Sunday, 25 January 2015

Research Design.. A Case Study


Hypothesis: “The more information people have about their (national or local) government’s activities, the more likely they are to trust the government.”

The operational hypothesis is the higher the percentage of citizens who watch television on government’s development activities, the higher the percentage of citizens’ trust in the government. The independent variable is the level of information about the government’s actions, while the dependent variable is trust in government. The independent variable is thought to cause the dependent variable, and it is this causality that this research attempts to measure. The unit of analysis is individuals (watching or not watching television programmes on government’s activities). An opinion research will be conducted to ascertain whether an individual watches national television programmes on government’s activities, and whether or not s/he trusts the government. The true experimental and quasi-experimental designs would be used to undertake this research in Liberia.

True Experimental Design
A Social Science 102 class comprising 100 students will constitute the experimental study group. A total of 100 tickets will be printed with number one written on 50 of them, and number two on the rest.  All the tickets will then be mixed in a sealed box. During Wednesday’s class, each student will be asked to pick only one ticket from the box containing the 100 tickets. This process constitutes the randomized selection in the true experimental design. The randomization assumes that the two experimental groups selected for the research would be statistically identical in their distribution on both the dependent variable and all other variables whether or not they are being measured.
Following the random selection of tickets, all 50 students with number one tickets would be assembled in a class room to be used as the treatment group to which the stimulus (television news on government’s activities) will be administered. The remaining 50 students with the number two tickets will be the control group. They will sit in a separate class room while the treatment group watches two hours television programme on government’s development activities (the stimulus). At the end of the television programme, all the 100 students will be requested to complete a questionnaire about trust and confidence in the Liberian government. This questionnaire would measure the posttest results. The level of trust in government as measured from students who watched the television programme will be compared with that of the students who did not.  If higher number of students who watched the television programme expresses trust in the government, it would then be concluded that the independent variable (more information about government’s activities) causes the dependent variable (trust in government).

Quasi-Experimental Design
Using the same 100 students, the dependent variable would be tested before and after the independent variable has been applied (Monroe 2000, p.38).  Students will not be assigned in groups.  All 100 students would be asked to complete a questionnaire about trust in the Liberian government. This questionnaire is intended (as a pretest) to measure the level of trust all of them have in the Liberian government. After this pretest, the students would be requested to watch two hours television programme on the government’s development activities. This will serve as a stimulus after which a posttest measuring the extent to which a change in the level of trust in government has occurred. All the students will be asked again to complete a questionnaire measuring their trust in government (posttest). The posttest would be carried out irrespective of whether or not all the students watched the television programme. If the level of trust expressed in the government in the posttest exceeds that of the pretest, this change (increased trust in government) would be attributed to the stimulus thereby leading to the conclusion that high level of information about government’s activities leads to trust in the government.

Validity       
The same unit of analysis (individuals) will be used to measure the hypothesis. These two designs have high face validity because the measure looks right to the researcher. The internal validity of the true experimental design is higher compared to the quasi-experimental because the manipulation and randomization of the experimental groups ensured that no other variable interfered. We can therefore claim causality in the true experimental research compared to the quasi-experimental one under which the experimental group was not manipulated or controlled. The content validity and construct validity are low in both designs. For example, this research cannot be content valid when it does not actually consider all the causal elements that increase trust in the government. It is possible that citizens may be adequately informed about their government, but yet have less trust in it, or this could be the reverse. The external validity of these two designs is low because generalizable claims cannot be made across the general population, in all settings and at all times. Regarding construct validity, it cannot be claimed with absolute certainty that inferences from this research can become a functioning reality (Bollen 1989). Finally, the credibility of the television programme is another factor to consider. An independent television would have an objective analysis of government’s activities compared to a state owned television. Hence the ownership or control of the television programme could affect validity.

Causality
We can claim causality because these two designs have accounted for correlation, time order and nonspuriousness (Monroe 2000, p.32). The posttest comparisons in both designs indicate that there is a correlation between the independent and dependent variables due to the change observed in the level of trust as an effect of the stimulus. Nonspuriousness is ensured in the true experimental design because the treatment and controlled groups were considered to be exactly the same in all aspects due to the use of the randomized selection process, but nonspuriousness in the quasi-experimental design is questionable because the pretest of the students must have aroused their interest in the study thus affecting their responses in the posttest. Also, outside influences (previous exposure, maturation and conversations during the interval between the pretest and posttest) must have affected their responses. It is therefore difficult to attribute change to the stimulus in the quasi-experiment as the students were not controlled, randomized and equal in all aspects. Regarding time order, this criterion has been addressed because the stimulus was applied prior to the posttest measurement of the dependent variable in both experimental designs.

Strengths & Weaknesses
The artificial setting and manipulation of students in the true experiment is problematic because humans cannot be controlled like laboratory animals. Therefore, the results from the true experiment may not be obtained in real life situation. However, we can use this design to study general features of a small group of people. These designs lack genuine representative sample to back claims for generalizability and external validity. It would be difficult to conduct such research nationwide or amongst a group of countries. In the quasi-experimental design, it is uncertain that all other causal variables have been eliminated (Shively 2013, p.85). These designs are cost and time efficient, but replicating them nationwide would be cost intensive.

References
Bollen, A 1989, Structural equations with latent variables, Wiley & Sons, USA.

Shively, W 2013, The craft of political research, Pearson Prentice Hall, USA.

Monroe, A 2000, Essentials of political research, Westview Press, Boulder, Colorado, USA.

By Tom Kaydor
thkaydor@gmail.com



No comments:

Post a Comment