Paying to tie the knot: Does education make marriage more expensive? Evidence from Jordan Open Access

The relationship between education and marriage behaviour has been widely explored in studies examining life-course economic outcomes (Krafft and Assaad, 2020; Gebel and Heyne, 2016; Wang and Ou, 2024; Mysíková, 2015; Juárez and Gayet, 2014; Nguyen et al., 2025). Within this literature, the “marriage market” framework remains central, depicting marriage as a process of search and matching under incomplete information about partner quality (Adachi, 2003; Becker, 1973, 1974; Grossbard-Shechtman, 1995). Education, beyond its role in signalling future earning potential, may also influence marriage timing and expectations by raising individuals' perceived value in the market and their ability to finance increasingly expensive marriage ceremonies and dowries. In this context, rising educational attainment may contribute to escalating marriage costs, particularly in settings like Jordan, where marriage often involves substantial financial outlays.

This paper examines the relationship between educational attainment and the increasing financial demands associated with marriage in Jordan. Within the traditional male breadwinner–female homemaker model, societal norms dictate that men must demonstrate financial readiness prior to marriage (Hoodfar, 2023). This expectation, coupled with the escalating costs of engagement ceremonies, wedding expenditures, and dowry payments, has imposed a significant financial burden on prospective grooms and their families (Chen and Pan, 2023). Concurrently, the decline in access to stable, well-compensated employment has compromised young men's capacity to fulfil these societal expectations (Assaad et al., 2010; Monga and Lin, 2015a, b; Salehi-Isfahani and Egel, 2010; Salem, 2016). These challenges have contributed to a phenomenon of delayed marriage, or “waithood,” characterised by extended transitions to adulthood, economic uncertainty, and social discontent (Dhillon and Yousef, 2009; Singerman, 2007; Kuhn, 2012).

In Jordan, the age of first marriage has been steadily rising for both men and women, with the trend particularly pronounced among women with higher levels of education (Leesch and Skopek, 2023). This shift represents a significant demographic and cultural transformation in a society where early and often arranged marriages have traditionally been the norm (Tahir, 2021). Women pursuing tertiary education tend to delay marriage until after completing their studies, typically between the ages of 23 and 30, a period during which concerns about declining fertility and reduced chances of having larger families become increasingly salient (Bharati et al., 2023; Howlader et al., 2023). In some cases, this delay is linked to the social concept of “spinsterhood”, a term used to describe women who remain unmarried beyond a socially constructed ideal age (Hamamra and Uebel, 2025; Sadigov, 2020). According to the Department of Statistics (2016), the number of never-married women in Jordan rose significantly from 40,790 in 2000 to 67,743 in 2015. One contributing factor to this delay is the rising cost of marriage. Expenditures related to engagement ceremonies, wedding celebrations, and dowries have increased markedly, especially in urban areas and among middle-income households. These growing financial demands can deter families from proceeding with marriage arrangements promptly.

In Western countries, delaying marriage is often seen as a choice linked to freedom and self-discovery (Arnett, 2000). However, in Jordan, it signifies a challenging and often involuntary period. Marriage is a crucial milestone in Jordanian society, marking the transition to adulthood and serving as a prerequisite for legitimate sexual relationships and reproduction, due to legal prohibitions on premarital relationships (Egel and Salehi-Isfahani, 2010; Rashad et al., 2005). Most young people only leave their parental home after marriage, which prolongs economic dependency and family control (DeJong et al., 2005; Singerman, 2007). Additionally, while the west increasingly separates first marriage from parenthood, in Jordan, childbirth typically follows marriage closely, often within a year (Gebel and Heyne, 2014).

Given the central role of marriage in Jordanian society, alongside the growing challenges young people face in navigating an extended transition to adulthood, this study addresses a critical research question: What factors influence individual-level marriage costs? While much of the existing literature has explored the timing of key life transitions, such as first marriage and first childbirth, emphasising the influence of educational attainment and labour market outcomes (Zhang and Liang, 2023; Krafft and Assaad, 2020; Gebel and Heyne, 2016). This study extends that line of inquiry by focusing specifically on how educational attainment shapes the financial dimensions of marriage. This question is particularly salient in Jordan, where young people often encounter prolonged school-to-work transitions, high youth unemployment, and low female labour force participation (Janta et al., 2015). Within this context, we examine how bride educational levels contribute to rising marriage costs paid by groom. By analysing the relationship between education and the economic burden of marriage, this study offers a nuanced understanding of how socioeconomic factors shape marital transitions in Jordan.

There remains limited empirical evidence on how educational attainment shapes marriage costs in the MENA region. This study seeks to fill that gap by examining three key research questions in the Jordanian context: (1) How does a bride's level of education influence overall marriage-related expenditures that paid by groom? (2) To what extent does education affect specific components of marriage costs, such as spending on engagement and wedding ceremonies and the gold value of dowry? (3) How does this relationship vary across key dimensions, including urban and rural residence, household wealth, age at marriage, and marriage cohort? In addition, we explore gender differences in this relationship, recognising that in Jordan's socio-cultural setting, education is often viewed as an essential signal of economic readiness for men. At the same time, similar expectations for women are less pronounced.

We base our empirical analysis on Jordanian data for two primary reasons. First, the Jordan Labor Market Panel Survey (JLMPS) offers high-quality, nationally representative panel data that tracks households in 2010 and 2016, capturing detailed information on marriage costs through internationally standardised survey instruments (Krafft and Assaad, 2018). Second, household structures in Jordan continue to be deeply influenced by enduring cultural and social norms. Our analysis centres on educational attainment, using years of schooling as the key variables of interest. We control for a range of bride and demographic characteristics, and results from the ordinary least squares (OLS) show a consistent positive relationship between education and marriage-related costs.

To identify causal effects, we leverage exogenous variation by using the bride's parental education as an instrument for the bride's educational attainment. This approach is grounded in the idea that a bride's parental education, influenced by broader societal and policy factors, shapes children's education, but is unlikely to affect the marriage costs directly, which are paid by the groom. The instrumental variable (IV) estimates support the OLS findings, revealing a robust and positive relationship. Moreover, Heckman's model, double machine learning and IV quantile regression results are consistent with the 2SLS estimates, further confirming the robustness of the observed relationship between bride's education and marriage cost paid by the groom.

The paper is organised as follows: Section 2 describes the data and sample selection; Section 3 outlines the identification strategy; Section 4 presents the results; and Section 5 concludes.

The study uses data from the 2010 and 2016 waves of the Jordan Labor Market Panel Survey (JLMPS), conducted by the Jordanian Department of Statistics (DOS) in collaboration with the Economic Research Forum (ERF). The DOS oversaw sampling and fieldwork activities. The JLMPS provides an opportunity to examine the impact of education on cost of marriage. The first wave was collected in early 2010, while the second wave took place between December 2016 and April 2017. After applying geographic sampling weights, the data are nationally representative (OAMDI, 2018a, b).

In the 2010 wave of JLMPS, a representative sample of 5,102 households and 25,953 individuals was surveyed. The sample covered both urban and rural areas across Jordan's three main regions—North, Middle, and South—and was stratified by 12 governorates. The second wave in 2016 built upon the original 2010 panel by recontacting households and adding both split households (new households formed by individuals from the 2010 sample) and a refresher sample. This resulted in a total of 7,229 households, including 3,058 from the original sample ^[1], 1,221 split households, and 2,950 refresher households, encompassing 33,450 individuals overall. The refresher sample introduced in 2016 was designed to oversample neighbourhoods identified in the 2015 population census as having high concentrations of non-Jordanian households. It was stratified by governorate, urban/rural/camp status, and the share of non-Jordanian residents (Krafft and Assaad, 2018). The 2016 sample weights were adjusted using the 2015 census to account for the original sampling design, the inclusion of the refresher sample, and attrition at both the household and individual levels ^[2].

The data include details of education as well as the cost and timing of marriages. We use the education data to estimate the impact on marriage costs, allowing for heterogeneity by place of residence (urban/rural), wealth, and marriage cohort (before or after 2010s). Governorates are the first level of administrative geography. Our sample consists of couples aged 15–40 at the time of first marriage in each survey year.

In Jordan, marriage costs are traditionally borne primarily by the groom and his family (Sieverding et al., 2019), though sometimes the bride's family contributes to specific expenses such as furniture or household items. The survey's reported marriage cost specifically refers to the amount paid by the groom, and our sample includes only cases where the groom covers these costs. The survey also asks about the shares of the bride's and groom's families in the house, furniture, and appliances related to the marriage. For our analysis, we exclude any observations where the bride or her family or the groom's family contribute at least 1%, focusing instead on cases where the groom alone bears 100% of these costs. This exclusion may overlook important socio-cultural factors about who bears marriage expenses. Moreover, it affects the validity of our IV approach: if the parents of either spouse pay the costs, then the bride's parental education, used as an instrument, becomes a direct explanatory factor rather than a valid instrument. Thus, identifying strong instruments requires careful consideration of cultural traditions and who bears the costs.

Our primary outcome of interest is the total cost of marriage, which captures the full financial burden of getting married, as reported by grooms in the JLMPS, and is measured in Jordanian Dinar. This includes a wide range of expenses, such as the dowry (mahr), wedding and engagement ceremony costs, gold or jewellery, clothing, gifts, and housing-related expenses, including furniture or rent. In addition, we consider several disaggregated components: (1) costs paid specifically for engagement and wedding ceremonies, (2) costs of the matrimonial home (land/apartment purchase, construction, and finishing), (3) the amount of dowry (gold only) registered in the marriage contract and (4) the amount of dowry (both in cash and gold).^[3] This detailed classification enables a more nuanced analysis of marriage costs, thereby enhancing the robustness of the study's empirical findings.

Our analysis of education outcomes builds on the works of Assaad et al. (2021) and Krafft and Assaad (2020). As the main covariate in the analysis, we use completed years of schooling, ranging from 0 to 25 years, as a continuous measure of education. In our analysis, we focus on the impact of a bride's education on the marriage costs reported by the groom. To effectively link spouses within the survey data, we use the household ID, along with data indicating the spouse's presence and sex of each individual. This approach allows us to accurately identify which individual is the husband and which is the wife. Once we have established the connections, we progressively match the wife's education and her parents' education level to the husband's records using the identifiers we have gathered. This method enables us to compile comprehensive observations that include details from both spouses. As a result, we can explore how the bride's educational background influences the financial responsibilities of the groom, particularly as reflected in the expenses he reports related to the marriage.

We include a comprehensive set of control variables to account for bride, familial, and contextual factors that may confound the relationship between education and the cost of marriage. bride-level controls include age and its square to capture potential non-linear life-cycle effects, along with categorical indicators for age at first marriage to reflect timing effects on marital expenditures. Employment status is included to account for the bride's economic standing at the time of marriage. Family structure is proxied by the number of sisters and brothers for bride, which may influence financial obligations and marriage expectations within the household, potentially leading to competition for resources. We also control for geography, distinguishing between urban and rural residences to capture spatial variation in social norms and market prices and include household wealth quintiles as a proxy for long-run economic resources.

Additionally, a binary indicator for consanguinity (i.e. marriage to a relative) is included to account for cultural practices that may significantly reduce dowry and marriage-related costs. These covariates enhance the robustness of our estimation by mitigating omitted variable bias and allowing for a more precise assessment of the effect of bride's education on marriage costs paid by groom, with fixed effects for governorates to account for regional differences in infrastructure, culture, and economic opportunities.

The following equation describes the relationship between education and marriage cost:

where $Mci$ is a variable capturing the marriage cost and i identifies an individual. $Edui$ is vector representing the education level; $Xi$ is a row vector of control variables that also affect marriage cost; $β0,β1andβ2$ are column vectors of parameters to be estimated; We include province fixed effects by adding dummy variables for each province (with one omitted category to avoid perfect multicollinearity), these province dummies are included in $Xi$⁠; together with a binary indicator for the year 2016; $εi$ is the error term.

The relationship likely suffers from endogeneity issues, as cultural norms can influence both education and marriage cost. Reverse causality occurs when anticipated marriage costs (or social expectations around marriage) may influence decisions about education. For example, in some cases, women may drop out of school early because their families expect them to get married at a young age (Attanasio and Kaufmann, 2017; Lloyd and Mensch, 2008). Omitted variable bias may arise from unobserved factors such as hidden cultural expectations which can influence the connection between women's education and marriage costs. In some communities, people link higher education to a higher social standing or modern values. This connection might create pressure for more expensive marriage ceremonies, larger dowries, or more impressive displays. These cultural norms, which are often not captured in household surveys, can increase both educational levels and marriage costs. As a result, there may seem to be a positive link between education and marriage costs, even if education itself does not directly cause those higher costs (Bühler et al., 2024; Gore and Carlson, 2010; Buttenheim and Nobles, 2009; Singerman, 2007). We estimate a two-stage least squares (2SLS) specification to address this endogeneity. The first stage regression is given as:

where $Xi$ is the same row vector of covariates as in equation (1) $IVi$ is the set of instrumental variables; $α0$⁠, $α1$ and $α2$ are column vectors of parameters to be estimated; $νt′$ is the error term. The second-stage regression is now given as:

where, $Eduiˆ$ is the predicted education, estimated from equation (2), $β1$ the main parameter of interest; $β0$⁠, $β1$ and $β2$ are column vectors of parameters to be estimated; We include province fixed effects by adding dummy variables for each province.

These province dummies, along with a binary indicator for the year 2016, are included in $Xi$⁠; $εi′$ is the error term. For the validity of the instrumental variable, we require two assumptions, first $α1≠0$⁠, second $Ε(IViεi′)$ = 0. All statistical analyses conducted using Stata 18. Marriage costs correspond to the actual year of each marriage. To account for changes in price levels over time, all costs were adjusted for inflation using Jordan's Consumer Price Index (CPI), with 2010 as the base year.

Beyond the simultaneity issue we look to address using an instrumental variable (IV) approach, other potential sources of endogeneity must also be considered. A key concern in this context is the non-random selection into high marriage cost relationships and omitted variable bias. To mitigate these risks, we incorporate an extensive set of control variables that account for bride, household, and societal characteristics. However, we acknowledge that unobserved factors may still influence the results.

The IV model is implemented with robust standard errors to address potential autocorrelation at the unit level, ensuring more consistent estimates across various forms of correlation (Cameron and Trivedi, 2010). Additionally, incorporating governorate-fixed effects controls for income disparities and unobserved cultural differences across provinces. Year-fixed effects are also included to account for time-varying factors that affect all governorates (Weidner and Zylkin, 2021). This methodology helps isolate the effects of the variables of interest and minimises bias from unobserved governorate differences and temporal trends (Millimet and Bellemare, 2023).

Our analysis addresses potential sample selection bias arising from the non-random availability of the outcome variable (e.g. the cost of marriage observed only for married individuals). Ignoring this bias could lead to inconsistent and biased parameter estimates. To correct for this, we apply Heckman's two-step selection estimator (Heckman, 1979), which explicitly models the marriage decision (selection equation) jointly with the cost of marriage (outcome equation), thereby correcting for possible correlation between unobservable factors affecting both the likelihood of being married and the cost itself. The two-step estimator uses instrumental variables that influence the probability of marriage but not the cost directly, helping to identify the model and correct for selection bias.

Furthermore, to enhance robustness against model misspecification and mitigate the effects of high-dimensional confounding, we implement a double machine learning (DML) estimator for partially linear model and linear IV model (Ahrens et al., 2024a). This approach integrates machine learning algorithms, such as Lasso regression, for the estimation of nuisance parameters while employing cross-fitting techniques to eliminate regularisation bias, thereby facilitating valid inference for treatment effects. The DML framework enables flexible control of confounders while specifically targeting the causal parameter of interest, which significantly strengthens the credibility of our estimates (Ahrens et al., 2024b).

Finally, to explore heterogeneous effects across the outcome distribution and address endogeneity concerns, we implement IV quantile regression. This method extends traditional IV approaches by estimating the impact of endogenous regressors at different quantiles of the dependent variable, thereby uncovering how effects may vary across the conditional distribution rather than only at the mean (Amemiya, 1982; Chernozhukov and Hansen, 2006; Kaplan and Sun, 2017).

Our analysis uses the bride's parental education levels as instrumental variables for the bride's education. This choice is motivated by economic literature, which shows that in developing societies, such as Jordan, parents' education can have a significant influence on their children's education (Holmlund et al., 2011). In such contexts, education is widely perceived as a key pathway out of poverty and toward a more stable and dignified life (Edeji, 2024). Families with more educated parents are more likely to invest in their children's schooling, making parental education a strong predictor of children's educational attainment. Furthermore, since parental education primarily affects children's outcomes through intergenerational knowledge transmission and resource allocation, rather than directly influencing later-life outcomes like marriage costs, it plausibly satisfies the exclusion restriction (Black et al., 2005; Oreopoulos et al., 2006). This instrumental variable approach builds on established findings that parental education influences children's human capital accumulation without directly altering cultural preferences or marriage market conditions.

Parental education serves as a valuable and credible instrument for estimating the causal impact of education on marriage-related outcomes. In socio-economic contexts, where education is highly valued as a means of social and economic mobility, parents with higher education levels are more likely to invest in their children's schooling (Erola et al., 2016). However, bride's parental education itself is unlikely to directly affect the marriage costs or timing of their children once other family and community characteristics are controlled for (Milovanska-Farrington, 2022; Jennings et al., 2012). This makes it a strong, exogenous predictor that can isolate the variation in education needed to identify its causal effect on outcomes such as age at marriage or marriage expenses.

One potential concern regarding our instrumental variable, particularly parental education, is the possibility of correlation with the error term if children's educational attainment is systematically linked to unobserved household characteristics that independently influence marriage costs. The exclusion restriction could be violated if, for example, more educated parents also transmit cultural values or social networks that directly affect marriage market outcomes beyond their impact through children's education. To address this concern, we incorporate comprehensive control variables, including the father's occupation ^[4], household wealth, and governorate fixed effects, which help account for potential confounding factors that might correlate with both parental education and marriage costs. This approach reduces the influence of unobserved heterogeneity and strengthens the validity of our IV assumptions. Furthermore, we rely on established empirical evidence (Black et al., 2005; Oreopoulos et al., 2006) suggesting that parental education primarily affects children's socioeconomic outcomes through human capital accumulation rather than through direct cultural or financial channels unrelated to schooling. This supports our assumption that parental education satisfies the exclusion restriction.

Regarding years of education, approximately 71.6% of the brides in the sample had completed 12 or more years of education. Table 1 shows a significant positive correlation between our instrumental variable (bride's parental education) and the bride's education.

Table 2 presents the key components of marriage costs in logarithmic form. The average total marriage cost (logged) is 9.25. Ceremony-related expenses, including engagement and wedding costs, average a log cost of 7.15. The cost of establishing the matrimonial home, which includes land and construction with a mean logged cost of 8.05. The logged average for brideprice is 8.26, highlighting its importance in marriage, while the average for dowry jewellery is 7.18. Overall, housing and dowry expenses are the main cost drivers in marriages. brides in the sample had an average education of 9.85 years.

Educational attainment among mothers, 61% were illiterate, and 24% could only read and write, indicating limited educational access for women in earlier generations. In contrast, 37% of fathers were illiterate, and another 38% could read and write. Both parents had similar rates of basic education (7–9%), but fathers were more likely to have completed secondary (8 vs. 5%), post-secondary (3 vs. 2%), or university education (5 vs. 1%). Regarding employment, 73% of respondents were employed, while 27% were unemployed or outside the labour force. On average, brides had 3.9 sisters and 3.9 brothers. Marriage occurred between the ages of 15–19 for 6% of respondents, 20–24 for 36%, and 25–29 for 39%. Additionally, 19% of respondents married after the age of 30. The average age of respondents was 44. About 26% lived in rural areas. In terms of wealth, 18% were in the poorest quintile and 18% in the richest. Consanguineous marriage was reported by 29% of participants. Fathers' occupations were nearly evenly divided between skilled 43% and unskilled 57% roles, showing a range of socioeconomic backgrounds ^[5].

Table 3 presents the OLS estimates of equation (1). The results show that higher education for females is positively associated with a marriage cost. Specifically, a one-year increase in bride education is associated with a 5% increase in total marriage costs paid by her groom. Hence one standard deviation increases in education (4.85 years) is associated with a 0.24 increase in the marriage cost, equivalent to roughly a 25% higher real marriage cost compared to the average. This association holds even after controlling for household characteristics, parental background, and geographic variables.

Table 4 shows the instrumental variable estimates, which confirm the OLS estimates. Specifications 1, 3 and 5 presents the first-stage results of equation (2), illustrating how the bride's parental education level in each household has a positive effect on the increased bride education level. The instrument is a powerful predictor of years of education, as reflected by the F-statistic for all specifications.

The second-stage IV results, presented in Specifications 2, 4, and 6, show a positive and statistically significant effect of bride education on marriage costs. Specifically, a one-year increase in bride education is associated with a 11.5, 11.8, and 11.5% point increase in marriage cost paid by her groom, respectively. These IV estimates align with the findings from the OLS model but are notably larger in magnitude, suggesting a stronger causal effect among compliers. The results remain robust after controlling for a range of covariates. The validity of the IV approach hinges on the assumption that bride's parental education influences the educational attainment of their bride, who are now of marriageable age. This is particularly relevant in the context of Jordan, where education is highly valued as a path to social mobility and improved life prospects. Families with more educated parents tend to invest more in their children's education. Conditional on the covariates included in the regression, bride's parental education is assumed to affect marriage costs paid by groom only through its effect on the bride's education level. One potential concern with this strategy is that parental education may also reflect broader aspects of socioeconomic background, which could independently influence marriage-related expenses.

To mitigate concerns about potential confounding factors, we include a comprehensive set of control variables reflecting parental background, household characteristics, and geographic variation. First, to capture disparities in economic development across Jordan's 12 governorates, we control for urban versus rural residence and household wealth. Second, we adjust for bride and household demographics, including the number of sisters and brothers, the respondent's age and age at first marriage, and whether the husband is a blood relative. Third, we include the father's occupation as a proxy for household-level human capital and socioeconomic status. Lastly, all models account for unobserved time-invariant regional factors and temporal shocks by including governorate and survey-year fixed effects.

Overall, these findings strongly support a positive and statistically significant relationship between bride years of education and marriage cost covered by groom.

Table 5 presents IV estimates examining the impact of education on various components of marriage costs. The results show that increased years of schooling are significantly associated with higher costs for ceremonies, brideprice and jewellery. Specifically, a one-year increase in bride education raises costs borne by the groom by 9.4%, 9.2 and 14.3%, respectively. In contrast, the effects of education on matrimonial home costs are not statistically significant. Bride's parental education remains a strong predictor across all outcomes: both father's and mother's bride education levels are positively and significantly associated with their bride education in each category of marriage cost. The F-statistics for the first stage are well above conventional thresholds, indicating strong instruments, and the overidentification tests suggest the instruments are valid across most specifications. These findings suggest that while education increases some dimensions of marriage expenses, particularly visible and symbolic ones, such as ceremonies, brideprice and jewellery, but it does not uniformly affect all components of marriage costs.

This section presents a series of robustness checks to assess the stability and validity of the main results. We begin by examining heterogeneity in the effect of education across key subgroups. This is followed by estimation using an alternative econometric specification, the Heckman selection model of a two-step estimator, and the double machine learning estimator of a partially linear model and a partially linear IV model, and lastly the IV-quantile regression.

First, our analysis reveals significant heterogeneity in the effect of bride education on marriage costs by place of residence. The results indicate that education has a statistically significant and positive effect on marriage costs among urban residents, whereas this effect is significant at 10% in rural areas. These findings suggest that urban settings may act as a potential mechanism through which education increases marriage costs, possibly due to stronger signalling effects in more competitive and formal marriage and labour markets, which are typically found in urban areas.

Next, we further explore how the effect of education on marriage costs varies across higher and lower wealth quintiles. As shown in Table 6, the positive association between bride's education and marriage cost is statistically significant for family in the lower and higher wealth groups, with the largest effect observed among the high-wealth group. These findings suggest that the number of years of schooling a bride has plays a more prominent role in shaping marriage costs among higher and lower wealth households.

We also examine whether the timing of marriage affects the relationship between the bride's education and marriage costs. Panel B of Table 6 shows the results by age at marriage. The estimates reveal a statistically significant effect of education on marriage costs for couples marrying before and after age 25. Additionally, we analyse differences by marriage cohort and find that the effect of education on marriage costs is strongest and significant for those married in the 2010s compared to earlier cohorts. These findings suggest that marriage expenses have become more important over time, likely driven by changing social norms and increased competition in marriage markets. The growing influence of education may reflect its rising role as a symbol of economic security and prestige in marital arrangements.

The results from the two-step Heckman selection model in Table 7 indicate that selection bias is not statistically significant in marriage, suggesting that unobserved factors influencing the decision to marry are not correlated with unobserved factors affecting marriage costs. This supports the validity of the 2SLS estimates obtained from the sample of married women, indicating they provide consistent estimates of the causal impact of education on marriage costs. The findings highlight the strong positive influence of education on marriage expenses, with significant coefficients observed for the bride's education as well as her father's and mother's education levels (Dolgikh and Potanin, 2024, 2025). Our sample of households where the groom bears all marriage expenses is not randomly selected. To correct for this, we employ a two-step Heckman selection model: the first stage estimates the probability of being married, and the second stage estimates the probability that the groom bears all expenses (see Table A1 in the  Appendix).

Table 8 reports the results of the Double Machine Learning estimation of the partially linear model and the partially linear IV model (Ahrens et al., 2024a, b). The outcome variable is the cost of marriage, and the key regressor is the wife's years of schooling. Estimation uses 5-fold cross-fitting with 3 resamples, applying cross-validated Lasso for nuisance function estimation. To improve the reliability of our causal estimates and address possible model errors and complex confounding. In this study, we examine the effect of the bride's education on the marriage costs paid by the groom. The DML results show a significant positive effect, with coefficients of 0.054 in the partially linear model and 0.129 in the IV model. These findings are consistent with the OLS estimate (0.051) and the 2SLS estimate (0.120), both showing a positive impact of education on marriage expenses. However, unlike OLS, which omitted variables may bias, the DML approach flexibly controls for many confounders and accounts for endogeneity using instruments. This makes our results more reliable and provides a clearer understanding of how a bride's education influences marriage costs. Compared to OLS and 2SLS, DML better handles model errors and confounding, making it suitable for complex socioeconomic data like ours.

Finally, we explore potential distributional heterogeneity in the impact of education on marriage costs, we estimate an Instrumental Variable Quantile Regression consistent with the 2SLS. This method, following the approaches outlined in Amemiya (1982), Chernozhukov and Hansen (2006), and Kaplan and Sun (2017), allows us to assess how the effect of education varies across different points of the marriage cost distribution. Table 9 presents estimate at the 25th, 50th, and 75th quantiles. The results suggest that the effect of education is not uniform across the distribution. Specifically, education has a positive effect on total marriage costs at the three percentiles. For individual components, such as ceremony cost, jewellery and brideprice, the estimates are consistently positive and significant, especially at higher quantiles, suggesting that bride education plays a stronger role in driving up ceremonial and symbolic marriage expenditures in wealthier or higher-cost matches.

This paper examines the impact of educational attainment on marriage cost outcomes in Jordan, using individual-level data from the Jordan Labor Market Panel Survey (JLMPS). We hypothesise that bride's parental education influences bride educational outcomes, which may reinforce traditional gender roles and, in turn, affect marriage-related expenditures. To identify causal effects and address potential endogeneity, we employ several econometric techniques, including OLS, IV estimation, Heckman's model, double machine learning, and IV quantile regression. Building on earlier studies, such as Farahzadi and Rahmati (2020) on female labour participation in Iran, Wu and Wu (2015) on intergenerational influences and gender in self-employment, Kossova et al. (2020) on marriage and male wages in Russia, Kossova and Potanin (2022), and Barg and Beblo (2009) on returns to marriage versus cohabitation, our study offers a methodological contribution by exploiting microdata from Jordan, our approach provides a more comprehensive examination of household-level determinants of marriage costs and the selection into marriage and payment responsibilities. This allows us to incorporate governorates fixed effects, thereby controlling for unobserved heterogeneity and enhancing the robustness of our findings on how education influences marriage costs in the Jordanian context.

Our findings show that bride higher education levels in Jordanian families are associated with increased marriage-related expenses, including greater spending on engagement and wedding ceremonies, as well as higher dowry or mahr values, often paid in gold from groom side. These effects are observed even after controlling for employment status, household composition, age at first marriage, urban residency, wealth status, consanguinity, and father's occupation. These results support recent research documenting the complex relationship between education and marriage outcomes. For instance, Ahn and Winters (2025) find that higher education reduces the likelihood of marriage among younger individuals and increases the probability of never marrying, even at older ages, highlighting the evolving role of education in marriage markets. Conversely, while Wang and Ou (2024) find that compulsory education reforms in China did not significantly influence the likelihood of marriage, they do report that higher female education levels increased husbands' income and education, particularly in more educated provinces. Our findings complement this by showing that higher education levels are associated with increased marriage costs in Jordan, highlighting how education influences not only partner selection but also the financial aspects of marriage.

Building on this, future research could examine how household structure and family composition affect parental investments in education and their willingness or capacity to incur higher marriage expenses. In particular, the shift from a focus on child quantity to child quality may depend not only on the opportunity cost of women's time but also on caregiving norms, economic dependency, and multi-generational household dynamics. Understanding these relationships can shed light on broader demographic and economic behaviours tied to marriage.

Ultimately, this study highlights the significance of incorporating household and educational factors into policies designed to mitigate inequality in marriage markets and foster inclusive development. By linking education to rising marriage expenditures, our findings suggest that public policy, particularly in areas such as access to education, gender equity, and family support, should consider how rising educational attainment reshapes traditional expectations and financial pressures associated with marriage, especially in culturally conservative and economically diverse societies like Jordan.

This study has several limitations. First, the analysis relies on household survey data from 2010 and 2016; therefore, marriage cost information beyond 2016 is not available, which prevents us from capturing more recent trends. Second, despite controlling for many household and individual characteristics, unobserved factors, such as cultural norms or informal family agreements, may still affect marriage expenditures. Finally, the findings are specific to the Jordanian context and may not generalise to countries with different social, cultural, or economic norms.

We would like to express our gratitude to Professor Mohsen, the editor of the Journal of Economic Studies, as well as the two anonymous reviewers for their insightful and constructive feedback. Their comments significantly enhanced the paper and bolstered its contribution to the field.