Constructs, Independent and Dependent Variables
Did you ever think about what is actually being measured by a questionnaire? When someone decides to write a questionnaire, also known as an “instrument”, the questions all need to point to a construct. A construct is something unobservable that we want to measure. Since it is something we cannot see, we instead measure various behaviors that point to that construct. For example, if you want to study whether patriotism motivates people to vote at election time, you would write questions that ask about a person’s patriotic behaviors. After all, we can’t measure the construct of “patriotism” directly, but we can measure things that represent it. Questions such as “I feel a loyalty to the values of my country” and “I always fly my country’s flag on Memorial Day”, when answered in the affirmative, point to a person who is likely to have a higher level of patriotism.
Variables, on the other hand, are the numerical representations of constructs or characteristics that vary from person to person. When people fill out the patriotism instrument, their answers to those questions will either be averaged or summed and that will give a single value. This value will be the “patriotism” variable, representing the construct of patriotism. In this particular example, it will be the independent variable, usually signified as the “IV”. The IV is the variable in research that is expected to have some effect on another variable. This other variable is the dependent variable, usually signified as the “DV”, and is the variable that represents the outcome. In this example, patriotism (the IV) is expected to have some effect on the outcome of voting behavior (the DV). By measuring people’s levels of patriotism (a construct), and asking them how likely they are to vote in elections (a characteristic), we can then use statistical analysis to find out if their patriotism is associated with their tendency to vote. This would most directly be done with a correlation, but other analyses may also be done to answer questions of prediction or group differences.
We should be mindful, however, that a list of questions doesn’t necessarily make up an instrument. If the questions aren’t all written to measure various aspects of a single construct, then it is merely a list of questions. It is the relatedness of the questions that make it an instrument, and there are ways to test this relatedness in a statistical fashion, such as inter-item correlations or, in some cases, factor analysis (later!). The construct is the center of an instrument, the star of the show you might say, and a useful thing to keep in mind as you design your research study. So should you write your own instruments for your research study? Occasionally this might be necessary, but in most cases instruments already exist that measure the constructs you might have in mind.