Thus, a high adjusted correlation coefficient leads to the result that the zero-intercept (unexplained difference between means of actual data and predicted data) should also be small. Similarly a low correlation coefficient should reflect a large zero-intercept, since the unexplained variance is large. A regression showing a high correlation coefficient and a low zero-intercept provides high confidence in the regression specification. Similarly, a regression with low correlation and high zero-intercept provides high confidence. Where correlation and zero-intercept are both high or both low, the specification offers only low confidence.