Well, to the extent that omitted variables introduce bias to the underlying model, yes...but you can test/correct for those issues relatively easy. If you are modeling the relationship/treatment effects, you are not trying to capture all the variables that have any interaction with your independent variable...you're trying to select the best group of variables that explain the most amount of variance in your target variable(CPI) in a statistically/robust way, given the time continuity of your series. You aren't calculating the actual impact - you're estimating it..and it's just hard to imagine that a variable with nearly no correlative relationship that explains, at most, 5% of the variance in CPI even when using lagged effects would ever give you any significant effect on inflation (because profits are a product of consumption...consumption isn't a product of profits....at least not directly). Either way, you shouldn't use after-tax profits as a dependent variable anyways...lol