Introducing a Third Variable - Direction of Causation Issue
Introducing a Third Variable - Direction of Causation Issue
If a causal link between two variables is thought to exist, a reasonable question is: which variable is the causal (or independent) variable and which is the "caused" (or dependent) variable? Such a question arose when the task was to distinguish between:
INTERVENING VARIABLE SPURIOUS ASSOCIATION
BETWEEN A AND I BETWEEN A AND I
A I
A->U-+1 and \/
U
In some situations there is a reciprocal causal relationship. Attitude-can influence purchase, for example. However, the act of purchasing and using a product or service can affect the attitude, which again affects purchase. Even when a reciprocal causal relationship exists, it might be useful to determine the direction of the dominant flow of influence. Does the attitude-to-purchase direction have a greater effect than the purchase-to-attitude direction, for example?
One approach to determining the direction of causation is to draw on logic and previous theory, as was done in arguing that the usage variable was a common cause of advertising recall and intentions:
In this context it is useful to observe whether one of the variables is relatively fixed and unalterable. Variables like sex, age, and income are relatively permanent. If, for example, an association is found between age and attendance at rock concerts, it would be unrealistic to claim that attendance at rock concerts causes people to be young. In this case age could not be a "caused" variable because it is fixed in this context. However, it could be that age is an important determinant of who attends rock concerts.
A second approach is to consider the fact that there is usually a time lag between cause and effect. If such a time lag can be postulated, the causal variable should have a positive association with the effect variable lagged in time. When intervally scaled variables are involved, the approach is termed cross-lag correlation.
If a causal link between two variables is thought to exist, a reasonable question is: which variable is the causal (or independent) variable and which is the "caused" (or dependent) variable? Such a question arose when the task was to distinguish between:
INTERVENING VARIABLE SPURIOUS ASSOCIATION
BETWEEN A AND I BETWEEN A AND I
A I
A->U-+1 and \/
U
In some situations there is a reciprocal causal relationship. Attitude-can influence purchase, for example. However, the act of purchasing and using a product or service can affect the attitude, which again affects purchase. Even when a reciprocal causal relationship exists, it might be useful to determine the direction of the dominant flow of influence. Does the attitude-to-purchase direction have a greater effect than the purchase-to-attitude direction, for example?
One approach to determining the direction of causation is to draw on logic and previous theory, as was done in arguing that the usage variable was a common cause of advertising recall and intentions:
In this context it is useful to observe whether one of the variables is relatively fixed and unalterable. Variables like sex, age, and income are relatively permanent. If, for example, an association is found between age and attendance at rock concerts, it would be unrealistic to claim that attendance at rock concerts causes people to be young. In this case age could not be a "caused" variable because it is fixed in this context. However, it could be that age is an important determinant of who attends rock concerts.
A second approach is to consider the fact that there is usually a time lag between cause and effect. If such a time lag can be postulated, the causal variable should have a positive association with the effect variable lagged in time. When intervally scaled variables are involved, the approach is termed cross-lag correlation.
Comments
Post a Comment