Exploratory. Descriptive. Causal. Predictive.
Often the initial step, exploratory research helps us become familiar with the problem situation, identify important variables, and use these variables to form hypotheses that can be tested in subsequent research. The hypothesis will be a statement about a variable or the relationship between variables—for example, “Production will increase if we switch the line locations of operators A and B.” The hypothesis may not be true, but it is useful as an assertion that can be examined through the collection of sample data for a test period during which the operators have traded line positions.
Exploratory research can also be of a qualitative nature. One such method is the focus group interview, in which a moderator leads a small-group discussion about a topic and the client watches and listens from behind a one-way mirror.
As might be expected, descriptive research has the goal of describing something. For example, a survey by the National Association of Home Builders found that the average size of single-family homes ballooned from 1500 square feet in 1970 to nearly 2400 square feet in 2004.
In causal research, the objective is to determine whether one variable has an effect on another. In examining data for broken utility poles, the Duquesne Light Company found that about 30% fewer poles were being damaged after Pennsylvania’s stricter drunken driving laws took effect in 1983. According to a company spokesperson, “It may just be a coincidence, but before '83, we averaged 1000 broken poles a year. After that, it dropped to abort 700 a year. Most of the accidents involving our poles occur between 1 and 4 A.M., about the time when most bars are closing and the people are going home."
Regarding causal studies, it should be pointed out that statistical techniques alone cannot prove causality. Proof must be established on the basis of quantitative findings along with logic. In the case of the telephone poles in the preceding paragraph, it would seem obvious that causation was not in the reverse direction (i.e.., reduced pole damage causing the stricter drunk driving laws). However, we must consider the possibility that one or more other variables might have contributed to the reduction in pole damage—for example, the company may have begun a switch to underground wiring at the same time that the stricter drunken driving legislation was enacted.
Predictive research attempts to forecast some situation or value that will occur in the future. A common variable for such studies is the expected level of future sales. As might be expected, forecasts are not always accurate. For example, in 1984, the Semiconductor Industry Association predicted a 22% increase for 1985, a year in which sales actually fell by about 17%.
With any forecast, there will tend to be an error between the amount of the forecast and the amount that actually occurs. However, for a good forecasting model, the amount of this error will be consistently smaller than if the model were not being used.