When you fit a time-series regression, you are assuming that the coefficients are stable over time. estat sbcusum tests that assumption. It bases its result on whether the time-series abruptly changes in ways not predicted by your model. Said more technically, it tests for structural breaks in the residuals.
estat sbcusum uses the cumulative sum of recursive residuals or the cumulative sum of OLS residuals to determine to test whether there is a structural break. Under the null hypothesis, the cumulative sum of residuals will have mean zero.
The command also graphs the cumulative sum with confidence bands, which allows you to see whether the series behaves as the null hypothesis would predict.