameter Models

Time-varying parameter (TVP) models often assume that the TVPs evolve according to a random walk. This assumption, however, might be questionable since it implies coefficients change smoothly and in an unbounded manner. assumption is relaxed by proposing flexible law of motion for large-scale vector autoregressions (VARs). Instead imposing restrictive walk evolution latent states, hierarchical mixture priors on state equation are carefully designed. These effectively allow discriminating between periods which times where better characterized stationary stochastic process. Moreover, this approach capable introducing dynamic sparsity pushing small changes towards zero if necessary. The merits model illustrated means two applications. Using synthetic data these modeling techniques yield precise estimates. When applied US data, reveals interesting patterns low-frequency dynamics forecasts well relative wide range competing models.