MEN WOMEN TIPPING STUDY
This study is described before the data is collected so "more" is the hypothesis prediction. In this
statement the correlational studier is making the guess that men will tip more than
women. It is a hypothesis prediction since this problem statement has not had
any data collected yet to test the hypothesis. How did I invent this example to
explain the difference between the true experiment's independent variable and a
correlational study's pseudo-independent variable? I base this hypothesis
prediction on the inductive information from people who had worked as servers telling me that men tip more than
women. It is inductive since I used the information from a number of individual
observations to come up with the testable hypothesis: "Men will tip more than
women." Another way I could have come up with this on my own was by using deduction.
If I took as a major premise that "Wealthier people tip more than
poorer people." Then continued with the minor premise: "Men are paid more than Women."
Then I could have concluded therefore, "Men will tip more than
Women." Of course if the data disconfirmed this, if I found there was no
difference between men and women in tipping or women tipped more than men I
would need to re-examine my theory to see if I could figure out where my
reasoning was off.
True Experiments vs. Correlational Studies: Confounds
The main difference between correlational studies and true experiments is that it is
ambiguous what the results mean in a correlational study whereas in a true
experiment it is clear the independent variable caused the results. Let me give
you an example of this issue.
When the name introduction in Garrity & Degelman led to higher tipping and the tip tray with a credit card logo led to higher tipping there was no doubt, it was unequivocal, that those specific manipulations led to higher tips; there was no other alternative explanation for the result. That is not true for pseudo-independent variables like Gender. Here is how it could be ambiguous. Assume that men do tip more than women. But it really wasn't a sex difference that this hypothetical study detected but instead a wealth difference. In other words, when the data showed men tipped more than women maybe what it really showed was richer people tip more than poorer people and the men/women cause was bogus. That would mean if you were able to regroup the participants in the study by wealth so one group was the wealthier (both men and women) and the other group was the poorer (both men and women) there would be no difference within those groups between men and women's tips.
Pseudo-independent variables, like Gender, may have the problem of group membership being associated with other differences between the groups. That problem is called "confounding". So we could say that gender is confounded with wealth since on the average men make more than women. That makes it difficult and debatable as to what conclusions can be reached from correlational studies since they use pseudo-independent variables.
For example, in racial profiling individuals are singled out to be searched by race and supporters of this crime fighting approach justify the tactic by pointing at crime statistics that show minorities are disproportionately represented in prisons. However, others point out that race itself does not predict being a criminal; actually poverty predicts being a criminal and minorities are more likely to be poor. Here race is confounded with poverty.
Another example of the confounding problem is in a study (Meddock & Osborn, 1968)I did as an undergraduate. Inspired by S.A. Barnett's book The Rat, we decided to see if the greater fear of new objects in wild rats compared to lab rats would be true for mice too. We trapped wild mice around the Oxford, Ohio area and borrowed some lab mice from the Miami University colony and tested them against each other, finding that our wild mice showed greater fear of new objects than lab mice. We replicated Barnett's findings, but we were not really able to claim that we supported Barnett's conclusion that these results showed genetic changes had occurred between wild and lab mouse species. Why? Here the pre-experimental environment, particularly the wild mice's experience of living in homes or a dairy barn then being trapped and kept in cages was confounded with the genetic factor. We could not tell if it was the difference in experiences or difference in genetics that caused the wild mice to show more neophobia than the lab mice.
The key point is when you are thinking about what you can conclude from a correlational study be ready to look for alternative explanations. This is specially important when the results are presented for advocacy purposes; the investigator may not bring up alternative explanations that should be considered.
On the "Men tip more than women" proposed study it turns out that at least some research considered by American Demographics concludes women say they tip slightly more than men. A guess about "why"; women more frequently say they tip because of the effect it has on others, because people depend on the money or it's expected of them. Men more frequently say they tip for reasons that personally benefit them. Since men's reasons for tipping are less likely to be met, that could be why women tip more. However, my idea that rich people tip more than poor people was supported by the research reported in American Demographics. This suggests that the wealth difference between men and women is not strong enough to overcome the "reasons why people tip factor." However, these articles did not sufficiently distinguish between what people said they did and what they actually did to reach a firm conclusion. As with many research questions, more research would need to be done to reach a more definitive conclusion.