Nutritious Partner: Association Between Educational Assortative Marriage and Healthy Eating in Chinese Older Adults

Abstract

Introduction:

Promoting healthy dietary patterns among older adults is a crucial strategy for mitigating health equity disparities. As a key social determinant, educational assortative marriage (EAM) profoundly shapes health behaviors in later life via interpersonal influence and shared household dynamics. Despite its importance, the specific impact of EAM patterns on the diet quality of Chinese older adults remains significantly underexplored.

Methods:

We utilized cross-sectional data from the 2017 to 2018 wave of the Chinese Longitudinal Healthy Longevity Survey (CLHLS). A Healthy Eating Index (HEI) was constructed based on self-reported intake of 13 food groups, which were further categorized into plant-based and animal-based components. EAM was classified into homogamy, hypogamy, and hypergamy. Ordinary least squares (OLS) regression models were estimated, and the entropy balancing method was applied to adjust for covariate imbalance.

Results:

Compared to homogamy marriages, hypogamy marriages were significantly associated with lower HEI scores (adjusted and balanced β = −1.44, p < 0.01). Conversely, hypergamy marriages were significantly associated with higher HEI scores (adjusted and balanced β = 1.38, p < 0.01). In extended analysis, the association between EAM patterns and HEI persists predominantly among couples without higher education. Gender heterogeneity analysis revealed that the association between EAM and HEI was more pronounced among men than women.

Conclusion:

This study demonstrates that EAM is a significant social determinant of diet quality among Chinese older adults. Public health interventions promoting healthy aging should consider the educational dynamics within couples, targeting nutritionally vulnerable marital subgroups with tailored strategies.

Introduction

Against the backdrop of accelerating global population aging, improving health outcomes among older adults and promoting healthy lifestyles have emerged as critical public health priorities to facilitate successful aging. A balanced and nutritious diet serves as a fundamental determinant of health and well-being in later life stages by ensuring adequate intake of essential macronutrients. The promotion of healthy dietary patterns among older adults carries significant implications for addressing health equity disparities.¹

Educational assortative marriage (EAM), where individuals marry partners with similar educational backgrounds, have significant implications for health outcomes among older adults. In China, where men have long dominated higher education, women traditionally tend to marry men with equal or higher education levels (homogamy or hypergamy, respectively), while men typically marry less-educated women (hypogamy).² This pattern is known as “mén dāng hù duì” (well-matched marriages) in traditional Chinese culture. Current research has primarily revealed how EAM affect mortality, cognitive health, cardiovascular outcomes, as well as overall well-being and health behaviors in older adults.^3–6 In terms of health-related behaviors, studies in China found that individuals with highly educated spouses show greater compliance with dietary guideline.⁷ Partner’s educational attainment influences one’s own health by shaping various health behaviors.⁸ Similarly, studies in the United States confirmed that the husband’s education is associated with healthy eating and other health behaviors.^9,10 Evidence in the Netherlands also suggested that having a less-educated partner increases the likelihood of engaging in health-risk behaviors such as smoking and alcohol consumption.¹¹ Partners’ educational attainment appeared more critical for preventing unhealthy behaviors than for promoting positive health behaviors.¹² A study in Japan revealed that a couple’s combined educational attainment is crucial for the success of smoking cessation interventions.¹³ In general, the current body of literature on EAM and health behaviors has concentrated on younger populations, with less attention paid to older adults. Given that older adults are a key health-vulnerable group, promoting and maintaining their dietary health is essential for the broader goal of realizing health equity. The specific effects of EAM patterns on healthy eating among older adults behaviors warrant further investigation.

Investigating relationships between EAM and diet holds significant implications for developing targeted nutritional education programs and public health interventions for vulnerable marital groups, ultimately enhancing the health and well-being of older adults’ families and promoting healthy aging. We utilized data from the Chinese Longitudinal Healthy Longevity Survey (CLHLS) 2017–2018 survey. We first classified older adults into three EAM patterns based on the educational levels of both respondents and their spouses as recorded in CLHLS. Subsequently, we constructed a Healthy Eating Index (HEI) using dietary data from CLHLS and employed ordinary least squares (OLS) regression models to examine the association between EAM patterns and HEI scores. To address potential confounding factors and strengthen causal inference, we applied entropy balancing methods to balance the covariates. In additional analyses, we further investigated the differential effects of nine distinct EAM matching patterns on healthy eating among older adults.

Materials and Methods

Data

We utilized data from 2017 to 2018 wave of the CLHLS, a nationally representative longitudinal study on aging conducted by the Peking University Center for Healthy Aging and Development. Covering 23 provinces in China, CLHLS is the earliest and longest-running social science survey in the country, with its most recent wave completed in 2018.

To ensure data reliability, rigorous preprocessing and cleaning procedures were implemented. We focused on the older adults aged 60 and above. From the original dataset (N = 15,874), to most accurately examine the effect of EAM on healthy eating,¹⁴ we restricted our analytical sample to participants who were currently married and living with their spouse (N = 6135). Consequently, individuals who were separated (N = 276), divorced (N = 52), widowed (N = 9004), never married (N = 140), or identified as other outliers (N = 407) were excluded from the analysis. After excluding cases, age under 60 or with missing values in HEI, EAM, and other covariates (N = 2066), the final analytical sample comprised 4069 eligible participants. The oldest participant in the sample was 109 years old. Supplementary Table SA1 also reported the specific coresidence pattern of the older adults.

Ethical statement

The CLHLS study we used was approved by the Research Ethics Committee of Peking University (IRB00001052-13074), and all participants or their proxy respondents provided written informed consent.

Healthy Eating Index

During the follow-up of the CLHLS, participants reported their current frequency of intake for 13 specific foods: fruits, vegetables, meat, fish, eggs, bean-based foods, salt-preserved vegetables, sugar, tea, garlic, nuts, mushrooms or algae, and dairy products. The food frequency questionnaires provided four or five response options for each item.

Healthy eating was assessed using the HEI, which included the same 13 food components and had a maximum score of 50. Higher HEI scores reflect healthier eating patterns. Based on prior studies examining the health relationships of these food groups,^15–17 pickled vegetables and sugar were reverse-scored for each participant. The specific scoring values assigned to each food group are detailed in Table 1. For analytical purposes, we further stratified the HEI into two distinct components: An Animal-Based Healthy Eating Index (ABHEI) encompassing meat, fish, eggs, and dairy products (16 points total),¹⁸ and a Plant-Based Healthy Eating Index (PBHEI) comprising all remaining (mushroom and algae were also included, though they do not fall under plant from a strictly theoretical standpoint) dietary components (34 points total).¹⁹

Table 1.

Score of Healthy Eating Index

Food	Score
Fruit	“Almost every day” = 3 “Quite often” = 2 “Occasionally” = 1 “Rarely or never” = 0
Vegetable
Meat	“Almost every day” = 4 “At least once a week” = 3 “At least once a month” = 2 “Occasionally” = 1 “Rarely or never” = 0
Fish
Eggs
Food made from beans
Tea
Garlic
Nut
Mushroom or algae
Milk products
Salt-preserved vegetables	“Almost every day” = 0 “At least once a week” = 1 “At least once a month” = 2 “Occasionally” = 3 “Rarely or never” = 4
Sugar

Educational assortative marriage

EAM was determined based on the educational disparity between respondents and their spouses. According to China’s national education system, we categorized older adults’ years of schooling into four levels: illiterate (0 years), primary school (1–6 years), high school (7–12 years), and college or above (≥13 years). EAM patterns were classified into: homogamy (spouses with equal education levels), hypogamy (respondent more educated than spouse), and hypergamy (spouse more educated than respondent). China’s education system operates as a multilevel screening process governed by highly standardized examinations.²⁰ As such, progression to a higher level of education strongly reflects the quality of an individual’s schooling. For this reason, educational attainment (measured by years of schooling) is considered a primary proxy for education quality.²¹

Covariates

The selection of our covariates was guided by a conceptual framework acknowledging that the determinants of healthy eating operate at both the individual and collective levels.²² At the demographic level, we adjusted for age (continuous), sex (male/female), and race (Han/non-Han). Socioeconomic controls included individual educational attainment (categorized as illiterate, primary school, high school, and college or above) and income (quartiles). Household characteristics comprised residential type (urban/rural) and whether having any children (yes/no), housework participation (more than once a week/others), and number of coresidents (more than one/others). Health-related factors incorporated self-rated health status (good or so good/others). We also included older adults’ social activities participation (more than once a week/others). Geographic controls accounted for regional variations (East, Central, Northeast, and West China).

Statistic methods

We employed OLS regression models to examine the association between EAM and HEI. To enhance the national representativeness of the sample, we also applied survey weights provided by the CLHLS. These weights were adjusted based on China’s 2015 national 1% population sample survey (also known as the “mini-census”).²³ Furthermore, given that marriage involves socioeconomic self-selection,²⁴ the observed relationship between EAM and HEI may be subject to confounding. To address potential confounding factors, we implemented entropy balancing method that adjusts control group weights to match the covariate distribution of the treatment group. This technique directly incorporates balance constraints into the weighting function while preserving the full sample. Empirical evidence suggests that entropy balancing outperforms alternative methods in reducing bias and minimizing mean squared error.²⁵ We defined the hypergamous group as the treatment group, with the hypogamous and homogamous groups serving as the control. The justification for this approach lies in the dynamics of the Chinese marriage market, where hypergamy is often a product of self-selection. Individuals in this group, possessing what is contextually termed “higher bargaining power,” may leverage other assets (youth, higher earnings, or a greater contribution to housework) to secure an educationally upward marriage.²⁶ The analysis was conducted using the EBALANCE package in Stata,²⁷ with detailed balancing results presented in Figure 1. The results of the procedure confirmed its effectiveness: after balancing, the standard mean difference for every covariate was below 0.1%, signifying that an excellent covariate balance was achieved.

FIG. 1.

Entropy balancing analysis result. While we operationalized years of schooling and income as ordered categorical variables in the covariate set, for the purpose of entropy balancing, we utilized their original continuous versions to enhance the precision of the results.

Result

Descriptive analysis

Table 2 presents the descriptive characteristics of the study sample stratified by EAM type—homogamy, hypogamy, and hypergamy. Significant differences were observed across EAM groups in terms of dietary outcomes, sociodemographic characteristics, and health status.

Table 2.

Descriptive Analysis

Variable	Homogamy (N = 1985)	Hypogamy (N = 1371)	Hypergamy (N = 713)	p ^a
HEI, mean (SD)	27.52 (7.52)	28.06 (7.12)	27.26 (7.39)	0.03
PBHEI, mean (SD)	17.51 (5.10)	18.06 (5.02)	17.53 (4.95)	<0.01
ABHEI, mean (SD)	10.00 (3.55)	10.00 (3.37)	9.72 (3.65)	0.19
Age, mean (SD)	77.01 (9.17)	77.27 (9.10)	75.78 (8.22)	<0.01
Education, n (%)				<0.01
Illiterate	731 (36.83)	0 (0.00)^b	419 (58.77)
Primary school	705 (35.52)	701 (51.13)	240 (33.66)
High school	438 (22.07)	538 (39.24)	54 (7.57)
College and above	111 (5.59)	132 (9.63)	0 (0.00)
Sex, n (%)				<0.01
Female	830 (41.81)	180 (13.13)	569 (79.80)
Male	1155 (58.19)	1191 (86.87)	144 (20.20)
Race, n (%)				0.89
Non-Han	114 (5.74)	84 (6.13)	43 (6.03)
Han	1871 (94.26)	1287 (93.87)	670 (93.97)
Income (quantile), n (%)				0.03
1	607 (30.58)	402 (29.32)	206 (28.89)
2	406 (20.45)	341 (24.87)	180 (25.25)
3	476 (23.98)	312 (22.76)	147 (20.62)
4	496 (24.99)	316 (23.05)	180 (25.25)
Residence, n (%)				0.11
Urban	1216 (61.26)	790 (57.62)	427 (59.89)
Rural	769 (38.74)	581 (42.38)	286 (40.11)
Having child(ren), n (%)				0.87
No	21 (1.06)	14 (1.02)	9 (1.26)
Yes	1964 (98.94)	1357 (98.98)	704 (98.74)
Self-reported health, n (%)				<0.01
Good or so good	847 (42.67)	671 (48.94)	320 (44.88)
Others	1138 (57.33)	700 (51.06)	393 (55.12)
Housework participation, n (%)				<0.01
More than once a week	1295 (65.24)	775 (56.53)	543 (76.16)
Others	690 (34.76)	596 (43.47)	170 (23.84)
Social activities, n (%)				0.48
More than once a week	168 (8.46)	102 (7.44)	53 (7.43)
Others	1817 (91.54)	1269 (92.56)	660 (92.57)
Number of coresidents, n (%)				0.70
More than one	830 (41.81)	589 (42.96)	309 (43.34)
Others	1155 (58.19)	782 (57.04)	404 (56.66)
Region, n (%)				<0.01
East	1043 (52.54)	687 (50.11)	356 (49.93)
Central	387 (19.50)	310 (22.61)	171 (23.98)
Northeast	109 (5.49)	44 (3.21)	32 (4.49)
West	446 (22.47)	330 (24.07)	154 (21.60)

The Kruskal–Wallis test was employed to assess differences in continuous variables, while Pearson’s chi-squared test was used for categorical variables to examine variations across sample groups.

For illiterate older adults, there is no possibility for them to have a hypogamy marriage, as illiterate is the lowest level of education. The same for college and above in hypergamy group.

ABHEI, Animal-Based Healthy Eating Index; HEI, Healthy Eating Index; PBHEI, Plant-Based Healthy Eating Index; SD, Standard deviation.

HEI varied significantly across EAM types (p = 0.03). Older adults in hypogamous marriages exhibited the highest mean total HEI score (28.06 ± 7.12), followed by those in homogamous (27.52 ± 7.52) and hypergamous marriages (27.26 ± 7.39). This pattern was driven by differences in plant-based HEI subscores (p < 0.01), while no significant difference was observed for animal-based components (p = 0.19). Additional differences were also found across age (p < 0.01), education (p < 0.01), sex (p < 0.01), income quantiles (p = 0.03), self-rated health (p < 0.01), housework participation(p < 0.01), and geographic region (p < 0.01).

OLS results

Table 3 presents the associations between EAM and HEI, PBHE, and ABHEI, estimated via OLS regression. Initially, in the unadjusted model, hypogamy was positively associated with HEI scores when compared with homogamy (β = 0.54, p < 0.05), while hypergamy was not (β = −0.26, p > 0.1). This association was reversed after controlling for covariates. Upon a stepwise regression check, the reversal occurred because the initial finding was confounded by the respondent’s own education level. Lower-educated individuals are more likely to be in hypergamous marriages and higher-educated individuals in hypogamous ones. The true effect of EAM became apparent only after this confounding factor was statistically controlled. Across all adjusted specifications, the results indicate a strong and statistically significant similar association between EAM and HEI. After adjusting for covariates and applying entropy balancing, hypogamy is significantly associated with poorer dietary outcomes compared with homogamy. For the full sample, hypogamy is linked to a 1.44 point decrease in the total HEI score (β = −1.44, p < 0.01). This deficit is observed across both PBHEI (β = −0.95, p < 0.01) and ABHEI (β = −0.49, p < 0.01). Conversely, hypergamy is consistently associated with higher HEI scores. Hypergamous unions are linked to a 1.38 point increase in total HEI (β = 1.38, p < 0.01). This positive association is evident for both PBHEI (β = 1.05, p < 0.01) and ABHEI (β = 0.34, p < 0.01).

Table 3.

Association Between Educational Assortative Marriage and Healthy Eating Index

	Unadjusted and unweighted coef. (robust 95% CI)			Adjusted and unweighted coef. (robust 95% CI)			Adjusted and census weighted coef. (robust 95% CI)			Adjusted and entropy balancing weighted coef. (robust 95% CI)
	HEI	PBHEI	ABHEI	HEI	PBHEI	ABHEI	HEI	PBHEI	ABHEI	HEI	PBHEI	ABHEI
EAM
Homogamy	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Hypogamy	0.54** (0.04–1.05)	0.55*** (0.20–0.89)	0.00 (−0.25 to 0.24)	−1.30*** (−1.80 to −0.81)	−0.65*** (−1.00 to −0.30)	−0.65*** (−0.90 to −0.41)	−1.23*** (−1.84 to −0.63)	−0.60*** (−1.03 to −0.16)	−0.64*** (−0.94 to −0.33)	−1.44*** (−2.28 to −0.60)	−0.95*** (−1.56 to −0.33)	−0.49** (−0.91 to −0.07)
Hypergamy	−0.26 (−0.89 to 0.37)	0.02 (−0.41 to 0.45)	−0.28* (−0.58 to 0.02)	1.48*** (0.91–2.05)	1.08*** (0.67–1.49)	0.40*** (0.11–0.68)	0.89** (0.18–1.59)	0.77*** (0.28–1.27)	0.11 (−0.25 to 0.47)	1.38*** (0.76–2.00)	1.03*** (0.61–1.46)	0.34** (0.02–0.67)
Education
Illiterate				Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Primary school				2.73*** (2.19–3.27)	1.64*** (1.26–2.03)	1.09*** (0.82–1.36)	2.35*** (1.64–3.05)	1.38*** (0.88–1.88)	0.97*** (0.61–1.32)	2.58*** (1.85–3.31)	1.75*** (1.25–2.26)	0.82*** (0.45–1.20)
High school				5.75*** (5.08–6.42)	3.65*** (3.16–4.13)	2.10*** (1.77–2.44)	5.49*** (4.64–6.35)	3.63*** (3.03–4.23)	1.86*** (1.42–2.31)	6.36*** (5.33–7.40)	4.10*** (3.36–4.85)	2.26*** (1.76–2.75)
College and above				7.28*** (6.26–8.30)	4.57*** (3.84–5.30)	2.71*** (2.20–3.22)	6.34*** (5.11–7.57)	3.96*** (3.05–4.88)	2.38*** (1.81–2.96)	7.15*** (5.58–8.71)	4.72*** (3.53–5.90)	2.43*** (1.73–3.13)
Sex
Female				Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Male				1.17*** (0.70–1.64)	0.84*** (0.50–1.17)	0.33*** (0.09–0.56)	1.10*** (0.53–1.67)	0.75*** (0.34–1.16)	0.35** (0.06–0.64)	1.40*** (0.73–2.08)	0.94*** (0.46–1.43)	0.46*** (0.12–0.80)
Race
Non-Han				Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Han				0.88** (0.05–1.72)	0.59* (−0.00 to 1.19)	0.29 (−0.13 to 0.71)	1.10** (0.14–2.05)	1.09*** (0.41–1.77)	0.01 (−0.49 to 0.51)	−0.31 (−1.39 to 0.78)	−0.24 (−1.04 to 0.57)	−0.07 (−0.72 to 0.58)
Age				−0.02* (−0.05 to 0.00)	−0.05*** (−0.07 to −0.03)	0.03*** (0.01–0.04)	−0.02 (−0.06 to 0.02)	−0.04*** (−0.07 to −0.02)	0.02*** (0.01–0.04)	−0.03 (−0.07 to 0.01)	−0.05*** (−0.08 to −0.03)	0.02** (0.00–0.04)
Income
Q1				Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Q2				1.20*** (0.66–1.74)	0.44** (0.06–0.83)	0.75*** (0.48–1.02)	1.01*** (0.33–1.70)	0.27 (−0.22 to 0.76)	0.75*** (0.39–1.10)	1.53*** (0.70–2.37)	0.68** (0.11–1.25)	0.85*** (0.41–1.29)
Q3				1.98*** (1.42–2.54)	0.82*** (0.42–1.22)	1.16*** (0.88–1.44)	1.71*** (1.00–2.42)	0.62** (0.11–1.13)	1.09*** (0.75–1.44)	2.27*** (1.41–3.14)	0.87*** (0.27–1.48)	1.40*** (0.96–1.84)
Q4				3.98*** (3.37–4.58)	1.97*** (1.54–2.40)	2.01*** (1.71–2.31)	3.61*** (2.82–4.40)	1.63*** (1.07–2.19)	1.98*** (1.59–2.37)	4.11*** (3.17–5.06)	1.83*** (1.19–2.47)	2.28*** (1.79–2.78)
Residence
Urban				Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Rural				−1.72*** (−2.15 to −1.29)	−1.11*** (−1.42 to −0.81)	−0.61*** (−0.82 to −0.39)	−1.70*** (−2.23 to −1.16)	−1.07*** (−1.45 to −0.69)	−0.62*** (−0.89 to −0.35)	−2.07*** (−2.71 to −1.42)	−1.34*** (−1.78 to −0.91)	−0.72*** (−1.06 to −0.38)
Have child(ren)
No				Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Yes				1.38 (−0.47 to 3.23)	1.07 (−0.25 to 2.40)	0.31 (−0.62 to 1.23)	1.28 (−2.21 to 4.77)	1.12 (−0.61 to 2.85)	0.16 (−1.95 to 2.27)	−0.12 (−2.36 to 2.12)	0.32 (−1.30 to 1.95)	−0.44 (−1.48 to 0.60)
Self-reported health
Others				Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Good or so good				1.20*** (0.81–1.59)	0.79*** (0.51–1.07)	0.41*** (0.21–0.60)	1.28*** (0.79–1.77)	0.87*** (0.52–1.22)	0.41*** (0.17–0.66)	1.18*** (0.56–1.79)	0.81*** (0.39–1.23)	0.36** (0.05–0.68)
Housework participation
Others			.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
More than once a week				0.44* (−0.00 to 0.88)	0.35** (0.03–0.66)	0.09 (−0.13 to 0.31)	0.64** (0.11–1.18)	0.37* (−0.02 to 0.76)	0.28* (−0.00 to 0.55)	−0.15 (−0.88 to 0.58)	−0.11 (−0.59 to 0.37)	−0.04 (−0.44 to 0.35)
Social activities
Others				Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
More than once a week				2.15*** (1.43–2.88)	1.54*** (1.03–2.06)	0.61*** (0.25–0.97)	2.44*** (1.62–3.26)	1.68*** (1.09–2.28)	0.75*** (0.38–1.13)	2.18*** (0.92–3.44)	1.50*** (0.60–2.40)	0.68** (0.11–1.24)
Number of coresident
Others			.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
More than one				−0.41** (−0.81 to −0.01)	−0.06 (−0.35 to 0.22)	−0.34*** (−0.54 to −0.14)	−0.48* (−0.98 to 0.03)	−0.04 (−0.40 to 0.33)	−0.44*** (−0.69 to −0.19)	−0.50 (−1.13 to 0.13)	−0.06 (−0.50 to 0.38)	−0.44*** (−0.77 to −0.11)
Region
East				Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Central				−1.05*** (−1.55 to −0.54)	0.05 (−0.32 to 0.41)	−1.09*** (−1.35 to −0.84)	−1.04*** (−1.68 to −0.39)	0.05 (−0.41 to 0.51)	−1.09*** (−1.41 to −0.77)	−1.04*** (−1.80 to −0.27)	0.07 (−0.46 to 0.59)	−1.10*** (−1.50 to −0.71)
Northeast				1.54*** (0.58–2.50)	1.20*** (0.52–1.89)	0.34 (−0.14 to 0.81)	1.11* (−0.03 to 2.25)	1.12*** (0.31–1.93)	−0.01 (−0.66 to 0.65)	1.93** (0.36–3.50)	1.12* (−0.08 to 2.31)	0.81** (0.07–1.56)
West				−1.21*** (−1.72 to −0.69)	−0.38** (−0.75 to −0.01)	−0.82*** (−1.08 to −0.57)	−1.11*** (−1.76 to −0.45)	−0.18 (−0.65 to 0.28)	−0.92*** (−1.25 to −0.59)	−1.20*** (−2.01 to −0.40)	−0.38 (−0.93 to 0.17)	−0.82*** (−1.24 to −0.40)
Constant	27.52*** (27.19–27.84)	17.51*** (17.29–17.73)	10.00*** (9.85–10.16)	22.26*** (19.30–25.22)	16.46*** (14.34–18.57)	5.80*** (4.33–7.28)	22.41*** (17.47–27.35)	15.98*** (13.05–18.91)	6.43*** (3.65–9.20)	23.18*** (18.68–27.68)	17.69*** (14.40–20.97)	5.49*** (3.39–7.59)

***p < 0.01, **p < 0.05, *p < 0.1.

ABHEI, Animal-Based Healthy Eating Index; EAM, Educational Assortative Marriage; CI, Confidence Interval; Coef., Coefficient; HEI, Healthy Eating Index; PBHEI, Plant-Based Healthy Eating Index; Ref., Reference.

Overall, a score difference of 2.82 points on the HEI emerged between the hypogamy and hypergamy groups. An examination of the covariate coefficients helps to contextualize the practical significance of EAM’s impact on the HEI. Older adults with a college and above education had an HEI score 7.15 points higher than those who were illiterate (p < 0.01). Men scored 1.40 points higher than women (p < 0.01). The highest income quartile group scored 4.11 points higher than the lowest income quartile group (p < 0.01), and rural residents scored 2.07 points lower than their urban counterparts (p < 0.01). Self-rated health, social participation, and region of residence also significantly influenced HEI scores. The magnitude of the EAM coefficient was similar to that of these covariates, indicating that EAM has a substantial impact on shaping HEI disparities among the older adult population.

Figure 2 presents the comprehensive matching analysis results. The association between EAM patterns and HEI, PBHEI, and ABHEI identified in the baseline regression persists predominantly among couples without higher education. Specifically, coefficients above the diagonal (indicating hypogamous marriages) were consistently lower, while those below the diagonal (indicating hypergamous marriages) were higher. This educational gradient showed a progressive enhancement with increasing educational attainment, with the most pronounced effects observed at the threshold of higher education completion.

FIG. 2.

Extended EAM analysis. Adjusted and entropy-balanced coefficients in grid. ABHEI, Animal-Based Healthy Eating Index; EAM, educational assortative marriage; HEI, Healthy Eating Index; PBHEI, Plant-Based Healthy Eating Index.

Gender heterogeneity analysis

Given Chinese women’s predominant role in household dietary decisions, EAM may demonstrate gender-specific associations with healthy eating among married older adults.²⁸ Upon analyzing the Gender–EAM interaction in Figure 3, men display significantly stronger marginal effects on all HEI across both hypergamy and hypogamy marriage types. Women, conversely, are less influenced by the EAM–HEI association and show particular resilience to the negative effects of hypogamy. This confirms that the link between EAM and HEI is characterized by significant gender heterogeneity.

FIG. 3.

Gender heterogeneity in association between EAM and HEI. Adjusted and entropy-balanced model with 95% confidence interval. ABHEI, Animal-Based Healthy Eating Index; Coef., coefficient; EAM, educational assortative marriage; HEI, Healthy Eating Index; PBHEI, Plant-Based Healthy Eating Index.

Sensitivity analysis

While legally distinct from being married in the Chinese context, widowhood does not erase the lasting influence of a marital partnership that may have spanned a significant portion of an individual’s life.²⁹ Concurrently, restricting analysis to intact marital dyads risks introducing a significant selection bias toward a healthier, more advantaged cohort, thereby limiting the generalizability of the findings. In light of this, we performed a sensitivity analysis to test the robustness of our findings to the inclusion of the widowed subsample. The CLHLS contains information on the educational attainment of the widowed participants’ deceased (last) spouse. We leveraged this information to define the EAM patterns for widowed participants. As demonstrated in Supplementary Table SA2, the widowed cohort differs systematically from the married cohort, particularly in age and other key characteristics. However, when these individuals were included in our regression models (Table 4), the primary associations between EAM and HEI remained robust and statistically significant.

Table 4.

Sensitivity Analysis (Including Widowed Older Adults)

	Unadjusted and unweighted coef. (robust 95% CI)			Adjusted and unweighted coef. (robust 95% CI)			Adjusted and census weighted coef. (robust 95% CI)			Adjusted and entropy balancing weighted coef. (robust 95% CI)
	HEI	PBHEI	ABHEI	HEI	PBHEI	ABHEI	HEI	PBHEI	ABHEI	HEI	PBHEI	ABHEI
EAM
Homogamy	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Hypogamy	1.87***(1.53–2.22)	1.41***(1.18–1.65)	0.46***(0.29–0.63)	−1.58***(−1.96 to −1.20)	−0.92***(−1.18 to −0.66)	−0.66***(−0.86 to −0.46)	−1.30***(−1.82 to −0.78)	−0.74***(−1.11 to −0.36)	−0.57***(−0.83 to −0.31)	−1.44***(−2.03 to −0.85)	−0.92***(−1.34 to −0.50)	−0.52***(−0.81 to −0.22)
Hypergamy	0.89***(0.48–1.30)	0.65***(0.37–0.93)	0.24**(0.04–0.44)	1.18***(0.82–1.55)	0.78***(0.53–1.03)	0.40***(0.21–0.59)	0.85***(0.28–1.41)	0.72***(0.32–1.11)	0.13(−0.16 to 0.42)	1.29***(0.90–1.68)	0.85***(0.58–1.11)	0.45***(0.24–0.65)

***p < 0.01, **p < 0.05, *p < 0.1.

ABHEI, Animal-Based Healthy Eating Index; Coef., Coefficient; CI, confidence interval; EAM, educational assortative marriage; HEI; Healthy Eating Index; PBHEI, Plant-Based Healthy Eating Index; Ref., reference.

Furthermore, coresidence patterns could be another significant confounding factor. The conclusions may be biased in cases where the couple’s diet is determined by other household members. Therefore, we also conducted a sensitive heterogeneity analysis based on the older adults’ coresidence type (whether they lived with other household members). Supplementary Figure SA1 illustrates the results. The association between EAM and HEI remains stable across different coresidence patterns.

Discussion

To our knowledge, this study is the first investigation examining the association between EAM and healthy eating among Chinese older adults. Our findings demonstrate that compared with elders in homogamous marriages, those in hypogamous unions exhibit significantly lower HEI scores, whereas individuals in hypergamous marriages show higher HEI scores. These results align with existing literature documenting the positive relationship between spousal education and health behaviors,¹¹ while extending previous findings in several important dimensions. Specifically, our cross-sectional data confirm and generalize the results from prior single-city studies in China that found associations between spousal educational advantage and healthier oil/salt intake patterns.⁷ Moreover, the current study expands the evidence base by examining a more comprehensive range of dietary indicators through the HEI framework, thereby providing a more nuanced understanding of how marital educational gradients influence nutritional outcomes in older populations.

The association between EAM and dietary quality may operate through multiple mechanisms. Marriage is a process of deep lifestyle assimilation.³⁰ With increasing duration of cohabitation, spouses’ dietary habits experience a process of socialization characterized by mutual influence, adjustment, and convergence. Higher spousal education typically correlates with greater health literacy and nutritional knowledge.³¹ They are more likely to be exposed to scientific information on diet and nutrition and to interpret it accurately. In turn, this knowledge can be subtly transmitted to the less-educated spouse through daily communication and joint decision-making.³² Older adults with lower education may benefit from their more educated partners’ health knowledge and subsequently adopt similar healthy eating habits through marital socialization. Conversely, individuals in hypogamous marriages were potentially due to the necessity of compromising personal habits within intimate partnerships.³³ Therefore, in marriages with an educational mismatch, the couple’s dietary quality often converges to an intermediate level between their initial habits.³⁴ The specific level of this convergence, however, is contingent upon the balance of various factors, including bargaining power in household, couple members’ perceptions of health threats, and preferences for outcomes.³⁵ The EAM pattern is also strongly associated with the family’s overall socioeconomic status, particularly household income.³⁶ Educational homogamy or certain types of hypergamy often lead to higher household income, providing the financial resources to afford more nutritious and diverse foods.³⁷ Furthermore, EAM may influence diet by fostering different forms of social capital and peer environments. A spouse with higher education may have social networks that place a greater value on diet. These external social norms can then be introduced into the household, acting as a catalyst for changing the family’s dietary habits.³⁸ Extended analysis reveals that highly educated older adults represent a distinct subgroup in China, with couples where either partner attained higher education being colloquially termed “well-cultivated families.” This demographic demonstrates greater behavioral autonomy,³⁹ appearing less susceptible to spousal influences in hypogamous pairings across all dietary dimensions. Gender heterogeneity analyses revealed a distinct pattern, reflecting China’s traditional gendered division of household labor, where men primarily engage in income-generating work while women oversee domestic responsibilities, including food preparation.⁴⁰ Consequently, husbands’ dietary knowledge and food preferences often mirror their wives’ domestic practices, making male respondents particularly susceptible to spousal influences regardless of the educational gradient. Women demonstrate relative resilience to the negative dietary impacts in hypogamy, potentially because, as primary food gatekeepers, highly educated women may resist adopting less healthy eating patterns from lower-educated spouses, a phenomenon consistent with the “moving down” avoidance hypothesis.¹¹ Gender differences in meat consumption (typically higher among men) may further explain why ABHEI components show stronger associations in male subgroups.⁴¹

This study has several notable strengths. First, we developed a comprehensive HEI for older adults and stratified it into plant-based and animal-based components, representing the first investigation of their associations with EAM. Second, our application of entropy balancing methodology effectively addressed potential selection bias and strengthened causal inference. Third, we comprehensively examined the impacts of all 16 possible EAM patterns on dietary quality. However, several limitations should be acknowledged. The CLHLS dietary data were self-reported and thus subject to recall bias.⁴² Although we employed entropy balancing, several unmeasured factors may persist. First, marriage is a phenomenon characterized by a high degree of self-selection. Partner selection in the marriage market is influenced by an individual’s own characteristics, their family background, surrounding environment. Particularly in a multiethnic nation such as China, marital practices are also shaped by diverse cultural influences.⁴³ Marriage often implies an exchange of social status and resources.⁴⁴ These unobserved selection bias may still diminish the reliability of the study’s findings. Second, as a health behavior variable, the HEI is simultaneously subject to multiple levels of influence, including individual, household, environmental, and cultural factors. These factors are difficult to fully capture, could potentially introduce bias into the research findings. At the household level, as Supplementary Table SA1 has shown, although our study was restricted to older adults coresiding with a spouse, a certain proportion of these individuals also live with their children or other relatives. While we conducted a sensitivity analysis, a limitation remains in that we still cannot identify the primary person responsible for dietary decisions within the household from the CLHLS data; this could systematically affect our conclusions. Furthermore, our baseline regression only included older adults who were married and living with their spouse, which may lead to insufficient sample representativeness; focusing only on couples who are still married and both in the survey likely captures a healthier, more selective group. Although widowed older adults were also included in the sensitivity analysis, the heterogeneity of effects between these two groups still requires further investigation.

Health Equity Implications

Our findings have significant implications for nutritional interventions and demonstrate that decontextualizing individuals from their family relationships is a suboptimal approach. We strongly advocate for future nutritional interventions to shift from being person-centered to family-centered by implementing couple-based dyadic interventions.⁴⁵ Such interventions facilitate communication between partners, establish shared health goals, and provide mutual emotional support to overcome barriers to behavioral change.⁴⁶ Furthermore, future public health measures must cease to reinforce the traditional stereotype of women as the sole party responsible for household nutrition. Instead, program design should actively challenge this gendered division of labor, fostering shared nutritional responsibility and equitable decision-making within the family, rather than merely transmitting knowledge unilaterally to women. In the context of China’s rapidly aging population and persistent educational disparities across generations, this research provides a valuable foundation for designing socially and culturally informed policies that promote equitable and sustainable dietary improvements among older adults.

Footnotes

Author’s Contribution

B.M.: Writing—original draft, conceptualization, methodology, data curation, and formal analysis.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

No funding was received for this article.

Supplementary Material

Abbreviations Used

References

Chen

, Maguire

, Brodaty

, et al. Dietary patterns and cognitive health in older adults: A systematic review. J Alzheimers Dis, 2019; 67(2):583–619; doi: 10.3233/JAD-180468

Dong

, Xie

. Trends in educational assortative marriage in China over the past century. Demography, 2023; 60(1):123–145; doi: 10.1215/00703370-10411058

Guner

, Kulikova

, Llull

. Marriage and health: Selection, protection, and assortative mating. Eur Econ Rev, 2018; 104:138–166; doi: 10.1016/j.euroecorev.2018.02.005

Zhao

, Zhang

. Gendered outcomes? Educational assortative marriages, gender ideology, and depressive symptoms in China. J Affect Disord, 2025; 372:463–469; doi: 10.1016/j.jad.2024.12.033

. Educational mobility through marriage and risk of cognitive impairment in late life. Innov Aging, 2021; 5(Supplement_1):706–707; doi: 10.1093/geroni/igab046.2645

Moafi-Madani

, Baig

, Roberts

, et al. Abstract P573: Educational assortative mating and cardiovascular outcomes. Circulation, 2023; 147(Suppl_1):AP573; doi: 10.1161/circ.147.suppl_1.P573

Wang

, Yamashita

, Xu

, et al. The spouse’s level of education and individuals’ dietary behaviors in China. Int J Behav Med, 2015; 22(4):461–470; doi: 10.1007/s12529-014-9444-0

, Fu

, Zhao

, et al. Influence of spousal education on partner’s self-rated health: Cross-sectional study among 1382 married couples in Shanghai, China. Asia Pac J Public Health, 2013; 25(5):398–408; doi: 10.1177/1010539511420417

. Spousal education and cognitive functioning in later life. J Gerontol B Psychol Sci Soc Sci, 2020; 75(7):e141–e150; doi: 10.1093/geronb/gbz014

10.

Homish

, Leonard

. Spousal influence on general health behaviors in a community sample. Ajhb, 2008; 32(6):754–763; doi: 10.5993/AJHB.32.6.19

11.

Monden

CWS

, van Lenthe

, De Graaf

, et al. Partner’s and own education: Does who you live with matter for self-assessed health, smoking and excessive alcohol consumption? Soc Sci Med, 2003; 57(10):1901–1912; doi: 10.1016/S0277-9536(03)00055-8

12.

Chen

, Olsen

, Lamu

. The influence of parents’ and partner’s education on own health behaviours. Soc Sci Med, 2024; 343:116581; doi: 10.1016/j.socscimed.2024.116581

13.

Takagi

, Kondo

, Takada

, et al. Differences in spousal influence on smoking cessation by gender and education among Japanese couples. BMC Public Health, 2014; 14(1):1184; doi: 10.1186/1471-2458-14-1184

14.

Song

, Hu

, Zhang

. Educational assortative mating and couples’ linked occupational trajectories in China. Soc Sci Res, 2025; 126:103114; doi: 10.1016/j.ssresearch.2024.103114

15.

Kennedy

, Ohls

, Carlson

, et al. The healthy eating index. J Am Diet Assoc, 1995; 95(10):1103–1108; doi: 10.1016/S0002-8223(95)00300-2

16.

Yan

, Li

, Zou

, et al. Healthy eating and all-cause mortality among Chinese aged 80 years or older. Int J Behav Nutr Phys Act, 2022; 19(1):60; doi: 10.1186/s12966-022-01280-6

17.

Zhang

, Zhou

, Wang

, et al. Association between healthy eating and depression symptoms among Chinese older adults: A cross-sectional study based on the Chinese Longitudinal Healthy Longevity survey. Prev Med Rep, 2024; 38:102616; doi: 10.1016/j.pmedr.2024.102616

18.

Petersen

, Flock

, Richter

, et al. Healthy dietary patterns for preventing cardiometabolic disease: The role of plant-based foods and animal products. Curr Dev Nutr, 2017; 1(12):cdn.117.001289; doi: 10.3945/cdn.117.001289

19.

Marchese

, McNaughton

, Hendrie

, et al. A scoping review of approaches used to develop plant-based diet quality indices. Curr Dev Nutr, 2023; 7(4):100061; doi: 10.1016/j.cdnut.2023.100061

20.

Howlett

. Gaokao warriors: Diligent struggle in China’s College Entrance Exam. Anthropology & Edu Quarterly, 2023; 54(1):22–39; doi: 10.1111/aeq.12442

21.

Guo

, Huang

, Zhang

. Education development in China: Education return, quality, and equity. Sustainability, 2019; 11(13):3750; doi: 10.3390/su11133750

22.

Raine

. Determinants of healthy eating in canada: An overview and synthesis. Can J Public Health, 2005; 96(Suppl 3):S8–S15; doi: 10.1007/BF03405195

23.

Guo

, Gietel-Basten

, Gu

. The lowest fertility rates in the world? Evidence from the 2015 Chinese 1% sample census. China Popul Dev Stud, 2019; 2(3):245–258; doi: 10.1007/s42379-018-0012-1

24.

Liu

, Mu

. Mate selection in China: Trends and determinants. Fam Relat, 2023; 72(3):755–773; doi: 10.1111/fare.12718

25.

Hainmueller

. Entropy balancing for causal effects: A multivariate reweighting method to produce balanced samples in observational studies. Polit Anal, 2012; 20(1):25–46; doi: 10.1093/pan/mpr025

26.

Wang

. Age-discrepant marriages and educational assortative mating in urban China: The exchange of youth for status. J of Marriage and Family, 2025; 87(2):636–658; doi: 10.1111/jomf.13042

27.

Hainmueller

, Xu

. Ebalance: A stata package for entropy balancing. SSRN Journal, 2011; doi: 10.2139/ssrn.1943090

28.

Amugsi

, Lartey

, Kimani-Murage

, et al. Women’s participation in household decision-making and higher dietary diversity: Findings from nationally representative data from Ghana. J Health Popul Nutr, 2016; 35(1):16; doi: 10.1186/s41043-016-0053-1

29.

Wilcox

, Evenson

, Aragaki

, et al. The effects of widowhood on physical and mental health, health behaviors, and health outcomes: The Women’s Health Initiative. Health Psychol, 2003; 22(5):513–522; doi: 10.1037/0278-6133.22.5.513

30.

Epstein

, Lindner Pomerantz

. Assimilation through Marriage. Rev International Economics, 2013; 21(2):191–203; doi: 10.1111/roie.12030

31.

Schafer

, Schafer

, Dunbar

, et al. Marital food interaction and dietary behavior. Soc Sci Med, 1999; 48(6):787–796; doi: 10.1016/S0277-9536(98)00377-3

32.

Fazli

, Yazdani

, Mohebbi

, et al. Oral health literacy and socio-demographics as determinants of oral health status and preventive behavior measures in participants of a pre-marriage counseling program. PLoS One, 2021; 16(11):e0258810; doi: 10.1371/journal.pone.0258810

33.

Cazenave

. Social structure and personal choice. J Fam Econ Iss, 1979; 2(3):331–358; doi: 10.1007/BF01083713

34.

Tucker

, Anders

. Social control of health behaviors in marriage. J Applied Social Pyschol, 2001; 31(3):467–485; doi: 10.1111/j.1559-1816.2001.tb02051.x

35.

Lewis

, McBride

, Pollak

, et al. Understanding health behavior change among couples: An interdependence and communal coping approach. Soc Sci Med, 2006; 62(6):1369–1380; doi: 10.1016/j.socscimed.2005.08.006

36.

Eika

, Mogstad

, Zafar

. Educational assortative mating and household income inequality. J Polit Econ, 2019; 127(6):2795–2835; doi: 10.1086/702018

37.

Pesando

. Educational assortative mating in Sub-Saharan Africa: Compositional changes and implications for household wealth inequality. Demography, 2021; 58(2):571–602; doi: 10.1215/00703370-9000609

38.

Flatt

, Agimi

, Albert

. Homophily and health behavior in social networks of older adults. Fam Community Health, 2012; 35(4):312–321; doi: 10.1097/FCH.0b013e3182666650

39.

Sun

, Dong

, Ding

. The impact of dietary knowledge on health: Evidence from the China Health and Nutrition Survey. Int J Environ Res Public Health, 2021; 18(7):3736; doi: 10.3390/ijerph18073736

40.

Maume

ZZL

, David

, Bellas

. Chinese husbands’ participation in household labor. J Comp Fam Stud, 2000; 31(2):191–215; doi: 10.3138/jcfs.31.2.191

41.

Rosenfeld

, Tomiyama

. Gender differences in meat consumption and openness to vegetarianism. Appetite, 2021; 166:105475; doi: 10.1016/j.appet.2021.105475

42.

Subar

, Freedman

, Tooze

, et al. Addressing current criticism regarding the value of self-report dietary data12. J Nutr, 2015; 145(12):2639–2645; doi: 10.3945/jn.115.219634

43.

. Marriage of matching doors: Marital sorting on parental background in China. DemRes, 2016; 35:557–580; doi: 10.4054/DemRes.2016.35.20

44.

, Pan

. Beauty and status: Matching and exchange within Chinese marriages. Chin J Sociol, 2023; 9(3):375–407; doi: 10.1177/2057150X231189027

45.

Arden-Close

, McGrath

. Health behaviour change interventions for couples: A systematic review. Br J Health Psychol, 2017; 22(2):215–237; doi: 10.1111/bjhp.12227

46.

Burke

, Giangiulio

, Gillam

, et al. Physical activity and nutrition programs for couples: A randomized controlled trial. J Clin Epidemiol, 2003; 56(5):421–432; doi: 10.1016/S0895-4356(02)00610-8

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.11 MB