Determining the percentage change among the study group between the pre- and post-survey, and comparing it to analogous percentage change among the control group, can be done in five calculations once the data is entered:

For example (to make the example simple), let's assume you interviewed 3 people from your study group and 3 from your control group and got the following scores on "how many times the respondent gave blood in the last 12 months".

	Pre-survey	Post-survey
Study group
person 1	3	4
person 2	1	0
person 3	5	7

	Pre-survey	Post-survey
Control group
person 1	6	6
person 2	2	4
person 3	2	2

1) determine mean¹ (average) scores for pre-survey and post-survey among study group and among control group. The mean scores for the study group pre-survey is 3 = (3+1+5)/3. The mean score for the study group post survey is 3.7=(4+0+7)/3.

The mean scores for the control group pre-survey is 3.3 = (6+2+2)/3. The mean score for the control group post survey is 4=(6+4+2)/3.

2) determine the percent improvement of the study group. The study group's mean score went up from 3 in the pre-survey to 3.7 in the post survey, for a percentage increase of 23.3%= ((3.7-3.0)/3).

3) determine the percent improvement of the control group. The control group's mean score went up from 3.3 in the pre-survey to 4 in the post survey, for a percentage increase of 21.2%= ((4-3.3)/3.3).

4) you compare the percentage increase of the study group against the percentage increase of the control group. Here the study group is increasing 23.3% vs. 21.2% among the control group. That means that the control group is increasing 9.9% faster, which equals ((0.233-0.212)/0.212).

5) you need to compare this percentage increase against a confidence interval given the size of the sample. You would have to consult a statistics book or a local statistician to tell you if this relative increase among the study group relative to the control group was statistically significant, i.e., whether you could reasonably conclude that it was real.

Importance of control groups: This little example shows you why you have comparison groups. Looking only at the study group shows a 23.3% increase in giving blood. Without a control group, the program would conclude that it is making amazing progress. With the control group (that also shows strong improvement), you realize that much of the increase likely stems from factors having nothing to do with the program since the control group didn't have the advantage of participating in the program. Maybe there is a factor at work like the local Red Cross adopting a vigorous outreach and PR campaign to get people to give blood more often.

You want to determine average (mean) responses for each survey question, and compare the percentage improvement or decline for each social capital question in the survey among the study group relative to the control group.

Return to Step 4c Return to Step 6

1. Note: there are many other ways to summarize the distribution of data across the different respondents. If you are working with someone sophisticated in numerical analysis, you may want to discuss with her whether means are the best way to summarize the data, and what the relative advantages and disadvantages of other ways of summarizing the data are, such as: frequency distributions, mode, median, standard deviation, range, etc. These terms are summarized in the glossary.