Resource theory and the distribution of power between husband and wife: a critical evaluation.

Mukhtar, Hanid ; Mujahid-Mukhtar, Eshya

I. INTRODUCTION

Due to its impact upon households' consumption and expenditure patterns, power distribution within a family occupies a position of considerable importance in socio-economic and marketing research. An analysis of family power structure is capable of identifying the 'target group/individual' for any demand management (or marketing) policy. Further, one cannot ignore its importance for sociologists and anthropologists.

This paper is limited to the study of power distribution between husband and wife in the family expenditure decisions, with particular reference to the (in) validity of the resource theory of power distribution) To furnish better insights, the family power structure is analyzed in the context of certain socio-economic and cultural conditions e.g., age, income, education, family structure and ethnicity.

Popular works related to this topic have mainly focused on either of the following theories:

(i) The cultural theory, (2) which suggests that culture pre-determines the roles of husband and wife in the family decision process; and

(ii) The resource theory, (3) which emphasizes the influence of individual's resources e.g., income, education, social status etc., in determining the power balance in favour of the more 'resourceful' spouse.

There is no doubt that culture plays a prominent role in the allocation of power between husband and wife. For centuries, traditional societies have relied upon the patriarchal dominance of family system. However, with the recent advancement of women in the fields of education and employment, certain developed societies can claim a more egalitarian decision-making process in the family.

Acknowledging the validity of both the theories, this paper, however, does not accept the applicability of only one individual theory; rather it attempts to highlight that for most societies in the developing countries, an interactive approach seems more appropriate. Individual resources certainly exert an impact upon the power distribution in the family, but they are constrained within a well-defined set of cultural values.

In the context of Pakistan, as in most traditional societies, it may not be possible to distinguish between the impacts of resource factors and cultural forces. For example, a Pakistani woman is subjected to several socio-cultural constraints which, very often, deny her the free access to certain important 'resources', such as education, employment and income. The dependence of such a woman upon the decision of her relatives e.g., parents, uncles, brothers, husband etc., hinders her from deciding about her own education and employment [Shah (1986)]. This not only deprives her of the resource-based 'competence' but also reduces her ability to contribute in subsequent decision-making, within or outside the family.

Besides ignoring the cultural constraints, the resource theory suffers from other limitations:

(i) Its inability to explain the power enjoyed by children in certain family decisions e.g., recreation etc. In all societies, whether developed or not, this group of apparently 'resource-less' individuals often wield considerable authority in certain family decisions;

(ii) Its ignorance of the fact that decision-making within the household is a sub-set of a larger decision-making process which confronts an individual at any given moment in his/her life cycle. Increased involvement in one aspect of this decision process may reduce the individuals involvement in some other spheres of this process. Extra time and effort spent in decision-making outside the household, which at time follows an increase in individual's resources, may lead to an apparent reduction in his/her decision-making authority within the household; and

(iii) Its failure to account for the fact that certain decisions, including the expenditure decisions, are often made on the basis of inherent advantage rather than acquired 'resource' competence. For example, the day-to-day expenditure decisions are usually made by the women only because they are considered to be 'better shoppers', a term synonymous with better hagglers in countries like Pakistan.

II. METHODOLOGY AND DATA

One of the most important decisions in a family is economic i.e., the expenditures of the household. This study focuses on the relative powers enjoyed by the husband and wife in the expenditure decisions, and attempts to quantify the impact of various resource and socio-cultural factors on this power by employing regression (logit models) technique.

The dependent variables for the series of regressions to be estimated are dummy variables with values 1 (if the decision is taken by the husband/wife) and 0 (if the decision is taken by the spouse).

The following (one each for husband/wife) regression has been estimated:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)

Where:

[POW.sub.i] : Decision taken by husband/wife; i = husband or wife;

FLNO : Number of female children;

MLNO : Number of male children;

HINC : Monthly income of the husband (in rupees);

WINC : Monthly income of the wife (in rupees);

HEDU : Educational level of the husband (years of schooling);

WEDU : Educational level of the wife (years of schooling);

HAGE : Age of the husband (years);

YAGE5 : Dummy variable for young child(ren) in the household; (4)

 = 1 if the youngest child is less than five years of age,
 0 otherwise;

XFAM = 1 if household was an extended family system, 0
 otherwise;

URDU = 1 if head of the household is Urdu speaking, 0 otherwise;

PUNJABI = 1 if the head of the household is Punjabi, 0 otherwise;

SINDHI = 1 if the head of the household is Sindhi, 0 otherwise;

PATHAN = 1 if the head of the household is Pathan, 0 otherwise;
 and

BALUCH = 1 if the head of the household is Baluchi, 0 otherwise.

Each set of these regressions was estimated for both pucci and katchi abadis separately.

As the dependent variable is a dummy variable with values 1 and 0, therefore the Maximum Likelihood Method (MLE) of estimating the logit equation is used. With three possible outcomes of the decision-making process i.e., husband, wife or both husband and wife, being the decision-maker(s), there was a choice whether to apply multimonial logit technique to a given sub-sample once (5) or apply binomial logit technique twice, once each for husband's and wife's power variable. (6) For simplicity, the latter technique was preferred. (7)

The probability (of husband or wife making expenditure decisions in the given household) can be expressed as:

[P.sub.i] = 1/1 + exp (-X [B.sub.i]) ... ... ... (2)

i = husband or wife depending on the dependent variable

Where:

[P.sub.i] = The probability vector of ith person (i.e., husband or wife) making expenditure decisions;

X = Matrix of regressors; and

[B.sub.i] = Regression coefficients.

It should be noted that the (partial) marginal change in probability, [P.sub.i], due to a unit change in an independent, say [x.sub.j], is given by:

[partial derivative] [P.sub.i]/[partial derivative] [X.sub.j] = [B.sub.ij] [P.sub.i] (1 - [P.sub.i] (3)

i.e., for 0 < [P.sub.i] < 1, the partial derivative is just a fraction of the regression coefficient. As [B.sub.ij] is constant, therefore the partial derivative has a maximum (absolute) value (= one fourth of the coefficient) at [P.sub.i] = 0.5 and declines asymptotically on both sides of this probability value and reaches zero at [P.sub.i] equal to 1 or 0.

In accordance with the purpose of the paper the marginal impact of each regressor on [P.sub.i] relative to its impact on [P.sub.j] is determined; where [P.sub.j] is either the probability of the spouse (of person i) being the decision-maker for the same sub-sample of households or is the same person (i.e., husband or wife) in a different sub-sample. This could be achieved by either assuming the initial level of [P.sub.i] to be the same as [P.sub.j] and by simply comparing the regression coefficient, or by calculating the (sub) sample proportions [P.sub.i]. and [P.sub.j] with estimated regression coefficients and comparing the partial derivatives. As the selection between the two modes of comparison is subjective and is dictated by the purpose of the study, both techniques were used for illustration and explanation of regression results.

Finally, as the data provide only three possible outcomes of the decision-making process i.e., the husband or wife or both being the decision-maker(s) and if POWh, POWw and POWj represent husband's, wife's and joint power respectively, then:

POWh + POWw + POWj = 1 ... ... ... ... (4)

therefore,

[partial derivative] POWh/[partial derivative] [X.sub.i] + [partial derivative] POWw/[partial derivative] [X.sub.i] + [partial derivative] POWj/[partial derivative] [X.sub.i] = 0 ... ... ... (5)

Similar relationships (Equations 4 and 5) hold for the expected powers i.e., probabilities, which imply that all three 'powers' (probabilities) cannot increase (or decrease) with a unit change in any variable, [X.sub.i]. If the first term in Equation 5 is positive (negative) and the second term negative (positive), then the third term will be negative or positive depending upon the relative magnitude of the first two terms. This should be kept in mind while interpreting the results of the regressions.

The study is based upon household data for the Karachi metropolitan area. Given the ample opportunities it provides for gainful employment, Karachi has, of recent, been the centre of attraction for people from all over the country. This makes this data set fairly representative of the national population yet it clearly suffers from an urban bias. In an attempt to reduce this bias, data on households living in 'katchi abadis' (non-regularised localities) were included in the sample.

III. RESULTS AND CONCLUSIONS

The results obtained from the estimations of the regressions are presented in Table 1. In general, it is interesting to observe, that in pucci abadis, contrary to the resource theory arguments, the power of the wife in expenditure decisions is positively influenced not only by improvements in her own income (WING) and education (WEDU), but also by those in her husband's income (HINC) and his education (HEDU). Further, it may be noted that, generally, these variables exert a negative impact on the power of the husband. In the katchi abadis, wife's income and husband's education tend to redistribute power in favour of the husband.

The significant, and at times with the opposite sign, effects of resource variables indicate a deficiency in both the theories i.e., culture and resource, of power distribution. This seems to suggest that although culture provides the husband with a large initial share (57.3 percent and 66.4 percent in pucci and katchi abadis respectively) in the distribution of decision-making (regarding household expenditures) but the wife's (husband's) power deviates positively (negatively) around her resource endowment. This is largely due to the reason that the wife's income is usually considered to be her own asset, out of which she can spend according to her own will. Also, her ability to work outside the house gives her the experience of dealing with the outside world and thus she is considered capable of taking expenditure decisions.

On the other hand, an increase in husband's income is usually accompanied by an increase in demand for his time and effort outside the house, leaving him with less time to concentrate on decision processes within the household. This leads to an increase in the power of the wife at the expense of the husband's power, which is being exercised elsewhere.

To highlight the impact of various variables on all three probabilities i.e., probability of husband or wife or both being the decision-maker(s), partial derivatives of the probability functions (Equation 2) for both husband and wife are calculated at the mean probability level (Table 2). The impact of these variables on the probability of joint decision-making is determined by using Equation 5. These derivatives indicate that joint decision-making is adversely affected in the event of a small child i.e., of less than 5 years of age. This could be attributed to a reduction in communication between the spouses as the child demands more of parent's (especially that of the mother's) time. No such explanation could be provided to the negative partial derivative of number of male children in the family and could only be explained as a cultural peculiarity.

In both pucci and katchi abadis, number of female children seems to have the strongest positive impact on joint decision-making followed by education (of both husband and wife). The extended family system has a strong positive impact on joint decision-making in the katchi abadis, while in the pucci abadis its impact is negative.

With respect to the socio-cultural variables in pucci abadis, the power of the wife increases with the number of female children (FLNO) but declines with the number of male children (MLNO); for the power of the husband, the reverse pattern is obtained. One possible reason for this could be that with more daughters (sons), the family demand pattern tilts in favour of goods purchased for female (male) consumption e.g., clothes, shoes, etc., and the wife (husband) is the better judge for such expenses. On the other hand, in katchi abadis, the power of the wife increases (and that of the husband decreases) with increase in the number of male children. As expected, a young child (age less than 5 years) in the family lowers the wife's power in both the pucci and katchi abadis, the extended family system tends to increase the wife's power (and decrease husband's power) in the pucci abadis. For the katchi abadis, however, the family system does not appear to be an important variable.

The ethnicity dummies indicate that while in the Urdu-speaking and Punjabi families the family power structure is not significantly different from the excluded category ('other ethnic groups'), the Pathan and Baluch families are much more conservative in family power distribution with husbands enjoying more (and the wives having less) power compared to their counterparts in the excluded category.

In conclusion, one may state that empirical evidence for Pakistan reveals that neither cultural nor resource theory alone is capable of explaining the power distribution in the Pakistani society. In fact, factors pertaining to both these theories, along with a set of other socio-economic variables, need to be included for a meaningful analysis of household power distribution,

There is no denying that culture plays an important role in defining the household power structure. This is evident from the high percentage of male decisions. However, contrary to the culturalist beliefs, individual resources, such as education and income, does influence family power structure significantly. Their impact, however, is not in conformity with the predictions of the resource theory.

Comments on "Resource Theory and the Distribution of Power between Husband and Wife: A Critical Evaluation"

Basing the analysis on the response to a question relating to the pattern of decision-making with respect to family expenditure from a sample of households from Karachi metropolitan area, the authors test the validity of the cultural approach as well as the resource theory in the distribution of power between spouses in Pakistan. Notwithstanding the importance of the topic, the way the authors have handled the methodological issues cast some doubts on the robustness of the derived conclusions. My comments relate to the discussion of some of the methodological issues and the impact that an improper handling of the issues may have had on the major conclusions of the study.

First, the approximation of power structure with the decision-making power makes the measurements of the phenomenon easy but may not be very helpful in properly understanding the various dimensions of the power structure. Attributes other than the pattern of decision-making are also important and need to be considered in the study of the prevailing model of power in any study.

Second, the decision-making power measured as it was by the authors from a sole question on family expenditure could have been measured alternatively from responses to some other questions or a set of a number of questions. The pattern of decision-making measured from answers to different questions often varies. It is essential to have an idea about the extent of such variability. If it is large, the validity of cultural and/or resource theory in any particular context can be a function of the type of questions asked and included in the analysis.

Third, the nature of responses to any question are related to the type of respondent. Wives, husbands and their siblings have different perceptions of the decision-making authority and often there are considerable discrepancies between their responses to the same questions. It is not clear from the paper as to who were the respondents in the survey.

Fourth, a query can also be raised as to whether the responses to the question could be gathered more effectively through survey studies or observation techniques often used by anthropologists or by a combination of both methods of analysis. It is often thought that it is easy to lie in response to questions than to have a put on job with respect to behaviour. If so, observation techniques could generate more reliable information.

Fifth, it is no wonder that using regression analysis the authors have come to the conclusion that both culture as well as command over the resources are the major determinants of allocating the decision-making powers between the two sexes in Pakistan. Regression equations unless derived from a theoretical model hardly throw up meaningful information on the nature of the explanatory factors in any empirical setting.

Sixth, notwithstanding the importance of the topic, the way it has been handled appears therefore to be very simplistic. To begin with one should have presented a good deal of discussion regarding the methodology wherein a dependent variable, i.e., "who decides what" has to be explained. A concrete example may explain this. The spending pattern of the households is circumscribed by the availability of the resources. Who goes to market depends on the nature of the market and the cultural milieu of the household. It is not clear from the paper as to what extent the decisions reported to be made by females or males were in fact individual decisions. In a study like this, one should be very careful about the biases introduced by the response error. In the sphere of the household it is not always true that either the female decides or the male decides. There can be a consensus at the same time and there may be a good deal of discussion among both husband and wife. Correspondingly, some of the results, such as the effect of the resource variables may be due to the misspecification and misunderstanding of the responses. In addition, it must be pointed out that decision-making pattern varies across different social-economic groups. It would have been better had the authors concentrated on a single strata and tried to understand the decision-making pattern.

Seventh, understanding of the relative role played by different members of the family in expenditure decisions would need to be supplemented with the understanding of the interrelationship between distribution of decision-making power and the intrahousehold allocation of food.

In conclusion, it needs to be emphasized that future work in this important area in Pakistan should concentrate in the evolution of a theoretical framework to guide the multivariate econometric analysis of causal factors and on the collection of valid data on different dimensions of the power structure.

Sarfraz Khan Qureshi

Pakistan Institute of Development Economics, Islamabad.

REFERENCES

Blood, R. O., and D. M. Wolfe (1960)Husbands and Wives: The Dynamics of Married Living. New York: The Free Press.

Conklin, George H. (1981) Cultural Determinants of Power for Women within the Family: A Neglected Aspect of Family Research. In G. Kurian and R. Ghosh (eds) Women in the Family and the Economics." An International Comparative Survey. Westport and London: Greenwood Press.

Foss, D., and M. Straus (1977) Culture, Crisis and Creativity of Families in Bombay, San Juan and Minneapolis. In Lenero-Otero (ed)Beyond the Nuclear Family Model: Cross Cultural Perspectives. Beverly Hills and London: Sage Publications.

Fox, G. L. (1973) Another Look at Comparative Resources Model: Assessing the Balance of Power in Turkish Families. Journal of Marriage and Family.

Kim, Kyong-Dong (1964) A Study of the Confucian Values in Korea: Content Analysis of Children's Textbooks (in Korean). Eulyoo: Seoul.

Kim, O. J. L., and K. Kim (1981) A Casual Interpretation of the Effect on Mother's Education and Employment Status on Parental Decision-making Role Patterns in the Korean Family. In G. Kurian and R. Ghosh (eds) Women in the Family and the Economics: An International Comparative Survey. Westport and London: Greenwood Press.

Kolenda, Pauline (1967) Regional Differences in Indian Family Structure. In Robert Crane (ed) Regions at Regionalism in South Asian Studies: An Exploratory Study. Durham, N. C.: Monograph Number of Five of the Duke University Programme Southern Asia.

Lupri, E. (1969) Contemporary Authority Patterns in the West Ger Family: A Study in Cross National Validation. Journal of Marriage and Family.

Mukhtar, H. (1985) Essays in Consumer Behaviour: Evidence from Pakistani Data. Unpublished Ph.D. Dissertation. Boston University, Boston, USA.

Oppang, C. (1970) Marital Power and Resources: An Urban African Experience. Journal of Marriage and Family.

Rodman, H. (1967) Marital Power in France, Greece, Yugoslavia and the United States. Journal of Marriage and Family.

Safilios-Rotschild, C. (1969) Family Sociology or Wives' Family Sociology? A Cross-cultural Examination of Decision-making. Journal of Marriage and Family.

Shah, N. (1986) Pakistani Women: A Socioeconomic and Demographic Profile Islamabad: Pakistan Institute of Development Economics, and Honolulu, Hawaii: The East-West Population Institute, East-West Center.

Straus, M., and D. Winkelman (1969) Social Class, Fertility and Authority in Nuclear and Joint Households Bombay. Journal of Asian and African Studies.

(1) The present study is a part of a larger study conducted at the Applied Economics Research Centre, University of Karachi.

(2) See, for example, Kim (1964); Kim and Kim (1981).

(3) See Blood and Wolf (1960); Foss and Straus (1977); Straus and Winkelman (1969); Conklin (1981); Fox (1973); Kolenda (1967); Lupri (1969); Mukhtar (1985), Oppang (1970); Rodman (1967) and Safilios-Rotschild (1969).

(4) Wife's (and youngest child's) age were strongly correlated with the age of the husband (HAGE) and therefore were not included in the regression.

(5) The dependent variable taking the value '0, 1 or 2 for (say) when decisions are made jointly, by husband or by wife respectively.

(6) The husband's (wife's) power variable is defined as a dummy variable which takes a value 1 if husband (wife) is the decision-maker and value 0 otherwise.

(7) The additional benefit in applying binomial logit to husband's and wife's power variable separately is that the error term in the two equations can have a Weibul distribution with different variances, a much weaker assumption than required for multinomial Weibul distribution.

HANID MUKHTAR and ESHYA MUJAHID-MUKHTAR *

* The authors are Senior Research Economists at the Applied Economics Research Centre, University of Karachi.

Table 1
Estimated Logit Equations for Husbands' and Wives' Power

 Pucci Abadis

 POWw POWh

Constant -1.516760 0.9918870
(T-Statistics) (-6.250) (4.375)

No. of Female Children 0.0524249 -0.099969
(T-Statistics) (1.607) (3.233)

No. of Male Children -0.0260130 0.0301383
(T-Statistics) (1.607) (1.175)

Husband's Income 0.0000380 -0.000050
(T-Statistics) (3.116) (-3.744)

Wife's Income 0.0001003 -0.000272
(T-Statistics) -1.126000 (-2.694)

Husband's Education 0.0229070 -0.030421
(T-Statistics) (2.376) (-3.371)

Wife's Education 0.0102102 -0.015601
(T-Statistics) (0.988) (-1.596)

Husband's Age 0.006583 -0.006978
(T-Statistics) (1.676) (-1.889)

Dummy for Youngest Child -0.1489430 0.1790550
(T-Statistics) (-1.927) (1.662)

Extended Family (Dummy) 0.2807130 -0.219821
(T-Statistics) (2.927) (-2.402)

Urdu Speaking (Dummy) 0.0053958 0.1275790
(T-Statistics) (0.041) (1.032)

Punjabi (Dummy) 0.0277761 0.1304120
(T-Statistics) (0.168) (0.844)

Sindhi (Dummy) -0.225819 0.3594780
(T-Statistics) (-0.865) (1.503)

Pathan (Dummy) -0.501441 0.9577970
(T-Statistics) (-1.836) (3.700)

Baluch (Dummy) -1.2712300 1.6025800
(T-Statistics) (-3.043) (3.268)

Chi-Squared Statistics 84.848 168.09

 Katchi Abadis

 POWw POWh

Constant -1.206360 1.163470
(T-Statistics) (-4.435) (4.464)

No. of Female Children 0.105150 -0.105154
(T-Statistics) (3.364) (3.489)

No. of Male Children 0.057027 -0.050913
(T-Statistics) (2.011) (-1.871)

Husband's Income 0.000089 -0.000095
(T-Statistics) (2.362) (-2.609)

Wife's Income -0.000155 0.000143
(T-Statistics) (-2.079) (2.087)

Husband's Education -0.024376 0.011714
(T-Statistics) (-2.01) (1.025)

Wife's Education 0.067163 -0.061122
(T-Statistics) (3.950) (-3.735)

Husband's Age 0.003168 -0.005144
(T-Statistics) (0.690) (-1.179)

Dummy for Youngest Child -0.257209 0.305368
(T-Statistics) (-2.211) (2.740)

Extended Family (Dummy) -0.109480 0.040277
(T-Statistics) (-0.994) (0.386)

Urdu Speaking (Dummy) 0.0730350 -0.206332
(T-Statistics) (0.498) (0.386)

Punjabi (Dummy) -0.158467 -0.077361
(T-Statistics) (-0.945) (-0.483)

Sindhi (Dummy) -0.769790 0.669625
(T-Statistics) (-3.224) (2.991)

Pathan (Dummy) -0.721592 0.753539
(T-Statistics) (-3.535) (3.820)

Baluch (Dummy) -0.128788 0.148923
(T-Statistics) (-0.656) (0.780)

Chi-Squared Statistics 86.617 105.57

Table 2
Partial Derivatives of the Probability Function

 Pucci Abadi

 Husband Wife Joint

FLNO -0.02450240 0.01100923 0.01349317
MLNO 0.00738689 -0.00546273 -0.00192416
HINC -0.00001225 0.00000798 0.00000427
WINC -0.00006666 0.00002106 0.00004560
HEDU -0.00745618 0.00481047 0.00264571
WEDU -0.00382380 0.00214414 0.00167966
HAGE -0.00171030 0.00138249 0.00032781
YAGE * 0.04388638 -0.03127803 -0.01260835
XFAM * -0.05387813 0.05894973 -0.00507159
URDU * 0.03126961 0.00113311 -0.03240273
PUNJ * 0.03196398 0.00583298 -0.03779696
SIND * 0.08810806 -0.04742200 -0.04068606
PATN * 0.23475600 -0.10530260 -0.12945340
BLCH * 0.39279230 -0.26695830 -0.12583400

 Katchi Abadi

 Husband Wife Joint

FLNO -0.02359656 0.02119824 0.00239830
MLNO -0.01424880 0.01149664 -0.00007176
HINC -0.00002131 0.00001794 0.00000335
WINC 0.00003208 -0.00003124 -0.00000084
HEDU 0.00262862 -0.00491420 0.00228558
WEDU -0.01371578 0.01354006 0.00017571
HAGE -0.00115431 0.00063866 0.00051564
YAGE * 0.06852458 -0.05185334 -0.01667125
XFAM * 0.00903816 -0.02207117 0.01303301
URDU * -0.04630090 0.01472386 0.03157704
PUNJ * -0.01735981 -0:03194695 0.04930676
SIND * 0.15026380 -0.15518970 0.00492581
PATN * 0.16909420 -0.14547300 -0.02362122
BLCH * 0.03341832 -0.02596366 -0.00745466

The figures are not the partial derivatives but the difference in
power (probability) between the given ethnic group and the excluded
group, all other variables being the same.