CORRECTING THE UNDERESTIMATED FREQUENCY OF THE HEAD-OF-HOUSEXILD EXPERIENCE FOR WOQMN FARMERS
Department of Anthropology University of Florida
Paper presented at the Conference of Gender Issues in Farming Systems Research and Extension that was organized by the Women in Agriculture Program, University of Florida, from 26 February to 1 March 1986.
Women contribute to agricultural production in many ways. These contributions are being increasingly recognized and documented) as attested by many papers at this Conference of Gender Issues in Fanning Systems Research and Extension. One dimension of women as farmers is still being disputed; that is the extent to which women make independent decisions about their farming systems, especially when the wcmen are wives and their husbands are in residence. When women are heads of their own households it is generally accepted that the women make independent decisions. Because the frequency ith which women act as heads of households is a widely accepted indicator of women as decisionmakers, this paper focuses on how that frequency is determined and suggests that current methods underestimate the significance of the head-of-household experience for women farmers. Corrected frequencies have important implications for extension and credit policies and practices. The conceptual and methodological changes that are suggested in this paper also illuminate some of the discrepancies between how farming systems research and extension is described in theory and applied in practice.
Before proceeding, I want to make 2 disclaimers. (1) Focusing on the head-ofhousehold experience does not deny that women make many independent decisions when married and living with their husbands, but the status of head of household is an
important A key indicator. Information about this status and its occupant is usually collected during surveys and the head of households often the only (or preferred) individual to be interviewed during surveys. (2) Focusing on the household as a unit does not mean that the household is the best, or only, social and production unit that fanning systems researchers should be studying. In reality people allocate land, labor, and capital (key"factors of agricultural production) within households and among households, and rny dccizicno tha.t inf!unc a ^u l pr..u.tion.aro c ... .... i.trz and decisions that
influence agricultural production are made because of intrahousehold and suprahousehold relationships. However, this paper has a restricted focus on the household and the head of household because these are currently the accepted standard units of data collection and analysis.
My argument is that the currently accepted way to estimate the frequency of women-headed households underestimates the occurrence and importance of this experience. Current estimates are almost always based on surveys that note status at one time; this is static information. Even when historical changes or trends are desired the same static surveys are usually used to create a time series. Frequencies or relationships are compared that come from different surveys that were conducted at different times. This method may easily result in misrepresenting social dynamics. For instance, this practice estimates the frequency of wonen as household heads at any given time but being a household head need not be a permanent
Becoming a household head may be a phase that many married women experience
during their lifetimes as they divorce, or are widowed, or as their husbands migrate away for shorter or longer periods of time. Surveys could show that one-fifth of all households were headed by women every year. The easily drawn conclusions are that
(1) one-fifth of the women have this statu, of household head and this experience of independence and disadvantage, and (2) four-fifths of women are wives and do not experience being head of household. The missing data, however, refer to the turnover of status, i.e.) the extent to which women change from being mebers
tflChof .headed households to being heads of their own households, or vic~ersa. If the turnover rate is rapid and continual, then many mere than one-fifth of the women will experience being head of household over a period of years.
The number of households headed by women at any given time may not be the
important statistic. What may be much more important from a standpoint of policy, reseat'ch, and extension may be the percentage of adult women who experience being heads of households over a ten year period, or the likelihood that an adult women will experience being head of household during her lifetime. These statistics recognize that more women will be heads of households over time than at any given time. If women-headed households share any characteristics that affect their farming practices and standards of living, then recognizing that more women cycle through this phase increases the importance of working with women who are not now heads of
household but might occupy that decisionmaking status in the future.
Changing the method of measuring' the frequency of women-headed households has irr lications for the importance of ensuring that wives as well as husbands receive training and experience with agricultural innovations and that wives as well as husbands establish credit records and contacts with change agents and institutions. Wives with more training and experience and with better credit records and contacts may be able to maintain higher levels of productivity and living standards when they cycle through the disadvantaged head-of-household phase.
DISADVANTAGES OF WOEN-HEADED HOUSEHOLDS
The farming systems research and social science literatures are increasingly documenting the disadvantages of women-headed households. At minimum, there is the disadvantage of reduced household labor since women-headed households rely in many cases on a single adult, the mnan, while households headed by men usually have available the labor of at least two adults, the husband and wife. Labor shortages and capital shortages are intimately linked because in almost all societies today, and definitely in all agricultural societies today, labor is a comnodity. People are able to earn money through their labor, and money may be used to hire labor for agriculture. Capital is not as short for all wcren-headed households. Often the availability of capital is associated with the existence of an absent husband who still sends remittences to his wife. Women-headed households with remittances from absent husbands may be more similar in terms of capital availability (and in terms
of hiring labor) to men-headed households than to waomen-headed households without remitting husbands.
Another disadvantage common to many women-headed households is reduction in the
land that is cultivated. This may be due to shortages of labor and capital rather than land, that is, the land may still be controlled by the woman but she does not control the recources to cultivate it. Land may actually be reduced for several reasons:
(1) the access to land may be through a larger suprahousehold kinship group that reallocates land away from a household that is reduced in labor ard consumption units; (2) a household containing both a wife and husband may be able to acquire land through both familieswhereas upon divorce or widowhood the woman may have access to land only through her own family; and (3) in many societies, the husband-and wife cultivate land controlled only. by the husband or his :family, and upon divorce or widowhood the woman must return to her family that then has to reallocate its land in order to allow her access to land to cultivate. These reasons all apply in societies in which the-mnarried couple do not have joint freehold title to farmland.
Another disadvantage is the reduced access that nany women household heads
experience in their relationships, or lack of relationships, with governmental and other institutions. Women are often discriminated against by extension agents, who are predominantly men, and by credit programs. This discrimination is sometimes overtly sexual in that male agents do not want to interact in professional ways with women farmers. Discrimination may be based upon the othbr characteristics of womenheaded household that were already mentioned: their shortages of labor, land, and capital.
Governmental policies, however, may reduce or eliminate the disadvantages that
are experienced by women-headed households. Just as shortages of labor are tranformed into shortages of capital, so may the availability of capital be to hire labor.
Just as access to training) credit, and extension advice is currently restricted because most governments have not made a major issue out tf contacting wonen farmers, a change in government policies, for instance, a directive that male agents, who are in most Atries the oneyho work with basic production inputs work with
women farmers would reduce that disadvantage.
The real question is why governments and their extension and credit agencies
should change their policies toward women. Although morality and equity have a place in this argument I personally think that economics and production are more effective. The argument has to be made that women need extension and credit resources in order to make (or continue to make) important contributions to national agricultural production. This argument is being made and may be heard in many papers here at this conference. My own paper is designed to strengthen that argument and extend it. by demonstrating that the head-of-household experience is much more common for women than is I presently recognized.
A ZAMBIAN CASE
Surveys usually note household composition and marital status at the time of
the survey. This synchronic, or ahistoric, data may easily mislead Ob servors into underestimating social dynamics. My first practical awareness of the discrepancy between aggregate statistics and social dynamics occurred in the early 1970s when I was studying socioeconomic change in smallholder ccnnities in northwestern Zambia. In order to learn rates of change in demographic and migratory factors, the same territorial population was surveyed twice. A 100 percent sample of 117 matrilineallybased villages (called matri-villages here) was surveyed in 1971 and again, 12 months later, in 1972. These matri-villages comprised approximately half of one rural settlement stretching for more than a mile along a road.
Both surveys shoWedthe same total population of 1223 people, including adults and chilrenpin the same 117 matri-villages (see Table I). Apparently, this was a stable population to judge by the aggregate statistics. The appearance was illusory.
Since the surveys were part of a larger ethnographic study, personal names and relationships of all the people were collected, and the named individuals were all traced over the year to learn about social dynamics: marriages, divorces, births) deaths, migrations, etc. Although a total of 1223 was present both times, only 1006 people were present for both surveys: 217 people had emigrated in the intervening 12 months, and an equal number immigrated. Thus, only 82 percent of the 1971
population remained in 1972. 1440 people in all were residents during one or both suveys -434 being there for only one survey. If individual names had not been collected and compared, aggregated statistics would have camouiflaged the movement of people and projected a misleading inmge of stability.
There were biological reasons for some instability: 70 of the 434 transitory people represented births into and deaths out of the population. Social and economic reasons were more important. Among these was the changing of marital status by women: 22 percent of the movement (97 cases) was due to women entering and leaving the locality because of marriage, divorce, or widowhood. This underestimated the extent of marital change because the 97 did not include the common case of women who remained within the 117 matrilineallybased villages while changing their marital status. The statistic only notes movement in ar4out of the 117 matri-villages. Unfortunately, my data did not note whether the 97 individual women who changed marital status also changed the status of their own households from men-headed to women-headed, although this was undoubtedly the most common outcome. This cannot be verified, however, because status of household head in 1972 was not collected for those who left the locality in the interim.
To place the 97 when in perspective, there were only 466 women of marriaable age in the locality in 1972 marriageablee age being defined as 15 years of age and older). The high turnover of marital status is common in this matri lineal area of southern Africa. This case does not infer that high rates of marital instability are common around the world. What the case does show is the extent to whibh aggregate statistics and ahistoric data may mislead observors into underestimating the frequency with which women in this area experienced changes in marital status, an important causal factor in changes in head-of-household status.
A ILLAWIAN CASE
Data from Malawi that were collected as part of a farming systems research program show this high women's mobility in a neighboring country. This research experience confirmed my awareness of the discrepancy between actual social dynamics and the estimates of dynamics that are based on aggregated statistics from time series surveys. Before the Malawi data can be presented, some background information is needed about survey procedures in that country.
Much of rural Malawi is incorporated into integrated rural development
projects that are routinely surveyed annually for evaluation purposes. Each project has its own evaluation unit that conducts these annual-surveys of smallholder
agriculture during the cropping season, essentially from October through May. Sampling is both stratified and random. Each project is divided into extension planning areas (EPAs), which were subdivided into enumeration areas (EAs) for the purpose of the 1977 national population census. Each year each project randomly selects a number of its EPAs and within the selected EPAs a number of EAs. Within each of the selected EAs an enumerator collects a list of all households and 20:-25 households
(plus some alternatives) are randomly selected for that years survey. This total procedure results in randomly selected clusters of households. Each year the EPAs, EAs, and households are newly randomly selected. This excessive reliance or trust in random sampling absorbs all survey resources, so no samples are followed longitudinally in order to examine actual social and agricultural dynamics.
The National Sample Survey of Agriculture (NSSA) of 1980-1981 caused a change in this procedure. The NSSA is conducted at approximately 10-year intervals over the entire countryside where there are smallholders and includes nonproject as well as project areas. In the projects the NSSA absorbs the evaluation survey staffs while special staff are hired for the nonproject areas. The 1980-1981 NSSA was elaborate and required project evaluation staff to continue collecting NSSA data through December 1981. This meant that the evaluation staffs had no time in which to select
new random samples for the 1981-1982 cropping season surveys of project areas. Thus,
the same EPAs and EAs were used) as well as half of the 20 NSSA households in each EA. The other 10 households in each EA were newly randomly selected from the NSSA lists of households. This gives analysts the opportunity to examine longitudinally
for two consecutive years some large smallholder populations.
One large project covers much of the Lilongwe plain and extends south and west
of the national capital. This Lilongwe Rural Development Project (LRDP) has been the
object of several studies (Lele 19). The NSSA surveyed 540 households in LRDP or #?7 WV of ?0. I'm he LADW evalmstj,N j7f,4frf .e rveged hal Yof fhOrompte or.o 10 A .o4w'/ds i. e a-I
27 clusters, plus a newly selected 10 households in each cluster. Thus, 270 households
were surveyed for two years in a row in IDP.
Two research units joined forces to conduct a follow-up study of these LRDP
households. One unit was the national farming systems research unit that I
directed -the Adaptive Research Program of the Department of Agricultural Research,
Ministry of Agriculture -and the other unit was the Women in Agricultural Development
Project, which was directed by Dr. Anita Spring. Among the purposes for this study were the unprecedented opportunity to "piggyback" on these two years of survey data
and the desire to demonstrate to other Malawian researchers the advantages of longitudinal analysis and the savings in governmental research resources when
agencies shared survey data and piggybacked on each other's samples.
By comparing the names of individuals and the coded household numbers, 267 of the
potential 270 households were identified that had been surveyed for two years. In
1980-1981, 22 percent of these households (58 of 267) were women-headed (see Table 2).
In 1981-1982) 23 percent (60 of 266) were women-headed. The observor is led to
believe that wcman-headedness is a fairly stable factor and is found in one of
every four or five households. A further conclusion might be that 22-23 percent of
women experience being heads of their households.
Comparing individual households from one year to the next gives another impression. Instead of two households changing (58 to 60)) 17 changed sex of head-lO men-headed becoming women-headed, and 7 vice versa. Thus, one-sixth of the women who were household heads in 1981-1982 became heads only during the preceding year (see Table 2). A total of 67 households (50,10,7) were noted as being headed by women at some time in the slightly more than one year of the two surveys) or 25 percent of the 267 households. Assuming that this change from one year to the next is a normal occ4ence, more than 22.23 percent Of households are headed by women at some time, and more women experience being heads of household than are indicated by the frequency of women heads at any given time.
If this social dynamic is confirmed by testing against other Malawian data, then the woman-head-of-household experience is more important a factor in farming systems research and agricultural policy than is currently recognized. The LRDP data show approximately 3 percent of households shift each year in each direction (woman to nan, and man to woman). More testing and more longitudinal information are needed to distinguish the number of "repeaters", or households that shift sex of household head more than once, the number of "stables", or households that remain headed by the same sex for many years, and the extent and variability in these percentage shifts. If the number of repeaters are low and the number of stables high, then the annual combined shift of 6 percent in both directions means a rapid increase in the cumulative number of women who experience being head of household over time, and this status is accepted as one that entails agricultural (farming systems) decisionmaking.
A subsample of the 267 households in LRDP was chosen for more intensive study during 1982. Seventeen of the 27 EAs were randomly selected, and 6 o4 the 10 resurveyed households in each of the 17 EAs; this created a subsample of 102, of which 101 were successfully located and interviewed approximately six months after
the 1981-1982 survey.
As would be expected, the subsample resembles very closely the larger
sample fran which the subsample was drawn. In the subsample, wanen-headedness appears to be fairly stable from one year to the next (21 and 20 percent) and to appear at approximately the same levels as in the larger sample. Again as in the larger sample, the apparent stability conceals the actual dynamics of shifting household-headedness, which occurs at the annual rate of 2 percent each way (m n to wonanj and vice versa) in the subsample.
The subsample did reveal an unexpectedly large shift in sex of household head between the 1981-1982 cropping season and the resurvey six months later in the dry, nonagricultural season. Only 16 percent of households in the subsample were headed by women during the 1982 dry season survey (see Table 2). In the intervening six months since the 1981-1932 survey, 13 percent of households had changed the sex of household head, most of them (9 percent) changing from women headed to nan headed. These shifts were real, i.e., only one household was a repeater showing more than one change over the three surveys (see Table 3).
The move toward men-headedness probably expressed to a great extent the returning of husbands who were away from their wives during the agricultural season. These
"s~oA 40 ~~~~. ~v e~ 'Ac --w~ i
husbands are probably~Tatistic of wcmen-headedness is stable from one rainy season to the next' but varies wiA81y from the dry season frequency. This seasonal variability was not discovered in the usual surveys conducted during the cropping season. Still to be examined is the extent to which the presence of a resident husband during the dry season affects the agricultural decisionmaking process, or, stated in another ways the extent to which agricultural and farming systems decisions are made during the actual cropping season by the woman.
The additional changes in household-headedness shown in the subsample (see
Tables 2 and 3) mean that although no more than 21 percent of the households were headed by women in any given survey, 27 percent of the households in the subsample had a woman head over the 18 month period from the first survey in 1980-1981 to the
third survey in 1982. The larger sample showed 25 percent of the households with women heads over the 12 month period from the first to second surveys. Both sample and subsample confirm the underestimates of women-headedness given by surveys that only report the frequency at any given time.
The Zambian and Malawian cases document the discrepancy between the frequency of wanomen-headed households at any given time and the cun ulative number of women who experience being the heads of households over time. Shifts from one season to another and one year to another mean that many more wanmen experience heading households during their life mes than are indicated by the current survey methods that only measure the frequency at one time.
Women-headed households are a category of disadvantaged households aw in
terms of labor, land, capital, and governmental services. Correcting the under'iue~ nIf-ha* ke reS oitz~ *df wv %ue,,r cpe,4e~see estimated frequency of the woman-head-of-household experiencethe disadvantages for longer or shorter periods of time. Recognizing that waen-headedness is a phase that many households may pass through as well as a permanent category for a smaller number of households draws our attention also to the ways that the category's disadvantages might be ameliorated by preparing women before they phase into being household heads.
Instead of believing that wives are dependent upon their husbands in men-headed households and do not need any independent or separate instruction or access to services wives may be seen as potentially independent heads of households ae4-de
tbc Mn ..a +--....tontiA.l.Y had f h..... -d who need to receive their
own training and experience with innovations and establish their own credit records
and contacts with extension and credit agents. Gaining these skills and advantages while still in a household headed by someone else means that, if these women cycle through the normally disadvantageous phase of the woman-head-of-household experience
lhc~re wopmch o .i e heffer jire to 'core W;ItN Ve ~.d~i~~4
Ate WJe to COjtipl(e -t mqore poduc.M~e aric4/f(UVe.
The current farming systems research and extension model utilizes a simple household unit that underestimates many social factors. Although, in theory, farming systems are biosocial systems that include social factors) in practice these factors are ignored or treated as parameters while research concentrates on biological and biotechnical relationships. Although farming systems research and extension is supposed to be conducted by multidisciplinary terms, in practice the disciplines and interests that are represented are more technical than social.
Perceiving the head-of-household experience as a phase rather than a permanent status for many wcmen is a conceptual shift that integrates farming systems research and extension, with its agricultural-production-technology focus and reliance on a simple synchronic household unit, and the rich social science literature on life and domestic cycles, marital forms and instability, and the variable forms of distribution of wealth, labor and responsibility between husband and wife. This kind of integration is needed to rescue farming systems research and extension from collapsing into a micro-and locality-specific Green Revolution activity..
The specific argument in this paper cannot be generally applied without
assesing localized variation in several social variables. Frequencies of wcmen headed households at any given time, especially during the cropping or other active seasons; frequencies of shifting from one household head to another, both annually and seasonlly) disadvantages of women-headed households in terms of labor, land, capital, ifnd access to services'and the homogeneity or types of women-headed households--these
variables need to be defined in a locality or region before applying this argument. Matrilineal areas in southern Africa, such as the Zambian and Malawian cases referred to here, have relatively high rates of marital instability and male labor migration, which make this argument particularly forceful.
The argument in this paper on women who are not currently heads of households. These women need to receive training and services as if they were independent decisionmakers. Obviously) women who are currently heads of households also need to
receive training and resources. Although this argument continually focuses on the women who are or will become heads of households, this focus was chosen because the household-head status is a widely accepted indicator of decisionmaking power and responsibility. It is not implied that wcmen only make decisions when they are
household heads. What is argued is that the extent to which women experience being heads of their households, and thereby experience the power and responsibility of independent decisionmaking, is underestimated. and that current one-time survey methods inherently underestimate social dynamics.
TABLE 1 MARITAL CHANGES AS EXPRESSED IN MIGRATION
OF WNOEN IN AND OUT OF RURAL AREA OF NORHWFSTERN
ZAMBIA, 1971 TO 1972
IOPUIATION OF 117 MATRI-VILLAGES 1223 1223
INCLUDING ADULTS AND CHILDREN
PEOPLE PRESENT AT BOTH SURVEYS 1006
EMIGRANTS FROM AREA IN YEAR 217
IMMIGRANTS INIO AREA IN YEAR + 217
TOTAL PEOPLE SURVEYED 1440
REASONS FOR MOVING
IWMEN MARRYING AND JOINING HUSBAND 42
0@EN DIVORCING OR BEING WIDOWED + 55
MARITAL SUBIOTAL 97
BIRTHS AND DEATHS 70
CHILDREN MOVING WITH RELATIVES 131
NOTE: 1972 POPULATION (N=1223) CONTAINED 466 O4EXN OF MAPRIAGABLE
AGE (15 YEARS AND OLDER).
ABE2 CHANGES IN GENDER OF HOUSEHOLD HEAD IN LIIDJNGWE
RURAL DEVELOPMENT PROJECTr MAIAWI, 1980/81 TO 1982
A. LARGER SUBSAMPLE 1980/81 1981/82
lMEN HOUSEIHDLD HEADS TRIAL 58 (22%) 60 (23%)
- UNCHANGED OVER YEAR 50
- CHANGED FROM MEN 10
MEN HOUSEHOLD HEADS TOTALL 209 (78%) 206 (77%)
- UNCHANGED OVER YEAR 199
- CHANGED FIRO WOMEN 7
TOTAL HOUSEHOLDS 267 (100%) 266 (100%)
B. INITNSIVELY SURVEYED SUBSAIX.. 1980/81 1981/82 1982
OMEN HOUSEHOLD HEADS TOTAL 21 (21%) 20 (20%) 16 (16%)
- UNCHANGED OVER YEAR 18 12
- CHANGED FROM MEN 2 4
MEN HOUSEHOLD HEADS 'IUAL 80 (79%) 80 (80%) 85 (1
- UNCHANGED OVER YEAR 78 76
- CHANGED FRO WOMEN 2 9
ITOTAL HOUSEHOLDS 101 (100%) 100 (100%) 101 (101%)
OTE: MISSING 1981/82 DATA ON ONE HOUSEHOLD THAT WAS HEADED BY A OMAN IN
1980/81 AND 1982
NOE: INTERVAL BETWEEN 1980/81 AND 1981/82 WAS 12 MONTHS; DATA COLLECTED
DURING RAINY AGRICULTURAL SEASON (Novenber through April). INTERVAL BETWEEN 1981/82 AND 1982 WAS 6 MONTHS; DATA FOR 1982 COLLECTED DURING
DRY NON-AGRICULTURAL SEASON.
TABLE 3 INDIVIDUAL PATTERNS OF CHANGES OF GENDER OF HOIDUSEHOLD HEADS IN
INTENSIVELY SURVEYED SUBSAMPLE (N= 100-101) OVER THREE SURVEY TIMES
HOUSEHOID IN 1980/81 HOUSEHOLD IN 1981/82 HOUSEHOLD IN 1982
MAN MAN WOMAN N-4 (25%)
MAN OMAN WMAN N=1
(2 CHANGES) MAN WOAN MAN N=1
OMAN WAN M AN N=8 (50c)
WOMAN MAN MAN + N=2
TrAL HOUSEHOLDS CHANGED 16 (100%)
ALTHOUGH 20-21% OF HOUSEHOLDS ARE HEADED BY A WMAN DURING ANY
GIVEN RAINY, AGRICULTURAL SEASON, 2% OF HOUSEHOLDS HAD A WOMAN HEAD OVER 18 MONTHS. THE LARGER SUBSAMPLE CONFIPMS THE INCREASE
IN WOMEN HOUSEHOLD HEADS OVER TIME AND SHOWS A LARGER PERCENTAGE INCREASE OVER 12
MONTHS: 25% OF HOUSEHOLDS (N = 266) HAVE A WO1"AN HEAD OF HOUSEHOLD
IN AT LEAST ONE OF TWO QDNSECUTIVE RAINY SEASONS.