It is not necessarily possible to avoid measurement error when using such ways of measuring variables. The reason for this is that the researcher is not always in control of the data that are gathered. Therefore, the researcher cannot be sure that the data are correct.
When doing observations, the researcher has a relatively large degree of control over the data that are gathered. The researcher is the one who is observing behaviors on the part of the subjects and is deciding what data to record. In such a situation, the researcher can avoid measurement error by being very careful to observe subjects in an objective way. For example, a researcher who wants to determine what percentage of teachers’ interactions with their students are positive must enter the experiment with a clear way to determine what interactions are positive. This will help prevent the researcher from being misled by such things as their personal opinions of a given teacher or student.
However, when using the other ways to measure variables, the researcher has much less control. In an interview, the researcher can be very careful to ask questions that are worded correctly. The researcher can be sure to ask the questions in the same way to each subject. He or she can avoid giving any hint as to the answer they expect. All of these things will help to avoid measurement error. However, the researcher cannot prevent the respondent from answering incorrectly, whether from faulty memory, from a desire to mislead the interviewer, or any other cause. When using self-reports, the researcher is similarly at the mercy of the respondent. The researcher can try to reduce measurement error by making very clear what sort of responses are appropriate, but the researcher can once again do nothing if respondents give incorrect answers. Finally, the researcher can be very careful to be accurate when copying available records. They can be very careful that they are only looking at records that are germane to their subject. But once again, they cannot ensure that the person who kept the records did so correctly.
Thus, when using such methods, it will be very difficult for the researcher to avoid measurement error because he or she is not in control of what data are recorded.