Figure A5 An increase in Z causes an increase in Y at any value of X An increase in Z causes a decrease in Y at any value of X

good way to determine how a graph will shift is to perform a simple experiment like this: Put your pencil tip anywhere on the graph labeled June—let's say at point C. Now ask the following question: If I hold advertising constant at \$6,000, do I expect to sell more or less ice cream as temperature rises in July? If you expect to sell more, then the amount of sales corresponding to \$6,000 of advertising will be above point C, at a point such as C'. From this, we can tell that the graph will shift upward as temperature rises. In September, however, when temperatures fall, the amount of sales corresponding to \$6,000 in advertising would be less than it is at point C. It would be shown by a point such as C". In that case, the graph would shift downward.

The same procedure works well whether the original graph slopes upward or downward and whether it is a straight line or a curved one. Figure A.5 sketches two examples. In panel (a), an increase in some third variable, Z, increases the value of Y for each value of X, so the graph of the relationship between X and Y shifts upward as Z increases. We often phrase it this way: "An increase in Z causes an increase in Y, at any value of X." In panel (b), an increase in Z decreases the value of Y, at any value of X, so the graph of the relationship between X and Y shifts downward as Z increases.