Female Labour Force Participation in Indonesia: Why Has it Stalled?

Abstract

This paper examines and disentangles the factors that have led to the largely unchanged participation (about 51%) of women in Indonesia’s labour force in the past two decades. We use data from the National Socio-economic Survey (Susenas) and Village Potential Statistics (Podes) from 1996 to 2013 in order to conduct a cohort analysis that distinguishes the effects of time and age on female labour force participation. We find that the raw labour market participation figures, which show little change over time, mask changes that offset one another in the current population. Evidence suggests that social norms are changing to support female participation, but this is offset by the changing industrial structure. Our projections show that the government’s current policies are unlikely to allow Indonesia to reach its G20 goal of decreasing the gender gap in labour force participation by 25% between 2014 and 2025.

Tulisan ini menelaah dan menguraikan faktor-faktor yang membawa kepada tidak adanya perubahan partisipasi wanita di angkatan kerja Indonesia (sekitar 51%) pada dua dekade terakhir. Penulis menggunakan data dari Survei Sosial Ekonomi Nasional dan Statistik Potensi Desa dari tahun 1996 hingga 2013 untuk membangun analisis kohor yang membedakan efek dari waktu dan usia pada partisipasi angkatan kerja wanita. Penulis menemukan bahwa angka kasar partisipasi angkatan kerja, yang menunjukkan hanya sedikit perubahan seiring waktu, sesungguhnya mengaburkan berbagai perubahan di dalam populasi saat ini, yang saling mengimbangi satu sama lain. Bukti-bukti mengungkapkan bahwa norma-norma sosial telah berubah dan mendukung partisipasi wanita, namun perubahan ini diimbangi oleh pergantian struktur industri. Proyeksi penulis menunjukkan bahwa kebijakan pemerintah saat ini kurang menunjang pencapaian tujuan G20 untuk mengurangi sebanyak 25% celah gender pada partisipasi angkatan kerja antara 2014 dan 2025.

Keywords:

• female labour force participation
• labour markets
• gender
• Indonesia
• cohort analysis
• female employment

JEL classification:

• O12
• O15
• J16
• J21

Notes

1 Figures compiled by AIPEG, as cited at http://www.bbc.com/indonesia/indonesia-42428508.

2 This core questionnaire is supplemented by modules covering about 60,000 households. These modules collect additional information on areas such as health care and nutrition, household income and expenditure, and labour force experience.

3 Sakernas identifies only the number of children under age 10 in a household.

4 For 1996, 2000, and 2011, we matched the data of the corresponding years of Podes and Susenas. For 2007, we matched data from the 2007 Susenas with data from the 2008 Podes, because there was no Podes for 2007. For 2013, we matched data from the 2013 Susenas with data from the 2011 Podes, as this was the closest year to 2013. However, the 2013 Susenas does not include the unique village identifiers that are available in other years. The smallest geographical unit reported is the district, so we calculated the main source of income in a district, using data from the 2011 Podes, and supplemented it with data from the 2013 Susenas. To check for robustness, we re-estimated our main results, without the 2013 data. The results were similar, so the 2013 data were retained for the estimation.

5 The years studied reflect the data to which we had access, and our desire to study a reasonably long period. As the AFC occurred in 1997–98, we studied data from 1996 and 2000, but not from 1997, 1998, or 1999. We then used data from 2007 because it is roughly midway between 1996 and 2013.

6 We used 49 dummy variables for ages 15–64 (the omitted category was age 15) and 49 dummy variables for birth cohorts—one for each birth year from 1943 to 1992 (the omitted category was 1943).

7 Another approach is to omit the year/period effect under the assumption that it is negligible. For example, Goldin and Mitchell (2017) argue that if period effects influence all individuals in a year, independent of their age, then the cohort and life-cycle effects will dominate.

8 In addition, we estimated equation 1 for provinces in Java and Bali and separately for the outer islands (all provinces except those in Java and Bali). The results are presented in the appendix.

9 The effect of upper secondary and tertiary education on labour force participation dropped between 1996 and 2000. This probably reflects the result of the AFC in 1998, which disproportionately affected better-educated workers. Unemployment rates were high among better-educated workers in 2000.

10 The sample averages of the explanatory variables were applied to the regression coefficients and added to the age and birth cohort effects to present the impact of different ages and birth cohorts on labour force participation.

11 The age effect also shows a considerable difference between married and unmarried women. For unmarried women, labour force participation peaked by age 25 (similar to the peak participation age in the Netherlands, as presented in figure 3). However, for married women, labour force participation peaked at about age 50. An alternative approach is to estimate a Oaxaca decomposition of the changes in FLFP between 1996 and 2013 in order to identify the characteristics associated with increased participation. Such an analysis shows that increases in participation by married women are predominantly driving the changes across time, ceteris paribus.

12 Mining/quarrying also showed improvement across the younger cohorts, but the variability in these results and the age effect were likely due to the smaller sample size of women in this sector. For example, there were only 7,795 observations for this sector, compared with 47,312 observations for the processing/industry sector and 123,007 for the large trade/retail sector.

13 To calculate the national FLFP, we estimated the model using both urban and rural samples, and we included a control variable for urban areas. The results are presented in table A3 in the appendix.

14 We compared our projections for the percentage of population by age group with the UN’s forecasts. They were similar. Our projected urbanisation rate was also similar to the UN’s rate.

15 Note that both the predictions show an increase in FLFP compared with BPS’s official FLFP estimate for 2015. BPS uses data from Sakernas to calculate FLFP.