Female labour force participation: Measurement and data quality

Official data revealed a sharp decline in female labour force participation in India between 2004-05 and 2011-12, despite fast economic growth in the country. Examining the measurement of women’s work and data quality issues, this article identifies three explanations for the low observed female labour force participation: inconsistent treatment of non-market work, more women in higher education, and the disproportionate time spent by women on childcare.

The female labour force participation rate (FLFPR) in India has sparked widespread curiosity and debate among researchers over the past decade. The conversation gained momentum when the National Statistical Survey Organisation (NSSO) reported a sharp decline in FLFPR between 2004-05 and 2011-12, a period of rapid GDP (gross domestic product) growth. Consequently, several explanations emerged for this decline, including rising enrolment in education among younger women, a lack of suitable employment opportunities for women, women withdrawing from formal work due to rising family income, and the influence of social norms. These may serve as potential explanations, but only if the data are of high quality and the measurements of FLFP are accurate – factors that have not been thoroughly examined.

We assess measurement and quality of data as well as questions and issues pertaining to FLFPR in India (Bhalla, Bhasin and Das 2024). Our findings challenge assumptions of a sharp decline in FLFPR, significantly lower FLFPR compared to other countries, and slow job creation between 1999-2000 and 2022-23.

Employment data and its discontents

While increased school and college enrolment contributed to some decline in FLFPR between 2004-05 and 2022-23, potential issues with data quality and the definition and measurement of FLFP continue to obscure the true trends. The classification of unpaid family workers, especially within the context of labour market informality, requires closer scrutiny. Their ‘definitional’ movement into and out of the labour force between 2004-5 and 2011-12 led to most of the observed decline in FLFPR during that time. Lack of clarity on definitions, combined with potential data inconsistencies, suggests that the observed decline in FLFPR could be largely due to measurement errors rather than systematic shifts in socioeconomic factors.

Continuing with the role of measurement errors, and in line with the International Labour Organisation (ILO), we find that the 2017-18 and 2018-19 Periodic Labour Force Survey (PLFS) data, the primary official source for labour market statistics in India in recent years, may also not be of “good” quality. This is what the ILO said on its website in November 2023: "In the model of labor force participation, the PLFS observations for 2018 (2017-18) and 2019 (2018-19) have been excluded as they appear to present limited comparability with both the previous NSS results and the newer PLFS results.¹" Stated differently, NSS and PLFS data from earlier (1983 to 2011) and subsequent years (post 2018-19) appear to be of internationally comparable, and reasonably “good” quality.

Definition of work and measurement inconsistencies

Labour force estimates follow directly from the definition of work and unemployment. Work was defined by the ILO at the 13^th International Conference of Labor Statisticians, as work undertaken for “pay or profit”. This definition is challenging in low- and lower-middle income countries due to the presence of a large household enterprise sector. These small enterprises undertake home production of goods and services. The strict application of this definition implies that home production for sale in the market (whether paid or unpaid) counts as work but home production for home consumption is not treated as work. Thus, it seems the issues with FLFPR estimates in India stem, at least partly, from inconsistent definitions of work in NSSO (1983–2011-12) and PLFS (2017-18 onwards) (Kapsos et al. 2014).

Roy Model and U-shaped FLFPR

To deepen understanding of the issue, we build a structural model based on the Roy Model (1951) to compare returns to education in the household sector (HH) and the labour market (LM). This reveals a U-shaped relationship between FLFPR and education, consistent with Goldin’s (1994) celebrated U-shape curve. The U-shaped relationship essentially implies an initial decline in FLPR as years of education increase and after a while, additional years of education result in an increase in FLPR.

The standard Roy Model focuses on occupational choices based on potential earnings. We adapt this model to examine a woman's choice between working in LM or HH. Economically, like any worker, a woman chooses LM if it offers higher potential earnings but remains in HH if the implicit returns are greater there (Robbins 1930, Heckman and Killingsworth 1986). Factors such as education, work experience, area of residence (rural/urban), number of children, family earnings from male members, and the proportion of females in the household influence returns in both sectors, thereby shaping overall FLFPR.

An interesting feature of the Roy model is that, with an appropriate ‘exclusion restriction’, it allows for the estimation of returns of these determinants in HH, even though its earnings are unobserved. A variable that affects LM earnings but not HH earnings can serve as such an exclusion restriction (Heckman and Honore 1990, French and Taber 2011). We argue that college degree holding status can serve as such an exclusion restriction. Once years of education and other academic credentials (major indicators of human capital) are taken into consideration, the degree holding status or diploma becomes merely a signaling device. It only indicates that the holder can complete the programme as opposed to someone who has the same human capital but was unable to complete it.

Our analysis reveals that education impacts rewards from work in a ‘non-linear’ manner (that is, initial returns to education are limited at lower years of education and as the years of education increase, so do the marginal returns to education), both in LM and HH. These impacts also differ substantially between LM and HH. This non-linearity and differential impact generate the U-shaped relationship between education and FLFPR (see Figure 1).

Figure 1. Simulated female work force participation rate, 2022-23

Note: (i) Age group is 25-64 years. (ii) Earnings are only available for LM.

Source: PLFS, 2022-23.

Trends in female labour force participation rate

The World Bank’s South Asia Development Update (SADU) (2024) compares India’s FLFPR with respect to other emerging markets and reaches the conclusion that FLFPR in India is lower than it should be. Such assessments are not straightforward and require a determination of comparable countries. Advanced and formerly Eastern Europe economies have a considerably higher average education and income level than India and therefore cannot be part of the comparison group (since both education and income are important (positive) determinants of FLFP). Countries in Middle East and North Africa are not (broadly) comparable given that their social norms for female work force are considerably different than in India. Sub-Saharan African economies are much poorer than India and other emerging economies.

Hence, for our analysis, we select 61 countries in Asia (East and South Asia) and Latin America. From this list of 61, we eliminate economies whose population was less than 5 million in 2022 (small economies). Further, we eliminate the following countries from our sample: China, because its very large population can affect weighted population averages; Afghanistan, Cuba, Myanmar, North Korea and Venezuela. This leaves us with 29 countries as the comparator group hereafter referred to as AsiaLA-29 (rest of the world comparable to India).

Table 1. Labour force participation rate (in %): A cross-country perspective

	Ages >=15 years			Ages >=25 years
	India		AsiaLA-29	India		AsiaLA-29
	Usual	CWS	CWS	Usual	CWS	CWS
All
1999	61.6	58.3	62.3	67.0	63.5	67.3
2011	55.9	53.2	62.4	63.0	60.1	67.8
2022	57.9	54.1	61.0	65.1	61.5	66.1
Females
1999	38.9	33.8	44.2	42.1	36.9	47.1
2011	31.2	27.1	46.4	35.0	30.5	49.9
2022	37.0	31.3	46.2	41.9	36.0	49.5
Males
1999	83.6	82.0	80.3	91.4	89.8	87.7
2011	79.8	78.5	78.6	91.1	89.8	86.2
2022	78.5	76.8	76.2	88.6	87.5	83.5

Notes: (i) Usual Status: worked at least 30 days in the last year. (ii) Current Weekly Status (CWS): worked at least 1-hour last week. (iii) AsiaLA-29 are defined in main text.

Source: NSS, PLFS, and ILO data on current weekly status.

Table 1 contains a summary of the results for age-groups >=15 and >=25 years for the three years 1999-2000, 2011-12, and 2022-23. The AsiaLA-29 data does not contain labour market data for Usual Status (US); we believe usual status is more representative (as is the >=25 age group because it is uncontaminated by educational enrolment). While the discussion will be on US for India compared to Current Weekly Status (CWS) for AsiaLA-29, the tables report the data for CWS (ILO-modelled data) for AsiaLA-29, and both CWS and US data for India. One major result is that there is not much difference in the aggregate all-workers (women and men) LFPR rates between India and AsiaLA-29 for the last 20 years or so.

This result must be qualified: the comparison is between US for India and CWS for Asia and Latin America. CWS is very likely representative of labour market developments in most of the countries of Asia and Latin America, while US is more representative for India. If CWS is compared with CWS, then in 2022-23, a 4.6 percentage point gap remains – 61.5 % for India and 66.1% for AsiaLA-29 (>=25 years) It may be noted here that for 2023-24, India’s CWS LFPR has increased to 63.4 %.

Childcare and female labour force participation

To further examine the drivers of FLFP in India, we analyse time-allocation decisions using Time-Use Survey (TUS) data for individuals aged 15-64 years. The gender difference in average time spent on family care, especially childcare, is particularly relevant for analysis of FLFP. Higher time allocation to certain domestic activities restricts the number of hours available for market work. Using TUS data from OECD (Organisation for Economic Co-operation and Development), we find that Indian men spend about 15 minutes on childcare per day which is consistent with the global average (Table 2); however, Indian women spend 70 minutes on childcare, which is substantially higher than the global average (Figure 2). Polish women come a close second in terms of the time spent on childcare (55 minutes) while German women spend the least (25 minutes).

Table 2. Time spent on childcare, Time-Use Surveys, 2019

Country	Time spent on childcare (minutes)
	Men	Women
Austria	20	45
Belgium	14	29
Canada	18	37
Estonia	18	42
Finland	13	31
France	13	29
Germany	11	25
Greece	13	25
Hungary	18	47
Italy	16	33
Japan	7	32
Korea	11	44
Luxembourg	12	28
Mexico	14	48
Netherlands	14	28
New Zealand	16	44
Norway	14	34
Poland	22	55
Spain	23	42
Sweden	19	31
Turkey	10	43
UK	15	37
USA	17	37
India	15	70

Source: OECD data, and authors’ computation using Indian Time Use Survey.

The effect of time allocated by Indian women towards children’s education is substantial. For example, if the time devoted by women to human capital investments in children (instructing, teaching, training and helping children) exceeded one hour a week (to conform to the definition of CWS for classification as work) then the FLFPR increases by 4 percentage points. Therefore, it appears that almost the entire observed gap in India’s FLFPR with AsiaLA-29 is on account of greater expansion of female enrolment in higher education in India (due to catch-up) or a higher time allocation to non-market work such as childcare. However, it is difficult to determine whether this significantly higher investment in childcare is due to choice or a result of limited access to suitable formal employment or flexible job options – factors which ultimately contribute to a lower FLFP in India.

Figure 2. Female labour force participation rate and time spent on childcare

Source: World Bank, OECD Data, and authors’ computation using Indian Time Use Survey.

Conclusion

India’s low female LFPR has been extensively discussed over the last decade with considerable concern regarding the relatively low contribution of women to the Indian labour market. Our findings suggest that the relatively low observed FLFPR in India can be explained by inconsistent treatment of non-market work by NSS, an expansion of women in higher education, and disproportionate time spent by women on non-market activities such as childcare.

The consistent treatment of home production significantly decreases the drop in FLFPR between 2004-05 and 2011-12. Furthermore, the expansion of women in higher education and the closure of the gender gap in higher education is a net positive as females who withdrew from workforce to acquire a college degree are more likely to work post completion of their degree relative to those without a degree. An interesting paradox is that Indian men spend more time on market work relative to other countries while Indian women spend more time on non-market work. This is reflected in the aggregate labour force participation rates for both genders. Consequently, Indian men would have to share the burden of non-market work to allow for a greater expansion of FLFPR in India.

Note:

PLFS survey provides employment statistics at a higher frequency relative to the erstwhile EUS surveys. The shift to PLFS is in line with rapid structural change of India’s economy and labour market which necessitates more frequent measurement of labour market conditions.