Marriage, dowry, and women’s status in rural Punjab, Pakistan

2.1 Literature on dowry

Although empirical evidence is scarce, economists have actively conducted theoretical studies on dowry (Becker 1991; Botticini and Siow 2003; Anderson 2003, 2007; Boserup 2007). The two main interpretations of dowry see it as (1) a price determined in the marriage market (price model) and (2) the pre-mortem inheritance (bequest model). The definition of dowry has been the subject of discussions that aim to define its nature. In line with the price hypothesis, dowry is often defined as a transfer of cash, gifts, or both from the bride’s parents to the groom and his parents at the time of marriage, which is expected or even demanded by the groom and his family. However, the bequest hypothesis considers dowry as property taken by the bride to her new home or given to her during the marriage rituals by her parents.

These two interpretations of dowry are often considered mutually exclusive (Zhang and Chan 1999; Anderson 2003; Arunachalam and Logan 2016), and existing studies focus on the question of whether dowry is, in fact, a price or a bequest. Such a definitive definition is considered important because the two hypotheses are believed to lead to opposite policy implications. If dowry is a pre-mortem bequest voluntarily offered by the bride’s parents, laws prohibiting dowry practice might not be necessary. If dowry is a price that can be driven up in the current marriage market where the bride’s income-earning ability or higher human capital accumulation does not seem to be fairly evaluated as compared with the groom’s, it potentially decreases girls’ or wives’ chances of survival; thus, its prohibition by law may be effective in enhancing women’s welfare.

Although these two models are often considered mutually exclusive, typical empirical studies construct an estimation model to test the price or the bequest model independently; thus, such studies cannot reject one over the other even if they claim their results support such a conclusion. Those testing the price model typically regress the dowry amount on the bride’s and the groom’s characteristics (Behrman et al. 1995, 1999; Deolalikar and Rao 1998; Mbiti 2008), and most of them find no, or at most weak, evidence supporting the price model.⁹ Those testing the bequest model usually regress women’s welfare measures on the amount of dowry (Zhang and Chan 1999; Brown 2009). They claim that the finding that dowry enhances women’s welfare supports the bequest model; however, the finding does not necessarily reject the price model because women’s welfare could increase with a higher dowry amount under the price model. Possibly, dowry is determined in the marriage market, but the bride’s parents may also voluntarily pay a higher amount so that the groom and his parents will treat their daughter better after marriage.

The current study is not the first to empirically investigate the relation between dowry and women’s welfare and empowerment. Zhang and Chan (1999) and Brown (2009) use East Asian datasets and show that dowries have positive effects on several measures of women’s welfare. Jejeebhoy (2000) demonstrates that dowry size is positively associated with women’s decision-making power in north India but not in the south. Bloch and Rao (2002) and Srinivasan and Bedi (2007), both using data from a few specific Indian villages, indicate that women with higher dowry amounts are less likely to suffer domestic violence from their husbands, while Suran et al. (2004), using data from rural Bangladesh, find a completely opposite effect. These mixed empirical results may be due to the fact that dowry has different meanings and roles across both time and space. Chan (2014) uses data from Karnataka, India to consider dowry’s heterogenous nature and decomposes it into two components, i.e., a bequest dowry over which the wife has control and a groom price over which she does not; it is only the former that enhances women’s status in the marital household. In sum, the relation between dowry and women’s welfare remains an open empirical question because its precise nature likely varies depending on the social context.

2.2 Problems with existing data

The biggest obstacle to conducting empirical studies on dowry is the lack or inadequacy of data. Because dowry is legally banned in India and Bangladesh, people usually hesitate to reveal accurate dowry amounts. The standard question on dowry in the Indian dataset asks about community-based dowry. For example, the India Human Development Survey asks “Generally in your community for a family like yours, what are the kind of things that are given as gifts at the time of the daughter’s marriage?” A community-based dowry is not necessarily the same as an individual dowry, which is paid by the bride’s parents at the time of marriage. Alternatively, the question regarding dowry allows only a binary answer, i.e., whether or not a positive amount of dowry was paid (e.g., Survey on the Status of Women and Fertility both in India and Pakistan). A binary answer, of course, does not provide much additional information. The norm of whether a positive dowry amount is provided corresponds to, and is largely explained by, ethnic, religious, and caste backgrounds in South Asia.

Since dowry is not legally banned in Pakistan, the amount personally paid by the female respondent’s parents can be asked without reservation in a Pakistani dataset such as the Pakistan Rural Household Survey. Although Pakistani interviewees may not intentionally conceal true information on dowry practices, the survey may nonetheless contain recall errors because respondents must recall the dowry amount paid by their parents several years ago. Figure 1a plots the predicted amount of real dowry values measured in Pakistani Rupees in 2004 onto marriage year using data from the Pakistan Rural Household Survey. Because the consensus is that real dowry amounts are increasing, or at least exhibit a non-declining trend, the figure implies the general tendency of recall errors. In other words, the longer the gap between the interviewees’ marriage and the recall time, the more likely they are to overestimate the dowry amount.

Open image in new window

3.1 Survey

When conducting our survey between June and October 2013, we intended to capture the heterogeneous aspects of the Punjab province in Pakistan. We divided Punjab (36 districts) into five regions: Pothohar (or North), Central, East, West, and South Punjab. Climate, culture (including marriage/inheritance practices), and socioeconomic conditions differ across regions but are similar within each region. We randomly selected one district from each region, namely, Rawalpindi, Mandi Bahauddin, Narowal, Muzaffargarh, and Bahawalnagar (Fig. 2). We used the district census for 1998–1999, the latest census available in Pakistan as of 2013, to randomly select six villages in rural areas in each of the five districts. We restricted sampling villages to those with a population of at least 1000 at the time of the census. In each village, we selected 22 households, following a stratified random sampling methodology. First, with assistance from the village chief, we made a list of households in the village and categorized them into a stratum. The strata are kammees¹⁰ (i.e., traditional service or lower caste, with annual income ≤ PKR 200,000, > PKR 200,000) and zamindars (i.e., landowning farmers with land < 5, 5–12.5, and > 12.5 acres). Eligible households in our survey are defined as those with an economically active husband and wife aged 15–65. Second, we performed stratified random sampling so that the share of each stratum in our sample corresponds to the share of each stratum of the village population (= households).

Open image in new window

On the basis of the sampling process explained above, we conducted questionnaire-based structural interviews. Ninety-nine percent of those selected responded to our interviews. The questionnaire was carefully designed to comprehensively understand marriage practices in rural Punjab, Pakistan. The questionnaire consists of two parts; the first contains questions for the husband, and the second has questions for the wife. Because the second part contains sensitive questions to assess the wife’s status in the marital household, we attempted to maintain the wife’s privacy as much as possible, for example, by requesting a separate interview room so that the wife could answer without feeling pressure from her husband.

3.2 Marriage practices

The first part of the questionnaire asks husbands about their marriage practices, including dowry amount. Because dowry is well known to consist of various items that enable a young couple to start their married life immediately after marriage, we asked about the value of each item within the dowry. The actual question is “Generally, in your community, for a family like yours, what is the approximate value of each item given as dowry at the time of the daughter’s marriage?” We explicitly asked about dowry as observed in their community, or jati (= sub-caste), because we are interested in the practice of dowry itself as well as the personal amount of dowry their parents paid. Their answer to the question about dowry in the community is likely to convey precise information about dowry practices. Their reply might report how much they would pay for each item if they are currently in the process of providing a dowry to their daughter. In either case, the answer is likely to provide more precise measures on the itemized value of dowry. Presumably, remembering every single item of their personal dowries from several years ago would be difficult. The second part of the questionnaire asks wives about their personal dowries and other marriage expenses, i.e., “How much dowry (bride price) did your parents (in-laws) provide at the time of your marriage?” These personal amounts are used in the main empirical analysis. The correlation coefficient between the personal dowry amount and community-based dowry amount is 0.65 (Fig. 3). The community-based dowry amount is higher than the personal dowry amount paid by the wives’ parents, which is not surprising given the alleged increasing trend in the real amount of dowry.

Open image in new window

Panel B of Table 1 shows the itemized average dowry values generally provided by a daughter’s parents at the time of her marriage. Contrary with our expectation, the amount of cash is not very large, and cash included in a dowry, especially cash given to the groom, seems a token payment, with an average negligible amount of PKR 3759. The average value of gold/jewelry offered to the bride by her parents is the largest among all items, PKR 76,651. Both cash and gold/jewelry are offered by approximately 90% of brides’ parents. Also, somewhat to our surprise, the average value of electronics, furniture, and kitchenware offered to the bride by her parents is high, and these items are offered by an even higher percentage of the brides’ parents (95 to 100%). Although each item has an average value lower than that of gold/jewelry, the average value of furniture, electronics, and kitchenware combined amounts to PKR 136,390—much greater than that of the gifted gold/jewelry. Although we should carefully interpret the amount of gold/jewelry offered to the bride by her parents as gifts to the bride because such items can easily be converted into cash and might be taken by the groom and his parents, items such as furniture, electronics, and kitchenware can be safely interpreted as gifts to the bride by her parents. In India, the groom’s parents are reported as often asking the bride’s parents for dowry to prepare future dowries for their daughters (circulating dowry). However, in rural Pakistani Punjab, the largest share of dowry being furniture/electronics/kitchenware makes it difficult to support the hypothesis of circulating dowry. Looking at items’ values in detail, dowry seems a trousseau voluntarily offered by the bride’s parents to their daughter at the time of her marriage and is at her disposal once in the marital household.

Although dowry expenses incurred by the bride’s parents are notoriously known and, in fact, are the single greatest expense in marriage, the expenses incurred by the groom’s parents are far from negligible. Panel B also shows the average value of marriage expenses generally incurred by both sides. In addition to the ceremony expense,¹⁴ the groom’s side also bears the cost of gifts to the bride called bari, an indispensable part of the ceremony. Bari typically consists of jewelry and clothing offered to the bride and her female relatives, and it can be considered a customary bride price. These two major marriage expenses incurred by the groom’s side, i.e., the ceremony expense and bari, together are equivalent to the amount of dowry incurred by the bride’s side. This indicates that marriage expenses in Pakistani Punjab are not disproportionally borne by the bride’s parents.

Also, dowry is sometimes partially incurred by the groom’s side. We observed that in some communities, the groom’s side customarily bears 50–60% of the dowry expense. Although groom-side households bearing half the expense of dowry account for only 7% of the sample, and therefore, the average value is only PKR 10,713, this custom is far from negligible because it does not fit into either the price hypothesis or the bequest hypothesis. Under the price model, dowry payment should be one-sided, made by the side gaining from marriage or oversupplied in the marriage market (Zhang and Chan 1999). Under the bequest model, dowry should be paid entirely by the bride’s parents (Botticini and Siow 2003). The fact that the groom’s side bears approximately half the dowry expenses can be consistent with the idea that dowry serves as a resource to help the new couple start their marital life. It also fits into the interpretation of dowry as trousseau, implied by the itemized average value of dowry.

Marriage is a contract for Muslims, and mehr (Islamic bride price), not dowry, is required to conclude the marriage. Mehr consists of two parts: (1) moajel, the immediate transfer at the time of marriage from the groom’s to the bride’s side and (2) non-moajel, a deferred transfer promised for payment at the time of divorce. A substantial amount of mehr is non-moajel, PKR 32,336 on average, though only 17% of those interviewed answer that they generally specify non-moajel in the marriage contract. Moajel is considered to become a token payment in contemporary rural Punjab, Pakistan, and a negligible amount (PKR 7253 on average) is reported.¹⁵ Because the amount of mehr is written into the marriage contract and is binding, it might enhance women’s status in the marital household because their husbands cannot obtain a no-fault divorce without incurring substantial costs corresponding to the amount of mehr.¹⁶ We do not find any strong correlation between the practice of writing non-moajel into the marriage contract and household status (whether zamindars or kammees) or household wealth (quality of living and size of land ownership).

We asked wives why the real amount of dowry differed among siblings.¹⁷ Excluding “emotional attachment” to one daughter and an “upward trend” in dowry amounts, which are less meaningful answers in our study, dowry offered to the bride by her parents tends to be higher when (1) the groom is of higher quality (higher education, higher earning ability), (2) the groom’s family’s status/economic condition is better, (3) the bride’s parents are in a better financial condition (compared with the groom’s family and/or to the time of their other daughters’ marriages), and/or (4) the marriage is arranged out of biradari (literally “brotherhood,” a group of male kin). As determinants of the dowry amount, these factors seem to reinforce the importance of groom quality and the bride’s parents’ financial capacity to pay. The idea that a higher quality groom increases the dowry amount is better explained by the price model, while the idea that the financial capacity of the bride’s parents increases the dowry amount is closer to the bequest model. This seems to indicate that these two major hypotheses about dowry, the price and bequest hypotheses, are not exclusive. To suggest effective policies aimed at improving women’s welfare, examining the relation between dowry and women’s welfare in the given context could be more useful rather than discussing which hypothesis best captures the true nature of dowry.

3.3 Measures of women’s status

The questions designed to gauge women’s status/empowerment are all answered by wives and summarized in Panel C of Table 1. The first measure is women’s decision-making power; we asked wives who has the most say in decision making regarding (1) what to cook on a daily basis, (2) whether to buy an expensive item such as a television or refrigerator, (3) how many children to have, (4) what to do if a child falls sick, and (5) whom the children should marry. Each variable for women’s decision making equals one if the wife has the most say in deciding each item and zero otherwise. As expected, the majority of wives, 75 and 63%, have the most say on what to cook and what to do when a child falls sick, respectively. In contrast, only a small fraction of wives, 18, 26, and 26%, make decisions on major household expenses, fertility, and children’s marriages, respectively. We construct the decision-making index by principal component analysis allowing for correlations across decision-making variables (see Gorsuch 2003). The index variable equals the only factor having an eigenvalue greater than one.

The second measure is women’s autonomy; we asked the wife whether she has to ask for her husband’s permission to go to (1) the local health center, (2) the home of relatives/friends in the village, and (3) the neighborhood shop. Each variable for women’s autonomy equals one if the wife does not need to ask permission from her husband and zero otherwise. Wives in Pakistani Punjab seem, on average, to have autonomy. Approximately 70% of wives do not have to ask permission from their husbands to go to the local health center, and approximately 75% of wives can visit their relatives/friends and a neighborhood shop without asking for permission. The reason for their modest autonomy could be partially because of the prevalence of village endogamy, implying that village people are relatives, and the practice of purdah¹⁸ is relatively relaxed within the village. As in the case of the decision-making index, the autonomy index is constructed using principal component analysis. Figure 4 demonstrates the kernel densities of these two indices.

Open image in new window

4.1 Estimation model

As the dowry amount is presumably endogenous, we address the endogeneity concern in three ways. First, we include a rich set of observed characteristics such as the bride’s natal family’s income at the time of marriage and the bride price paid by the groom’s parents. Then, we follow the procedure developed by Altonji et al. (2005) to indicate how large the effects of unobservables should be in order to remove the effects of dowry. Second, we include disaggregated marriage expenses, namely, cash/gold/jewelry, furniture/electronics/kitchenware, ceremony expenses paid by the bride’s household, and bride price (immediate and deferred). The objective here is to compare the association between women’s status and a component of dowry, either cash/gold/jewelry or furniture/electronics/kitchenware, among the households who can afford the same level of ceremony expenses. Third, we conduct a series of reduced-form regressions of women’s status on three different sets of dowry and bride price paid to others. We do not impose an implausible exclusion restriction.¹⁹ Instead, we would like to show that women enjoy higher status in the marital household when they come from a family that tends to exchange higher dowries; when they come from a higher-dowry village; or when they come from a higher-dowry community.

In a series of reduced-form regressions, we make the variables of dowry and bride price that reflect the marriage market, excluding the bride’s and groom’s personal unobservables by construction. In the literature, this is known as the “−i method”, which is used to construct an instrument when no instrument exists (Aizer 2010; Vogl 2013). In Punjab, Pakistan, the marriage market is defined among relatives rather than within the village. Thus, the first set considers the wife’s brothers’ average amount of dowry paid by the brothers’ in-laws (\( \overline{D_{i_{br}j}} \)) and the wife’s sisters’ average amount of bride price paid by the sisters’ in-laws (\( \overline{B_{i_{sis}j}} \)). Because of the prevalence of endogamy and strong assortative mating in rural Pakistani marriages, the average dowries (bride prices) that the wife’s brothers’ (sisters’) in-laws paid are presumably positively associated with the dowry (bride price) amount that the wife personally received from her parents (in-laws). A drawback of using \( \overline{D_{i_{br}j}} \) and \( \overline{B_{i_{sis}j}} \), is the reduction in sample size and the possibility of selection bias given that only women with at least one sister and one brother can be included in the sample. We address this concern in the estimation. The second set considers village-level average amounts of dowry (bride price) excluding the wife’s own. This is closer to the original −i method proposed by the existing literature (Aizer 2010; Vogl 2013), arguing that the instruments based on a leave-out village average do not include household unobservables by construction. In addition, they neither reduce the sample size nor induce sample selection bias. However, the second set also poses drawbacks. That using the second set, \( \overline{D_{- ij}} \) and \( \overline{B_{- ij}} \), cannot be estimated with village fixed effects because this set captures the village average excluding the own dowry (bride price) amount and, thus, is spuriously negatively correlated with the own amount when the village average is controlled. Instead, we control for the village-level wealth measure, which is constructed by principal component analysis using information on the condition of housing, namely, roof, walls, and flooring. The third set considers the community-based dowry (bride price) as a proxy for the personal dowry (bride price) as in Chan (2014). Here, community means jati (= sub-caste), sharing the same standard of living and a common marriage market. The questionnaire to the husband asks for the community-based amount of dowry and bride price typically paid at the time of marriage, as discussed in Section 3.

4.2 Main results

One might argue that positive association between dowry and women’s status simply reflects household affluence, and women in more affluent families are usually better treated and, thus, more empowered. However, if this argument holds, the association between immediate bride price and women’s status should be positive and stronger because it directly reflects the wealth of the groom’s household and, thus, the marital household. Although not statistically significant, the amount of bride price that the wife received from her in-laws at the time of marriage decreases her status in the marital household. Overall, these results imply that the greater the dowry, the more likely that the wife has decision-making power and autonomy, even after controlling for household wealth.

In contrast, the amount of non-moajel mehr (deferred bride price) shows a positive and significant association with the wife’s autonomy. This is in line with Ambrus et al. (2010), by suggesting that mehr increases the cost of unilateral divorce from the husband’s side and functions to protect the wife in rural Punjab, Pakistan.

As expected, the higher a wife’s education, the higher her decision-making power. The significantly negative effects of literacy are unexpected, which might simply capture the fact that lower-caste wives are less oppressed than upper-caste wives and act relatively independently of their husbands in South Asia (Chakraborty and Kim 2010; Bidner and Eswaran 2015). Interestingly, endogamy, both marriage within the same village and marriage between cousins, is significantly associated with the wife having lower decision-making power.²¹

Following Altonji et al. (2005), we check how large the effects of unobservables would need to be in order to completely eliminate the dowry’s effects. Without a set of observables, the dowry effect decreases by 0.35 and 0.1 percentage points in terms of women’s decision making and autonomy, respectively.²² These estimates imply that the effects of unobservables should be at least 1.6 times larger than those of observables.

4.3 Discussion and robustness checks

In determining the women’s status in the marital household, the effect of the net dowry (dowry minus bride price) may matter. The estimation procedure is repeated by replacing dowry and bride price with net dowry (Appendix Table 9, Panel A). Note that the amount of deferred mehr is included separately from the net dowry because the former is never paid unless divorce occurs and cannot be considered assets brought into marriage. The results suggest that the assets brought into the marriage (i.e., bride’s net dowry amount) enhance her decision-making power and autonomy. These positive associations are not surprising given that the associations of bride price paid at the time of marriage run counter to those of dowry in the main estimations.

Possibly, a greater age or education difference might weaken the wife’s status in the marital household. Including age difference as well as education difference (replacing the husband’s age and education) does not affect the main outcomes, and the coefficients of these variables are not significant (Appendix Table 9, Panel B).

One might argue that the number of years since marriage is important because the effect of dowry (or bride price) might be greater soon after marriage. Including years since marriage and its interaction term with the amount of dowry (replacing woman’s age at marriage) magnifies the base effect of dowry in women’s autonomy but decreases its magnitude and the significance level in women’s decision making (Appendix Table 9, Panel C). This implies that a larger dowry increases the wife’s autonomy in the marital household immediately after marriage; however, the association is not significantly affected by the passage of years since marriage.

No measure of gender ideology is included in the main estimation because of endogeneity concerns. One might argue that inclusion of such measures may undermine the association between dowry and women’s status in the marital household. Although we cannot deny the possibility that preferences and ideological beliefs concerning gender relations are not intrinsic and change over the marital life, the estimation procedure is repeated by including two measures for son preference: the first variable equals one when the wife’s ideal number of sons is greater than that of daughters and the second equals one when the wife would like to provide higher education to sons rather than to daughters. Controlling for gender ideology does not substantially affect the main estimation results (Appendix Table 9, Panel D). Interestingly, women’s decision making is positively associated with son preference. However, it is negatively associated with the preference for providing higher education to sons as compared with daughters. A possible interpretation is that highly educated women may be against gender inequality in education and be more empowered in the marital household but prefer having more sons than daughters. This seems consistent with some existing studies indicating that son preference is more prevalent among highly educated women rather than uneducated women in South Asia (Das Gupta 1987; Muhuri and Preston 1991).