The conditions in which the parameters of a model estimated separately for two sets of data show no substantial or statistically significant differences. This provides some assurance that relationships are stable, but it does not ensure that they will be stable over the forecast horizon. (See Chow test.)