THE ELABORATION MODEL

The elaboration model is used to understand the relationship between two variables through the simultaneous introduction of additional control variables.

I. History

During WWII, Samuel Stouffer headed a special social research unit to study morale of soldiers during the war. The researchers attempted to confirm empirically some commonly held propositions:

1. Promotions surely affected soldiers' morale, and soldiers serving in units with low promotion rates would have relatively low morale.

2. Given racial segregation and discrimination in the South, black soldiers being trained in Northern training camps would have higher morale than those trained in the South.

3. Soldiers with more education would be more likely to resent being drafted into the army as enlisted men than would soldiers with less education.

Although each of these propositions made sense, none proved true!!

1. The Military Police, where promotions were few and slow, created a condition where few soldiers had experiences of a less qualified buddy being promoted before them. They were in a relative sense, and, by the standards of their reference group, pleased with the fairness of the promotion system. On the other hand, the Army Air Corps, where promotions were many and fast, more soldiers reported unfairness in their experience, relative to their reference group.

2. The general morale of black soldiers serving in Northern training camps and of those serving in Southern training camps seemed to differ little.

3. Less educated soldiers were more likely to resent being drafted into the Army than were soldiers with more education.

Stouffer suggested that soldiers evaluated their position in life not in accord with absolute, objective standards, but on the basis of their relative position as compared to others around them. The introduction of the Control Variables helped to clarify what was otherwise an anolomy.

The elaboration model is a quasi-experimental method designed to answer questions about data relationships in contingency table analysis. A third variable is introduced into a bivariate relationship and its effect on the relationship measured.

The effect is measured by looking at the measures of statistical significance and at the measures of association for the "controlled" relationships.

To begin, you must know whether the CV is antecedent (prior in time) to the other two variables or whether it is intervening between them.

Example: Intervening

IV --- CV --- DV

Stouffer found that the IV (educational level) affected the intervening CV (having friends deferred), which in turn affected the DV (acceptance of induction).

Example: Antecedent

IV

Test Variable(CV)

DV

The correlation between death by drowning and the sale of ice cream cones, both variables are the result of a third variable (season of the year) which occurred before both of the observed variables.

After introducing the CV into a bivariate relationship, you look at the control tables to assess its effect on the initial relationship, in one of four ways:

1. Replication: Whenever the partial relationships (control tables) are essentially the same as the original bivariate relationship, the term replication is used to describe the result, regardless of whether the CV is antecedent or intervening. The original relationship can be characterized as replicated if the relationship, which is typically statistically significant for the bivariate tables, remains so for all levels of the test variables.

2. Explanation: This is the term used to describe a spurious relationship, one in which the original relationship is explained away by the introduction of the CV. Two conditions are required for this to occur. (1) The CV must be antecedent to IV and DV. (2) The partial relationships must change significantly (statistical significance disappears in the partial tables, and measures of association drop significantly).

3. Interpretation: Similar to explanation, except for the time placement of the CV and the implications that follow from that difference. The CV is intervening in a time order sense (e.g. educational level and resentment of induction).

4. Specification: Sometimes the introduction of the third variable (CV) produces partial relationships that differ from each other (e.g. on remains significant, the other does not). Such a relationship is called specification, because we have specified the conditions under which the original relationship occurs. This is true whether the CV is antecedent or intervening.

The elaboration model is a method for evaluating the possible outcomes that can occur when a CV is introduced into a bivariate relationship. The way one characterizes the outcomes depends upon the time order of the (CV). The elaboration model is a method of analysis designed to answer questions in a quasi-experimental framework. SOCSTATSIM is a computer program designed for practice of the elaboration model method. You will be asked to do some runs using this program to internalize these concepts.