Introducing APC Bounding Analysis: An Example of the Long-Term Change of Subjective Social Status in Taiwan
本文以臺灣社會個人主觀社會地位評分的長期變動為例，介紹新近發展的APC界限設置分析（Bounding Analysis）方法。所謂APC分析是探究三個時間面向—個人層次的年齡（Age），群體層次的時期（Period）和世代（Cohort）—如何影響研究者關心的應變項。由於A=P−C，當這三個解釋變項同時納入分析時，會有共線性問題。相對於其他APC分析方法，新近發展的界限設置分析應是較佳處理認定問題（identification problem）的策略。此策略能以最少的先決條件，呈現APC三者影響效應之正負方向及上下限。本研究以此APC分析方法，探析1984至2019年臺灣社會變遷基本調查資料中，以1至10分測量之主觀社會地位的長期變化。初步分析結果顯示，主觀社會地位在30至35歲之間達到高峰；在不同時期上，則呈現隨社會經濟局勢的情況而上下波動的趨勢。在世代效果上，可以發現1960–1964以後出生世代的主觀社會地位是高於先前出生的世代。
This paper introduces a newly developed APC (Age, Period, Cohort) bounding analysis to untangle the temporal dynamics that shape social change. Researchers have long grappled with the challenge of disentangling the three distinct temporal forces influencing social change: the impact of individual life experiences (Age Effect), the consequences of specific historical events (Period Effect), and the enduring influence of generational experiences (Cohort Effect). However, the inherent problem of perfect collinearity, represented by Age=Period−Cohort, has compelled researchers to seek various statistical decomposition methods to identify these effects separately. Commonly used APC analyses often seek point estimation and impose certain arbitrary assumptions, potentially compromising the reliability of the estimates. In contrast, Fosse and Winship (2019a, 2019b) introduced the bounding analysis approach as a superior strategy to address the identification problem. This method provides estimates of the lower and upper bounds for age, period, and cohort effects with minimal statistical constraints and assumptions. Researchers can effectively disentangle the complex relationships among age, period, and cohort by applying orthogonal transformation techniques. This study uses data from the Taiwan Social Change Survey (TSCS). This nationally representative dataset surveys modules of similar questions every five years to revisit major research topics like social inequality and civic behaviors. Our research analyzes 23 waves of TSCS data collected between 1984 and 2019, specifically focusing on the long-term patterns of change in subjective social status (SSS) in Taiwan. After filtering out respondents younger than 20 and older than 65 and deleting missing data, the analytical sample comprises 44,743 cases. The results of our study underscore the efficacy of the APC bounding analysis in profiling the complex and evolving dynamics of SSS. Notably, our findings reveal a non-linear relationship between age and SSS, with individuals’ SSS generally peaking around 30 to 35 before gradually declining as they advance beyond 60. Cohorts also emerge as significant contributors to variations in SSS, with those born after 1964 reporting higher SSS compared to their predecessors born before the 1960s. Our analysis of the period effect highlights the dynamic nature of SSS, showing fluctuations in response to prevailing political and economic conditions during survey years. Notably, events such as the missile crisis of 1995 and the Asian financial crisis of 1997 left a noticeable imprint on individuals’ SSS during the years 1995–1999. Similarly, the global financial crisis of 2008 had a discernible impact on SSS evaluations.While the bounding analysis offers a more plausible solution to the identification problem, it primarily provides descriptive insights. Building on the findings of the bounding analysis, researchers can further explore the causal mechanisms underlying the relationships between outcome variables and temporal factors.