Determinants of Female Labour Force Participation Dynamics: Evidence From 2000 & 2007 Indonesia Family Life Survey

Transcription

1 Determinants of Female Labour Force Participation Dynamics: Evidence From 2000 & 2007 Indonesia Family Life Survey Diahhadi Setyonaluri PhD Student Australian Demographic and Social Research Institute (ADSRI) Australian National University Paper presented at NATSEM HDR Workshop, 13 September 2012

2 2 Introduction Female s share of the total labour force is increasing, with more young women joining formal sectors between Lower labour force participation rate of married women (compared to unmarried) due to age, presence of young children, and tendency of employers in formal sectors to hire young, single women. Other studies found a tendency for women to move to other type of employment that accommodate women s roles at work and home. This study explores the role of changing family status in determining women s labour force dynamics.

3 3 Literature Review Previous studies found strong effect of marital status, childbirths and presence of young children in reducing female s employment continuity Flexibility in part-time jobs reduces the incompatibility between work and childcare (Drobnic et al.,1999) Presence of adult women in the household reduce the negative effect of young children on women s work continuity (Cerruti, 2000) Family status is important, but not a sole determinant. Other factors, particularly education and previous job characteristics are also found to determine women s employment transitions (Taniguchi and Rosenfeld, 2002)

4 Data: Indonesia Family Life Survey (IFLS) 2000 & 2007 Multilevel longitudinal survey (individual, household, community) fielded in 13 provinces*. Baseline sample (1993) was 7,224 households & 22,000 individuals, with 95% re-contact rate of original households (Strauss, 2004) IFLS 2000 & 2007 documents: Yearly retrospective work history that covers the period Marriage, childbirths, migration, education histories. This study: Combine life histories from 2000 & 2007 IFLS into a long-format life history data Select women who have ever worked throughout and were aged years old in Note: *IFLS was fielded in 1993, 1997, 2000 and It was also fielded in 1998, with a selected sample from previous wave Background - Research Question - Data - Method - Variables - Result - Summary 4

5 Definitions of work & labour force participation in IFLS IFLS defined work as broadly includes formal and informal, full-time and part-time, and seasonal and yearround labour. For current employment, IFLS probes the labour force participation using several questions: work or not, job searching, try to work/help to earn income at least 1 hour, have a job/business but temporarily not working, work at family-owned business. In retrospective work history, IFLS collects information whether a respondent work or not in a particular year. Background - Research Question - Data - Definition - Sample - Method - Variables - Result - Summary 5

6 Method: Discrete Time Event History Analysis Dependent variable Labour Force Exit conditional probability of leaving the labour force, given that a woman was working from the beginning of spell. Labour Force Re-entry conditional probability of entering the labour force, given that a woman had left the workforce Measure Censor Dummy variable of 0 if a woman work 1 If a woman did not worked Women who were continuously working up to 2007 are censored at that point. Dummy variable of 0 if a woman did not work 1 If a woman worked Women who had left the labour force up to 2007 are censored at that point. Sample Women who were unmarried at the beginning of first work spell (n=2,204) Women who had left the labour force (n=1,125) Models are estimated with logistic regression (Singer and Willet, 2003) Background - Research Question - Data - Method - Variables - Result - Summary 6

7 Independent variables (family dynamics) Variable Measure Time- Varying? Exit Model Re-entry First marriage Single = 0 Had the first marriage = 1 Yes First birth Had no child = 0 Had the first birth = 1 Yes Family status 1. Unmarried 2. Married, no children 3. Married, 1 child 4. Married, 2+ children Yes Background - Research Question - Data - Method - Variables - Result - Summary 7

8 Independent variables (cont d) Migration Variable Education attainment Urban/rural Region Measure Whether had migrated during the work spell (reference: not migrated) Highest level education attained at the last survey (reference: no schooling/elementary) Whether live in urban or rural area during the spell (reference: rural) Whether live in Java or outside Java (reference: outside Java) Time- Varying? Age Age in years (continuous) Yes Employment of the last jobs 1. Self-employment (ref.) 2. Government workers - professionals, managers, clerical 3. Government workers - others 4. Private workers professionals, managers, clerical 5. Private workers manual, sales, service 6. Unpaid family workers 7. Casual workers Presence of other adult females Presence of females age 15 and older in the household where a woman reside (measured in 2000) (reference: no adult females in households) Background - Research Question - Data - Definition - Sample - Method - Variables - Result - Summary Yes No Yes Yes No No 8

9 Descriptive Table 1. Mean age of events and duration of spells by working pattern Mean age and duration Working Pattern Continuous Interrupted Age at first marriage Age at first birth Age at start of work spell Age at labour force exit n.a Age at labour force reentry n.a Duration of work spell Duration of labour force break n.a. 2.9 Background - Research Question - Data - Method - Variables - Result - Summary 9

10 Table 2: Estimated coefficients from transition out of the labour force models Variables Model 1 Model 2 Model 3 First marriage (0=never married; 1= had 1st marriage) 1.447*** 1.394*** 0.847*** First birth (0=never had a birth; 1=had 1st birth) 0.878*** 0.821*** 0.829*** Education (last survey) (ref: no schooling/elementary) Junior secondary 0.235** 0.215** Senior secondary 0.369*** 0.335*** Tertiary 0.382*** 0.315** Presence of other adult females (0=none 1=1+) Previous employment (ref: self-employment) Government workers - professional, managerial, clerical *** *** Government workers - others * * Private workers - professional, managerial, clerical 0.529*** Private workers - manual, sales/service workers 0.982*** 0.463** Unpaid family workers 0.504*** 0.575*** Casual workers 0.381* Interaction terms for first marriage and: government workers - professional, managerial, clerical government workers - others private workers - professional, managerial, clerical 0.37 private workers - manual, sales/service workers 0.965*** unpaid family workers casual workers 1.169*** *** p<0.01; ** p<0.05; ***p<0.1 *) These models are estimated using logit regression that includes random effects since rho s value for model 1 to model 3 is more than zero. **) Model 2 and 3 are also controlled by age, migration, urban/rural, and Java or Non-Java residences. Source: IFLS 2000 & 2007, authors calculation 10

11 Predicted probability margins of previous employment type and marital status of women *Predicted margins with 95% confidence interval. *Interaction between last job status and marital status is significant at 1% Source: IFLS 2000 & 2007, author s calculation based on the result of model with interaction estimation Background - Research Question - Data - Method - Variables - Result - Summary 11

12 Table 3: Estimated coefficients of transitions back into the labour force Family status (ref: never married) Variables Model 1 Model 2 Married - no children *** *** Married - 1 child *** *** Married - 2+ children Education (last survey) (ref: no schooling/elementary) Junior secondary Senior secondary Tertiary 0.392** Presence of other adult females (0=none 1=1+) Previous employment (ref: self-employment) Government workers - professional, managerial, clerical ** Government workers - others Private workers - professional, managerial, clerical *** Private workers - manual, sales/service workers *** Unpaid family workers *** Casual workers *** p<0.01; ** p<0.05; ***p<0.1 *) These models are estimated using pooled logit regression since the rho value from the estimation results using logit regression that include the random effects is equal to zero. The models are estimated using individual ID code (PIDLINK) as a cluster variable. **)Model 2 is also controlled by age, migration, urban rural and Java-non Java residences. Source: IFLS 2000 & 2007, authors calculation 12

13 Discussion This study confirms that family status is important, but not a sole determinant of female labour force dynamics: Stronger impact of marital status than childbearing on labour force exit indicates that decision to have a child is jointly determined with decision to marry. Effect of marital status on labour force exit is stronger for blue collar private workers, indicate low job protection, or women choose this type of work for a transit before marriage. Positive effect of education on labour force exit and reentry indicate high-educational homogamy, higher value of child s quality, better access to job information. Background - Research Question - Data - Method - Variables - Result - Summary 13

14 Feedbacks: Source: 14

15 In Indonesia, females adjust their labour force participation during childbearing years Figure 1a & 1b. Age-specific LFPR for males and females, Indonesia Males Females Source: Sakernas 1980, 1985, 1990 and 2000; author s calculation using Sakernas August 2007 data tape Background - Research Question - Data - Definition - Sample - Method - Variables - Result - Summary 15

16 16 Occupation: Professionals, managers & clerical workers Scientist, engineers, air force & navy, medical doctors, vets, medical workers, statisticians, mathematicians, economist, accountants, lawyers, lecturers, religion related workers, journalists, writers, athletes, artists, art workers. Legislatives, managers. Government clerks; bookkeeping, accounting & auditing clerks, billing & posting clerks, payroll & timekeeping clerks, tellers, transportation & telecommunications operators.

17 17 Occupation: Sales & service workers Retail, insurance, land & building, and manufacturing products sales owners, managers, supervisors & workers; Food preparation & serving workers, building & cleaning workers, maid & housekeeping cleaners; personal care workers; security guards. Occupation: Manual workers Constructions & extraction workers; installation, maintenance & repair workers; mechanics; production workers, assemblers; transportation & material moving workers; agriculture workers.

18 18 Formal Jobs Self-employed with permanent workers Government workers Private workers Informal Jobs Unpaid family workers Self employed Self employed with unpaid/temporary workers Casual workers (agriculture & nonagriculture)

19 Limitations of the data (1) IFLS collected retrospective work history for 5 or 9 years prior to a survey, not from the start of the first job. Start year of work in the work history might not be the start year of the first job. Respondents were women aged 15+ at the time of survey, not at the first year of work history. Each respondent has a different start year/age, making a cohort comparison is difficult. Overlap years which potentially generate a seam bias problem. Retrospective variables are limited. Survey Year * Year Observed in Work History Background - Research Question - Data - Definition - Sample - Method - Variables - Result - Summary 19

20 Consequences of data limitations Left-censoring in event history analysis: Actual start year of the first job is unknown It can be linked to year starting first full-time employment, however there are gaps between start year and first year being observed in WH Cannot examine important effect of husband s income and presence of other adult females in household. Background - Research Question - Data - Definition - Sample - Method - Variables - Result - Summary 20

21 Table 2: Estimated coefficients from transition out of the labour force models First marriage Variables Model 1 Model 2 Model 3 (0=never married; 1= had 1 st marriage) 1.447*** 1.394*** 0.847*** First birth (0=never had a birth; 1=had 1 st birth 0.878*** 0.821*** 0.829*** Education (last survey) (ref: no schooling/elementary) Junior secondary 0.235** 0.215** Senior secondary 0.369*** 0.335*** Tertiary 0.382*** 0.315** Migration (0 never migrated 1=migrate) 0.696*** 0.695*** Urban/rural (0=rural 1=urban) Region (0=non Java 1=Java) *** *** Presence of other adult females (0=none 1=1+) Age *** *** *** Previous employment (ref: self-employment) Government workers - professional, managerial, clerical *** *** Government workers - others * * Private workers - professional, managerial, clerical 0.529*** Private workers - manual, sales/service workers 0.982*** 0.463** Unpaid family workers 0.504*** 0.575*** Casual workers 0.381* Interaction terms for first marriage and: government workers - professional, managerial, clerical government workers - others private workers - professional, managerial, clerical 0.37 private workers - manual, sales/service workers 0.965*** unpaid family workers casual workers 1.169*** *** p<0.01; ** p<0.05; ***p<0.1 *) These models are estimated using logit regression that includes random effects since rho s value for model 1 to model 5 is more than zero. Source: IFLS 2000 & 2007, authors calculation 21

22 Table 3: Estimated coefficients of transitions back into the labour force Variables Model 1 Model 2 Family status (ref: never married) Married - no children *** *** Married - 1 child *** *** Married - 2+ children Education (last survey) (ref: no schooling/elementary) Junior secondary Senior secondary Tertiary 0.392** Migration (0 never migrated 1=migrate) 0.685*** Urban/rural (0=rural 1=urban) Region (0=non Java 1=Java) Presence of other adult females (0=none 1=1+) Age Previous employment (ref: self-employment) Government workers - professional, managerial, clerical ** Government workers - others Private workers - professional, managerial, clerical *** Private workers - manual, sales/service workers *** Unpaid family workers *** Casual workers *** p<0.01; ** p<0.05; ***p<0.1 *) These models are estimated using pooled logit regression since the rho value from the estimation results using logit regression that include the random effects is equal to zero. The models are estimated using individual ID code (PIDLINK) as a cluster variable. Source: IFLS 2000 & 2007, authors calculation 22

23 Descriptive statistics of independent variables in transition out of the labour force model Variable Mean SD Minimum Maximum Marital status Had first birth Education no schooling/elementary junior secondary senior secondary Tertiary Migration status Urban/rural Island Last employment characteristics Self-employment Government workers - professional, managerial, clerical Government workers - others Private workers - professional, managerial, clerical Private workers - manual, sales/service workers Unpaid family workers Casual workers Presence of other females Age

24 Descriptive statistics of independent variables in transition into the labour force model Variable Mean SD Minimum Maximum Family status Never married Married without children Married, with 1 child Married, with 2+ children Education no schooling/elementary junior secondary senior secondary Tertiary Migration status Urban/rural Island Last employment characteristics Self-employment Government workers - professional, managerial, clerical Government workers - others Private workers - professional, managerial, clerical Private workers - manual, sales/service workers Unpaid family workers Casual workers Presence of other females Age

25 Characteristics at the end of work spell Work Pattern Continuous Interrupted Total Marital status Unmarried Married Number of children Education No schooling/elementary Junior secondary Senior secondary Tertiary Other Migration status Never migrate Ever migrate Urban/Rural residence Rural Urban Island Outside Java Java Last employment type Self-employment Government workers - professional, managerial, clerical Government workers - others Private workers - professional, managerial, clerical Private workers - manual, sales/service workers Unpaid family workers Casual workers Missing Presence of other adult females in household No adult females At least one adult female Number of sample (n) 949 1,255 2,204 25

26 Questionnaire of current employment section in IFLS 26

27 Questionnaire of retrospective work history in IFLS 27

28 Definition of the first work and break spells Work spell/episode is the time span a woman spends in a working state. break spell/episode is the time span a woman spends not in a working state, given she had worked before N N N W W W W N N W W W 1 st work spell 1 st break spell Start 1 st spell Left workforce Re-enter workforce Background - Research Question - Data - Definition - Sample - Method - Variables - Result - Summary 28