About The ARDA | Tools | FAQs | Contact Us
Independent and Dependent Variables

Many theories of human behavior distinguish cause from effect. For example, being poor may cause one to affiliate with a religious sect, perhaps because doing so gives one a sense of self-esteem and respect from fellow members, or otherwise to compensate for one’s disadvantages in the wider society. However, belonging to a sect can instill discipline, support hope, and provide a social network of people who can be helpful. Thus: poverty can cause sect membership, but sect membership can cause prosperity. To the extent this is true, it is correct to speak of cause-and-effect relations between the two variables, but those relations are complex, even contradictory.

In statistical research we often speak of independent variables rather than causes, and dependent variables rather than effects. In an experimental study, researchers manipulate the independent variable, for example dividing research subjects into two groups at random, and treating them differently. The treatment is clearly the cause of any significant differences in the dependent variables across the two groups. In analyzing existing questionnaire data, we do not have the luxury of changing anything, so we can seldom be sure which variables are causes of changes in other variables. Thus, when we speak of independent variables, we refer to those variables that we are treating as if they were causes for the purposes of the particular analysis. Dependent variables, similarly, are being temporarily treated as if they were effects, but in another analysis of the same variables might be used as independent variables.

Often it is fairly easy to distinguish cause from effect, simply on the basis of understanding what each variable means, and having some familiarity with the conditions of human life. In the 1999 Arts and Religion Survey, OFT_ATT is a measure of frequency of attendance at religious services. 18.6 percent of all respondents say they never “go to church, synagogue, or some other place of worship” every week. The numbers are very different between males (23.2 percent) and females (14.5 percent), but no reasonable person would conclude that staying out of church makes a person become male. Clearly, for the majority of people, gender is an independent variable, locked in at conception and well-established at birth.

More subtle hypotheses arise when we examine religious attendance across major categories of religious affiliation. The exact same percentage of Protestants and Catholics never attend church – 12.8 percent, but the fraction is higher for the Jewish group, 26.6 percent. Consider this hypothesis: staying away from religious services tends to make a person become Jewish. While not logically impossible, this hypothesis would not make sense to anybody who knew anything about the major western religious traditions. Some non-religious people respect Judaism as their own ethnic heritage, or assuming that is what the survey researcher is asking about, check the “Jewish” box on the questionnaire. In addition, some Jewish families practice their faith within their family households, rather than in public. So there is ample scope for a research project on why so many people who consider themselves to be Jewish stay away from religious services, thus treating “Jewish” as an independent variable whose effect needs to be explained on the dependent variable of “never attending.” But treating “Jewish” as the dependent variable would not make a plausible study.

Among those who are agnostics, atheists, or express no preference, fully 63.6 percent never attend religious services. This may seem completely uninteresting, and one might say: “Of course, people who lack faith will never attend church, so why bother even mentioning that fact.” This explanation. however, is too simplistic. It assumes that being atheist is a permanently independent variable, such as being male or female. One could argue with some plausibility that the reverse is true: Failure to attend religious services over a period of years will tend to make people lose religious faith, to the point that some of them become atheists. A more likely explanation is that both lack of faith and lack of attendance are dependent variables, and the independent variable is something like being raised by parents who were not religious.

The ARDA page for OFT_ATT also gives the number by region of the United States. In the south, just 13.8 percent of respondents never attend religious services, compared with 23.8 percent in the west, and 18.6 percent nationally. To analyze this situation properly, one would need to combine a macrosociological analysis of the regions with analysis at the level of individual respondents, untangling many variables and hypotheses. If a region already has a distinctive religious ecology, people born there or who migrate to that region may conform to social pressures, and adopt the local customs quite apart from their own original religious preferences. The south is the region of the proverbial Bible Belt, which may provide a cultural basis for higher church attendance. A common explanation for low religious attendance in the west of the nation focuses on migration, suggesting that migration causes low church attendance because migrants lack strong social bonds to people who are already members of churches. One could also argue that people who attend church often develop strong social bonds with members of their congregation, so they are less likely to migrate away. Ideally, one would like to have data from a long-term panel study, measuring the variables at several points in time, spaced years apart. One could then see how church attendance in one year relates to the probability of migration in a later year, and how migration relates to church attendance in an even later year.

If one variable is measured much earlier in a person’s life than another variable, then in the absence of any other consideration it is reasonable to treat the first variable as the independent variable. Since most surveys are done only at one point in time, we must usually use a combination of common sense and professional logic to decide. Given a reasonable hypothesis about cause and effect, it is appropriate to be guided by the hypothesis in distinguishing independent form dependent variables.

In a simple comparison of percentages, as above, one should percentage in terms of the independent variable, such that, for example the, different levels of attendance for Jewish respondents add up to 100 percent. It would be much harder to understand the results if we looked at what percent of people who never attended religious services belonged to each religious tradition. Very sophisticated statistical methods exist – such as path analysis – for dealing with complex theories that weave many variables together, so that a given variable may be independent form one hypothesis in the theory, but dependent for another. It is usually wisest to start simple, and build complex causal models only after one is confident one understands all its pieces.

QuickSearch The Knowledge-Base

To search the knowledge-base, enter a term below:

Select a Theory below to learn more:

Select a Concept below to learn more:

TCM Contributors

Would you like to be considered for a position on the Theories, Concepts, and Measures contribution team? If so, click on the link below and complete the TCM Online Request Form.

TCM Online Request Form

If you are already a contributor for Theories, Concepts, and Measures and need assistance in managing the content, please click on the link below for instructions.

Site Administration Instructions