I was reading the answers to this question when, the one by Dawson Verley aroused my curiosity. Does every change in structural policy result in a structural break? What conditions are necessary and/or sufficient for break?
Any help would be appreciated.
Assuming I understand the question appropriately, the answer, generally, is yes - but in its application to times series analysis/forecasting not all "structural breaks" would end up being included in a given model. For example, if we are modeling labor supply, policy related to minimum wage would be hugely important (an increase can be seen as a structural break or a shock). However, a small tariff placed on sugar imports, though it may effect labor supply in some minute way in some region, would not constitute a structural break in consideration of our model (though, by definition, it is a break).
To answer the second part of your question related to what conditions would necessitate inclusion in a model, I would argue that it is an intersection of theory/logic and statistical relevance. If the influence of the break on other variables is virtually unobserved, or if its coefficients are far from significant then it need not be included (barring some other compelling case for its inclusion). Or, we could say during the modeling process some break/shock variable added no advantage to the model fit (thinking AIC or SIC/BIC here).
Hopefully that at least gets at your question.
