when modelling the % returns on a government bond, I use a model like this:
$$\Delta P_t / P_{t-1} = \sum_i KRD_i (-\Delta y_i)$$
Does it make sense at all to add the following "country oas" component?
$$\Delta P_t / P_{t-1} = \sum_i KRD_i (-\Delta y_i) + \text{spread duration} (-\Delta OAS_{country, t})$$
So if our gov bond is US, we take a US denominated bond index with comparable duration to our bond and take negative difference of its oas at each time point and then add this component to our model.
This change will be incredibly small, but for some reason when I incorporate it into my model I get that a large part of risk stems from this component.