Measuring the occupational segregation of males and females in Pakistan in a multigroup context

The paper examines occupational segregation—the uneven distribution of demographic groups, particularly men and women, across occupations—and its implications for gender disparities in wages, job quality, and employment outcomes. Theoretical explanations considered include: compensating differentials (women may trade wages for non-pecuniary job attributes), crowding (women concentrated in a few occupations depresses their wages), and devaluation (women’s work is institutionally undervalued). In Pakistan’s patriarchal context, women face low labour force participation, concentration in select sectors (notably agriculture), and persistent gender wage gaps. The study’s purpose is to provide a detailed analysis of occupational segregation by gender in Pakistan, moving beyond overall (binary) measures to analyze multiple subgroups (age, education, hours, and sector/ownership) using local segregation measures in a multigroup framework. This enables assessing whether occupational distributions differ similarly for men and women across key characteristics and identifying which subgroups drive overall segregation.

Classical segregation measures (e.g., Index of Dissimilarity by Duncan and Duncan; Karmel–MacLachlan; WE index; sex-ratio measures; Gini-based measures) largely address binary groups and summarize overall segregation. As societies diversify, multigroup indexes (Reardon and Firebaugh’s G, C, H, P, R; Theil–Finizza mutual information) capture evenness across multiple groups but often only at an aggregate level. To study specific groups’ situations, local segregation indexes by Alonso-Villar and Del Río (2010) quantify each subgroup’s segregation relative to the total employment distribution and can be aggregated consistently to overall measures. Empirical studies across countries show persistent occupational gender segregation with cross-country variation; women often cluster in low-paid occupations, and individual characteristics do not necessarily reduce segregation. Prior Pakistan-focused work (Ahmed and Hyder, 2008; Irfan et al., 2013) found high gender segregation using binary measures. This study fills a gap by applying local multigroup measures at a detailed occupational level and across subgroups in Pakistan.

Design and measures: The study applies local segregation measures (Alonso-Villar and Del Río, 2010) to quantify each target group’s segregation across occupations relative to the overall employment distribution. Local indexes used include: a Gini-variant G^s; a generalized entropy family Φ^α (with α as segregation aversion); and a multigroup variant of the dissimilarity index D^s. These range from 0 (no segregation) to 1 (complete segregation) or can be transformed to (0,1). Consistency with overall segregation is ensured: overall Karmel–MacLachlan/Silber IP (I^ps), multigroup Gini (G), and mutual information (M) can be expressed as weighted averages of local indexes using groups’ employment shares. This framework avoids pairwise group comparisons and permits decomposition of overall segregation into group contributions.

Data: Microdata come from three waves of the Pakistan Labour Force Survey (2013–14, 2014–15, 2017–18), covering paid employees aged 15–65, working full-time. Own-account and contributing family workers are excluded. After excluding missing values, the final sample is 64,946 observations; males constitute 86.46% of the sample. Occupations are coded at the ISCO-08 three-digit level. Key labour market indicators (2013–18) show consistently lower female employment-to-population ratios, participation rates, and higher unemployment than males. Descriptive occupation and gender composition tables document substantial clustering by occupation.

Subgroup analysis: Beyond overall gender segregation, the study estimates local segregation by:

• Region: urban vs rural (by gender)
• Age: 15–29 (young), 30–44 (middle-aged), 45+ (elderly), by gender
• Education: low (up to secondary) vs high (above secondary), by gender
• Ownership/type of organization: public vs private sector, by gender
• Hours of work: full-time vs part-time (<48 hours/week), by gender Segregation curves (local segregation curves) plot cumulative employment (x-axis) vs cumulative proportion of target group (y-axis) to visualize dominance relationships across subgroups.

Overall and group contributions:

• Despite females comprising about 14% of total employment, females account for approximately 83% of overall segregation by the mutual information index (M). By multigroup Gini (G) and multigroup dissimilarity (Ip), females contribute about 50% each to overall segregation, reflecting pronounced feminization/masculinization patterns.
• Local segregation indexes are markedly higher for females than males: female D^k ≈ 0.60 and G^k ≈ 0.64 versus male D^k ≈ 0.10 and G^k ≈ 0.11; female Φ indices are substantially larger than male values.

Region (Table 5):

• Females experience higher segregation than males in both regions. Female rural: D^k 0.63, G^k 0.72; female urban: D^k 0.64, G^k 0.74. Male rural: D^k 0.16, G^k 0.22; male urban: D^k 0.24, G^k 0.34. Segregation among males is higher in urban than rural areas; among females, rural women show slightly higher Φ at low α but broadly high in both regions. About 67.89% of females and 59.91% of males are rural.

Age (Table 6):

• Females: segregation rises with age; >45 years shows highest segregation (D^E 0.62, G^F 0.71), followed by 30–45 (D^E 0.61, G^F 0.66), and <30 (D^E 0.58, G^F 0.65).
• Males: elderly males have the highest segregation (D^E 0.19, G^F 0.26); younger and middle-aged males show lower segregation (e.g., <30 D^E 0.17, G^F 0.23; 30–45 D^E 0.12, G^F 0.17). Labour force shares: females <30 42.27%, 30–45 41.85%, >45 15.89%; males <30 43.05%, 30–45 39.20%, >45 17.75%.

Education (Table 7):

• Higher education is associated with more segregation, especially for females. Female high education: D 0.83, G 0.90; female low education: D 0.63, G 0.69. Male low education shows the lowest segregation (D 0.17, G 0.21); male high education is higher (D 0.60, G 0.70). Workforce shares: males high 16.79%, low 83.21%; females high 17.05%, low 81.95%.

Ownership/type of organization (Table 8):

• Public sector workers are more segregated than private sector workers for both genders, with the highest levels among public-sector females. Female public: D 0.76, G 0.81; female private: D 0.66, G 0.77. Male public: D 0.54, G 0.61; male private: D 0.27, G 0.32. Distribution: males public 28.10%, private 71.90%; females public 55.01%, private 44.99%.

Hours of work (Table 9):

• Part-time workers exhibit higher segregation than full-time workers, especially among females. Female part-time: D 0.63, G 0.67; female full-time: D 0.48, G 0.60. Male part-time: D 0.24, G 0.32; male full-time: D 0.16, G 0.21. About 77.36% of females work <48 hours/week versus 30.52% of males.

Additional patterns:

• Female segregation curves consistently lie farther from equality than male curves across subgroups, indicating constrained occupational distributions for women. Human capital (education) does not reduce segregation; in several cases, higher education is associated with greater segregation for both genders. Elderly workers, particularly elderly women, show the highest segregation levels.

The findings demonstrate that occupational segregation in Pakistan is substantial and is disproportionately driven by the distribution of female employment across occupations. Even with a small demographic share, women’s occupational clustering contributes most to overall segregation (notably by the mutual information measure). This segregation is pervasive across regions, age groups, educational levels, organizational types, and working hours, with elderly women and highly educated women experiencing particularly high segregation. The patterns suggest that human capital accumulation alone has not translated into broader occupational integration for women. These outcomes are consistent with partial support for both compensating differentials (women concentrating in jobs with shorter hours or perceived safer/compatible conditions) and devaluation perspectives (institutional and social undervaluation of women’s work). The segmentation observed in public sector employment and among part-time workers underscores structural and institutional factors shaping occupational sorting. The results are relevant for policies targeting gender norms, job design, and institutional practices to reduce occupational barriers and broaden women’s access to diverse, higher-valued occupations.

This study applies local multigroup segregation measures to Pakistan’s labour market using detailed occupational microdata, providing both overall and subgroup-specific assessments of gender segregation. Key contributions include: (a) documenting that female workers, despite a small employment share, account for a large portion of overall segregation; (b) revealing systematically higher segregation for women across regions, with the highest levels among elderly and highly educated women; (c) showing that public-sector and part-time employment are associated with greater segregation for both genders; and (d) demonstrating that higher education does not necessarily reduce occupational segregation. Policy implications include promoting gender-norm change, investing in women’s education and skills aligned with labour market demand, ensuring gender-friendly workplaces, and incentivizing employers to adopt inclusive practices; facilitating flexible, safe, and mobility-sensitive work options—particularly for rural women—may also help. Future research should: (1) compare formal and informal sectors; (2) incorporate vertical segregation (hierarchical positions) when data permit; (3) explore causal pathways between education, norms, and occupational sorting; and (4) examine industry- and occupation-specific institutional factors that sustain segregation.

The analysis focuses on paid employees in the formal sector, excluding own-account and contributing family workers, which may limit generalizability to the broader labour market. The study addresses horizontal (across-occupation) segregation and does not analyze vertical segregation due to data constraints. Cross-sectional survey waves and available variables limit causal inference on the drivers of segregation.