- Email: [email protected]
Beyond political socialization: New approaches to age, period, cohort analysis
Electoral Studies 33 (2014) 1–6
Contents lists available at ScienceDirect
Electoral Studies journal homepage: www.elsevier.com/locate/electstud
Editorial
Beyond political socialization: New approaches to age, period, cohort analysis Keywords Cohort analysis Empirical methods Political behavior
Research on political socialization has accumulated a large number of insights about how voters acquire their political attitudes. Yet for all that, we know relatively little about when and why socialization experiences lead to generational differences in how citizens perceive and evaluate politics or behave in the political arena. Recognizing that societies are constantly changing, it is important to identify generational features of the electorate both to understand the present and to make predictions for the future. Ryder’s (1965) seminal article on “the cohort as a concept in the study of social change” was a plea to think about the transformation of society in such a way by taking into account cohort changes and replacements. He famously noted that “society persists despite the mortality of its individual members, through processes of demographic metabolism and particularly the annual infusion of birth cohorts (.). Successive cohorts are differentiated by the changing content of formal education, by peer-group socialization, and by idiosyncratic historical experience” (Ryder, 1965: 843). Based on the importance to emphasizing cohorts he further added that “since cohorts are used to achieve structural transformation and since they manifest its consequences in characteristic ways, it is proposed that research be designed to capitalize on the congruence of social change and cohort identification” (Ryder, 1965: 843). Motivated by Ryder’s message, and drawing on a collection of six papers in Electoral Studies symposium “generational differences in electoral behavior”, Wouter van der Brug and Sylvia Kritzinger not surprisingly conclude that ”if one wants to understand political changes, one must not overlook generational differences” (2012: 248). However, despite the recognition accorded this point, studies of the make-up of political generations are still scarce. A major reason is the methodological challenge posed by questions involving generational turnover and replacement. Some of the papers in the recent Electoral Studies collection explicitly address these 0261-3794/$ – see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.electstud.2013.06.012
problems (Konzelmann et al., 2012; Bhatti and Hansen, 2012), while others acknowledge the limitations their study might have “in distinguishing life-cycle and generational effects” (Walczak et al., 2012: 282). The aim of the collection of papers in our follow-up symposium is to discuss diverse methodological approaches that address the problems that arise from such empirical analyses and provide solutions for overcoming them.2 This special issue hence brings together scholars in the field of political socialization and cohort analysis in an effort to explicate and advance various statistical approaches with reference to a variety of data. The wide availability of panel studies and repeated cross-section surveys, often covering several decades, as well as important methodological advances which have been made in demography, statistics, and sociology, have the potential to promote the importance of age, period, and cohort (APC) analyses and increase our confidence in their results. This paper symposium therefore focuses on new methods of identifying political generations and, more generally, of observing APC effects, which are applied to the area of political behavior and attitudes. The emphasis therefore is on the methods used in order to give political scientists interested in conducting theoretically interesting APC analyses and understanding of how such investigations can and should be conducted. The focus lies especially on cohort effects, as studies investigating these are still scarce in the political science literature or are often too tenuous to draw meaningful conclusions. To set the stage for the articles in this special issue, this introduction provides an overview of APC analysis in general. 1. Defining age, period, and cohort effects Research into the question of why an individual holds specific attitudes or behaves in a certain way might hold three different – but highly-related – factors accountable: aging, enduring intercohort experiences, and time (Yang and Land, 2013). Firstly, we might attribute differences in 2 For an excellent overview of cohort analysis and methods used in to estimate age, period, and cohort effects, see also Yang and Land, 2013.
2
Editorial / Electoral Studies 33 (2014) 1–6
attitudes or behavior to age. Empirical studies often confirm that young and old citizens differ considerably in their political outlook. So-called age effects refer to changes that are associated with basic biological processes or progression through the life-cycle as social roles change with age or as the accumulation of social experience increases. These aging, or life-cycle effects, are usually indexed simply by an individual’s age, though sometimes by a measure of their “place in the life-cycle” (e.g., parent of young children; retired person).3 Secondly, observed attitudes or behavior might be thought of as a function of the current political, economic, or societal situation and idiosyncratic events that produce fluctuations over time and affect all age groups simultaneously. These period effects are usually measured by the current time t, for example, the year of a survey. Thirdly, citizens might differ in their political attitudes because of different socialization experiences which manifest themselves in their belief systems. The resulting cohort effects or, as they are sometimes called, generational effects are defined as “enduring intercohort distinctions that are attributable to the common ‘imprinting’ of cohort members. With regard to attitudinal dependent variables, generational effects are often presumed to be the result of cohort members having shared similar socializing experiences, especially during late adolescence and early adulthood” (Markus, 1983: 718; cf. Mannheim, [1928] 1952; Ryder, 1965). This influential phase in an individual’s lifecycle is often labeled the formative or impressionable years. A cohort is very generally defined as a “number of individuals who have some characteristics in common” (Glenn, 2005: 2) or that “share experiences” (Fienberg and Mason, 1985: 51). Ryder (1965: 845) describes a cohort as “an aggregate of individuals” which has “a distinctive composition and character reflecting the circumstances of its unique origination and history.” Cohorts are most often operationalized by people’s birth years, but they are sometimes divided into equal time periods – such as fiveyear intervals – where the span of years for each cohort may be dictated by theoretical concerns or by data constraints. But cohorts may also be defined with reference to any of a number of variables (e.g., persons who came of age at the same time or individuals who finished high school in a particular year).4 The term cohort analysis is usually used to describe the systematic comparison of two or more cohorts in regard to one dependent variable or a set of related dependent variables (Glenn, 2005: 3). The studies presented in this special issue follow this logic. Fig. 1 illustrates a simple cohort analysis by plotting annual percentages opposing interracial marriage for four different birth cohorts from the
3 Biological processes and place in the life-cycle may not coincide; one could, for example, be the mother of an elementary school child in one’s mid-20s or in one’s mid-40s or beyond. In practice, however, it is rare that a sharp distinction is made between the two concepts. 4 The terms ‘cohort’ and ‘generation’ are often used interchangeably, though generations are usually thought of as connected by some shared historical experience such as having grown up during the Great Depression (Elder, 1974). The boundaries of such events are often imprecise; nevertheless, for purposes of analysis, generations are often operationalized in terms of specific birth years.
United States between 1972 and 2002.5 The idea of such an analysis is to explore whether these cohorts differ in their attitudes and, typically, whether the differences can be attributed to events or attitudes characteristic of the time at which each cohort matured. Here it would seem relatively straightforward. The oldest cohort – born before 1930 – was socialized in a highly racially divided country, whereas the cohort born after 1970 grew up after the turbulent times of the Civil Rights Movements of the 1950s and 1960s, when legal and quasi-legal racial discrimination was abolished.6 The changing historical legacies during the formative years of these four cohorts are assumed to have shaped racial attitudes. According to Fig. 1, the cohort born before 1930 consistently exhibits the highest anti-miscegenation attitudes, with as many as 50 percent opposing interracial marriage in the mid-1970s. Each cohort born and socialized later is less against interracial marriages.7 This simple graph reveals three findings. Firstly, we observe a period effect, as all cohorts seem to become less and less racially intolerant over time. Secondly, the declining, more-or-less parallel lines of each cohort confirm that clear differences exist regarding racial attitudes depending on the time a respondent was born and hence socialized. Thirdly, we note what some people call generational replacement. That is, the thick solid line, which plots the overall trend in anti-racial statements, is not just declining at the same rate as, for example, the cohort born before 1930, but more sharply. Note that after the mid-1990s, the overall trend line is lower than the average attitude among the cohort born in 1930–1950. The explanation for this observation is simply that the weight of the ‘older’ cohorts in the overall population is decreasing as members of these two groups are fewer in number, as they are getting older and eventually dying. Similarly, the graph shows how new cohorts are entering the population, with the post-1970 cohort first included in the General Social Survey in 1989. Overall, it is assumed that cohort analysis or APC analysis in general – as illustrated here – is a method for studying longitudinal patterns of change. 2. Age, period, and cohort analysis in political science Research on age, period, and cohort effects is not new in political science. However, the attention is often on only one of the three. The interest in cohorts evolves mainly around the question of the ‘making of a generation’ side of it,
5 The question wording was as follows: “Do you think there should be laws against marriages between Negroes/Blacks/African-Americans and Whites? – Yes or No.” Fig. 1 plots the percentages agreeing with this statement. The data was taken from the U.S. General Social Survey, which is available annually or bi-annually since 1972. The question was not included after 2002. 6 Among the most important actions for abolishing state-approved discrimination in public life were ratification of the 24thAmendment to the U.S. Constitution (outlawing poll taxes) and passage of the Civil Rights Act of 1964 and the Voting Rights Act of 1965. 7 The small and initially inconsistent difference between the 1951– 1970 and the post-1970 cohorts is likely due to a declining cohort effect but may also be affected by small numbers of respondents when the youngest cohort first entered the analysis.
3
40 20
Born<1930 1930-1950 All 1951-1970 Born>1970
0
Anti- Interracial Marriage (in %)
60
Editorial / Electoral Studies 33 (2014) 1–6
1970
1975
1980
1985
1990
1995
2000
2005
Survey year Fig. 1. Cohort differences regarding interracial marriage.
focusing on political socialization. The classic definition of political socialization was established by Hyman (1959: 25) as an individual’s “learning of social patterns corresponding to his societal position as mediated through various agencies of society.” Since then, a focus on learning has produced a large volume of research on the formation of political attitudes and behavior (Adelson and O’Neil, 1966; Easton and Dennis, 1969; Jennings and Niemi, 1974, 1981; Niemi and Junn, 1998; Sears and Valentino, 1997; Campbell, 2006; Zuckerman et al., 2007; Sherrod et al., 2010). An important implication of the research on political socialization is the assumption that it produces “relatively enduring orientations toward politics in general” (Merelman, 1986: 279; emphasis added). In the beginning of the scientific analysis of the formation of political ideas, the focus was mainly on young children, as it was assumed, more or less unquestioningly, that political attitudes were determined very early in life (cf. Easton and Dennis, 1969). Often it was assumed that what was learned prior to adulthood remained unchanged in later life.8 It was also frequently believed that what is learned earliest in life is most important, as it served as an unchanging value basis for later attitudes (Niemi and Hepburn, 1995). However, as it became evident that political ideas that develop during childhood are revised due to college experience and general maturation (Searing et al., 1973), the scientific discussion shifted the research focus away from early political learning to more in-depth studies of aging. Most especially, Marsh’s (1971) critique of earlier studies challenged the assumption that “adult opinions are in large part the end product of political socialization” (1971: 455). He noted that extensive carryover into adulthood may apply only to important personality variables, whereas the influence of early socialization on political attitudes remained uncertain. The research shifted accordingly from attitude stability to the conceptualization
8 There was, initially, an empirical basis for this belief in that analyses of recalled party identification suggested that partisanship, once formed, rarely changed (Campbell et al., 1960, ch. 7).
of socialization as a process of individual development and learning. However, even studies that confirmed that political learning extends well beyond childhood acknowledged that young adults change more rapidly than older adults (Jennings and Niemi, 1981; Alwin and Krosnick, 1991; Jennings et al., 1989, 2009; Niemi and Jennings, 1991). As a consequence, life-cycle and aging effects became more and more the center of attention. Moreover, the research focuses also shifted to possible period effects that alter political attitudes. Party identification is a central concept in the study of socialization, including both life-cycle and period effects, and analyses of partisanship served as the main battlefield for the different proponents. Some scholars thought of partisanship less as an identity – being stable over the life-cycle – and more of an attitude that comes about as a function of informed reactions to the performance of governments and opposition parties on a number of policy areas, most notably the economy (Ordeshook, 1976; Fiorina, 1981; Page and Jones, 1979; Franklin and Jackson, 1983; MacKuen et al., 1989). As governments and economic good times are never permanent, an individual’s affiliation with a political party is always subject to ‘rational updating’.9 Hence this research tries to uncover how the nature of the current time affects the direction and strength of certain political attitudes, including ones so seemingly basic as partisanship.10 Scientific debates about age and period effects as well as more broadly on the origins of political attitudes and behavior are often unconnected to one another. Cohort effects have been discussed rather sporadically, even though it was acknowledged that it is important to take
9 One-time events and suddenly changing circumstances can also exert relatively uniform effects on citizens, as in the mid-1960s when partisanship suddenly weakened in the U.S. and as African-Americans shifted their partisanship from Republic to Democratic (Abramson, 1975). 10 Of course researchers outside the United States have from early on viewed partisanship as reacting strongly to current policy and party preferences. See especially the influential article by Thomassen (1976).
4
Editorial / Electoral Studies 33 (2014) 1–6
these into account as well.11 Conover (1991: 130), for example, argued that life-cycle and cohort effects are interwoven, as “people change in political orientations throughout their life, (but) generations respond differently to the same events.” Also Inglehart shows in his famous studies on value change (Inglehart, 1977; Abramson and Inglehart, 1995) that later learning must overcome the inertia of pre-existing orientations. Jennings (1996: 249) adds to the discussion by noting that “what each cohort brings into political maturity has a good deal of continuity and provides a certain degree of stability in terms of what that cohort is likely to draw on as it moves through the rest of the life-cycle.” But as Niemi and Sobieszek (1977: 228) pointed out 20 years earlier, “sorting out the relative impact of life-cycle, generational, and period effects will no doubt prove to be extraordinarily complicated.” 3. The tenacious problem of age, period and cohort analysis The problem of sorting out age, period, and cohort effects is evident in Fig. 1, as our description of these results fails to take into account a possible aging effect. This shortcoming gets to the core of the problem of any APC analysis. The difficulty in making claims about any of the three factors, whether the focus is on cohort, age, or period effects, is the requirement of accounting for all three simultaneously. If we fail to do so, we cannot know whether an observed attitude or behavior (Yijt) of an individual i (i ¼ 1, ..., I) is because she belongs to a specific cohort j (Cj; j ¼ 1, ..., J) or because of her age (Ait) or because of the current time t (Pt; t ¼ 1, ..., T). This can be summarized as
Yijt ¼ f Ait ; Cj ; Pt
(1)
It is not possible to disentangle the effects of age, cohort, and periods in this generalized linear model (1) without some kind of restrictions on the function f (Mason and Fienberg, 1985: 3; cf. Mason et al., 1973). The problem results from the linearly dependent relationship between age and the birth cohort a respondent belongs to at any given time t – assuming that all three effects are measured in the same time units, for example, years – which is typically expressed as
Cj ¼ Pt Ait
(2)
As can be seen from Eq. (2), the three effects cannot be identified with survey data from one time point, since one phenomenon would be completely determined by the remaining two. For example, knowing that a survey respondent who was interviewed in 2000 was 30 years old, we can infer approximately when she was born or to which birth cohort she belongs to. Or put differently, knowing that
11 Studies that explicitly focus on APC analysis: Baker (1978); Abramson (1979); Claggett (1981); Markus (1983); Miller (1992); Tilley (2002) and Tilley and Evans (2011) – partisanship; Klecka (1971); Prior (2010) – political interest; Watts (1999); Lyons and Alexander (2000); Franklin et al. (2004) – turnout/participation; Cutler and Kaufman (1975) – ideology; Jennings (1996) – political knowledge; Jennings and Stoker (2004) – civic engagement; Mishler and Rose (2007); Neundorf (2010); and Mattes (2012) – democratic attitudes. This list is by no means exhaustive.
a respondent was born in 1970 when interviewed in 2000, we can deduce that this person was about 30 years old at the time of the interview. That is, once the values of any two factors, such as cohort and period, are known, the value of the third factor is completely determined. Hence, model (1) is not identified. To date, the suggested strategy for identification of APC models has been to specify some set of restrictions on the estimated age, period, and cohort parameters that allow the model to be identified. Firstly, best known is the grouping of birth years into five-year cohorts, which forces grouped birth-years to have identical effects (Fienberg and Mason, 1979; Mason et al., 1973). Secondly, it is also possible to remove one of either age, period, or cohort entirely on the assumption that it does not matter at all (Converse, 1976, chs. 2–3). Thirdly, one can examine higher order transformations of age, period, or cohort, such as age squared (Fienberg and Mason, 1985). Fourthly, one can use a ‘proxy’ variable approach that assumes the cohort or period effects are proportional to certain measured variables (Rodgers, 1982; Heckman and Robb, 1985; O’Brien, 2000). Of course, none of these approaches solves the underlying problem, and none has proven entirely satisfactory (Glenn, 2005). 4. This symposium: new opportunities in age, period and cohort analysis Today’s much expanded data availability allows a promising new range of statistical methods to conduct APC analyses. Researchers today have the happy prospect of using a broad range of data that is often freely available for our purposes. APC analysis is typically conducted on one of diverse types of data (Harding, 2009: 1450). Firstly, aggregated data containing means or counts of the outcome of interest for various age groups at different time points can be used. When such data is arranged in a period-by-age table, cohort differences can be followed on the diagonals. Most previous studies focus on such aggregated population-level contingency tables for conventional multiple classification. However, these descriptive methods do not allow one to account for any factors other than age, period, and cohort. Increasingly, however, individual-level data sets in the form of a series of repeated cross-section sample surveys or panel studies are available to political scientists. These create both new opportunities and challenges to APC analysis. New data availability and the development of innovative methods offer new paths to study theoretically relevant age, cohort-, and period-related research questions. The articles of this special issue cover several of these new methods. The aim of this collection of eight papers is to introduce diverse methodological approaches to illustrate the problems of APC analyses and propose possible solutions to varied research questions. Some papers focus on identifying cohort effects (Tilley and Evans, 2014; Pop-Eleches and Tucker, 2014; Grasso, 2014), while others try to single out factors that explain cohort differences (Dinas and Stoker, 2014; Smets and Neundorf, 2014; Stegmueller, 2014). Still others emphasize the age-factor in their APC analyses (Bartels and Jackman, 2014; Kroh, 2014). More specifically, Tilley and Evans provide an example how to combine different individual-level data sources
Editorial / Electoral Studies 33 (2014) 1–6
(panel studies and repeated cross-sectional data) and aggregated election results. Using the side-information of these diverse sources makes it possible to estimate true age and cohort effects. They apply their approach to the question of whether support for the British Conservative Party is driven by aging effects or generational replacement. Bartels and Jackman employ their mathematical model of political learning to partisanship, in this case in the US. However, their focus lies on age-effects, estimating the critical years of heightened impact of period-specific effects or historical shocks on a citizen’s party attachments. They show that individuals of different ages attach different weights to political events. Their paper is one of the first empirical tests of the exact ages of the so-called formative years. The idea of historical shocks is also utilized by the design-based approach by Dinas and Stoker. Focusing on cohort effects, they introduce the language of experiments to APC analysis. The authors propose a method to test the impact of the socialization environment (“treatment”) on long-lasting attitudes. In order to test whether specific treatments, such as changes in enfranchisement, affect female turnout propensities, Dinas and Stoker use control groups (men, for one) to aid in the identification of cohort effects while also accounting for age and period effects. Using their approach, researchers have to consider who got a historical “treatment” and who did not. The idea to single out factors that account for cohort differences is also discussed by Smets and Neundorf. Their paper is proposing an approach that enables researchers to test whether substantively interesting factors can account for any observed cohort heterogeneity. For this the variance of cohort differences in turnout propensities is modeled in a multilevel structure. Stegmueller further uses this approach and extends the cross-classified model by allowing random effects (cohort and period) to be time-structured. Doing so, his model links theoretical ideas about the nature of social change, which is – in the absence of strong shocks – gradual and evolutionary. By using time-structured priors for cohort and time-period random effects, he explicitly models dependence of new cohorts and periods on the past. As mentioned above, a widely used approach in cohort analysis is the categorization of the APC variables. The study by Pop-Eleches and Tucker categorizes historical socialization periods in a meaningful way that identifies aging and generational effects of experiences during Communism on people’s democratic and economic attitudes. Grasso further develops this approach by proposing non-parametric methods for testing the validity of these pre-defined categories. This approach moreover allows one to assess the functional form of the cohort effect, while at the same time accounting for possible life-cycle and period influences. Both papers discuss the problems arising from comparative research due to limited data availability (for example, only a few time points are covered by comparable data sources). All the papers mentioned use repeated cross-sectional surveys that often cover several decades. However, it is further possible to use individual panel data, which allows one to observe the same respondents belonging to different generations over time as they age. Kroh uses such a panel study that covers 27 years, applying growth curve modeling. Using his approach it is possible to exchange age
5
as a function of life years to more meaningful variables that measure age effect, as in his empirical example experience and exposure with and to the political system. In introducing these diverse methodological approaches, all papers use examples relevant in the field of political behavior and attitudes, including political interest, voter turnout, partisanship, and voting behavior. Acknowledgements We would like to thank the Department of Politics and International Relations as well as Nuffield College at the University of Oxford and the Department of Political Science at the University of Rochester for kindly funding a workshop meeting to put this paper symposium together. References Abramson, P., Inglehart, R., 1995. Value Change in Global Perspective. University of Michigan Press, Ann Arbor. Abramson, P., 1975. Generational Change in American Politics. Lexington Books, Lexington, MA. Abramson, P., 1979. Developing party identification: a further examination of life-cycle, generational, and period effects. American Journal of Political Science 23 (1), 78–96. Adelson, J., O’Neil, R., 1966. Growth of political ideas in adolescence: the sense of community. Journal of Personality and Social Psychology 4 (3), 295–306. Alwin, D.F., Krosnick, J.A., 1991. Aging, cohorts and the stability of sociopolitical orientations over the life span. American Journal of Sociology 97 (1), 169–195. Baker, K.L., 1978. Generational differences in the role of party identification in German political behavior. American Journal of Political Science 22 (1), 106–129. Bartels, L., Jackman, S., 2014. A generational model of political learning. Electoral Studies 33, 7–18. Bhatti, Y., Hansen, K.M., 2012. The effect of generation and age on turnout to the European Parliament – how turnout will continue to decline in the future. Electoral Studies 31, 262–272. van der Brug, W., Kritzinger, S., 2012. Generational differences in electoral behaviour. Electoral Studies 31 (2), 245–249. Campbell, A., Converse, P.E., Miller, W.E., Stokes, D.E., 1960. The American Voter. Wiley, New York. Campbell, D., 2006. Why We Vote: How Schools and Communities Shape our Civic Life. Princeton University Press, Princeton. Claggett, W., 1981. Partisan acquisition versus partisan intensity: lifecycle, generation and period effects, 1952–1976. American Journal of Political Science 25 (2), 193–214. Conover, P., 1991. Political socialization: where’s the politics?. In: Crotty, W. (Ed.), Political Science: Looking to the Future, vol. 3 Northwestern University Press, Evanston, pp. 125–152. Converse, P., 1976. The Dynamics of Party Support: Cohort-analyzing Party Identification. Sage, Beverly Hills. Cutler, S., Kaufman, R., 1975. Cohort changes in political attitudes: tolerance of ideological nonconformity. Public Opinion Quarterly 39 (1), 69–81. Dinas, E., Stoker, L., 2014. Age–period–cohort analysis: a design-based approach. Electoral Studies 33, 28–40. Easton, D., Dennis, J., 1969. Children in the Political System: Origins of Political Legitimacy. McGraw-Hill, New York. Elder, G., 1974. Children of the Great Depression: Social Change in Life Experience. University of Chicago Press, Chicago. Fienberg, S.E., Mason, W.M., 1979. Identification and estimation of age– period–cohort models in the analysis of discrete archival data. Sociological Methodology 10, 1–67. Fienberg, S.E., Mason, W.M., 1985. Specification and implementation of age, period, and cohort models. In: Mason, W.M., Fienberg, S.E. (Eds.), Cohort Analysis in Social Research: Beyond the Identification Problem. Springer, New York, pp. 45–88. Fiorina, M.P., 1981. Retrospective Voting in American National Elections. Yale University Press, New Haven. Franklin, C.H., Jackson, J.E., 1983. The dynamics of party identification. American Political Science Review 77 (4), 957–973. Franklin, M.N., Lyons, P., Marsh, M., 2004. Generational basis of turnout decline in established democracies. Acta Politica 39 (2), 115–151.
6
Editorial / Electoral Studies 33 (2014) 1–6
Glenn, N.D., 2005. Cohort Analysis, second ed. Sage, Thousand Oaks, CA. Grasso, M., 2014. Age, period and cohort analysis in a comparative context: political generations and political participation repertoires in Western Europe. Electoral Studies 33, 63–76. Harding, D.J., 2009. Recent advances in age–period–cohort analysis. a commentary on Dregan and Armstrong, and on Reither, Hauser and Yang. Social Science & Medicine 69 (10), 1449–1451. Heckman, J., Robb, R., 1985. Using longitudinal data to estimate age, period, and cohort effects in earnings equations. In: Mason, W.M., Fienberg, S.E. (Eds.), Cohort Analysis in Social Research: Beyond the Identification Problem. Springer, New York, pp. 137–150. Hyman, H., 1959. Political Socialization: A Study in the Psychology of Political Behavior. Free Press, Glencoe, IL. Inglehart, R., 1977. The Silent Revolution: Changing Values and Political Styles among Western Publics. Princeton University Press, Princeton. Jennings, M.K., Niemi, R.G., 1974. The Political Character of Adolescence. Princeton University Press, Princeton. Jennings, M.K., Niemi, R.G., 1981. Generations and Politics: A Panel Study of Young Adults and Their Parents. Princeton University Press, Princeton. Jennings, M.K., Stoker, L., 2004. Social trust and civic engagement across time and generations. Acta Politica 39 (4), 342–379. Jennings, M.K., 1989. The crystallization of orientations. In: Jennings, M.K., van Deth, J., et al. (Eds.), Continuities in Political Action. de Gruyter, Berlin and New York, pp. 313–348. Jennings, M.K., Stoker, L., Bowers, J., 2009. Politics across generations: family transmission reexamined. Journal of Politics 71 (3), 782–799. Jennings, M.K., 1996. Political knowledge over time and across generations. Public Opinion Quarterly 60 (2), 228–252. Klecka, W.R., 1971. Applying political generations to the study of political behavior: a cohort analysis. Public Opinion Quarterly 35 (3), 358–373. Konzelmann, L., Wagner, C., Rattinger, H., 2012. Turnout in Germany in the course of time: life cycle and cohort effects on electoral turnout from 1953 to 2049. Electoral Studies 31 (2), 250–261. Kroh, M., 2014. Growth trajectories in the strength of party identification: The legacy of autocratic regimes. Electoral Studies 33, 90–101. Lyons, W., Alexander, R., 2000. A tale of two electorates: generational replacement and the decline of voting in presidential elections. Journal of Politics 62 (4), 1014–1034. MacKuen, M.B., Erikson, R.S., Stimson, J.A., 1989. Macropartisanship. American Political Science Review 83 (4), 1125–1142. Mannheim, K., 1952. The problem of generations. In: Keckemeti, P. (Ed.), Essays on the Sociology of Knowledge. Routledge and Kegan Paul, London, pp. 276–322 [1928]. Markus, G.B., 1983. Dynamic modeling of cohort change: the case of political partisanship. American Journal of Political Science 27 (4), 717–739. Marsh, D., 1971. Political socialization: the implicit assumptions questioned. British Journal of Political Science 1 (4), 453–465. Mason, K.O., Mason, W.M., Winsborough, H., Poole, W., 1973. Some methodological issues in cohort analysis of archival data. American Sociological Review 38 (2), 242–258. Mason, W.M., Fienberg, S.E. (Eds.), 1985. Cohort Analysis in Social Research: Beyond the Identification Problem. Springer Verlag, New York. Mattes, R., 2012. The “born frees”: the prospects for generational change in post-Apartheid South Africa. Australian Journal of Political Science 47 (1), 133–153. Merelman, R., 1986. Revitalizing political socialization. In: Herman, M. (Ed.), Political Psychology: Contemporary Problems and Issues. Jossey Bass, San Francisco. Miller, W.E., 1992. Generational changes and party identification. Political Behavior 14 (3), 333–352. Mishler, W., Rose, R., 2007. Generation, age, and time: the dynamics of political learning during Russia’s transformation. American Journal of Political Science 51 (4), 822–834. Neundorf, A., 2010. Democracy in transition: a micro perspective on system change in post-socialist societies. Journal of Politics 72 (4), 1096–1108. Niemi, R., Hepburn, M., 1995. The rebirth of political socialization. Perspectives on Political Science 24 (1), 7–16. Niemi, R.G., Jennings, M.K., 1991. Issues and inheritance in the formation of party identification. American Journal of Political Science 35 (3), 969–988. Niemi, R.G., Junn, J., 1998. Civic Education: What Makes Students Learn. Yale University Press, New Haven. Niemi, R.G., Sobieszek, B., 1977. Political socialization. Annual Review of Sociology 3 (1), 209–233.
O’Brien, R.M., 2000. Age period cohort characteristic models. Social Science Research 29 (1), 123–139. Ordeshook, P.C., 1976. The spatial theory of elections: a review and a critique. In: Budge, I., Crewe, I., Farlie, D. (Eds.), Party Identification and Beyond. Wiley, London. Page, B., Jones, C.C., 1979. Reciprocal effects of policy preferences, party loyalties and the vote. American Political Science Review 73 (4), 1071–1089. Pop-Eleches, G., Tucker, J., 2014. Communist socialization and postcommunist economic and political attitudes. Electoral Studies 33, 77–89. Prior, M., 2010. You’ve either got it or you don’t? The stability of political interest over the life cycle. Journal of Politics 72 (3), 747–766. Rodgers, W.L., 1982. Estimable functions of age, period, and cohort effects. American Sociological Review 47 (6), 774–787. Ryder, N.B., 1965. The cohort as a concept in the study of social change. American Sociological Review 30 (6), 843–861. Searing, D., Schwartz, J., Lind, A., 1973. The structuring principle: political socialization and belief systems. American Political Science Review 57 (2), 415–432. Sears, D.O., Valentino, N.A., 1997. Politics matters: political events as catalysts for preadult socialization. American Political Science Review 91 (1), 45–65. Sherrod, L., Torney-Purta, J., Flanagan, C., 2010. Handbook of Research on Civic Engagement in Youth. Wiley, New York. Smets, K., Neundorf, A., 2014. The hierarchies of age–period–cohort effects: political context and the development of generational turnout patterns. Electoral Studies 33, 41–51. Stegmueller, D., 2014. Bayesian hierarchical age–period–cohort models with time-structures effects: an application to religious voting in the US, 1972–2008. Electoral Studies 33, 52–62. Thomassen, J., 1976. Party identification as a cross-national concept: its meaning in The Netherlands. In: Budge, I., Crewe, I., Farlie, D. (Eds.), Party Identification and Beyond. Wiley, London. Tilley, J., Evans, G., 2011. Political generations in Northern Ireland. European Journal of Political Research 50 (5), 583–608. Tilley, J., Evans, G., 2013. Ageing and generational effects on vote choice: combining cross-sectional and panel data to estimate APC effects. Electoral Studies 33, 19–27. Tilley, J., 2002. Political generations and partisanship in the UK, 1964– 1997. Journal of the Royal Statistical Society. Series A (Statistics in Society) 165 (1), 121–135. Walczak, A., van der Brug, W., de Vries, C., 2012. Long- and short-term determinants of party preferences: inter-generational differences in Western and East Central Europe. Electoral Studies 31 (2), 273–284. Watts, M., 1999. Are there typical age curves in political behavior? The “age invariance” hypothesis and political socialization. Political Psychology 20 (3), 477–499. Yang, Y., Land, K.C., 2013. Age–Period–Cohort Analysis: New Models, Methods, and Empirical Applications. CRC Press. Taylor & Francis Group, Boca Raton. Zuckerman, A.S., Dasovi c, J., Fitzgerald, J., 2007. Partisan Families: The Social Logic of Bounded Partisanship in Germany and Britain. Cambridge University Press, Cambridge.
Anja Neundorf* School of Politics and IR, University of Nottingham, Nottingham NG7 2RD, United Kingdom Richard G. Niemi1 Department of Political Science, University of Rochester, Rochester, NY 14627-0146, United States E-mail address: [email protected] Corresponding author. Tel.: þ44 (0)115 95 14795; fax: þ44 (0)794 694 7367. E-mail address: [email protected] 18 June 2013
1
Tel.: þ1 585 275 5364; fax: þ1 585 271 1616.
Contents lists available at ScienceDirect
Electoral Studies journal homepage: www.elsevier.com/locate/electstud
Editorial
Beyond political socialization: New approaches to age, period, cohort analysis Keywords Cohort analysis Empirical methods Political behavior
Research on political socialization has accumulated a large number of insights about how voters acquire their political attitudes. Yet for all that, we know relatively little about when and why socialization experiences lead to generational differences in how citizens perceive and evaluate politics or behave in the political arena. Recognizing that societies are constantly changing, it is important to identify generational features of the electorate both to understand the present and to make predictions for the future. Ryder’s (1965) seminal article on “the cohort as a concept in the study of social change” was a plea to think about the transformation of society in such a way by taking into account cohort changes and replacements. He famously noted that “society persists despite the mortality of its individual members, through processes of demographic metabolism and particularly the annual infusion of birth cohorts (.). Successive cohorts are differentiated by the changing content of formal education, by peer-group socialization, and by idiosyncratic historical experience” (Ryder, 1965: 843). Based on the importance to emphasizing cohorts he further added that “since cohorts are used to achieve structural transformation and since they manifest its consequences in characteristic ways, it is proposed that research be designed to capitalize on the congruence of social change and cohort identification” (Ryder, 1965: 843). Motivated by Ryder’s message, and drawing on a collection of six papers in Electoral Studies symposium “generational differences in electoral behavior”, Wouter van der Brug and Sylvia Kritzinger not surprisingly conclude that ”if one wants to understand political changes, one must not overlook generational differences” (2012: 248). However, despite the recognition accorded this point, studies of the make-up of political generations are still scarce. A major reason is the methodological challenge posed by questions involving generational turnover and replacement. Some of the papers in the recent Electoral Studies collection explicitly address these 0261-3794/$ – see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.electstud.2013.06.012
problems (Konzelmann et al., 2012; Bhatti and Hansen, 2012), while others acknowledge the limitations their study might have “in distinguishing life-cycle and generational effects” (Walczak et al., 2012: 282). The aim of the collection of papers in our follow-up symposium is to discuss diverse methodological approaches that address the problems that arise from such empirical analyses and provide solutions for overcoming them.2 This special issue hence brings together scholars in the field of political socialization and cohort analysis in an effort to explicate and advance various statistical approaches with reference to a variety of data. The wide availability of panel studies and repeated cross-section surveys, often covering several decades, as well as important methodological advances which have been made in demography, statistics, and sociology, have the potential to promote the importance of age, period, and cohort (APC) analyses and increase our confidence in their results. This paper symposium therefore focuses on new methods of identifying political generations and, more generally, of observing APC effects, which are applied to the area of political behavior and attitudes. The emphasis therefore is on the methods used in order to give political scientists interested in conducting theoretically interesting APC analyses and understanding of how such investigations can and should be conducted. The focus lies especially on cohort effects, as studies investigating these are still scarce in the political science literature or are often too tenuous to draw meaningful conclusions. To set the stage for the articles in this special issue, this introduction provides an overview of APC analysis in general. 1. Defining age, period, and cohort effects Research into the question of why an individual holds specific attitudes or behaves in a certain way might hold three different – but highly-related – factors accountable: aging, enduring intercohort experiences, and time (Yang and Land, 2013). Firstly, we might attribute differences in 2 For an excellent overview of cohort analysis and methods used in to estimate age, period, and cohort effects, see also Yang and Land, 2013.
2
Editorial / Electoral Studies 33 (2014) 1–6
attitudes or behavior to age. Empirical studies often confirm that young and old citizens differ considerably in their political outlook. So-called age effects refer to changes that are associated with basic biological processes or progression through the life-cycle as social roles change with age or as the accumulation of social experience increases. These aging, or life-cycle effects, are usually indexed simply by an individual’s age, though sometimes by a measure of their “place in the life-cycle” (e.g., parent of young children; retired person).3 Secondly, observed attitudes or behavior might be thought of as a function of the current political, economic, or societal situation and idiosyncratic events that produce fluctuations over time and affect all age groups simultaneously. These period effects are usually measured by the current time t, for example, the year of a survey. Thirdly, citizens might differ in their political attitudes because of different socialization experiences which manifest themselves in their belief systems. The resulting cohort effects or, as they are sometimes called, generational effects are defined as “enduring intercohort distinctions that are attributable to the common ‘imprinting’ of cohort members. With regard to attitudinal dependent variables, generational effects are often presumed to be the result of cohort members having shared similar socializing experiences, especially during late adolescence and early adulthood” (Markus, 1983: 718; cf. Mannheim, [1928] 1952; Ryder, 1965). This influential phase in an individual’s lifecycle is often labeled the formative or impressionable years. A cohort is very generally defined as a “number of individuals who have some characteristics in common” (Glenn, 2005: 2) or that “share experiences” (Fienberg and Mason, 1985: 51). Ryder (1965: 845) describes a cohort as “an aggregate of individuals” which has “a distinctive composition and character reflecting the circumstances of its unique origination and history.” Cohorts are most often operationalized by people’s birth years, but they are sometimes divided into equal time periods – such as fiveyear intervals – where the span of years for each cohort may be dictated by theoretical concerns or by data constraints. But cohorts may also be defined with reference to any of a number of variables (e.g., persons who came of age at the same time or individuals who finished high school in a particular year).4 The term cohort analysis is usually used to describe the systematic comparison of two or more cohorts in regard to one dependent variable or a set of related dependent variables (Glenn, 2005: 3). The studies presented in this special issue follow this logic. Fig. 1 illustrates a simple cohort analysis by plotting annual percentages opposing interracial marriage for four different birth cohorts from the
3 Biological processes and place in the life-cycle may not coincide; one could, for example, be the mother of an elementary school child in one’s mid-20s or in one’s mid-40s or beyond. In practice, however, it is rare that a sharp distinction is made between the two concepts. 4 The terms ‘cohort’ and ‘generation’ are often used interchangeably, though generations are usually thought of as connected by some shared historical experience such as having grown up during the Great Depression (Elder, 1974). The boundaries of such events are often imprecise; nevertheless, for purposes of analysis, generations are often operationalized in terms of specific birth years.
United States between 1972 and 2002.5 The idea of such an analysis is to explore whether these cohorts differ in their attitudes and, typically, whether the differences can be attributed to events or attitudes characteristic of the time at which each cohort matured. Here it would seem relatively straightforward. The oldest cohort – born before 1930 – was socialized in a highly racially divided country, whereas the cohort born after 1970 grew up after the turbulent times of the Civil Rights Movements of the 1950s and 1960s, when legal and quasi-legal racial discrimination was abolished.6 The changing historical legacies during the formative years of these four cohorts are assumed to have shaped racial attitudes. According to Fig. 1, the cohort born before 1930 consistently exhibits the highest anti-miscegenation attitudes, with as many as 50 percent opposing interracial marriage in the mid-1970s. Each cohort born and socialized later is less against interracial marriages.7 This simple graph reveals three findings. Firstly, we observe a period effect, as all cohorts seem to become less and less racially intolerant over time. Secondly, the declining, more-or-less parallel lines of each cohort confirm that clear differences exist regarding racial attitudes depending on the time a respondent was born and hence socialized. Thirdly, we note what some people call generational replacement. That is, the thick solid line, which plots the overall trend in anti-racial statements, is not just declining at the same rate as, for example, the cohort born before 1930, but more sharply. Note that after the mid-1990s, the overall trend line is lower than the average attitude among the cohort born in 1930–1950. The explanation for this observation is simply that the weight of the ‘older’ cohorts in the overall population is decreasing as members of these two groups are fewer in number, as they are getting older and eventually dying. Similarly, the graph shows how new cohorts are entering the population, with the post-1970 cohort first included in the General Social Survey in 1989. Overall, it is assumed that cohort analysis or APC analysis in general – as illustrated here – is a method for studying longitudinal patterns of change. 2. Age, period, and cohort analysis in political science Research on age, period, and cohort effects is not new in political science. However, the attention is often on only one of the three. The interest in cohorts evolves mainly around the question of the ‘making of a generation’ side of it,
5 The question wording was as follows: “Do you think there should be laws against marriages between Negroes/Blacks/African-Americans and Whites? – Yes or No.” Fig. 1 plots the percentages agreeing with this statement. The data was taken from the U.S. General Social Survey, which is available annually or bi-annually since 1972. The question was not included after 2002. 6 Among the most important actions for abolishing state-approved discrimination in public life were ratification of the 24thAmendment to the U.S. Constitution (outlawing poll taxes) and passage of the Civil Rights Act of 1964 and the Voting Rights Act of 1965. 7 The small and initially inconsistent difference between the 1951– 1970 and the post-1970 cohorts is likely due to a declining cohort effect but may also be affected by small numbers of respondents when the youngest cohort first entered the analysis.
3
40 20
Born<1930 1930-1950 All 1951-1970 Born>1970
0
Anti- Interracial Marriage (in %)
60
Editorial / Electoral Studies 33 (2014) 1–6
1970
1975
1980
1985
1990
1995
2000
2005
Survey year Fig. 1. Cohort differences regarding interracial marriage.
focusing on political socialization. The classic definition of political socialization was established by Hyman (1959: 25) as an individual’s “learning of social patterns corresponding to his societal position as mediated through various agencies of society.” Since then, a focus on learning has produced a large volume of research on the formation of political attitudes and behavior (Adelson and O’Neil, 1966; Easton and Dennis, 1969; Jennings and Niemi, 1974, 1981; Niemi and Junn, 1998; Sears and Valentino, 1997; Campbell, 2006; Zuckerman et al., 2007; Sherrod et al., 2010). An important implication of the research on political socialization is the assumption that it produces “relatively enduring orientations toward politics in general” (Merelman, 1986: 279; emphasis added). In the beginning of the scientific analysis of the formation of political ideas, the focus was mainly on young children, as it was assumed, more or less unquestioningly, that political attitudes were determined very early in life (cf. Easton and Dennis, 1969). Often it was assumed that what was learned prior to adulthood remained unchanged in later life.8 It was also frequently believed that what is learned earliest in life is most important, as it served as an unchanging value basis for later attitudes (Niemi and Hepburn, 1995). However, as it became evident that political ideas that develop during childhood are revised due to college experience and general maturation (Searing et al., 1973), the scientific discussion shifted the research focus away from early political learning to more in-depth studies of aging. Most especially, Marsh’s (1971) critique of earlier studies challenged the assumption that “adult opinions are in large part the end product of political socialization” (1971: 455). He noted that extensive carryover into adulthood may apply only to important personality variables, whereas the influence of early socialization on political attitudes remained uncertain. The research shifted accordingly from attitude stability to the conceptualization
8 There was, initially, an empirical basis for this belief in that analyses of recalled party identification suggested that partisanship, once formed, rarely changed (Campbell et al., 1960, ch. 7).
of socialization as a process of individual development and learning. However, even studies that confirmed that political learning extends well beyond childhood acknowledged that young adults change more rapidly than older adults (Jennings and Niemi, 1981; Alwin and Krosnick, 1991; Jennings et al., 1989, 2009; Niemi and Jennings, 1991). As a consequence, life-cycle and aging effects became more and more the center of attention. Moreover, the research focuses also shifted to possible period effects that alter political attitudes. Party identification is a central concept in the study of socialization, including both life-cycle and period effects, and analyses of partisanship served as the main battlefield for the different proponents. Some scholars thought of partisanship less as an identity – being stable over the life-cycle – and more of an attitude that comes about as a function of informed reactions to the performance of governments and opposition parties on a number of policy areas, most notably the economy (Ordeshook, 1976; Fiorina, 1981; Page and Jones, 1979; Franklin and Jackson, 1983; MacKuen et al., 1989). As governments and economic good times are never permanent, an individual’s affiliation with a political party is always subject to ‘rational updating’.9 Hence this research tries to uncover how the nature of the current time affects the direction and strength of certain political attitudes, including ones so seemingly basic as partisanship.10 Scientific debates about age and period effects as well as more broadly on the origins of political attitudes and behavior are often unconnected to one another. Cohort effects have been discussed rather sporadically, even though it was acknowledged that it is important to take
9 One-time events and suddenly changing circumstances can also exert relatively uniform effects on citizens, as in the mid-1960s when partisanship suddenly weakened in the U.S. and as African-Americans shifted their partisanship from Republic to Democratic (Abramson, 1975). 10 Of course researchers outside the United States have from early on viewed partisanship as reacting strongly to current policy and party preferences. See especially the influential article by Thomassen (1976).
4
Editorial / Electoral Studies 33 (2014) 1–6
these into account as well.11 Conover (1991: 130), for example, argued that life-cycle and cohort effects are interwoven, as “people change in political orientations throughout their life, (but) generations respond differently to the same events.” Also Inglehart shows in his famous studies on value change (Inglehart, 1977; Abramson and Inglehart, 1995) that later learning must overcome the inertia of pre-existing orientations. Jennings (1996: 249) adds to the discussion by noting that “what each cohort brings into political maturity has a good deal of continuity and provides a certain degree of stability in terms of what that cohort is likely to draw on as it moves through the rest of the life-cycle.” But as Niemi and Sobieszek (1977: 228) pointed out 20 years earlier, “sorting out the relative impact of life-cycle, generational, and period effects will no doubt prove to be extraordinarily complicated.” 3. The tenacious problem of age, period and cohort analysis The problem of sorting out age, period, and cohort effects is evident in Fig. 1, as our description of these results fails to take into account a possible aging effect. This shortcoming gets to the core of the problem of any APC analysis. The difficulty in making claims about any of the three factors, whether the focus is on cohort, age, or period effects, is the requirement of accounting for all three simultaneously. If we fail to do so, we cannot know whether an observed attitude or behavior (Yijt) of an individual i (i ¼ 1, ..., I) is because she belongs to a specific cohort j (Cj; j ¼ 1, ..., J) or because of her age (Ait) or because of the current time t (Pt; t ¼ 1, ..., T). This can be summarized as
Yijt ¼ f Ait ; Cj ; Pt
(1)
It is not possible to disentangle the effects of age, cohort, and periods in this generalized linear model (1) without some kind of restrictions on the function f (Mason and Fienberg, 1985: 3; cf. Mason et al., 1973). The problem results from the linearly dependent relationship between age and the birth cohort a respondent belongs to at any given time t – assuming that all three effects are measured in the same time units, for example, years – which is typically expressed as
Cj ¼ Pt Ait
(2)
As can be seen from Eq. (2), the three effects cannot be identified with survey data from one time point, since one phenomenon would be completely determined by the remaining two. For example, knowing that a survey respondent who was interviewed in 2000 was 30 years old, we can infer approximately when she was born or to which birth cohort she belongs to. Or put differently, knowing that
11 Studies that explicitly focus on APC analysis: Baker (1978); Abramson (1979); Claggett (1981); Markus (1983); Miller (1992); Tilley (2002) and Tilley and Evans (2011) – partisanship; Klecka (1971); Prior (2010) – political interest; Watts (1999); Lyons and Alexander (2000); Franklin et al. (2004) – turnout/participation; Cutler and Kaufman (1975) – ideology; Jennings (1996) – political knowledge; Jennings and Stoker (2004) – civic engagement; Mishler and Rose (2007); Neundorf (2010); and Mattes (2012) – democratic attitudes. This list is by no means exhaustive.
a respondent was born in 1970 when interviewed in 2000, we can deduce that this person was about 30 years old at the time of the interview. That is, once the values of any two factors, such as cohort and period, are known, the value of the third factor is completely determined. Hence, model (1) is not identified. To date, the suggested strategy for identification of APC models has been to specify some set of restrictions on the estimated age, period, and cohort parameters that allow the model to be identified. Firstly, best known is the grouping of birth years into five-year cohorts, which forces grouped birth-years to have identical effects (Fienberg and Mason, 1979; Mason et al., 1973). Secondly, it is also possible to remove one of either age, period, or cohort entirely on the assumption that it does not matter at all (Converse, 1976, chs. 2–3). Thirdly, one can examine higher order transformations of age, period, or cohort, such as age squared (Fienberg and Mason, 1985). Fourthly, one can use a ‘proxy’ variable approach that assumes the cohort or period effects are proportional to certain measured variables (Rodgers, 1982; Heckman and Robb, 1985; O’Brien, 2000). Of course, none of these approaches solves the underlying problem, and none has proven entirely satisfactory (Glenn, 2005). 4. This symposium: new opportunities in age, period and cohort analysis Today’s much expanded data availability allows a promising new range of statistical methods to conduct APC analyses. Researchers today have the happy prospect of using a broad range of data that is often freely available for our purposes. APC analysis is typically conducted on one of diverse types of data (Harding, 2009: 1450). Firstly, aggregated data containing means or counts of the outcome of interest for various age groups at different time points can be used. When such data is arranged in a period-by-age table, cohort differences can be followed on the diagonals. Most previous studies focus on such aggregated population-level contingency tables for conventional multiple classification. However, these descriptive methods do not allow one to account for any factors other than age, period, and cohort. Increasingly, however, individual-level data sets in the form of a series of repeated cross-section sample surveys or panel studies are available to political scientists. These create both new opportunities and challenges to APC analysis. New data availability and the development of innovative methods offer new paths to study theoretically relevant age, cohort-, and period-related research questions. The articles of this special issue cover several of these new methods. The aim of this collection of eight papers is to introduce diverse methodological approaches to illustrate the problems of APC analyses and propose possible solutions to varied research questions. Some papers focus on identifying cohort effects (Tilley and Evans, 2014; Pop-Eleches and Tucker, 2014; Grasso, 2014), while others try to single out factors that explain cohort differences (Dinas and Stoker, 2014; Smets and Neundorf, 2014; Stegmueller, 2014). Still others emphasize the age-factor in their APC analyses (Bartels and Jackman, 2014; Kroh, 2014). More specifically, Tilley and Evans provide an example how to combine different individual-level data sources
Editorial / Electoral Studies 33 (2014) 1–6
(panel studies and repeated cross-sectional data) and aggregated election results. Using the side-information of these diverse sources makes it possible to estimate true age and cohort effects. They apply their approach to the question of whether support for the British Conservative Party is driven by aging effects or generational replacement. Bartels and Jackman employ their mathematical model of political learning to partisanship, in this case in the US. However, their focus lies on age-effects, estimating the critical years of heightened impact of period-specific effects or historical shocks on a citizen’s party attachments. They show that individuals of different ages attach different weights to political events. Their paper is one of the first empirical tests of the exact ages of the so-called formative years. The idea of historical shocks is also utilized by the design-based approach by Dinas and Stoker. Focusing on cohort effects, they introduce the language of experiments to APC analysis. The authors propose a method to test the impact of the socialization environment (“treatment”) on long-lasting attitudes. In order to test whether specific treatments, such as changes in enfranchisement, affect female turnout propensities, Dinas and Stoker use control groups (men, for one) to aid in the identification of cohort effects while also accounting for age and period effects. Using their approach, researchers have to consider who got a historical “treatment” and who did not. The idea to single out factors that account for cohort differences is also discussed by Smets and Neundorf. Their paper is proposing an approach that enables researchers to test whether substantively interesting factors can account for any observed cohort heterogeneity. For this the variance of cohort differences in turnout propensities is modeled in a multilevel structure. Stegmueller further uses this approach and extends the cross-classified model by allowing random effects (cohort and period) to be time-structured. Doing so, his model links theoretical ideas about the nature of social change, which is – in the absence of strong shocks – gradual and evolutionary. By using time-structured priors for cohort and time-period random effects, he explicitly models dependence of new cohorts and periods on the past. As mentioned above, a widely used approach in cohort analysis is the categorization of the APC variables. The study by Pop-Eleches and Tucker categorizes historical socialization periods in a meaningful way that identifies aging and generational effects of experiences during Communism on people’s democratic and economic attitudes. Grasso further develops this approach by proposing non-parametric methods for testing the validity of these pre-defined categories. This approach moreover allows one to assess the functional form of the cohort effect, while at the same time accounting for possible life-cycle and period influences. Both papers discuss the problems arising from comparative research due to limited data availability (for example, only a few time points are covered by comparable data sources). All the papers mentioned use repeated cross-sectional surveys that often cover several decades. However, it is further possible to use individual panel data, which allows one to observe the same respondents belonging to different generations over time as they age. Kroh uses such a panel study that covers 27 years, applying growth curve modeling. Using his approach it is possible to exchange age
5
as a function of life years to more meaningful variables that measure age effect, as in his empirical example experience and exposure with and to the political system. In introducing these diverse methodological approaches, all papers use examples relevant in the field of political behavior and attitudes, including political interest, voter turnout, partisanship, and voting behavior. Acknowledgements We would like to thank the Department of Politics and International Relations as well as Nuffield College at the University of Oxford and the Department of Political Science at the University of Rochester for kindly funding a workshop meeting to put this paper symposium together. References Abramson, P., Inglehart, R., 1995. Value Change in Global Perspective. University of Michigan Press, Ann Arbor. Abramson, P., 1975. Generational Change in American Politics. Lexington Books, Lexington, MA. Abramson, P., 1979. Developing party identification: a further examination of life-cycle, generational, and period effects. American Journal of Political Science 23 (1), 78–96. Adelson, J., O’Neil, R., 1966. Growth of political ideas in adolescence: the sense of community. Journal of Personality and Social Psychology 4 (3), 295–306. Alwin, D.F., Krosnick, J.A., 1991. Aging, cohorts and the stability of sociopolitical orientations over the life span. American Journal of Sociology 97 (1), 169–195. Baker, K.L., 1978. Generational differences in the role of party identification in German political behavior. American Journal of Political Science 22 (1), 106–129. Bartels, L., Jackman, S., 2014. A generational model of political learning. Electoral Studies 33, 7–18. Bhatti, Y., Hansen, K.M., 2012. The effect of generation and age on turnout to the European Parliament – how turnout will continue to decline in the future. Electoral Studies 31, 262–272. van der Brug, W., Kritzinger, S., 2012. Generational differences in electoral behaviour. Electoral Studies 31 (2), 245–249. Campbell, A., Converse, P.E., Miller, W.E., Stokes, D.E., 1960. The American Voter. Wiley, New York. Campbell, D., 2006. Why We Vote: How Schools and Communities Shape our Civic Life. Princeton University Press, Princeton. Claggett, W., 1981. Partisan acquisition versus partisan intensity: lifecycle, generation and period effects, 1952–1976. American Journal of Political Science 25 (2), 193–214. Conover, P., 1991. Political socialization: where’s the politics?. In: Crotty, W. (Ed.), Political Science: Looking to the Future, vol. 3 Northwestern University Press, Evanston, pp. 125–152. Converse, P., 1976. The Dynamics of Party Support: Cohort-analyzing Party Identification. Sage, Beverly Hills. Cutler, S., Kaufman, R., 1975. Cohort changes in political attitudes: tolerance of ideological nonconformity. Public Opinion Quarterly 39 (1), 69–81. Dinas, E., Stoker, L., 2014. Age–period–cohort analysis: a design-based approach. Electoral Studies 33, 28–40. Easton, D., Dennis, J., 1969. Children in the Political System: Origins of Political Legitimacy. McGraw-Hill, New York. Elder, G., 1974. Children of the Great Depression: Social Change in Life Experience. University of Chicago Press, Chicago. Fienberg, S.E., Mason, W.M., 1979. Identification and estimation of age– period–cohort models in the analysis of discrete archival data. Sociological Methodology 10, 1–67. Fienberg, S.E., Mason, W.M., 1985. Specification and implementation of age, period, and cohort models. In: Mason, W.M., Fienberg, S.E. (Eds.), Cohort Analysis in Social Research: Beyond the Identification Problem. Springer, New York, pp. 45–88. Fiorina, M.P., 1981. Retrospective Voting in American National Elections. Yale University Press, New Haven. Franklin, C.H., Jackson, J.E., 1983. The dynamics of party identification. American Political Science Review 77 (4), 957–973. Franklin, M.N., Lyons, P., Marsh, M., 2004. Generational basis of turnout decline in established democracies. Acta Politica 39 (2), 115–151.
6
Editorial / Electoral Studies 33 (2014) 1–6
Glenn, N.D., 2005. Cohort Analysis, second ed. Sage, Thousand Oaks, CA. Grasso, M., 2014. Age, period and cohort analysis in a comparative context: political generations and political participation repertoires in Western Europe. Electoral Studies 33, 63–76. Harding, D.J., 2009. Recent advances in age–period–cohort analysis. a commentary on Dregan and Armstrong, and on Reither, Hauser and Yang. Social Science & Medicine 69 (10), 1449–1451. Heckman, J., Robb, R., 1985. Using longitudinal data to estimate age, period, and cohort effects in earnings equations. In: Mason, W.M., Fienberg, S.E. (Eds.), Cohort Analysis in Social Research: Beyond the Identification Problem. Springer, New York, pp. 137–150. Hyman, H., 1959. Political Socialization: A Study in the Psychology of Political Behavior. Free Press, Glencoe, IL. Inglehart, R., 1977. The Silent Revolution: Changing Values and Political Styles among Western Publics. Princeton University Press, Princeton. Jennings, M.K., Niemi, R.G., 1974. The Political Character of Adolescence. Princeton University Press, Princeton. Jennings, M.K., Niemi, R.G., 1981. Generations and Politics: A Panel Study of Young Adults and Their Parents. Princeton University Press, Princeton. Jennings, M.K., Stoker, L., 2004. Social trust and civic engagement across time and generations. Acta Politica 39 (4), 342–379. Jennings, M.K., 1989. The crystallization of orientations. In: Jennings, M.K., van Deth, J., et al. (Eds.), Continuities in Political Action. de Gruyter, Berlin and New York, pp. 313–348. Jennings, M.K., Stoker, L., Bowers, J., 2009. Politics across generations: family transmission reexamined. Journal of Politics 71 (3), 782–799. Jennings, M.K., 1996. Political knowledge over time and across generations. Public Opinion Quarterly 60 (2), 228–252. Klecka, W.R., 1971. Applying political generations to the study of political behavior: a cohort analysis. Public Opinion Quarterly 35 (3), 358–373. Konzelmann, L., Wagner, C., Rattinger, H., 2012. Turnout in Germany in the course of time: life cycle and cohort effects on electoral turnout from 1953 to 2049. Electoral Studies 31 (2), 250–261. Kroh, M., 2014. Growth trajectories in the strength of party identification: The legacy of autocratic regimes. Electoral Studies 33, 90–101. Lyons, W., Alexander, R., 2000. A tale of two electorates: generational replacement and the decline of voting in presidential elections. Journal of Politics 62 (4), 1014–1034. MacKuen, M.B., Erikson, R.S., Stimson, J.A., 1989. Macropartisanship. American Political Science Review 83 (4), 1125–1142. Mannheim, K., 1952. The problem of generations. In: Keckemeti, P. (Ed.), Essays on the Sociology of Knowledge. Routledge and Kegan Paul, London, pp. 276–322 [1928]. Markus, G.B., 1983. Dynamic modeling of cohort change: the case of political partisanship. American Journal of Political Science 27 (4), 717–739. Marsh, D., 1971. Political socialization: the implicit assumptions questioned. British Journal of Political Science 1 (4), 453–465. Mason, K.O., Mason, W.M., Winsborough, H., Poole, W., 1973. Some methodological issues in cohort analysis of archival data. American Sociological Review 38 (2), 242–258. Mason, W.M., Fienberg, S.E. (Eds.), 1985. Cohort Analysis in Social Research: Beyond the Identification Problem. Springer Verlag, New York. Mattes, R., 2012. The “born frees”: the prospects for generational change in post-Apartheid South Africa. Australian Journal of Political Science 47 (1), 133–153. Merelman, R., 1986. Revitalizing political socialization. In: Herman, M. (Ed.), Political Psychology: Contemporary Problems and Issues. Jossey Bass, San Francisco. Miller, W.E., 1992. Generational changes and party identification. Political Behavior 14 (3), 333–352. Mishler, W., Rose, R., 2007. Generation, age, and time: the dynamics of political learning during Russia’s transformation. American Journal of Political Science 51 (4), 822–834. Neundorf, A., 2010. Democracy in transition: a micro perspective on system change in post-socialist societies. Journal of Politics 72 (4), 1096–1108. Niemi, R., Hepburn, M., 1995. The rebirth of political socialization. Perspectives on Political Science 24 (1), 7–16. Niemi, R.G., Jennings, M.K., 1991. Issues and inheritance in the formation of party identification. American Journal of Political Science 35 (3), 969–988. Niemi, R.G., Junn, J., 1998. Civic Education: What Makes Students Learn. Yale University Press, New Haven. Niemi, R.G., Sobieszek, B., 1977. Political socialization. Annual Review of Sociology 3 (1), 209–233.
O’Brien, R.M., 2000. Age period cohort characteristic models. Social Science Research 29 (1), 123–139. Ordeshook, P.C., 1976. The spatial theory of elections: a review and a critique. In: Budge, I., Crewe, I., Farlie, D. (Eds.), Party Identification and Beyond. Wiley, London. Page, B., Jones, C.C., 1979. Reciprocal effects of policy preferences, party loyalties and the vote. American Political Science Review 73 (4), 1071–1089. Pop-Eleches, G., Tucker, J., 2014. Communist socialization and postcommunist economic and political attitudes. Electoral Studies 33, 77–89. Prior, M., 2010. You’ve either got it or you don’t? The stability of political interest over the life cycle. Journal of Politics 72 (3), 747–766. Rodgers, W.L., 1982. Estimable functions of age, period, and cohort effects. American Sociological Review 47 (6), 774–787. Ryder, N.B., 1965. The cohort as a concept in the study of social change. American Sociological Review 30 (6), 843–861. Searing, D., Schwartz, J., Lind, A., 1973. The structuring principle: political socialization and belief systems. American Political Science Review 57 (2), 415–432. Sears, D.O., Valentino, N.A., 1997. Politics matters: political events as catalysts for preadult socialization. American Political Science Review 91 (1), 45–65. Sherrod, L., Torney-Purta, J., Flanagan, C., 2010. Handbook of Research on Civic Engagement in Youth. Wiley, New York. Smets, K., Neundorf, A., 2014. The hierarchies of age–period–cohort effects: political context and the development of generational turnout patterns. Electoral Studies 33, 41–51. Stegmueller, D., 2014. Bayesian hierarchical age–period–cohort models with time-structures effects: an application to religious voting in the US, 1972–2008. Electoral Studies 33, 52–62. Thomassen, J., 1976. Party identification as a cross-national concept: its meaning in The Netherlands. In: Budge, I., Crewe, I., Farlie, D. (Eds.), Party Identification and Beyond. Wiley, London. Tilley, J., Evans, G., 2011. Political generations in Northern Ireland. European Journal of Political Research 50 (5), 583–608. Tilley, J., Evans, G., 2013. Ageing and generational effects on vote choice: combining cross-sectional and panel data to estimate APC effects. Electoral Studies 33, 19–27. Tilley, J., 2002. Political generations and partisanship in the UK, 1964– 1997. Journal of the Royal Statistical Society. Series A (Statistics in Society) 165 (1), 121–135. Walczak, A., van der Brug, W., de Vries, C., 2012. Long- and short-term determinants of party preferences: inter-generational differences in Western and East Central Europe. Electoral Studies 31 (2), 273–284. Watts, M., 1999. Are there typical age curves in political behavior? The “age invariance” hypothesis and political socialization. Political Psychology 20 (3), 477–499. Yang, Y., Land, K.C., 2013. Age–Period–Cohort Analysis: New Models, Methods, and Empirical Applications. CRC Press. Taylor & Francis Group, Boca Raton. Zuckerman, A.S., Dasovi c, J., Fitzgerald, J., 2007. Partisan Families: The Social Logic of Bounded Partisanship in Germany and Britain. Cambridge University Press, Cambridge.
Anja Neundorf* School of Politics and IR, University of Nottingham, Nottingham NG7 2RD, United Kingdom Richard G. Niemi1 Department of Political Science, University of Rochester, Rochester, NY 14627-0146, United States E-mail address: [email protected] Corresponding author. Tel.: þ44 (0)115 95 14795; fax: þ44 (0)794 694 7367. E-mail address: [email protected] 18 June 2013
1
Tel.: þ1 585 275 5364; fax: þ1 585 271 1616.