While looking into something else, I did the following regressions: ERA was the dependent variable while opponents' OBP & SLG, each with no runners on (NONE) and runners on base (ROB) were the independent variables. The data set included all pitcher seasons of 140+ IP from 2007-2013. There were 740 pitchers. Here is the equation
ERA = 3.21*NONESLG + 7.06*NONEOBP + 6.30*ROBSLG + 7.06*ROBOBP -4.40
The r-squared was .866 and the standard error was .30. The t-values for the independent variables were all at least 10. So it seems, at least on the surface, that SLG has twice the impact on ERA when there are runners on base as when there are no runners on. Maybe if a pitcher's ERA seems high for his underlying stats it could be he has a high OBP with runners on base.