Researchers who investigate human development or social trends face the difficulty of disentangling various effects that lead to change. In such analyses, scientists note that change may occur for three reasons. The first type of change involves social and environmental forces related to the passage of time. These changes are termed secular effects; in this context, secular simply refers to the passage of time and does not imply a contrast with religion or a lack of religion. The second type of change is due to age effects, reflecting physiological changes in individuals. The final type of change relates to cohort effects, the macro conditions that birth cohorts experience over the life span. Because of the close interconnection between age, period, and cohort effects, researchers have noted the importance of teasing apart the relative contributions of these effects. Each effect can contribute independently to change, and there may be interactions among the effects that are not predictable from age, period, and cohort effects individually.
Scientists have developed a number of approaches to identifying the varying causes of change. In some research, investigators may ignore one effect (e.g., age), instead focusing only on the other two (e.g., cohort and period). The obvious drawback here is that one cannot always safely assume that the effect being ignored has not influenced the outcome measure. In other cases, the investigators may examine two of the three effects successively, one effect being temporarily ignored. In this approach, one can identify the relative contribution to an outcome measure by noting whether an effect’s absence in one of the models changes the adequacy of a prediction of the outcome.
One frequent topic associated with age, period, and cohort effects is the prevalence of suicide. An example of such research that reveals the importance of period effects, differentiated from age and cohort effects, compared suicide rates in Australia, the United States, and Canada. John Snowdon and G. E. Hunt (2002) identified an increase in suicide rates for successive cohorts in the United States and Canada in the mid- and late twentieth century. A change in Australian legislation limiting the availability of sedatives led to a different pattern. The historical change regarding therapeutic drugs, a period effect, is important in understanding suicide rates in the population. When such period effects were taken into account, estimates of suicide rates among certain cohorts were comparable in the three countries.
In much of the research, the cohort identified is the birth cohort. Some research creates different categories within the birth cohort. Studies of suicide are again instructive in this regard. In the United States blacks have traditionally been at relatively lower risk for suicide than some other groups in the United States (e.g., whites). Since the early 1980s, however, the rate of attempted and completed suicide has risen among blacks. Research has revealed a higher rate among younger black cohorts, as compared to older cohorts perhaps, as Sean Joe (2006) suggests, because of changing religiosity and greater acceptance of suicide.
With any complex social dynamic, however, period effects are not sufficient to characterize all aspects of a phenomenon. For example, in assessing alcohol consumption across the life span, researchers have documented that drinking decreases as people get older. Some of this effect is attributable simply to age. At the same time there is also evidence that cohort effects are important because of socialization factors. Finally, period effects emerge as important; relevant factors can include, as Mary Gilhooly (2005) notes, availability of alcohol, changes in drinking age, the extent of discretionary time in which to drink, and price.
Psychological and sociological studies have used age-period-cohort analysis, but other disciplines also consider these effects. For instance, research has revealed a cohort-period interaction in the voting patterns of citizens of countries formerly part of the Soviet Union. The data showed that, with the fall of communism, older voters tended to resist change, voting for candidates from the old regime. Younger voters, in contrast, embraced a more liberal approach in their voting. In this case, the period effect related to the introduction of a new economic system. Age as a factor was important, but only, as Sara Schatz (2002) shows, as it interacted with different macro experiences of the various birth cohorts.
Ann Crouter and A. E. Pirretti (2006) observe that in this type of research there are inevitable methodological concerns involving validity of data. Most researchers rely on secondary analyses, using existing datasets. As such researchers studying age, period, and cohort effects have to rely on questions that may not be ideally worded for a given project. Furthermore slight changes in wording across time may have notable effects on results. For instance, in studies on happiness by the Gallup organization, respondents answered a question whose three responses included “very happy,” “fairly happy,” and either “not very happy” or “not at all happy.” When the final response was “not at all happy,” respondents chose “fairly happy” more frequently than when the third alternative was “not very happy.” People were more comfortable declaring that they were “not very happy” but more reluctant to assert that they were “not at all happy.” Such dynamics, as Norval Glenn (2005) notes, complicate the long-term study of behaviors in age, period, and cohort research.
SEE ALSO Alcoholism; Data, Longitudinal; Pollsters; Public Health; Religiosity; Research, Longitudinal; Socialization; Stages of Development; Suicide; Survey; Voting; Voting Patterns
Crouter, Ann C., and A. E. Pirretti. 2006. Longitudinal Research on Work and Family Issues. In The Work and Family Handbook: Multi-Disciplinary Perspectives, Methods, and Approaches, eds. Marcie Pitt-Catsouphes, Ellen Ernst Kossek, and Stephen Sweet. Mawhah, NJ: Lawrence Erlbaum.
Gilhooly, Mary L. M. 2005. Reduced Drinking with Age: Is It Normal? Addiction Research and Theory 13 (3): 267–280.
Glenn, Norval D. 2005. Cohort Analysis. 2nd ed. Thousand Oaks, CA: Sage Publications.
Joe, Sean. 2006. Explaining Changes in the Patterns of Black Suicide in the United States from 1981 to 2002: An Age, Cohort, and Period Analysis. Journal of Black Psychology 32 (3): 262–284.
Portrait, France, Rob Alessie, and Dorly Deeg. 2002. Disentangling the Age, Period, and Cohort Effects Using a Modeling Approach. Tinbergen Institute Working Paper no. 2002–120/3. Social Science Research Network, http://ssrn.com/abstract=360780.
Schatz, Sara. 2002. Age Cohort Voting Effects in the Breakdown of Single-Party Rule. Journal of Aging Studies 16 (2): 199–219.
Snowdon, John, and G. E. Hunt. 2002. Age, Period, and Cohort Effects on Suicide Rates in Australia, 1919–1999. Acta Psychiatrica Scandinavica 105 (4): 265–270.
Bernard C. Beins