Maybe to state it another way: The schools that have already closed (all of them) and businesses that have already closed or implemented work-from-home (many of them) have produced the currently observed rates of transmission, hospitalization, ICU/vent needs and deaths. So any projection we make using the current data has to make the assumption that all those things continue, UNLESS we make adjustments to the model account for future actions. In this model, they are assuming that we can do much more - close schools, force shutdowns of non-essential businesses, etc. - but in reality we can't because those were done weeks ago. That's why I think there's a lot of risk that it's an underestimation.
I see what you are saying, and agree that needs to be taken into account, and I think in most models it is.
I also think there are models that can or could eventually capture benefits from people changing behaviors due to the guidance despite no SIP, but I don't believe most do. And if they did look at Iowa specific rates, for example, I think it will take a long time until those changed behaviors are fully captured. I think rates in many of the models for non-SIP states are including the initial high rate of change ramp up of the disease when people were slow to change behaviors.
Conversely I think because there is a definitive start point, SIP rates moving forward are going to look good.
But I'm arguing that the rate of transmission moving forward in a place operating like Iowa vs if we had official SIP are not going to be that different. That's based not just on observed behavior changes but by looking at rate of change in key metrics in Iowa now that changes in behavior are more widely adopted