Part 1: Previous Behavior – What we did
7/16/13 / Matt Bruce
Sometimes we need to know about shopping behavior or drinking behavior or seatbelt use, and because of time or cost, we are unable to conduct a diary study or observational research, so instead we ask about previous behavior. This is a reasonable substitute for many types of behaviors because previous behavior is typically a strong predictor of actual behavior. In other words, we are likely to do again what we did before. If you eat breakfast one day, there is a good chance you will eat breakfast again the next day.
Two common ways participants report their previous behavior include estimating their typical (average) behavior or reporting their most recent behavior. Measuring typical behavior is beneficial in some situations—it can help control for known confounding factors such as seasonality. For example, if you ask someone which outdoor activity he or she most commonly pursues, you should get the same answer in January or in June. But measuring typical behavior also has disadvantages. Studies suggest that people are not very accurate in reporting their typical behavior. Sometimes this is due to poor memory, and sometimes it is because people want to give socially desirable answers (do you really floss as much as you tell your dentist?). A self-assessment of typical behavior is more likely to measure one’s perception of behavior; depending on your research goals that may or may not be helpful.
An alternative way to measure previous behavior is to ask about one’s most recent behavior. For example “where did you spend your most recent vacation?” or “when was the last time you voted?” While one’s most recent experience might be atypical (maybe their last vacation was a once in a lifetime trip to Antarctica!), when you combine all experiences from a large enough sample, you will see a wide variety of answers. One of the advantages of measuring most recent behavior is the ease with which participants can answer a question about one event (a relatively recent event), and they are not required to recall experiences from memory. It also provides a specific frame of reference from which to consider and evaluate other factors—such as attitudes. A disadvantage of measuring most recent behavior is the increased possibility of confounding variables. If you ask someone in August to tell you which outdoor leisure activity they most recently undertook, don’t expect a lot of folks to say skiing. The summer season is confounding because it influences the way participants respond, which would not be a problem if we measured typical behavior instead. Experience, foresight, and careful word-choice are needed to seize the opportunities and avoid the pitfalls of measuring previous behavior.
At Corona, we believe measuring previous behavior can be a fruitful pathway to revealing insights. As data scientists, we carefully consider the above mentioned opportunities and challenges when designing our surveys. We acknowledge that measuring previous behavior has practical limitations—marketing campaigns cannot influence previous behavior. And while research suggests previous behavior is a good predictor, it is not a perfect predictor, as clearly people do things differently over time. But when combined with other measurements, understanding previous behavior allows us to start putting the puzzle pieces together.
For our next post in this series, we intend to investigate behavioral intentions. Stay tuned.
Other blogs in this three part series: