HYPOTHESES - CAUSE AND EFFECT
Causal order means the sequence in which variables are placed. This sequence determines what is supposedly the "cause" (the independent variable) and what is the "effect" (the dependent variable).
If the hypothesis is that poverty causes crime, then poverty is the independent variable and crime is the dependent variable. The causal order is poverty ==> crime. If the hypothesis is that crime causes poverty then the causal order is crime ==> poverty.
Consider this example. A researcher is interested in how police respond to domestic violence. She feels that officer characteristics, such as years of formal schooling, may be associated with officer attitudes about domestic violence (better educated officers might be less likely to tolerate domestic violence.) The formal hypothesis is officer characteristics ==> officer attitudes about domestic violence. But instead of stating this hypothesis, the researcher limits herself to identifying the potential "causes" such as education, age, length of service, etc. Each is an independent variable, but she does not specify a dependent variable. This is a legitimate approach to a new area of study. Itīs then up to other researchers to specify the dependent variable(s) and to formulate and test a hypothesis.
But, even when a formal hypothesis is absent, even when a dependent variable is not specified, even when no data is collected, there is usually an implicit hypothesis. Merely identifying possible "causes" suggests the hypothesis that officer characteristics ==> officer attitudes about domestic violence. Otherwise, why even get started?
Dependent variables in one hypothesis can be independent variables in another. Perhaps officer attitudes about domestic violence determine how officers handle incidents of domestic violence. One can take the dependent variable officer attitudes about domestic violence and make it the "causal" or independent variable:
Officer attitudes about domestic violence ==> Officer response to incidents of domestic violence
In other words, the more concerned an officer is about the issue of domestic violence, the more likely he/she will arrest a suspected batterer.
How would data be collected? For the independent variable - officer attitude - we would administer each officer a questionnaire (the correct terminology is "instrument"). To measure the dependent variable, we would review the actual reports each officer filed after responding to a domestic violence call. Assumedly, officers who are more concerned about domestic violence would more frequently employ arrest as a way of dealing with suspects of domestic violence.
For more information on hypotheses, see week 1 of research methods.
Note: None of the below reflects the results of "real" research - it was all made up to serve as examples.
Hypotheses are frequently challenged by researchers who feel that the wrong independent variable(s) was identified as the "cause". For example, a common hypothesis is that poverty ==> crime. We can draw samples, then measure each person or "case" along two variables: the independent variable (income - low or high) and the dependent variable (whether a person was ever arrested - yes or no). [Terminology: things we measure, including people, are called "cases"]
Letīs say we randomly select 100 low income persons and 100 high income persons, then check their arrest records. As we assign persons to each income group - low and high - we are automatically "measuring" independent variable "income". To measure the dependent variable - criminal history - we review ech personīs police and court records. Here are the results: [Terminology - the overall group, and each constituent group, is called a "sample"]
Prior arrest No prior arrest Low income 80 20 High income 20 80
It seems that income is strongly associated with arrest history. Eighty percent of the low income sample (80/100) had prior arrests, while only 20 percent of the high income sample had prior arrests (20/100).
But another researcher claims that income does not cause arrests. Instead, she claims that the real cause is lack of education. Her hypothesis is education ==> crime. In her opinion, the reason income seems like a cause of crime is because income is closely associated with the real cause - education. She feels that the apparent relationship between income and crime does not exist. [Terminology: if this is true, income would be a "spurious" independent variable].
How can we tell? Simple. We code each of the 200 cases for education - either less than high school, or high school and above. For no particular reason, it turns out that 80 persons had less than a high school education, while 120 had at least a high school education. (Note: we need at least 30 cases in each category for statistical reasons. It is otherwise irrelevant how many persons wind up below high school or high school and above.)
We make two duplicates of the original table, keeping the original independent variable income. (Please note that the below tables are set up just like original table.) Into one of these duplicate tables we insert the 80 persons with less than a high school education, while in the other duplicate we insert the 120 persons with at least a high school education. [Terminology: This process is called "controlling" for a new independent variable].
Here are the results:
Less than high school
|
High school and above
|
These tables clearly support the new researcherīs view - that education is the real "cause" of arrests:
1. Of persons with less than a high school education, 60/80 (75%) had arrest records
2. In contrast, of those with at least a high school education, only 20/120 (17%) had arrest records
3. Within the new tables, compare the arrest arrest rates for low income and high income persons. They are identical! It seems that when we "control" educational level (hold it constant) the differences in arrest rates between income levels disappears!
Of course, this was an extreme example. Here is a more realistic outcome:
Less than high school
|
High school and above
Prior arrest No prior arrest Low income 20 40 High income 10 50
1. Persons with less than high school are still much more likely to have arrest records (50/80, or 63%) than those with at least a high school education (30/120, or 25%).
2. But this time, high-income persons in both groups are less likely to have arrest records than low-income persons. In the less than high school group, 30/40 (75%) of low-income persons have an arrest record, while 20/40 (50%) of the high-income persons have arrest records. In the high-school and above group, the distinction is even more pronounced: 20/60 (33%) of low-income persons have an arrest record, while 10/60 (17%) of high-income persons have an arrest record.
What can we conclude? Education might be the most proximate (closest) "cause" of crime but it is in turn affected by income. So both are important.
income ==> education ==> crime
Of course, there may be other independent variables that we havenīt considered. Our book is full of contradictory studies. Researchers often attack one anotherīs findings - that the appropriate independent variables were not used, or that they were used but were incorrectly measured. In other words - my variables are the "real" cause, while "your" variables only "seem" to be the cause.