Introduction
This research analyses the measurable effects of overt sexual orientation on the earnings of workers in Pakistan. The sample data used in this study is overwhelmingly male. Thus, this study addresses the question of whether self-identified LGBT status has any explanatory power over differences in individual earnings of male workers in the Pakistan labour market. The empirical portion of this research uses two sub-samples: a sub-sample of LGBT workers and a sub-sample of their occupational counterpart non-LGBT workers.
In the interest of precision and concision, we use the terminology LGBT to identify the sub-sample of individuals in our research whose characteristics satisfy three conditions:
-
1. The sexual orientation of every individual in the LGBT sub-sample in our research is either lesbian, gay, bi-sexual or trans-sexual.
-
2. The sexual orientation of the individuals in the sub-sample is self-declared.
-
3. The members of the sub-sample are all employed and their employers know their sexual orientation.
Here, when we refer generally to the LGBT population, or the LGBT community, in Pakistan we mean persons who overtly display their sexual orientation as lesbian, gay, bi-sexual or trans-sexual. To the extent that there are earnings differentials among workers based on the sexual orientation of those workers, a possible explanation is invidious employment discrimination. That question has been of interest in academia as well as among policymakers concerned with the measurable manifestations of cultural attitudes (Reference Bailey, Wallace and WrightBailey et al., 2013; Reference Brewer and LyonsBrewer and Lyons, 2016; Reference DrydakisDrydakis, 2020; Reference Sandfort, de Graaf and BijlSandfort et al., 2003).
Employers’ discriminatory attitudes towards members of the LGBT community in Western societies are well-documented (Reference Badgett, Garnets and KimmelBadgett 2003; Reference D’Amico, Julien and TremblayD’Amico et al., 2015; Reference Gao and ZhangGao and Zhang, 2017; Reference KöllenKöllen, 2016) However, in traditional and patriarchal Asian societies with deeply rooted religious influences, as in Pakistan, one can observe multiple dimensions of social stratification based on ethnicity, religion, caste, gender and sexual orientation (International Labour Organization [ILO], 2022; Reference Mamun, Heyden, Yasser and KöllenMamun et al., 2016). The multiplicity of those stratifications makes it difficult to disentangle the effect of sexual orientation from other sociological influences on earnings.
Many sociological studies have shown that in societies where traditional rules control most public and many private activities, multiple types of discrimination in the workplace occur. To the extent that they are pervasive, these prejudices confer considerable economic advantages on employees identified as belonging to selectively favoured sociological groups, such as high caste sub-populations or those displaying ostentatiously masculine behaviour. The economic handicaps experienced by a social group that is not highly favoured (e.g. women) are often exacerbated if its members belong to a religious minority or are publicly identified with the LGBT community (Reference Begum Sadaquat and SheikhBegum Sadaquat and Sheikh, 2011; Reference Delavande and ZafarDelavande and Zafar, 2013). In Pakistan, for example, members of the LGBT community do not enjoy all the rights of citizens (Reference Khan and WadleyKhan, 2014).
The disfavoured social status of the LGBT community in Pakistan society diminishes the economic opportunities of its members and marginalises their active participation in the life of that society. Many people who identify as LGBT in Pakistan do not display ‘biological dysmorphia’, but their gender expressions do not comply with the traditionally approved gender expressions in that society (Reference Sadiq and BashirSadiq and Bashir, 2015). Reference Shah, Rashid and AtifShah et al. (2018) suggest that persons whose conduct displays overt non-conformity with the traditional culture in Pakistan society are thereby ostracised. The measurable effects of ostracism based on sexual orientation may be manifested in earnings differentials. It is the quantification of such differentials that is the purpose of this study.
The article begins with an overview of published findings relating to the position of LGBT people in Pakistan society and labour markets, followed by an outline of the measurement methodology adopted in this study. Econometric findings are then outlined, followed by a discussion and conclusion, identifying the need to extend this study and address its implications.
Social and cultural attitudes related to LGBT persons in Pakistan
The LGBT community in Pakistan
The existing systems of social organisation in traditional religious societies in Pakistan are characterised by gross disparities in educational opportunities, employment opportunities and access to health care. A direct consequence of these disparities is the exclusion of people who identify as being in one (or more) of the unfavourable groups at multiple levels in society. Pakistan has the second lowest human development index in South Asia (United Nations Development Program [UNDP], 2022). The likely explanation for that statistic is not mysterious; multiple hierarchies pervade Pakistan society based on ethnicity, faith, caste, gender and sexual orientation.
The disapproving attitude towards people overtly displaying disfavoured sexual orientation begins in the family. Family relations impose extraordinary restrictions on the behaviour of those members who openly identify as LGBT. These restrictions include concealment of the LGBT of juveniles, withholding sympathy and affection, and offering unequal educational and recreational opportunities as compared to their heterosexual counterparts. LGBT persons are regarded as inimical to the fundamental values of Pakistan society. People known (or thought) to be a part of the LGBT community are often banished from their homes because their sexual orientation offends religious and cultural traditions. In small village communities, people who identify as LGBT are ostracised and humiliated (Reference Khan and WadleyKhan, 2014). That treatment often induces them to reside in their own parochial communities, which are usually situated in large urban areas (Reference Shah, Rashid and AtifShah et al., 2018). The cohesion observed among LGBT people in these communities converts their social exclusion into a homogenous community of like-minded persons (Reference Sadiq and BashirSadiq and Bashir, 2015).
The Pakistan government issues identity cards to all employees, including LGBT workers. However, these documents do not translate into equality of economic opportunity for LGBT employees (Reference Kazi, Anwar and RaniKazi et al., 2014). The prejudice directed against the LGBT community manifests in many ways, including restricted educational opportunities and exclusion from mainstream Pakistan society (Reference Sadiq and BashirSadiq and Bashir, 2015). To the extent that wage discrimination is imposed by public and private employers, many LGBT workers are concentrated in the lower end of the income distribution (Reference Mamun, Heyden, Yasser and KöllenMamun et al., 2016).
Non-economic manifestations of discrimination are pervasive; members of the LGBT community experience humiliation and bullying. An instance of the cultural folklore in Pakistan society is a pervasive belief that members of the LGBT community generally suffer from cognitive deficits and experience difficulties in communication (Reference Collumbien, Chow and QureshiCollumbien et al., 2008). A direct consequence of cultural discrimination is that members of the LGBT community often lack the professional experience and expertise found in the general population. Their unfavourable societal status deprives them of opportunities available for many kinds of employment. It has been documented that the general work environment is hostile to LGBT employees in several high-level and professional occupations. Consequently, they leave those organisations and gravitate towards occupations that are deprecated in Pakistan society.
Social isolation of LGBT persons
Notwithstanding the diversity of Pakistan society, one can observe gender-based patterns that are common to more homogeneous communities. In Pakistan, these patterns tend to be detrimental to LGBT individuals. Although there are no legal barriers for LGBT persons to participate in social activities, they are routinely excluded from funeral ceremonies, religious and social processes and occasionally from events organised by political parties. Reference Abdullah, Zeeshan and KamalAbdullah et al. (2012) and Reference Shah, Rashid and AtifShah et al. (2018) suggest that it is to this this social exclusion of LGBT people that any poor development of cognitive and social skills can be attributed (see also Reference AhmedAhmed, 2010 and Reference Kazi, Anwar and RaniKazi et al., 2014).
In Pakistan, democracy is evolving to be more inclusive of women in politics and in the workplace. However, the LGBT community does not seem to enjoy increasing political participation. Their constituency is small, and many of them have never been registered voters in elections (Reference Abdullah, Zeeshan and KamalAbdullah et al., 2012). A consequence of their social marginalisation and small numbers is that political parties have little interest in reaching out to engage them in political processes. Some members of the LGBT community have moved away from their ancestral home sites and now reside in insular communities, many of which have no permanent address. The significance of this phenomenon is that the electoral franchise in Pakistan is tied to a permanent address, and individuals without such identification are disenfranchised.
This social marginalisation presents challenges to data collection, and makes available analysis all the more pressing. The following ‘Methodology’ section outlines the approach to statistical sampling, surveying and data analysis used in this study.
Methodology
The statistical sub-sample drawn from the LGBT population
We assembled what we believe to be a unique data set. Our sample consists of 380 respondents. About 190 respondents satisfied the three criteria defined in section ‘Introduction’.Footnote 1 An equal number self-identified as heterosexual. At the initial stage of our data gathering we elicited LGBT status by identifying persons who use social media applications known to be used by members of the LGBT population in Pakistan. These include Blued, Grinder, Romeo and Disco. Most LGBT people in Pakistan are thought to use these applications for socialising and meeting. People so identified were recruited into the data base. The recruits from these sources were asked to corroborate their status by self-identifying as LGBT. If they denied that status, they were not included in the database of the LGBT sub-sample.
Some of the LGBT observations were identified using snowball sampling, in addition to web-based applications whose clientele consists mainly (or exclusively) of members of the LGBT community. Snowball sampling is a non-probability sampling technique in which the observations in the sample have traits that are rare in the general population, or difficult to find in that population. In our study, the difficulty is finding LGBT workers in a traditional religious society.Footnote 2 In the application of snowball sampling in our research, self-identifying LGBT workers referred us to other workers known by them to be a part of the LGBT population. The referrals were recruited into the sample population in this study. Almost all the persons in the LGBT sub-sample were male.
The first author of this study personally interviewed respondents in the sample self-identifying as LGBT, after obtaining prior approval from their leaders and the nearest police authority. This process of corroboration and verification took almost 2.5 years (from January 2018 to August 2020). The interviewer elicited information to be included in the dataset. During these interviews, we collected data on earnings, occupation, educational attainment, employment, ethnicity, family life and residential status.
The authors collected data relating to the earnings of counterpart heterosexual workers. Counterpart heterosexual workers are defined in this study as non-LGBT persons working at same kind of occupation, or in a very similar kind of occupation.Footnote 3
The data collected from the non-LGBT workers included the same data collected from the LGBT workers in the sample. Observations in both sub-samples consisted of employees in various private and public sector organisations. Most respondents in our sample were employed in the private sector. A major challenge in gathering data pertaining to employment in the public sector was the lack of a publicly declared social identity; the Pakistan government does not recognise members of the LGBT community as a distinct social group.
We conducted face-to-face interviews with all respondents. During the interviews, we collected data through questionnaires. A qualitative survey framework was designed in consultation with the respondents to explore the parameters systematically associated with wage differentials. The questionnaire elicited basic information, including social and family life activities, employment, along with focus group discussions, face-to-face interviews, discussions and roundtable talks. It consisted of two parts: (a) demographics of participants: income, occupation, job status, age, experience, education, ethnicity, etc.) and (b) social, family and workplace life. The sample population comprised rural-urban mixed individuals from each province of Pakistan.
The content and questions were discussed in detail before data was collected. We selected our sub-populations as individuals who resided in groups located in Lahore, Multan, Peshawar, Karachi, Rawalpindi and Islamabad. We divided each gross sample of respondents into employed persons and unemployed persons. Only employed respondents were included in the LGBT sub-sample and the non-LGBT sub-sample. This sample dichotomy is amenable to an econometric analysis of the factors systematically associated with LGBT wage differentials in relation to heterosexual workers in comparable employment.
The statistical sub-sample drawn from the non-LGBT population
The sub-sample drawn from the non-LGBT population was conducted posterior to the data collected from the LGBT population. The following data collection protocol was applied: For every occupation represented in the LGBT sub-sample, we drew from the non-LGBT population a vector of observational variables for a worker with the same or a very similar occupation. This was a relatively easy task of data collection because it did not entail sensitive issues of sexual orientation. In many instances, we drew observations from the non-LGBT population of workers employed by the same employer as their occupational counterparts in the LGBT population.
The econometric model
The econometric specification we applied is an adaptation of the model developed by Reference BlandfordBlandford (2007).
The glossary of variables is displayed below.
ln (wages) represents the natural logarithm of the earnings of persons in the sample age denotes the respondent’s age.
Edu represents the respondent’s education level. This is defined as the number of years of formal schooling completed by the respondent.
Sector is a binary-valued variable. It is assigned value of 1 for a respondent who is employed in the private sector and 0 otherwise.
LGBT is a binary variable; it is assigned the value 1 for a respondent who identifies as LGBT and 0 otherwise.
ethnicity is an integer-valued variable. It is defined to identify the ethnic group associated with each respondent in the sample: Pashtoon = 1, Sindhi = 2, Baloch = 3, Kashmiri = 4 and Saraiki = 5. Footnote 4 The ethnicity of every respondent was elicited by questionnaires and interviews.
Urban is a binary variable; it is assigned a value of 1 if the respondent is located in an urban area and 0 otherwise.
Fam is a binary variable; it is assigned a value of 1 if the respondent resides with a family and 0 otherwise.
ε represents the randomly distributed error term.
The specification in (1) is carried out in three different applications. The applications differ from each other only insofar as explanatory variables are added sequentially to permit an assessment of the marginal effects on the significance of the pre-specified explanatory variables.
Statistical properties of the sample
The size of the LGBT community in Pakistan
The percentage of the Pakistan population consisting of LGBT individuals as measured by the census is quite small. The enumeration in the census of LGBT persons is likely underreported because their enumeration in the census is based on self-reporting. Individuals who identify themselves as LGBT are issued government documents displaying that identification. Members of the LGBT community were issued identity cards for the first time in the history of Pakistan (Reference AliAli, 2018). Such identifications can have pernicious consequences.
Table 1 displays the population of Pakistan in 2017 (207.77 million) and the size of the LGBT community in Pakistan in the same year (10,418). The table shows that in year 2017, the self-reported members of LGBT population in Pakistan was only 5% of the total population.
Table 1. Population of LGBT persons in Pakistan in 2017.
Source: Pakistan Bureau of Statistics (Population census, 2017).
Table 1 also suggests that the population of LGBT persons is relatively small outside the largest urban environments in Pakistan, as cited by the population census 2017. Figure 1 below shows a long-term average unemployment rate of 10%–15% among heterosexual workers whereas the unemployment rate among LGBT workers ranges from 80% to 90%.
Figure 1. Unemployment rates in Pakistan.
Source: Pakistan Bureau of Statistic and authors’ calculations.
Determinants of earning differentials
Table 2 below displays the coefficient estimates of alternative specifications of equation (1) above.
Table 2. Determinants of earning differentials.
** and * denote significant levels at the 1% and 5% levels, respectively. The t-statistics are in parentheses.
The findings of Model-I in Table 2 indicate that an increase of 1 year of age, 1 year of additional education and residence in urban areas are statistically associated with increases in earnings of 2.3%, 2.1% and 6.7%, respectively. In Model I, the evidence as to the effect of ethnicity suggests that it is not significantly related to the earnings of the workers in the sample. In Model-I the coefficient of age squared is not significant. The findings in Model-I also suggest that employment in the private sector is associated with smaller earnings than in the public sector.
The dummy variable representing the LGBT observations in Model-I is of special statistical significance by reason of the fact that its t-statistic is much larger than that of any other explanatory variable. This finding leaves little doubt that sexual orientation is significantly associated with earnings differentials observed between LGBT workers and their heterosexual counterparts. The numerical value of the estimated coefficient of LGBT in Model-I (i.e. −0.131) implies that LGBT orientation is strongly associated with lower earnings in the sample.
As a general inference, the findings in Model-I imply that employees in the sample who are identified as LGBT will have lower earnings than their heterosexual counterparts in the sample, after controlling for age and level of education.
The specification of Model-II represents an attempt to measure the co-varying effect of sexual orientation and age on earnings. That effect, if any, is measured by the product of LGBT × age as well as LGBT × age2. The findings of Model-II suggest that the earnings differentials of LGBT employees in the sample tends to be lessened by age.
The specification of Model-III represents an attempt to separate the effects on earnings of age and education in the LGBT sub-sample. We attempt to capture this effect by specifying the variable defined as the variable LGBT × education. The coefficient of that variable is positive, which suggests that education may tend to lessen the adverse wage differentials experienced by respondents in the LGBT sub-sample. However, the effect of palliation, if any, does not appear to be statistically significant in the sample population.
It should be noted that the adjusted R 2 statistics were significant and reasonably stable across all models. That finding suggests that the specification explains approximately 45% of the variation in the logarithm of earnings in the sample. Since the sample data are cross-sectional, an R 2 statistic of that magnitude suggests that the specification captures robust systematic relationships among the explanatory variables and earnings differentials.
Conclusion
This study addressed the question of whether there are earnings differentials observed in labour markets in Pakistan that are systematically related to the overt sexual orientation of workers. The statistical sample in the study consisted of data collected from male LGBT workers in Pakistan and male heterosexual workers employed in the same occupations.
We conclude that self-identified LGBT workers experienced significantly smaller earnings than their heterosexual counterparts employed in the same occupations with the same educational credentials. The statistical evidence supporting this conclusion is the finding that an LGBT dummy variable is a significant determinant of earnings. That finding is consistent with a hypothesis of employment discrimination based on sexual orientation. However, because of the small sample size in this research the evidence is only suggestive as to the question of employment discrimination. A larger, carefully controlled study is needed to establish the incidence and impacts of employment discrimination.
The methodological limitations and the conclusion in this research suggests a possible direction for future research. One such direction might be the development of a testable theory and its empirical validation which will explain observed earnings differentials between LGBT workers and non-LGBT workers in the Pakistan labour force.
The empirical validation of an explanatory theory will be likely to require an assembly of a larger and richer data base than was used in this research. The richness of that data base will permit a more refined and probing explanation for the earnings differentials. Employment discrimination is a hypothesis that might help to explain earnings differentials, but there are other hypotheses (not necessarily incompatible with discrimination) which might also have explanatory power.
Acknowledgements
We are grateful to Muhammad Riaz Deputy Census Commissioner, Pakistan Bureau of Statistics for the provision of secondary data and to Gurus of LGBT community (leaders of LGBT community living in Lahore, Multan, Peshawar, Karachi, Rawalpindi and Islamabad) for their support in primary data collection process especially in arrangements of Focused Group Discussions (FGDs) and interviews with their subordinates. The authors acknowledge the thoughtful and encouraging comments of the reviewers and the Editor-in-Chief. Their comments increased the merits of the research enormously.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
