Social mobility / reproduction Ascription and Achievement in Occupational Attainment in Comparative Perspective: a comparison of 42 nations, 1900-2000 Harry B.G. Ganzeboom Current Occupation Donald J. Treiman Russell Sage University Working Group on Social Inequality University of California-Los Angeles Father’s January 25-26 2007 occupation 2 Blau-Duncan’s SAT model - I BD’s theoretical contributions • Decompose social reproduction into different stages. Father’s – Different parts of the model can be expected education to follow different trends Education � There is no expectation about trends in Current social reproduction Occupation – Contextual influences are most likely to have their impact at career beginnings. � Look at first job or career beginnings for First Father’s comparative analyses occupation occupation 3 4 (Early) comparative hypotheses: Comparative hypotheses: modernization Institution arrangements • Political intervention by communism: • BD: Trends show decreasing ascription and – Limited accumulation of and transfer of property. increasing achievement – Direct discrimination against ‘class’ backgrounds in education • Lenski/Treiman: Similar expectations for and labor market. – Organized education/occupation link. industrializing (modernizing) societies: • Welfare state provisions: – Shift towards employment with skill requirements – Free education, equalized income distribution lowers ascription – Shift from property owning occupations • Educational arrangements: – Increased communication – Higher dispersion of education, in terms of level (duration) and specificity (vocational tracks) promotes achievement – Increased international and internal migration – Educational expansion raises age of entry into the labor market – Value change towards meritocracy and lowers ascription – Increased wealth � resource equality 5 6 1
Why large-scale comparative Our approach studies are so rare • As many data sources as we can get: high N, high • High quality data were first produced in the late number of countries. 1970’s, but they have become available in great • Replicated studies per country: allows for separating cohort and experience effects. numbers in the 1990’s • Combine over-time and cross-national comparisons (XT) • Immense work to harmonize these data. � NDF. • Stratification research has turned away form • Reduced (elementary) three-variables SAT model. continuous variable representation and OLS, • Model SAT in such a way that we look primarily at effects at career beginnings. and turned towards loglinear models for ‘class’ • Continuous measures, OLS estimation of micro-model. reproduction • Meta-analysis of estimated OLS coefficients using � Return to (detailed, complicated) bivariate analyses. XTGLS. 7 8 The SAT model – II (reduced) The SAT model – III (elementary) Education Education Current Current Occupation Occupation First Father’s Father’s occupation occupation occupation 9 10 Data Design • International Stratification and Mobility File • Data are arranged by 1047 contexts: – Nation (42) [ISMF]: – Entry cohort (1900..2000), five year wide (21) – 331 sample surveys from 42 countries, collected – Experience group (5..45), ten year wide (5) 1947-2003. • Micro-model estimated in each context: – 36 nations have replicated studies (with different years). – ISEI = B0 + B1* EDUC + B2*FISEI • The micro-model coefficients are subjected to a – Only men, 21-64, with valid occupation, education, father’s occupation codes. macro-level meta-analysis, using (1/se**2) as weights. – N=374.093. • More to come: see website. 11 12 2
LEVEL: socio-economic Measurement development • Multiple indicator measure of socio-economic • Occupation: FISEI and ISEI, derived from development. ISCO-68 and ISCO-88 (average). • Sources: – Bank’s Social and Political Indicators 1815-1973. • EDUC: level of education, expressed in – World Bank World Development Indicators 1960-2005 ‘virtual’ years of education. • Indicators: – Urbanization • Macro: Socio-economic LEVEL using 14 / – Share of farm, industrial employment 10 macro-indicators. – Energy consumption, GNP/GNI – Roads, telephones, newspapers, radio’s, mail • Macro: communist regime. – Literacy, school enrollment at primary, secondary, tertiary level. – Physicians 13 14 Socio-economic development in LEVEL - construction five countries • Each indicator interpolated for missing .997317 USA USA USA USA USA USA USA USA USA USA USA USA USA USA USA USA information in time series (war years, incomplete NET NET NET NET NET NET NET NET NET NET NET NET NET NET USA USA USA USA USA USA USA USA USA USA series). NET NET NET NET USA USA USA USA NET NET NET NET USA USA USA USA USA USA USA USA USA USA POL POL POL POL POL • Match between Banks and WB data made at USA USA USA USA USA USA USA USA USA POL USA USA USA USA NET NET NET NET NET POL POL POL POL percent of level USA USA USA NET NET NET NET POL POL POL POL indicator level, by calibrating the mean and USA USA NET NET NET NET NET USA USA POL POL POL POL NET NET NET NET standard deviation in overlapping interval (1965- NET NET NET NET USA NET NET NET NET NET NET BRA BRA NET NET POL POL POL POL BRA BRA 1973). USA NET NET NET NET POL POL POL BRA BRA BRA POL POL POL POL BRA BRA BRA BRA BRA BRA BRA BRA POL POL POL BRA BRA BRA BRA BRA • Composite expressed in 0..1 rank score (no POL POL POL POL POL POL BRA BRA BRA BRA BRA BRA BRA BRA BRA BRA BRA BRA IND IND IND IND IND IND outliers). 0: India in 1940, 1: US in 1995. IND IND IND IND IND IND IND IND IND IND IND IND .021856 IND -1 1 coh 15 16 XTGLS-Model Hypotheses • Ascription (direct effect of FISEI � ISEI) • Macro-level analyses with XTGLS panel- – Declines over cohorts model that takes into account: – Declines with socio-economic development – Panel: country – Is smaller in communist countries, in particular in – Time: cohorts orthodox times. – Heteroskedasticity between panels allowed • Achievement (direct effect of EDUC � ISEI) – Weights (variable within panels) allowed – Increases over cohorts – Increases with socio-economic development • When conditioning upon experience – Is stronger in communist countries – Common serial correlation between panels 17 18 3
Analysis: XTGLS B_fisei Analysis: XTGLS B_educ B0 .315 .305 .340 B0 1.281 1.361 1.613 (34.7) (31.1) (23.2) (23.9) (24.0) (17.8) EXP/10 -.011 EXP/10 -.153 (3.2) (5.7) EYR/10 -.010 -.017 EYR/10 .085 -.004 (5.0) (5.8) (5.8) (0.2) LEVEL -.190 -.168 -.175 LEVEL 1.818 1.648 1.613 (14.6) (11.5) (11.9) (21.4) (17.6) (17.4) COMM -.063 -.115 -.113 COMM .587 .842 .857 (9.9) (10.1) (5.9) (11.7) (10.7) (4.4) COMM_EXP 015 COMM_EXP .016 (1.9) (0.2) COMM_EYR .038 .048 COMM_EYR -.249 -.249 (6.4) (5.9) (5.4) (4.0) 19 20 Results: Ascription Results: Achievement • Effect declines by almost 2/3: • One century of socio-economic development had doubled the achievement effect. – Between lowest and highest level of development. • Communism had increased the achievement – Over 100 years effect as well, but to a lesser extent. • Communism has reduced ascription • Again, communism had a sharp effect at its considerably, by almost 1/3 at its start, but the beginnings, but this has much softened over effect has dampened over history. cohorts (history). • Communism at labor market entry remains • The influence of communism at career beginning important over the life cycle, level of persists during the life-cycle, but the effect of development at career beginnings looses its economic development at career beginning effects later. withers away. 21 22 Problem 2: continuous versus Problem 1: serial correlation discrete • Data are longitudinal in two dimensions: • Micro-model pays no attention to discrete (labor market entry) cohort and life-cycle turn in stratification analyses. (experience) • Possible solution: conditional logit model • At present serial correlations can only be at the micro level. estimated when conditioning by experience,-- but they are found to be small. 23 24 4
Problem 3: Paucity of macro-level measures • To be added: – Social democratic regimes, welfare regimes – Immigration – Income inequality – Educational arrangements • Early selection • Dispersion • Hard to find complete time-series for so many countries. 25 5
Recommend
More recommend
