Gender Differences in Adult Children's Support of Their Parents in Taiwan

Journal of Marriage and Family 65 (February 2003): 184–200184

I-FEN LIN Bowling Green State University

NOREEN GOLDMAN Princeton University*

MAXINE WEINSTEIN Georgetown University**

YU-HSUAN LIN Department of Health, Taiwan***

TRISTAN GORRINDO Vanderbilt University****

TERESA SEEMAN University of California—Los Angeles*****

l

Gender Differences in Adult Children’s Support of Their

Parents in Taiwan

This paper examines the patterns and determi-nants of four types of support provided by adultchildren to their parents, with particular attentionto differences in the helping behaviors of sons anddaughters. The data come from the 1989 wave ofthe Survey of Health and Living Status of the El-derly in Taiwan. The analysis is based on 12,166adult children from 2,527 families. We find thatusually only one child in a family provides helpwith activities of daily living (ADLs) or instru-mental activities of daily living (IADLs), but for

Department of Sociology, 217 Williams Hall, BowlingGreen State University, Bowling Green, OH 43403([email protected]).

*Office of Population Research, 243 Wallace Hall, Prince-ton University, Princeton, NJ 08544.

**Center for Population and Health, 312 Healy Hall, Box571197, Georgetown University, Washington, DC 20057.

***Survey Research Center for Population and Health, Bu-reau of Health Promotion, Department of Health, Taiwan.

****Vanderbilt School of Medicine, Vanderbilt University,Nashville, TN 37232.

*****Department of Epidemiology, Box 951772, School ofPublic Health, University of California—Los Angeles.

Key Words: gender differences, provision of support, Tai-wan.

financial or material support the responsibility islikely to be shared among siblings. Sons generallycarry the major responsibility for taking care oftheir older parents, and daughters fulfill the son’sroles when sons are not available.

Taiwan is one of the newly industrialized societiesthat have experienced rapid demographic, eco-nomic, and social changes over the second half ofthe last century. Between 1952 and 2000, life ex-pectancy at birth increased by about 20 years,whereas total fertility dropped by nearly fivebirths (Population Reference Bureau, 2001). As aresult of this demographic transition, the percent-age of people who are 65 years or older has al-most quadrupled from 2.5 to 9 and is expected torise to 14% of the population by the year 2020(Li, 1994). Taiwan has been transformed from arural, agricultural society to a highly urbanized,industrial one (Hermalin, Liu, & Freedman,1994): For example, the percentage of the popu-lation living in cities has tripled, the percentageof the labor force engaged in agriculture has de-creased from 56.1 to 13.7, and per capita incomehas grown more than ninefold over this period.

Despite these far-reaching changes, most par-ents in Taiwan continue to live with their adultchildren, particularly their married sons (Chang,1999; Ofstedal, Knodel, & Chayovan, 1999; Sun

185Parental Support in Taiwan

& Liu, 1994; Weinstein, Sun, Chang, & Freed-man, 1990). The direction of financial flows is stilldominated by transfers from adult children to theirparents (Lee, Parish, & Willis, 1994; Sun & Liu,1994), in part because placing parents in a nursinghome has been perceived as a violation of tradi-tional filial obligations (Kao & Stuifbergen,1999), and in part because the government pro-vides minimal protection from potential economichardship. The preference for family assistance andthe lack of institutional support underscore the im-portance for older persons of help provided bymembers of social networks, particularly adultchildren.

Although a number of studies have examinedintergenerational transfers in Taiwan (e.g., Chat-topadhyay & Marsh, 1999; Hermalin, Ofstedal, &Chang, 1996; Hermalin, Ofstedal, & Lee, 1992;Ofstedal et al., 1999; Sun & Liu, 1994), most aredescriptive or reflect the perspective of the agingparents, rather than the children. Previous studieshave examined the prevalence of support amongolder adults, the number and type of kin availablefor support, and factors determining whether ag-ing parents receive support, but little is knownabout the underlying mechanisms by which adultchildren provide support to their parents (an ex-ception is Lee et al., 1994).

The objective of the present study is to extendprior research by examining the empirical evi-dence in Taiwan for a set of five explanations ofwhy adult children provide support to their agingparents. A related goal is to compare and contrastthe helping behaviors of sons and daughters.There are two salient aspects of Taiwanese societythat motivate this detailed examination of genderdifferences in the provision of support by chil-dren. First, Taiwan is a patriarchal society inwhich sons bear the primary responsibility of con-tinuing the descent line of the father’s family. Al-though a daughter belongs to her father’s familybefore marriage, she joins the descent line of herhusband’s family upon marriage (Wolf, 1972).Unlike American families in which adult children,particularly married ones, rarely live with theirolder parents, traditional Taiwanese families placea strong emphasis on parent-son coresidence as animportant form of fulfilling filial obligation (Wein-stein et al., 1990). Older parents who have morethan one son often rotate among their sons’ resi-dences. In contrast, married daughters are expect-ed to live with their parents-in-law. A secondcharacteristic of Taiwanese society that has impli-cations for differences in the support provided by

sons and daughters relates to gender differencesin human capital and economic resources. UnlikeAmerican families that typically have no morethan a few children, prior to the past few decadesTaiwanese couples generally had large families. Inview of budget constraints involved in raisingmany children, and because of gender differencesin expectations regarding returns from sons anddaughters, Greenhalgh (1985) has argued that Tai-wanese parents generally invested more in theirsons’ than in their daughters’ education. The dif-ferences in literacy and education level result ingender differences in economic resources andskills later in life that are likely to affect the leveland type of support that sons and daughters areable to provide to their older parents.

In this study, we examine the patterns and de-terminants of four types of support provided byadult children: assistance with essential activitiesof daily living (ADLs), household chores, financ-es, and material goods. To provide insight intowhich children contribute to parental support, ourwork uses the adult child as the fundamental unitof analysis but also incorporates parental charac-teristics—such as their age, marital status, andhealth status—that have been shown to affect sup-port (Hermalin et al., 1992; Lee et al., 1994; Sil-verstein, Parrott, & Bengtson, 1995; Spitze & Lo-gan, 1990; Stoller, 1983).

THEORY AND PREVIOUS RESEARCH

Researchers have proposed a number of explana-tions for how adult children fulfill their obliga-tions to their parents; however, much of the ex-isting research has been conducted in Westernsocieties. Few systematic attempts have beenmade to examine the division of labor in Asianfamilies. In this paper, we use results from statis-tical models of the provision of support by chil-dren in Taiwan to examine five proposed theoret-ical explanations for the allocation ofresponsibilities. In the following section, we sum-marize each of the five explanations and compareearlier findings in American and Taiwanese fam-ilies.

Hierarchical Compensation

The hierarchical compensation explanation pro-poses that family members make themselvesavailable for help in serial order: If one individualis not available to help, another will step in (Can-tor, 1975; Johnson, 1983; Shanas, 1979). In the

186 Journal of Marriage and Family

United States, the preferred provider of support isthe spouse. When a spouse is not available,daughters are usually preferred over sons (Wolf &Soldo, 1988), and when no daughter is available,sons are likely to substitute and provide assistanceto their parents (Horowitz, 1985; Wolf, Freedman,& Soldo, 1997; but see Spitz & Logan, 1990).American daughters spend more hours per monthhelping their parents than sons do (e.g., Matthews& Rosner, 1988; Stoller, 1983; Wolf et al., 1997),and married daughters are more likely than mar-ried sons to live with unmarried, older mothers(Wolf & Soldo, 1988).

Historically in Taiwan, sons have borne themajor responsibility for regular care and supportof parents. Even today, Taiwanese sons providemore instrumental assistance and financial supportto parents than daughters do (Hermalin et al.,1992). Parents are more likely to live with theirmarried sons than with their married daughters(Chang, 1999; Ofstedal et al., 1999; Sun & Liu,1994; Weinstein et al., 1990), and have more fre-quent contact with sons (Hermalin, Ofstedal, &Chi, 1992). When a parent has more than one sonand has not yet divided family property, the eldestson typically takes the major responsibility of car-ing for his aging parents (Sung, 1981; Wolf,1968). Daughters are likely to fill the sons’ roleswhen sons are not available (Hermalin et al.,1992; Lee et al., 1994).

Gender Ideology

Gender ideology argues that family labor is divid-ed on the basis of gender-based expectations thatare internalized by individuals through socializa-tion (Ross, 1987). Because men are socialized tothe role of provider and women to the role ofcaregiver, sons are frequently involved in financialsupport, whereas daughters often assist withhousehold chores or personal care. This conjectureis supported by findings in the United States butnot by earlier research in Taiwan. Americandaughters spend more time on instrumental activ-ities of daily living (IADLs), such as transporta-tion, parental care, and meal preparation, thansons (Horowitz, 1985; Matthews & Rosner, 1988;Stoller, 1983), whereas sons are slightly morelikely to deal with financial tasks than daughters(Montgomery & Kamo, 1989). In Taiwan, re-searchers find that sons are more important pro-viders than daughters, regardless of the form ofsupport (Hermalin et al., 1992). However, gender-specific division of filial tasks may exist between

married sons and their spouses, rather than be-tween married sons and married daughters. In Tai-wan, a daughter’s responsibility to her own par-ents usually ends at marriage, at which time herresponsibility is transferred to her husband’s fam-ily (Greenhalgh, 1985).

Competing Commitments

The competing commitments explanation positsthat demand attributable to multiple roles deter-mines the time available for family assistance(Ross, 1987). On this basis, one would expect em-ployed children and married children to provideless support than their counterparts. In the UnitedStates, this position is supported by comparisonsof the amount of support provided by working andnonworking daughters (Lang & Brody, 1983;Matthews, Werkner, & Delaney, 1989). In addi-tion, employment is a significant predictor of per-ceived caregiver burden for daughters but not sons(Montgomery & Kamo, 1989), and daughters aremore likely than sons to perceive role conflicts(Finley, 1989). Married daughters spend less timethan unmarried daughters in providing parentalcare (Lang & Brody), and married children—bothsons and daughters—provide less help than never-married children (Stoller, 1983).

In Taiwan, counter to the competing commit-ments explanation, working children are morelikely than nonworking children to provide assis-tance with ADLs, financial support, and materialsupport (Hermalin et al., 1992). Moreover, mar-ried children are more apt to provide parental sup-port than unmarried children.

Reciprocity

Reciprocity suggests that the extent to which adultchildren help their parents is based on the parents’past investment in them (Gouldner, 1960; Thibaut& Kelley, 1959): Children who received a greaterinvestment, such as schooling or financial trans-fers from their parents, are more likely to provideparental support. In the United States, this conjec-ture is supported by the work of Cox and Rank(1992) who show that children who receive finan-cial transfers are more likely than children whodo not receive this assistance to have visits andtelephone contact with parents and to provide par-ents help. In Taiwan, Hermalin, Ofstedal, and Lee(1992) find that children who received a junior orsenior high school education are more likely thanchildren with at most a primary school education


to provide financial and material support. Lee etal. (1994) show that sons are more likely to sup-port their parents when the parents have alreadydivided their property.

External resources

The external resources explanation proposes thatthe relative resources of family members, such asincome and education, determine their power innegotiating the family division of labor (Ross,1987). This explanation implies that siblings withgreater financial resources may negotiate with lesswell-off siblings to exempt themselves from co-residing with parents or providing time-intensiveservices. Thus children with more resources mayprovide additional financial support but less assis-tance with daily activities. There is little evidencein either the United States (Finley, 1989) or inTaiwan (Lee et al., 1994) to support this expla-nation, but poor operationalization of this conceptin previous studies may be the cause. Previousstudies have examined children’s absolute levelsof education but overlooked the potential impor-tance of relative levels in determining negotia-tions among family members.

In sum, the hierarchical compensation expla-nation predicts that in Taiwan sons are more likelythan daughters to provide support to their olderparents, but that daughters will substitute for sonswhen they have no brothers. Gender ideology sug-gests that sons are more likely than daughters toprovide financial or material support, whereasdaughters are more likely than sons to provide as-sistance with ADLs or IADLs. The competingcommitments explanation implies that employedor married children are less likely to help thannonworking or unmarried children, regardless ofchildren’s gender. Reciprocity suggests that adultchildren with higher levels of schooling within afamily and children whose parents have dividedtheir property are more likely than their counter-parts to provide support to their aging parents. Theexternal resources explanation predicts that sib-lings with a higher relative level of education arelikely to provide financial and material support,whereas siblings with a lower relative level of ed-ucation are likely to provide assistance with ADLsor IADLs. Although both the reciprocity and ex-ternal resources explanations concern adult chil-dren’s educational attainment, the former predictsthe amount of support, whereas the latter predictsthe type of support.

METHOD

Data and Sample

Our data are drawn from the Survey of Health andLiving Status of the Elderly in Taiwan (TaiwanProvincial Institute of Family Planning, 1989).Face-to-face interviews were conducted in 1989with a national probability sample of 4,049 per-sons aged 60 and over. One older adult from eachhousehold was selected for interview (a proxy wasused when the adult was too ill to be interviewed).The response rate for the survey was 92%. Thedata are richer than most other information on old-er adults in three respects. First, the full popula-tion is represented in the survey, including insti-tutionalized persons. Second, the survey collectsdetailed information on all household membersand on potentially important social ties, includingan extensive set of questions on instrumental andfinancial exchanges in which the aging parent isthe provider or recipient. Finally, the survey col-lects a wide range of detailed information aboutmarital, residential, and occupational histories,household composition, social and economic ex-changes, emotional and instrumental support anddemands, and physical and mental health.

Although the data include detailed informationabout the relationships between the parent andeach of the coresident and nonresident membersof the family, the survey does not ask informationabout relationships among coresident and nonres-ident members. For instance, we may know thata parent has two married sons and two daughters-in-law living outside the household, but lack in-formation that would allow us to identify whichdaughter-in-law is married to which son. Similar-ly, we are not able to identify the parent of agrandchild who is in the household. Because ofthis limitation, we exclude support provided fromchildren-in-law and variables related to the pres-ence of grandchildren from the analysis.

The sample for this analysis is selected ran-domly from the 1989 survey to comprise twothirds of the respondents (i.e., 2,713 parents). Thisprocedure is part of a split-sample design for alarger project that examines the relationship be-tween the social environment and health amongolder Taiwanese. The ultimate objective of thisstrategy is to use the two-thirds subsample to es-timate models, as we do in this paper, and theremaining subsample to appraise the models insubsequent analyses. This type of cross-validationprocedure provides protection against overfitting


TABLE 1. NUMBER AND GENDER COMPOSITION OF

CHILDREN IN FAMILIES IN 1989

Families Percentage

PercentageDistribution

of Sons

No child 5.68

1 child 6.270 sons1 son

45.2954.71

2 children 9.070 sons1 son2 sons

14.6359.3526.02

3 children 11.570 sons1 son2 sons3 sons

7.6435.0347.459.87

4 children 14.670 sons1 son2 sons$3 sons

4.7722.3638.9433.92


1.8112.0233.3352.83


2.1110.3325.3562.21


1.016.38

18.4674.16

8 or more children 9.800 sons1 son2 sons$3 sons

0.753.01

14.2981.95

Total number of familiesMean number of children per family

27134.80

Note: Column totals in each panel may deviate slightlyfrom 100% because of rounding errors.

models to the data (for the split sample design,see Berk, 1991; Mosteller & Tukey, 1977).

Measures

Family structure. Family structure is characterizedas the number and gender composition of living

children in a family. As shown in Table 1, theaverage number of living children per family is4.8, reflecting the high fertility in Taiwan prior tothe 1960s. About two thirds of the parents havebetween two and six children. Not surprisingly, asfamily size gets larger, the likelihood of having atleast one son present in the family increases.

Parents’ characteristics. Table 2 shows the distri-butions of characteristics for 2,527 older parents.The total number of older parents is less than2,713 because those with no children (n 5 154)or only young children (n 5 32) are excludedfrom the analysis. (Children aged 15 and youngerare excluded from all parts of the analysis, be-cause, in view of the 9 years of compulsory ed-ucation provided by the Taiwanese government,most will be attending school and are unlikely tobe providing support to their parents.) About twothirds of the respondents are in their 60s in 1989,with women being slightly older than men. Mostrespondents, especially women, are illiterate or re-ceived only a primary school education.

The population in Taiwan consists of four eth-nic groups: Aborigines, Fukienese, Hakka, andMainlanders. Aborigines are the earliest migrantsto Taiwan, followed by Fukienese (originatingfrom the Fukien Province in China), Hakka (orig-inating from the Kwangtung Province in China),and then Mainlanders (mainly the Nationalist par-ty army and its supporters who left Mainland Chi-na during the 1949 Civil War). Fukienese consti-tute the major ethnic group in our sample.Mainlanders are disproportionately men becausemost Mainlanders migrating to Taiwan were sol-diers. Hakka share a more similar culture with Fu-kinese than the Mainlanders.

More than 60% of the parents in the sampleare married. For this generation, marriage rarelyends in divorce. Most of the respondents reportexcellent, very good, or good health, althoughwomen are more likely than men to say theirhealth is fair or poor. On average, the parents inthe sample report less than one functional limita-tion. Men are more likely than women to be eco-nomically independent: That is, they are morelikely to own their residence, retain some prop-erty, work, and have an income. Over one thirdof all parents live in urban areas.

Adult children’s characteristics. Children’s char-acteristics are presented in Table 3. The informa-tion is based on the parent’s report. Given thatmost parents have more than one child, the total


TABLE 2. PARENTS’ CHARACTERISTICS IN 1989a

Characteristics

Percentageb/Mean

Total Men Women

Age (years)60–6970–79$80

64.2328.936.85

68.1026.015.89

59.4732.518.02

EducationNo formal education (illiterate or can read)Primary schoolJunior high schoolSenior high schoolCollege or higherMissing

50.7729.927.955.665.180.51

32.2638.6511.648.128.760.57

73.4819.213.442.640.790.44

EthnicityFukieneseMainlanderHakkaOther ethnicity (Aboriginal, foreign)Missing

63.5118.1615.121.821.39

57.0426.0814.151.651.08

71.458.46

16.302.031.76

Marital statusc

Has a coresident spouseHas a non-coresident spouseSeparated or divorcedWidowedNever married

64.462.302.49

30.310.44

78.092.663.66

15.010.57

47.751.851.06

49.070.26

Health statusReports excellent or very good healthReports good healthReports fair or poor healthMissing

37.5137.0021.533.96

44.6836.5714.873.88

28.7237.5329.694.05

Difficulties with activitiesMean number of functional limitations (bathing, climbing stairs,

walking, crouching, reaching up over head, grasping) 0.47 0.36 0.60Residence ownershipc

Owns current residenceDoes not own current residence

53.1546.85

62.7937.21

41.3258.68

Property divisionHas divided all or partial propertyHas not divided propertyHas no propertyMissing

25.9243.8929.600.59

21.4852.3025.720.50

31.3733.5734.360.70

Working statusc

WorkingNot working

26.8373.17

40.0159.99

10.6689.34

Monthly income (NT$)d

#5,0005,000–9,99910,000–19,999.19,999Missing

29.4820.4226.7518.724.63

20.6920.7630.0324.863.66

40.2620.0022.7311.195.81

Area of residenceRuralCityUrban

Number of cases (%)

34.5928.5336.88

2,527 (100)

34.1228.6637.21

1,392 (55)

35.1528.3736.48

1,135 (45)

aIn total, 186 parents (or families) are excluded from the analysis. Among these families, 154 have no children, 14 haveonly one child and the child is under age 16 years in 1989, and 18 have more than one child and all children are underage 16. bColumn totals in each panel may deviate slightly from 100% because of rounding errors. cBecause the total numberof missing cases is less than 10, missing values have been replaced by the modal value of the given variable according tothe respondent’s gender. dNT$ indicates New Taiwanese dollars. In 1989, one U.S. dollar was equal to 26.17 NT$.


TABLE 3. CHILDREN’S CHARACTERISTICS IN 1989

Characteristics

Percentagea

Total Men Women

Type of support provided to parentAny helpHelp with ADLsb

Help with IADLsc

Financial help (financial support)Food, clothing, and other goods (material support)

51.260.804.09

43.2720.47

66.621.065.45

62.9818.90

35.790.532.71

23.4122.04

Age (years)#2021–3031–4041–50$51Missing

1.9217.0443.0726.1810.451.34

1.9817.1843.5126.4410.140.75

1.8616.8942.6325.9210.761.93

Eldest childIs eldest childIs not eldest child

19.6480.36

19.6780.33

19.6080.40

Number of brothers01 or 2$3

7.7550.3341.92

8.8150.3840.80

6.6850.2743.05

Number of sisters01 or 2$3

9.1247.1143.77

8.5548.8142.64

9.7045.3944.91

Level of educationNo formal education (illiterate or can read)Primary schoolJunior high schoolSenior high schoolCollege or higherMissing

7.7237.8115.1922.5616.190.53

3.0834.1217.4624.7020.100.54

12.3941.5312.9020.4112.260.51

Relative educationHighestMiddleLowestSame or only childAt least one child’s education in the family is unknown

27.9918.2336.6815.771.34

36.7419.8225.8816.151.41

19.1716.6347.5515.381.27

Employment statusFull-time employedPart-time employedNot workingMissing

67.023.53

28.770.68

89.432.467.470.64

44.454.60

50.220.73

Marital statusMarriedFormerly married (separated, divorced or widowed)Never marriedMissing

81.371.88

16.290.46

77.741.64

20.340.28

85.022.13

12.210.64

Living arrangementCoresides with the parentLives in the same city or townLives in a different city or townLives overseasMissing

Number of cases (%)

24.7928.0243.433.150.61

12,166 (100%)

38.3624.9533.502.780.41

6,105 (50%)

11.1231.1253.443.510.81

6,061 (50%)

aExcept for the first panel, column totals in each panel may deviate slightly from 100% because of rounding errors.bADLs (activities of daily living) include bathing, dressing, and maintaining toilet functions. cIADLs (instrumental activitiesof daily living) include shopping, meal preparation, transportation, and managing finances.


number of children in this table (N 5 12,166) ismuch larger than the number of parents in Table2 (N 5 2,527). A small number of adopted chil-dren and stepchildren (n 5 41) are included in theanalysis. Most of the children were in their 30s or40s in 1989. More than 90% of the children haveat least one sibling.

Children’s educational attainment is measuredin two ways: the absolute level of education re-ceived and the relative ranking in a family. Chil-dren’s educational attainment is substantiallyhigher than their parents’—only 3% of the sonsand 12% of the daughters received no formal ed-ucation, and 45% of the sons and 33% of thedaughters attended senior high school or college.Although the level of education increases substan-tially between generations, daughters remain lesseducated than sons, and sons are twice as likelyas daughters to be the most educated child in thefamily. More than 90% of the sons are employedfull-time or part-time, but only half of the daugh-ters are employed. Approximately three quartersof the children are married. Finally, the proportionof sons who live with their parents is more thanthree times the number for daughters, reflectingthe norm for coresidence with sons in Taiwan. Al-though daughters are less likely than sons to livewith their parents, nonresident daughters are morelikely than nonresident sons to live in the samecity or town as their older parents.

Four types of support. We examine four types ofsupport that children provide to their older par-ents: (a) help with bathing, dressing, or maintain-ing toilet functions (i.e., ADLs); (b) assistancewith household chores, such as shopping, mealpreparation, transportation, or managing finances(i.e., IADLs); (c) provision of financial support;and (d) provision of material support, such asfood, clothing, or other goods. In the survey, par-ents were asked if they received help at the timeof the interview from each coresident and nonres-ident member of the household for each of thesetypes of support. The percentage distributions inthe first panel of Table 3 show that about half ofthe children provide at least one type of support.

Analytical Strategy

We use multivariate models to examine the effectsof children’s characteristics, family structure, andthe needs and resources of parents on the provi-sion of each of four types of support. We focuson the first two sets of variables because they form

the basis of the five theoretical propositions de-scribed earlier. We examine the consistency be-tween the resulting estimates and the predictionsof each explanation, rather than carry out formalhypothesis tests, for two reasons. First, given thatthese explanations are not mutually exclusive, itis likely that our results will be at least partly con-sistent with more than one explanation. Second,children’s decisions about providing support totheir parents have been shown to depend on theirprevious history of support provision as well ason current and previous assistance provided bytheir siblings (Checkovich & Stern 2002). We lackinformation that would permit us to adequatelymodel the interdependence in the formal analysis.

We use the child as the unit of analysis, in-cluding coresident and nonresident children. Astatistical concern arising from the inclusion of allchildren is that observations from the same familytend to be correlated, thereby violating the clas-sical assumption of independence among obser-vations. Statistical methods that ignore the nestedstructure of the data generally underestimate thevariance of the estimated coefficients. To addressthis problem, we use Huber-White estimators(Johnston & DiNardo, 1997) in the statisticalpackage STATA (StataCorp, 1999) to provide ro-bust standard errors of the coefficients in the pres-ence of clustering.

Another statistical problem concerns the poten-tial endogeneity of some of the correlates of sup-port provision, such as breadwinner status and liv-ing arrangements. For example, although livingarrangements are likely to be highly correlatedwith support provision, some of the same unmea-sured variables that determine whether childrencoreside with their parents (e.g., the quality of therelationship between children and parents) arealso likely to determine whether they provide sup-port. To avoid the biases introduced by incorpo-ration of endogenous variables, we exclude thesevariables from the statistical models.

Because Taiwanese society is highly stratifiedby gender (and was even more so at the time whenthe children in the sample were growing up), weexpect that the associations between children’scharacteristics and the likelihood that they providesupport differ by their gender. Thus we includeinteraction terms between each of the explanatoryvariables and the child’s gender. However, ratherthan present the coefficients for the interactionterms, we calculate separate coefficients for sonsand daughters in order to simplify the presenta-tion.


TABLE 4. DISTRIBUTION OF SHARED RESPONSIBILITY FOR PROVIDING SUPPORT TO OLDER PARENTS IN 1989

Any Help ADLsa IADLsb

FinancialSupport

MaterialSupport

Only one adult child in family (171 families)% who received help from a child 39.18 4.09 11.70 32.16 18.71

More than one adult child in family (2,356 families)% who receive help from a child

% only one child helps% more than one child but fewer than half of the children help% more than half of the children help% of all children help% total

79.8419.9423.0225.9431.10

100.00

2.5972.1319.676.561.64

100.00

13.8471.4717.487.983.07

100.00

74.7023.9227.4427.1021.5399.99c

37.1432.4623.8922.9720.69

100.01c

All families 77.09 2.69 13.69 71.82 35.89

aADLs (activities of daily living) include bathing, dressing, and maintaining toilet functions. bIADLs (instrumental activ-ities of daily living) include shopping, meal preparation, transportation, and managing finances. cColumn totals deviateslightly from 100% because of rounding errors.

RESULTS

Adult Children’s Participation in Support

Table 4 describes how responsibility for support-ing parents is shared among children in a family.The two most frequent forms of help that parentsreceive from their children are monetary and ma-terial support. Among families with only onechild, about 39% of the parents receive some formof help from that child, but among larger families,nearly double that percentage receive help. Thesharing of filial responsibility varies by the typeof support: For ADLs and IADLs, one child typ-ically provides the help, but for financial or ma-terial support, the responsibility is likely to beshared among siblings.

Factors Predicting the Likelihood That an AdultChild Provides Support

Table 5 presents the results from a series of lo-gistic regression models of the determinants ofchildren’s provision of parental support. Separatemodels are estimated for each type of support.Each model includes all children above age 15(excluding children-in-law) of the 2,527 parents.The numbers in the cells are odds ratios (OR); forcategorical variables, they have been calculatedrelative to an odds ratio of unity for the omittedcategory. Several variables have been droppedfrom two of the regression equations (ADLs andIADLs) because none of the children with thesecharacteristics provided the relevant form of sup-port.

Three types of statistical tests are shown along-

side the coefficients: (a) two-tailed t tests associ-ated with individual regression coefficients; (b)two-tailed t tests associated with the difference be-tween a given coefficient for sons and the corre-sponding coefficient for daughters; and (c) x2 testsassociated with an entire set of categorical vari-ables (for those consisting of at least two cate-gories in addition to the omitted one).

Adult Children’s Characteristics andFamily Structure

The estimates in Table 5 provide mixed supportfor each of the five proposed explanations of sup-port provision. Contrary to predictions related tohierarchical compensation, the oldest child is notmore likely to provide help than younger siblings(shown in the panel Birth Order). This result maybe partly the consequence of frequent rotations ofresidence among Taiwanese children so that theoldest children do not always live with their par-ents and, when they do, other siblings may be pre-sent in the household (Weinstein, Sun, Chang, &Freedman, 1994). In further negation of this ex-planation, the results indicate that, except for as-sistance with ADLs (OR 5 0.66 for sons and 2.06for daughters), birth order is equally important forsons and daughters. However, the results regard-ing the gender composition of siblings do supportthis explanation. In particular, the existence ofbrothers does not seem to affect the likelihood thata son provides support, but a daughter is muchmore likely to provide certain types of help in theabsence of a brother (shown in the panel Numberof Brothers). Specifically, the odds that a daughter


TA

BL

E5.

ES

TIM

AT

ED

OD

DS

RA

TIO

SFR

OM

LO

GIS

TIC

MO

DE

LS

OF

TH

EL

IKE

LIH

OO

DT

HA

TA

CH

ILD

PRO

VID

ES

SUP

PO

RT

TO

AN

OL

DE

RPA

RE

NT,

BY

TY

PE

OF

SUP

PO

RT

a

Exp

lana

tory

Var

iabl

es

Any

Hel

p

Sons

Dau

ghte

rs

AD

Lsc

Sons

Dau

ghte

rs

IAD

Lsd

Sons

Dau

ghte

rs

Fina

ncia

lSu

ppor

t

Sons

Dau

ghte

rs

Mat

eria

lSu

ppor

t

Sons

Dau

ghte

rs

Chi

ldch

arac

teri

stic

sA

ge .20

b

#20

Bir

thor

der

0.53

*0.

46*

7.35

e1.

341.

150.

48*

0.40

*0.

630.

32*

Seco

nd-

orla

ter-

born

Eld

est

Num

ber

ofbr

othe

rsx

2(d

f5

2)$

30

1.01

1.61

1.26

0.96

22.7

3‡

2.73

*

0.66

3.51

3.67

2.06

†5.

05

10.7

2

0.95

4.95

2.45

1.16

21.9

5‡

9.84

*†

0.97

1.47

1.18

0.94

31.1

5‡

4.86

*†

0.97

2.33

1.53

1.01

5.49

1.52

1to

2N

umbe

rof

sist

ers

x2

(df

52)

$3

0 1to

2

0.93

1.62

1.21

0.96

1.48

*†0.

63

1.09

1.00

0.58

6.00

‡

0.13

*0.

48

0.53

4.09

1.61

†0.

51

1.43

0.80

0.79

0.87

1.73

*10

.17‡

2.45

*1.

29

0.91

1.76

1.26

1.00

1.44

*†0.

73

1.14

1.05

1.16

0.65

1.17

0.98

1.30

*1.

09

0.93

0.88

Rel

ativ

eed

ucat

ion

x2

(df

53)

Low

est

Hig

hest

Mid

dle

Sam

eor

only

child

21.5

4‡

1.45

*1.

101.

14

0.58

1.03

†0.

940.

99

3.99

0.97

2.07

2.08

0.85

1.51

1.01

1.99

4.95

1.07

0.73

0.74

0.30

0.88

0.94

0.91

25.1

8‡

1.52

*1.

231.

16

7.09

1.19

0.89

†1.

27

2.03

1.12

1.01

1.14

0.10

0.99

1.00

1.05

Inte

ract

ions

x2

(df

53)

No

brot

her

*H

ighe

sted

ucat

ion

No

brot

her

*M

iddl

eed

ucat

ion

No

brot

her*

Sam

eed

ucat

ion

oron

lych

ildE

mpl

oym

ent

stat

usx

2(d

f5

2)

9.82

‡0.

972.

020.

5315

9.55

‡

9.68

‡1.

028.

15*

0.60

25.0

4‡

0.64

0.37 e

0.79

1.70

0.16

0.67 e

0.54

12.1

1‡

0.95

0.92

0.68

1.33

4.43

0.83

0.75

1.00

0.60

8.42

‡

10.2

5‡0.

992.

450.

5714

9.04

‡

18.8

0‡0.

919.

80*

0.29

*30

.16‡

0.04

0.96

1.02

1.01

9.84

‡

13.3

0‡0.

954.

77*†

0.71

6.94

‡N

otw

orki

ngFu

ll-tim

eem

ploy

edPa

rt-t

ime

empl

oyed

Mar

ital

stat

usx

2(d

f5

2)M

arri

ed

5.67

*4.

93*

6.23

‡

1.42

*†1.

75*†

14.1

1‡

2.26

3.96

5.98

3.41

*7.

50*

21.7

4‡

1.16

2.19

5.28

1.11

2.92

*57

.81‡

5.98

*4.

54*

6.44

‡

1.56

*†2.

01*†

36.1

1‡

1.82

*2.

08*

19.4

3‡

1.25

*1.

3516

.57‡

Form

erly

mar

ried

Nev

erm

arri

ed0.

820.

80*

1.01

1.53

*†3.

012.

97*

6.65

*14

.86*

†1.

100.

62*

3.87

*†6.

26*†

0.91

0.80

*1.

202.

08*†

0.62

0.59

*0.

58*

0.57

*

Pare

ntch

arac

teri

stic

sG

ende

rF

emal

eM

ale

1.00

0.91

2.29

*0.

781.

321.

031.

000.

900.

950.

89


TA

BL

E5.

CO

NT

INU

ED

Exp

lana

tory

Var

iabl

es

Any

Hel

p

Sons

Dau

ghte

rs

AD

Lsc

Sons

Dau

ghte

rs

IAD

Lsd

Sons

Dau

ghte

rs

Fina

ncia

lSu

ppor

t

Sons

Dau

ghte

rs

Mat

eria

lSu

ppor

t

Sons

Dau

ghte

rs

Age

x2

(df

52)

2.26

0.90

5.09

4.92

4.44

1.60

2.76

4.51

5.94

0.59

60–6

970

–79

$80

Edu

catio

nx

2(d

f5

2)N

ofo

rmal

educ

atio

n

1.16

1.23

8.70

‡

1.06

0.89

3.58

2.70

*2.

692.

30

3.49

1.19

5.97

0.76

1.27

5.06

1.34

1.17

0.11

1.17

0.91

12.5

8‡

1.03

0.62

1.90

1.33

*1.

420.

36

1.09

1.15

3.68

Prim

ary

orju

nior

high

scho

olSe

nior

high

scho

olor

colle

geE

thni

city

x2

(df

52)

Fuk

iene

seor

Hak

kaM

ainl

ande

rO

ther

0.84

0.58

*14

.21‡

0.57

*0.

86

1.14

†0.

829.

41‡

0.83

†2.

47*†

0.66

0.27

1.81 e

0.60

14.7

1*†

0.61 e

0.68

*0.

523.

96

0.51

0.69

1.08

1.04

14.7

7‡

0.40

*3.

31*†

0.78

*0.

53*

11.4

8‡

0.60

*1.

00

1.07

†0.

809.

43‡

0.88

3.12

*†

1.08

1.12

4.58

0.62

*1.

06

1.28

1.23

10.3

8‡

0.58

*1.

93M

arita

lst

atus

No

spou

seor

spou

seno

tpr

esen

tSp

ouse

pres

ent

Hea

lthst

atus

x2

(df

52)

Exc

elle

ntor

very

good

0.98

1.41

0.76

*4.

240.

510.

26*

4.39

0.64

*1.

820.

43*

9.22

‡1.

061.

130.

66*†

4.77

0.99

4.04

0.87

5.17

Goo

dFa

iror

poor

Dif

ficul

ties

with

activ

ities

Num

ber

ofA

DL

sR

esid

ence

owne

rshi

p

1.04

0.90

0.97

1.24

1.23

0.99

f f

2.46

*

3.63

10.9

2

1.72

*

0.96

1.26

1.04

0.55

*1.

19

1.14

1.08

0.96

0.94

1.21

1.38

*

1.05

1.24

0.97

1.01

1.33

*1.

31

0.94

Doe

sno

tow

ncu

rren

tre

side

nce

Ow

nscu

rren

tre

side

nce

Prop

erty

divi

sion

x2

(df

52)

Has

not

divi

ded

prop

erty

Has

divi

ded

prop

erty

0.86

5.09

1.08

0.79

0.30

1.02

1.51

9.47

‡

4.00

*

0.97

1.74

0.49

†

0.86

3.52

1.55

0.92

7.24

‡

1.07

0.91

3.80

1.06

0.94

2.65

0.94

0.90

24.9

1‡

1.37

*

0.70

*5.

67

1.07

No

prop

erty

todi

vide

Wor

king

stat

usN

otw

orki

ngW

orki

ngM

onth

lyin

com

ex

2(d

f5

3)

0.82

0.41

*13

.75‡

0.95

0.59

*†2.

76

1.08

0.66

16.7

6‡

0.56

0.60

3.03

1.39

0.70

3.78

0.54

†

0.82

3.63

0.84

0.40

*16

.43‡

1.18

0.54

*3.

64

0.68

*

0.71

*4.

91

0.75

0.77

0.86

.N

T$1

9,99

9g

,N

T$5

,000

NT

$5,0

00–N

T$9

,999

NT

$10,

000–

NT

$19,

999

1.38

*1.

73*

1.55

*

0.91

†0.

87†

1.09

†

2.24

0.07

*1.

11

6.25

5.57

†2.

04

1.30

0.85

1.05

1.47

0.83

1.05

1.41

*1.

85*

1.56

*

0.91

†0.

85†

1.13

1.54

*1.

311.

18

1.18

1.16

1.17


TA

BL

E5.

CO

NT

INU

ED

Exp

lana

tory

Var

iabl

es

Any

Hel

p

Sons

Dau

ghte

rs

AD

Lsc

Sons

Dau

ghte

rs

IAD

Lsd

Sons

Dau

ghte

rs

Fina

ncia

lSu

ppor

t

Sons

Dau

ghte

rs

Mat

eria

lSu

ppor

t

Sons

Dau

ghte

rs

Are

aof

resi

denc

ex

2(d

f5

2)11

.37‡

36.1

5‡4.

626.

07‡

2.94

6.35

‡12

.69‡

13.7

5‡14

.73‡

41.3

4‡R

ural

City

Urb

an0.

67*

0.87

0.76

*0.

52*†

3.18

2.95

*1.

403.

21*

0.83

0.71

1.79

*†1.

72*†

0.67

*0.

950.

790.

64*†

0.62

*0.

66*

0.57

*0.

45*†

N Log

likel

ihoo

d12

,166

27,

067.

5512

,166

227

1.26

12,1

662

1,75

6.72

12,1

662

6,49

0.87

12,1

662

5,72

8.56

Not

e:T

heom

itted

cate

gory

for

each

vari

able

issh

own

inbo

ldita

lics.

a Ase

ries

ofdu

mm

yva

riab

les

indi

catin

gm

issi

ngin

form

atio

non

child

’sag

e,em

ploy

men

tst

atus

,an

dm

arita

lst

atus

,an

dpa

rent

’sed

ucat

ion,

ethn

icity

,hea

lthst

atus

,mon

thly

inco

me,

and

prop

erty

divi

sion

are

incl

uded

inth

em

odel

.b W

eex

amin

eddi

ffer

ent

cuto

ffpo

ints

for

child

ren’

sag

ean

dfo

und

that

child

ren

who

are

age

20or

youn

ger

are

sign

ifica

ntly

diff

eren

tfr

omol

der

child

ren

inth

eir

supp

ort

beha

vior

,bu

tam

ong

child

ren

who

are

olde

rth

an20

,th

ere

islit

tleva

riat

ion

byag

e.c A

DL

s(a

ctiv

ities

ofda

ilyliv

ing)

incl

ude

bath

ing,

dres

sing

,an

dm

aint

aini

ngto

ilet

func

tions

.d I

AD

Ls

(ins

trum

enta

lac

tiviti

esof

daily

livin

g)in

clud

esh

oppi

ng,

mea

lpr

epar

atio

n,tr

ansp

orta

tion,

and

man

agin

gfin

ance

s.e T

heva

riab

leis

drop

ped

from

the

regr

essi

onbe

caus

eno

neof

the

adul

tch

ildre

nin

the

cate

gory

prov

ide

help

toth

eir

pare

nts.

f The

who

lese

tof

vari

able

sis

drop

ped

beca

use

none

ofth

ech

ildre

nin

the

refe

renc

egr

oup

prov

ide

help

toth

eir

pare

nts.

g NT

$in

dica

tes

New

Taiw

anes

edo

llars

.In

1989

,on

eU

.S.

dolla

rw

aseq

ual

to26

.17

NT

$.*p

,.0

5ba

sed

ontw

o-ta

iled

tte

sts.

†p,

.05

base

don

two-

taile

dt

test

sfo

rth

ege

nder

diff

eren

cebe

twee

nth

eco

effic

ient

s.‡p

,.0

5ba

sed

onx

2te

sts

for

the

entir

eva

riab

lew

hen

ther

eis

mor

eth

anon

eca

tego

ry.

without a brother provides parents help withIADLs and financial support are about 10 times(OR 5 9.84) and five times (OR 5 4.86), respec-tively, the corresponding odds for a daughter withthree or more brothers (i.e., the reference catego-ry). This result is consistent with the finding ofLee et al. (1994) that sons are usually the pre-ferred providers and that sisters substitute whenno brother is available. Not surprisingly, the ex-istence of sisters does not affect the likelihood thata son provides support (shown in the panel Num-ber of Sisters). However, a daughter without a sis-ter is much more likely to assist parents withIADLs than a daughter with three or more sisters(OR 5 2.45), presumably because of a sharing ofresponsibilities among daughters.

We observe only weak support for the com-peting commitment explanation. The results showthat working children are not less likely than non-working children to provide parental support(shown in the panel Employment Status). Contraryto the expectation that unmarried children wouldbe especially likely to provide assistance, we findthat never-married sons are less likely than theirmarried counterparts to provide help with IADLs,financial support, and material assistance (OR 50.62, 0.80, and 0.59, respectively). However, con-sistent with this explanation, never-marrieddaughters are more apt to provide parents helpwith ADLs, IADLs, and financial support thanmarried daughters (OR 5 14.86, 6.26, and 2.08,respectively), and formerly married daughters aremore likely than married daughters to providehelp with ADLs and IADLs (OR 5 6.65 and 3.87,respectively).

Because of large gender differentials in edu-cational attainment and the importance of brothersfor whether daughters provide parental support,we evaluate the external resources explanation byincluding a three-way interaction among chil-dren’s education, gender, and the presence of abrother (shown in the panel Interactions). To fa-cilitate interpretation of this higher order interac-tion term, we compute odds ratios for each com-bination of the relevant variables, as shown inTable 6. Contrary to the supposition, the interac-tion effects suggest that children with the mostschooling are not more likely than the least edu-cated siblings to provide financial support, andchildren with the least education are not morelikely than the most educated siblings to providehelp with ADLs and IADLs. The pattern remainsthe same regardless of the presence of a brother.

The estimates in Tables 5 and 6 provide only


TABLE 6. ESTIMATED ODDS RATIOS FOR THE INTERACTION TERMS INCLUDED IN THE LOGISTIC MODELS IN TABLE 5

Any Help ADLsa IADLsb

FinancialSupport

MaterialSupport

DaughtersNo brother, highest educationNo brother, middle educationNo brother, lowest educationNo brother, same education or only child

2.8720.91*2.731.62

10.85c

10.7211.52

6.499.259.845.37

5.2642.39*4.861.79*

1.437.251.521.13

$1 brother, highest education$1 brother, middle education$1 brother, lowest education$1 brother, same education or only child

2.552.332.482.46

2.311.551.533.04

2.402.572.732.48

2.902.172.443.10

2.282.302.302.42

SonsNo brother, highest educationNo brother, middle educationNo brother, lowest educationNo brother, same education or only child

1.772.801.260.76

1.32c

3.676.03

2.411.222.452.41

1.783.561.180.78

1.651.581.531.76

$1 brother, highest education$1 brother, middle education$1 brother, lowest education$1 brother, same education or only child

2.802.121.932.20

1.533.271.583.29

2.601.772.431.80

2.902.351.912.22

2.422.182.162.46

Note: Calculations are based on the models shown in Table 5.aADLs (activities of daily living) include bathing, dressing, and maintaining toilet functions. bIADLs (instrumental activ-

ities of daily living) include shopping, meal preparation, transportation, and managing finances. cThe interaction is notestimated because none of the adult children in the category (no brother, middle education) provide ADL help to theirparents.

*Indicates that the interaction term is significant relative to no brother, lowest education for the same gender at p , .05.

weak support for the reciprocity explanation.Children with higher education are not more likelyto provide support than those with lower educa-tion, except that sons receiving the highest edu-cation in a family are more likely to provide par-ents financial support than their siblings withlowest education (OR 5 1.52, shown in the panelRelative Education). Sons are more likely to pro-vide ADL support and material support to parentswho have already divided their property than toparents who have not done so, but property divi-sion does not have a significant effect on daugh-ters’ provision of support (shown in the panelProperty Division).

Data for evaluating the one remaining propo-sition—gender ideology—come from the top pan-el of Table 3. The frequencies of type of supportby the gender of the child are only partly consis-tent with the anticipated gendered division of sup-port provision. Overall, sons are almost twice aslikely to provide help as daughters. However, al-though sons are much more likely than daughtersto provide financial support and slightly morelikely to provide help with IADLs, they are aboutas likely as daughters to provide assistance withADLs and material support.

Parents’ Resources and Needs

In Table 5, the estimates also provide some in-sights into how parental characteristics and needsaffect children’s provision of support. Demo-graphic characteristics of parents are importantdeterminants of support provision by sons anddaughters. Sons are more likely to provide ADLsupport to their fathers than to their mothers (OR5 2.29), but the parent’s gender is unrelated tothe likelihood that daughters provide parentalcare. Sons are more apt to provide help withADLs and material support to older as comparedwith younger parents (shown in the panel Age)and to give financial support to parents withoutany formal education as compared with more ed-ucated parents (shown in the panel Education).The estimates for ethnicity indicate that sons ofMainlanders are less likely than sons of Fukieneseor Hakka to provide monetary or material support,and daughters of Mainlanders are less likely thantheir counterparts to provide help with IADLs andmaterial support (shown in the panel Ethnicity). Incontrast, daughters of Aborigines or foreign-bornparents are more likely to provide help withIADLs and financial support as compared with


daughters whose parents are of other ethnic ori-gins.

Parental need is another determining factorregarding support provision by children. Daugh-ters are less likely to assist with ADL or IADLactivities and to give financial support to parentswho live with a spouse in contrast to parentswithout a spouse present (shown in the panelMarital Status). Daughters are less likely to pro-vide help with IADLs but more likely to providematerial support to parents in better health(shown in the panel Health Status). Not sur-prisingly, children are more apt to assist withADL activities when parents have greater func-tional difficulty (shown in the panel DifficultiesWith Activities). Children are also more likelyto provide financial support to parents who arenot working (shown in the panel Working Sta-tus) and parents with low incomes (shown in thepanel Monthly Income) compared with theircounterparts. Finally, children are more likelyto provide financial and material support to par-ents living in rural as compared with urban ar-eas, but children are less likely to provide helpwith ADLs and IADLs for rural parents.

Inclusion of Help Provided by the Spouseof a Married Child

Although it would be informative to examine theextent to which the spouses of children provideassistance to their parents-in-law, we are unableto identify spouses in these data. However, wecan obtain some notion of the potential sensitiv-ity of our estimates to the inclusion of informa-tion from spouses by examining families withonly one married son and one daughter-in-lawand families with only one married daughter andone son-in-law. (These families may have un-married siblings in the same household.) It seemsplausible that, for the vast majority of such fam-ilies, the married child and the child-in-law forma married couple. Of these families with onemarried child and one child-in-law (data notshown), 92% of the older parents report receiv-ing any help from a married son or a daughter-in-law compared with 80% of the respondentsreporting receipt of any support from only a mar-ried son. The biggest difference pertains to sup-port for IADLs: About 70% of the parents re-ceive support with IADLs from the son ordaughter-in-law as opposed to only 13% of theparents receiving this support from the marriedson alone. The effects on the estimates of in-

cluding help from a son-in-law are much smaller(data not shown). The results suggest that the di-vision of familial tasks may be gender-specific,with married sons mainly responsible for finan-cial support and their spouses for assistance withIADLs.

DISCUSSION

This paper examines the patterns of four types ofsupport provided by adult children in Taiwan andidentifies the factors that are most important indetermining children’s provision of support totheir older parents. Although most explanationsabout intergenerational transfers have been eval-uated in the United States, Taiwan is an interestingsetting for this research because it differs from theUnited States in several important ways. First, theaverage number of children per family among old-er parents is much larger in Taiwan. Second, co-residence is an important form of fulfilling filialresponsibility in Taiwan but not in the UnitedStates where adult children rarely live with theirolder parents. Finally, whereas daughters are ma-jor caregivers in the United States, sons assumethe major responsibility for taking care of theirparents in Taiwan.

We examine five explanations about how thecharacteristics of adult children and their parentsmay affect the provision of parental support. Inview of the fact that these explanations are notmutually exclusive and that we are modeling thesupport behavior of a specific adult without fullinformation on patterns of support across siblingsand over time, it is perhaps not surprising that wefail to obtain strong support for any particular ex-planation. Rather, we find partial support for thenotion of hierarchical compensation—sons gen-erally carry the major responsibility for takingcare of their older parents, and daughters fulfillthe son’s roles when sons are not available—andrelatively little support for the remaining propo-sitions.

This result is not altogether unexpected in viewof the patriarchal nature of Taiwanese society, inwhich sons bear the primary responsibility forcontinuation of the family line. As noted earlier,married sons in Taiwan are expected to live withand support their parents in old age, whereasdaughters’ formal obligations to their own parentsgenerally terminate at marriage. These expecta-tions are manifested in the estimates derived fromthe multivariate models of support provision:Married sons are more likely than unmarried sons


to provide support to their parents, but the reverseis true for daughters. Thus the culture in Taiwa-nese society generates a very different pattern ofsupport provision from that found in Western so-cieties where the rules of reciprocity and negoti-ation generally dominate social interaction.

Our study has several limitations. One short-coming is that because of a lack of informationand the potential complexity of the analysis, thestatistical models do not account for current orprevious support shared among siblings or multi-ple forms of support given at one time. An addi-tional limitation is that information regarding chil-dren’s provision of support is obtained from theparent. Parents may recall their interactions withsome children better than those with other chil-dren. Another drawback is our inability to identifycouples and hence spouses of the adult children.Our results based on a sample of families withonly one married son and one daughter-in-lawstrongly suggest that married sons are typicallyresponsible for different types of support than aretheir wives. Thus if we were able to identify cou-ples we might find stronger support for a divisionof assistance based on gender ideology. A fourthlimitation is our inability to identify grandchil-dren, a constraint that prevents us from viewinggrandchildren as a competing responsibility thatmay reduce the flow of support from adult chil-dren to parents. An additional constraint is the ab-sence of detailed information about children’s em-ployment, such as the flexibility of the workschedule, which would also permit us to explorethe competing commitment explanation more ful-ly. Finally, we have only one indirect measure ofchildren’s negotiation power; that is, their relativeeducational attainment. Additional measures, suchas children’s income and social networks, mightprovide a stronger test of the external resourcesexplanation.

Nevertheless, this study provides two impor-tant insights for aging research in Taiwan. First,researchers need to be cautious about adoptingtheories developed in Western societies in under-standing Asian ones. Second, research based on asingle parent-child dyad cannot represent a com-prehensive picture of intergenerational support. Inparticular, this approach is inappropriate for study-ing Asian families, which until very recently havebeen considerably larger than Western ones, andfor analyzing types of support that typically in-volve multiple children.

The findings of this study affect both futureresearch and the formulation of social policy in

Taiwan. Future research on intergenerationaltransfers needs to examine how siblings within afamily organize themselves regarding the provi-sion of support to older parents, paying attentionto the substitutability and complementarity of sup-port from sons and daughters. Findings from thisstudy underscore the importance of including chil-dren’s spouses, especially daughters-in-law, in theanalysis and of distinguishing among differenttypes of support.

The rapid demographic changes that have tak-en place in Taiwan over the past few decades, par-ticularly declines in fertility and high rates of out-migration by younger generations, are increasingthe likelihood that today’s parents will live aloneor have only one child present during their olderyears. In response to population aging and de-clines in coresidence, it is critical for the Taiwa-nese government to provide a stronger safety netfor older adults. The recent experience of the so-cial security system in Japan, however, suggeststhat a stronger safety net may place considerablefinancial pressure on the government (Ogawa &Retherford, 1997). Thus it behooves researchersand policy makers alike to evaluate the extent towhich the private market and the government cansubstitute family networks in providing the req-uisite support for the older population in comingyears.

NOTE

An earlier version of the paper was presented at thePopulation Association of America 2000 Annual Meet-ings, Los Angeles, CA. We gratefully acknowledge thesupport of the Behavioral and Social Research Programof the National Institute on Aging under grants R01-AG16661 and R01-AG16790; the Graduate School ofArts and Sciences, Georgetown University; and theChiang Ching-Kuo Foundation. Computing supportcame from the Office of Population Research at Prince-ton University, which receives core support from theNational Institute of Child Health and Human Devel-opment (P30-HD32030). We thank Jennifer C. Cornmanand German Rodrıguez for their helpful advice andcomments.

REFERENCES

Berk, R. A. (1991). Toward a methodology for meremortals. In P. V. Marsden (Ed.), Sociological meth-odology (Vol. 21, pp. 315–324). Washington, DC:American Sociological Association.

Cantor, M. (1975). Life space and the social supportsystem of the inner city elderly of New York. TheGerontologist, 15, 23–27.

Chang, M.-C. (1999). A longitudinal study on livingarrangements of the elderly in Taiwan. In C. Chen,A. I. Hermalin, S.-C. Hu, & J. P. Smith (Eds.),


Emerging social economic welfare programs for ag-ing in Taiwan in a world context (pp. 43–64). Taipei,Taiwan: Institute of Economics, Academia Sinica.

Chattopadhyay, A., & Marsh, R. (1999). Changes in liv-ing arrangement and familial support for the elderlyin Taiwan: 1963–1991. Journal of Comparative Fam-ily Studies, 30, 523–537.

Checkovich, T., & Stern, S. (2002). Shared caregivingresponsibilities of adult siblings with elderly parents.Journal of Human Resources, 37, 441–478.

Cox, D., & Rank, M. R. (1992). Inter-vivos transfersand intergenerational exchange. Review of Economicsand Statistics, 74, 305–314.

Finley, N. J. (1989). Theories of family labor as appliedto gender differences in caregiving for elderly par-ents. Journal of Marriage and the Family, 51, 79–86.

Gouldner, A. W. (1960). The norm of reciprocity: Apreliminary statement. American Sociological Re-view, 25, 161–178.

Greenhalgh, S. (1985). Sexual stratification: The otherside of ‘‘growth with equity’’ in East Asia. Popula-tion and Development Review, 11, 265–313.

Hermalin, A. I., Liu, P. K. C., & Freedman, D. S. (1994).The social and economic transformation of Taiwan.In A. Thornton & H.-S. Lin (Eds.), Social change andthe family in Taiwan (pp. 49–87). Chicago: Univer-sity of Chicago Press.

Hermalin, A. I., Ofstedal, M. B., & Chang, M.-C.(1996). Types of supports for the aged and their pro-viders in Taiwan. In T. K. Hareven (Ed.), Aging andgenerational relations over the life course: A histor-ical and cross-cultural perspective (pp. 400–437).New York: Walter de Gruyter.

Hermalin, A. I., Ofstedal, M. B., & Chi, L. (1992). Kinavailability of the elderly in Taiwan: Who is availableand where are they? (Comparative Study of the El-derly in Asia Research Report No. 92-18). Ann Ar-bor, MI: University of Michigan.

Hermalin, A. I., Ofstedal, M. B., & Lee, M.-L. (1992).Characteristics of children and intergenerationaltransfers (Comparative Study of the Elderly in AsiaResearch Report No. 92-21). Ann Arbor, MI: Uni-versity of Michigan.

Horowitz, A. (1985). Sons and daughters as caregiversto older parents: Differences in role performance andconsequences. The Gerontologist, 25, 612–617.

Johnson, C. L. (1983). Dyadic family relations and so-cial support. The Gerontologist, 23, 377–383.

Johnston, J., & DiNardo, J. (1997). Econometric meth-ods. (4th ed.). New York: McGraw-Hill.

Kao, H.-F., & Stuifbergen, A. K. (1999). Family expe-riences related to the decision to institutionalize anelderly member in Taiwan: An exploratory study. So-cial Science and Medicine, 49, 1115–1123.

Lang, A. M., & Brody, E. M. (1983). Characteristics ofmiddle-aged daughters and help to their elderly moth-ers. Journal of Marriage and the Family, 45, 193–202.

Lee, Y.-J., Parish, W. L., & Willis, R. J. (1994). Sons,daughters, and intergenerational support in Taiwan.American Journal of Sociology, 99, 1010–1041.

Li, R. M. (1994). Aging trends—Taiwan. Journal ofCross-Cultural Gerontology, 9, 389–402.

Matthews, S. H., & Rosner, T. T. (1988). Shared filial

responsibility: The family as the primary caregiver.Journal of Marriage and the Family, 50, 185–195.

Matthews, S. H., Werkner, J. E., & Delaney, P. J. (1989).Relative contributions of help by employed and no-nemployed sisters to their elderly parents. Journal ofGerontology, 44, S36–S44.

Montgomery, R. J. V., & Kamo, Y. (1989). Parent careby sons and daughters. In J. A. Mancini (Ed.), Agingparents and adult children (pp. 213–230). Lexington,MA: Lexington Books.

Mosteller, F., & Tukey, J. W. (1977). Data analysis andregression: A second course in statistics. Reading.MA: Addison-Wesley.

Ofstedal, M. B., Knodel, J., & Chayovan, N. (1999).Intergenerational support and gender: A comparisonof four Asian countries. Southeast Asian Journal ofSocial Science, 27, 21–42.

Ogawa, N., & Retherford, R. D. (1997). Shifting costsof caring for the elderly back to families in Japan:Will it work? Population and Development Review,23, 59–94.

Population Reference Bureau. (2001). World PopulationData Sheet of the Population Reference Bureau.Washington, DC: Author.

Ross, C. E. (1987). The division of labor at home. So-cial Forces, 65, 816–833.

Shanas, E. (1979). The family as a social support systemin old age. The Gerontologist, 19, 169–174.

Silverstein, M., Parrott, T. M., & Bengtson, V. L. (1995).Factors that predispose middle-aged sons and daugh-ters to provide social support to older parents. Jour-nal of Marriage and the Family, 57, 465–475.

Spitze, G., & Logan, J. (1990). Sons, daughters, andintergenerational social support. Journal of Marriageand the Family, 52, 420–430.

StataCorp. (1999). Stata statistical software: Release6.0. College Station, TX: Stata Corporation.

Stoller, E. P. (1983). Parental caregiving by adult chil-dren. Journal of Marriage and the Family, 45, 851–858.

Sun, T.-H., & Liu, Y.-H. (1994). Changes in intergen-erational relations in the Chinese family: Taiwan’s ex-perience. In L.-J. Cho & M. Yada (Eds.), Traditionand change in the Asian family (pp. 319–361). Hon-olulu, Hawaii: The East-West Center.

Sung, L.-S. (1981). Property and family division. In M.Ahern & H. Gates (Eds.), The anthropology of Tai-wanese society (pp. 361–378). Stanford, CA: Stan-ford University Press.

Taiwan Provincial Institute of Family Planning and In-stitute of Gerontology, University of Michigan.(1989). 1989 survey of health and living status of theelderly in Taiwan: Questionnaire and survey design(Comparative Study of the Elderly in Asia ResearchReport No. 1). Ann Arbor, MI: University of Michi-gan.

Thibaut, J. W., & Kelley, H. H. (1959). The social psy-chology of groups. New York: Wiley.

Weinstein, M., Sun, T.-H., Chang, M.-C., & Freedman,R. (1990). Household composition, extended kinship,and reproduction in Taiwan: 1965–1985. PopulationStudies, 44, 217–239.

Weinstein, M., Sun, T.-H., Chang, M.-C., & Freedman,R. (1994). Co-residence and other ties linking couplesand their parents. In A. Thornton & H.-S. Lin (Eds.),


Social change and the family in Taiwan (pp. 305–334). Chicago: University of Chicago Press.

Wolf, D. A., Freedman, V., & Soldo, B. J. (1997). Thedivision of family labor: Care for elderly parents.Journal of Gerontology: Social Science, 52B, 102–109.

Wolf, D. A., & Soldo, B. J. (1988). Household com-

position choices of older unmarried women. Demog-raphy, 25, 387–403.

Wolf, M. (1968). The house of Lim: A study of a Chi-nese farm family. New York: Appleton-Century-Crofts.

Wolf, M. (1972). Women and the family in rural Tai-wan. Stanford, CA: Stanford University Press.

Documents

Gender Differences in Adult Children's Support of Their Parents in Taiwan