Are Household Production Decisions Cooperative? Evidence on Migration and Milk Sales from Northern Kenya

Transcription

1 Are Househol Prouction Decisions Cooperative? Evience on Migration an Milk Sales from Northern Kenya John McPeak* an Cheryl Doss** Paper to be Presente at the AAEA meetings, August 00 Long Beach, California *Department of Public Aministration, Maxwell School, Syracuse University **Yale Center for International an Area Stuies 34 Hillhouse P.O. Box 0806 New Haven, CT Copyright 00 by John McPeak an Cheryl R. Doss. All Rights Reserve. Reaers may make verbatim copies of this ocument for non-commercial purposes, provie that this copyright notice appears on all such copies. 1

2 Are Househol Prouction Decisions Cooperative? Evience on Migration an Milk Sales from Northern Kenya Abstract: Among the Gabra of Northern Kenya, men ecie where to migrate with the househol hers while women ecie whether to sell milk. We test three moels of househol ecision-making. The results suggest that househol ecisions are conteste. Husbans appear to use migration ecisions to resist their wives access to income. Acknowlegements Financial support for this stuy was provie by an International Preissertation Fellowship from the Social Science Research Council an the American Council of Learne Societies with funs provie by the For Founation, the Mellon Founation, an the Grauate School of the University of Wisconsin-Maison. Work on the manuscript was conucte as a result of the authors collaboration on the Pastoral Risk Management Project of the Global Livestock Collaborative Research Support Program, fune by the Office of Agriculture an Foo Security, Global Bureau, USAID, uner grants DAN-138-G an PCE-G The opinions expresse o not necessarily reflect the views of the U.S. Agency for International Development.

3 Introuction When new opportunities are introuce into societies, the benefits are often conteste. In particular, men an women frequently renegotiate their traitional roles an responsibilities with the avent of new opportunities. Some evience, much of it escriptive an anecotal, suggests that it may not be appropriate to moel househol ecisions regaring the use of these new opportunities as cooperative ecisions. Cooperative moels may overlook the conteste nature of intra-househol ecision making. Unerstaning the nature of such contestation is critical for those who seek to introuce new opportunities to a given society in the name of evelopment. What appears to be a beneficial intervention uner the assumption that househols act cooperatively may instea lea to unforeseen, potentially averse, outcomes if ecisions are conteste. One frequent source of new opportunities arises when market opportunities evelop for goos that have been traitionally prouce an consume within the househol. In most economic analyses, market institutions are treate as a fixe set of rules that guie economic behavior. It is, however, important to recognize that as markets evelop, new rules associate with the market must be reconcile with existing cultural rules. Observe economic behavior may be the result of a negotiation process over which rules, those of the market or those prevailing in the culture, are applicable. In this paper, we investigate intra-househol patterns of ecision making for the Gabra who are nomaic pastoralists in northern Kenya. Over the past thirty years, herers in northern Kenya have seen a rapi growth of milk marketing opportunities. What makes this situation intriguing is that among the Gabra, traitional cultural rules allocate the responsibility to ecie where to locate the househol to the husban, while the management of milk is the wife s 1

4 omain. As livestock-raising in the Gabra area requires frequent migration an milk marketing only takes place in the small market towns of the stuy area, the analysis of these two ecisions allows unique insight into intra-househol negotiation over new market opportunities. Simply put, we investigate how men use their ecision-making power on migration to influence their wife s milk marketing. As we will elaborate below, people in the area escribe four possible reactions by househols to the new opportunities presente by the evelopment of milk markets. The first possibility is that husbans an wives recognize the market opportunities an make joint ecisions on location an milk marketing to maximize househol welfare. We call this the cooperative solution. A secon possibility is that husbans take over the milk marketing an ecie iniviually regaring location an milk marketing. This is a specific case of a cooperative solution. 1 A thir possibility is that husbans continue to make location ecisions without consiering the impact on milk marketing. We call this the traitional solution. The final possibility is that husbans view wives use of milk markets with trepiation, as milk marketing allows wives to expan their control over househol milk to control over cash income. In this case, men may make location ecisions to limit their wife s ability to market milk. We call this the conteste solution. We formally moel these outcomes below an then empirically investigate the pattern of househol ecision-making using panel ata from Gabra pastoral househols. The outline of this stuy is as follows. Section two presents a brief review of the literature on intra-househol ecision-making. Section three escribes the nature of pastoral prouction in the stuy area. Section four presents information on the ata use in this stuy. Section five formalizes the three moels of ecision-making escribe above: the cooperative

5 moel, the traitional moel, an the conteste moel. In section six, results of empirical analysis of househol ecisions are presente. A concluing section iscusses the implications of these finings. Empirical Literature on Cooperative an Noncooperative Outcomes Most of the literature on househol ecision-making assumes that househols act cooperatively. Although the literature on noncooperative moels is expaning, there are still relatively few empirical examples where the outcome is noncooperative. Many empirical stuies that have examine the issue of whether some measure of bargaining power affects househol ecisions simply assume that the outcome is cooperative. Much of the intrahousehol literature focuses on whether or not the househol can appropriately be moele as a single ecision maker or whether the bargaining power of iniviuals within the househol affects outcomes. For example, using ata from Brazil, Thomas (1993) tests whether the istribution of nonlabor income among househol members affects expenitures an fins that nonlabor income controlle by women is associate with a larger share of expeniture on human capital an leisure. Thomas an Chen (1993) fin similar results for Taiwan. In Ghana, Doss (1999) fins that the share of assets owne by women is significant in explaining househol expeniture patterns. Quisumbing an Maluccio (000) test whether assets brought to marriage by each spouse have ifferential effects on househol-level an iniviual-level outcomes in Banglaesh, Ethiopia, Inonesia, an South Africa. The most consistent result that they fin is that when women control relatively higher shares of resources, a greater share of the househol buget is spent on eucation. Other outcomes that have been 3

6 explore are health an eucation outcomes for chilren. All of these examine which cooperative moel of the househol is appropriate. None of these results rejects the notion that the outcome is a cooperative one. Thomas an Chen (1993) explicitly test for Pareto Efficiency in consumption an o not reject it. Similarly, Bourguinon, et al. (1993), use ata on househols in which both aults work full time an in which there is at most once chil an claim that the behavior that they observe is consistent with a Pareto efficient outcome. Thus, iniviual preferences an power may affect the outcomes of househol ecisions, but the ecisions appear to be cooperative outcomes. Two examples of noncooperative outcomes come from stuies that examine prouction ecisions within the househol. Ury (1996) uses etaile agronomic ata from Burkina Faso an fins that crop yiels are ifferent on plots controlle by men from those controlle by women within the same househol in a given year. He also fins that househols coul achieve higher total output by reallocating labor an fertilizer from men s plots to women s plots. Pareto efficiency woul require that marginal prouctivity for an aitional unit of labor or fertilizer be the same across all plots owne by the househol. Thus, he rejects a cooperative outcome. Similarly, Jones (1983) rejecte a cooperative outcome in her stuy of labor allocation following the introuction of irrigate rice prouction in Northern Cameroon. Both men an women continue to grow sorghum after irrigate rice was introuce, even though the returns to labor from rice prouction were higher. Men an women jointly cultivate the rice fiels, whereas sorghum plots were iniviually cultivate. Women receive some compensation for working on rice plots, but the amount of compensation was conteste. Reallocating labor from sorghum to rice woul again have increase total househol prouction. 4

7 This example from Cameroon highlights a theme that is foun frequently in the literature the introuction of new technologies or opportunities often results in the gains being conteste. Descriptions of these contestations are wiesprea in the literature on women in evelopment, but relatively few instances are explicitly moele an teste. Similar to the example of the introuction of irrigate rice in Cameroon is that of the Gambia. Von Braun an Webb (1989) foun that with the introuction of irrigate rice in Cameroon, men took over rice cultivation, isplacing the women who ha traitionally grown rice. In aition, women began growing cotton an grounnuts, which were traitionally men s crops. The introuction of a mechanical maize sheller into a Nigerian village shifte the control of the shelling process from men to women (Lapio 1991). The men respone by contesting the women s right to charge for shelling an eventually some of the men seize the machine. Another way that researchers have examine intrahousehol ecision making is by examining risk sharing within househols. Two stuies suggest that househol members o not fully pool their risk with each other. Dercon an Krishnan (000) fin that poor southern househols in Ethiopia o not engage in complete risk sharing; women in these househols bear the brunt of averse shocks. They reject the collective moel of the househol which imposes Pareto efficiency on allocations. Doss (001) fins that in Ghana shocks to men s an women s incomes have ifferent affects on househol expeniture patterns. These stuies provie support for the iea that iniviual househol members may be concerne about their own long-term access to resources an that membership in a househol is one way, but not the only way, that they seek to ensure this access. Most of the moels look at these issues in a static framework. In a static framework, it is challenging to think about why househols woul not reach cooperative outcomes. However, in 5

8 the longer term, changing the relative positions of iniviuals within the househol will affect the istribution of resources. Although, at this stage, our moel is static in the sense that it oes not cover more than one time perio, the story for why we see noncooperative outcomes is ue to concern over the potential long-term changes within househols that coul result from women gaining control over income. Gabra Pastoral Prouction Gabra are nomaic pastoralists living in northern Kenya an southern Ethiopia. Gabra inhabit an extremely ari an variable environment. Mean rainfall is below 300mm for most of the Gabra rangelans. Rainfall is also highly variable, with a coefficient of annual variation of 58 in Chalbi. Gabra househols share access to their grazing area, an migrate throughout this area in reaction to changing pasture conitions. Gabra househols migrate with their hers of camels, cattle, goats an sheep an rely almost entirely on these animals to meet their househol nees. In Gabra culture, the husban is given the right to ecie when an where to move the househol an the househol her. Such moves can be over extremely long istances. Traitionally, upon the husban s ecision to migrate, the housing materials an all the househol belongings are loae onto camels an move to the new location he has selecte. It is the woman s responsibility to reconstruct the house when they reach the new location an the husban s responsibility to buil new night enclosures for the animals from thorny bushes. They will remain at this site until the husban ecies the time has come to move again. All things insie the hut are uner control of the wife. Gabra symbolism is rich with contrasts between that which is insie the hut (female) with what is outsie the hut (male). This 6

9 is playe out each evening in the ritual surrouning the milking of the her. After the animals return from grazing, they are place in their night enclosures an milke by a esignate milker (women are not allowe to milk camels, nor are sexually active men). The containers full of milk are then taken to where the husban sits outsie the oor of the hut. He inspects the milk, takes a ritual sip, an then passes it through the oor into the hut where his wife receives it. When it passes into the hut, it becomes the wife s an it is her responsibility to manage it. Traitionally, the management responsibility meant that the wife ecie how much to use for each meal, how much to conserve as fermente milk or ghee, an how much to give away to other househols. Increasingly, it means she ecies how much of the milk will be markete an how much will be consume by the househol. The marketing option has introuce a change in the nature of the management ecision. Marketing allows the transformation of milk prouce from the her into cash. As she will usually spen this money on goos before returning to the family in the evening, she is now presente with a new set of ecisions over how to spen this income. We use evience on how men use migration ecisions to influence their wife s milk sales to investigate the nature of the intra-househol negotiation over granting the wife this ecision-making power. Description of the Data This stuy uses longituinal ata gathere in two areas of Marsabit District, Kenya. Gabra pastoralists occupy the two areas stuie: the Chalbi area an the Dukana area. The Chalbi area is rier than the Dukana area, but has more water points as it lies along the lowlan Chalbi basin. Marketing is more evelope in Chalbi than in Dukana, since Dukana is more remote an less serve by transport; vehicles traveling to Dukana must first past through Chalbi. 7

10 The ata was gathere using a sampling methoology similar to a transect. Enumerators walke between the main towns of the stuy area an interviewe nomaic househols they met along the way. This approach was chosen as there is no population list of exclusively nomaic herers to sample from in the stuy area. The questionnaire was retrospective in nature, recoring information for four time perios per year for each of the years Within a year, the four time perios correspon to the bimoal rainfall pattern of the area: the long rains, the ry season following these rains, the short rains, an the ry season following these rains. Each perio is roughly three months in length. Responents were aske to report the following variables for each time perio: ages of househol members; househol size; starting perio househol her size an species composition; average milk prouction from the her per ay an total milk sales per perio; an other sources of househol income. Househol size was converte into an ault equivalent scale following the metho outline by Martin (1985). 3 Variables recoring her size are converte to total livestock units (TLU), following the metho of Schwartz et al (1991). 4 The results appear to be reliable, both in terms of responents ability to provie answers an in terms of the internal coherence of the results. Variables exogenous to the househol are also recore in the ata set. Four variables are use to recor rainfall characteristics of a given time perio; one measures total rainfall in the current three-month perio plus the last three-month perio, a secon measures the percent of this total that fell in the current three month perio, an the final two are ummy variables that recor whether the perio in question is a rainy season. A variable recors the tons of foo ai 8

11 elivere to the towns of the stuy area in a given time perio. 5 Table 1 presents summary statistics of variables use in later regressions. Milk marketing is relatively new in the Gabra area. It began to appear in the late 1960 s, as towns began to grow in the Gabra area. Now, milk marketing is wiesprea. In the Chalbi sample, 67% of the 39 surveye househols sol milk at some time between 1993 an In Dukana, 86% of the 49 surveye househols were involve in milk marketing over the same perio. Milk sales accounte for 11% of househol cash income on average in Chalbi an 14% in Dukana. In contrast, livestock sales provie the majority of househol cash income; 73% in Chalbi an 67% in Dukana. Moels of Househol Decision Making Three ifferent moels of househol ecision making are presente. In each, the househol ecies where to locate an how much milk to sell. 6 A) Cooperative Decision Making In this moel, the househol ecies on the istance to settle from town an the milk sales level in a cooperative manner. Here, we moel it as a joint ecision by the husban an wife. If instea, the husban took over the milk marketing an mae both istance an sales ecisions himself, the outcome woul still be a cooperative one 7. The outcome maximizes the joint househol utility function. For both the husban an wife, efine a logarithmic utility function. Utility is an increasing an concave function of consumption. Total househol utility is obtaine by summing the utility of the husban an the wife. Therefore total househol utility is efine by h w U( c) = αln ( c ) + (1 α)ln ( c ) (1) 9

12 where h represents the husban, w represents the wife, an α represents the weights for the iniviual consumption in the utility function. Consumption (c) inclues milk consume by the househol members, goos purchase with the income from milk sol, an goos purchase from the sale of livestock. Assume that ecisions over the sale of livestock occur prior to ecisions over househol location an milk sales, so that the her contribution to consumption is fixe at hc when the location an milk sales ecisions are mae. 8 Total milk prouction is m, milk sales occur at price p, 9 an milk sales are represente by s. Consumption can be represente c = hc+ (m s) + s p () The istance from town to the househol location is represente by. Milk markets are locate in towns. Therefore, the labor effort involve with marketing milk is an increasing function of milk sales an istance from town. Assume the labor cost of milk marketing can be represente by a multiplicative specification ω 1 s, where ω1 represents a parametric weight on milk marketing labor. Towns also are the centers of amenities, such as health centers, schools, news an communication centers, public security, an markets for consumption goos. Therefore, settling further from town provies isutility by reucing househol members ability to access these amenities. However, as other herers also esire to be near town to take avantage of these amenities, labor effort for hering increases the closer one settles to town. Represent these two countervailing influences by 1 ω ω 3 (3) where ω an ω 3 again represent parametric weights on istance. The househol thus solves the following problem. 10

13 1 ln( ( ) ) h w Max hc+ m s + s p + ln( hc+ ( m s) + s p) ωs ω ω 1 3 s, (4) The solution of this problem provies the following conitions: 1 3 = ω + ω1 s (5) ω m hc s = ω1 p 1 p 1 (6) Thus, in the cooperative moel, the two ecisions are mae simultaneously an each epens on the other. Househols choose the istance from town as a ecreasing function of milk sales. Househols choose milk sales as a ecreasing function of istance. B) The Traitional Moel. In this moel, we assume that the husban makes the location ecision without consiering how this influences milk sales. In this case, a husban acts an the wife reacts. The husban is still operating uner the traitional cultural rules, an has not yet introuce milk marketing as a strategic consieration in his ecision. Assume he views the procees of milk marketing as his wife s concern an oes not consume the proucts purchase from milk sales. He ecies where to locate the her base on his own consierations an leaves it to his wife to ajust her milk marketing accoringly. The husban ecies the istance from town variable by solving the problem Max h 1 ln ( hc + ( m s)) ω ω3 (7) while his wife takes the istance as given an solves: Max s ln w ( hc + ( m s) + s p) ω s ω (8) 1 11

14 The solutions to this problem are ω 1 3 = ω (9) an 1 m hc s = ω1 p 1 p 1 (10) In this case, istance is etermine inepenently of milk sales an the milk sales ecision is a ecreasing function of istance. C) The Conteste Moel. In this moel as in moel (B) we assume that women control the income from milk sales an that the procees from milk sales o not enter the husban s consumption. In contrast to moel (B), the husban has now unerstoo that the introuction of milk marketing has create a new ecision-making context. In this situation, the husban has realize that his power as first mover allows him some leverage to manipulate his wife s milk sales. 10 As the milk sales lea to less milk for him to consume, it is in his interest to reuce the wife s incentive to sell. Men thus again solve the following problem: Max h 1 ln ( hc + ( m s)) ω ω3 (11) while the wife is face with the problem: Max s ln w ( hc + ( m s) + s p) ω s ω (1) 1 Solving recursively, we arrive at the first orer conition for the wife from (1) above. Substituting this into the husban s ecision problem an maximizing gives us the following. 1

15 h 1 m hc 1 Max ( hc + m + + ) ω ω3 ω1 p 1 p 1 ln (13) Solving this problem gives us the following conition: = ω where, as in (10): * 1 ( ω ω hc + m ) s 1 ( (14) s 1 m hc = ω1 p 1 p 1 * (15) In this case, istance is an increase function of milk sales, an the milk sales ecision is a ecreasing function of istance. The comparison of the three moels is summarize as follows: Cooperative Traitional Conteste Distance Variable Decreasing in s Not a function of s Increasing in s Milk Sales Variable Decreasing in Decreasing in Decreasing in We expect the milk sales variable to be always ecreasing in istance. The istinction between the three moels epens on the sign an the significance of the milk sales parameter in the equation for istance. These results provie the founation for the empirical estimations that follow. Empirical Analysis In this section, we use observe values for the istance a househol settles from town in a given perio an the total amount of milk sol in the perio to investigate the relationship between these ecisions. Denoting the istance from town ecision by, the milk sales ecision by s, γ an β as parameters to be estimate, X as matrices of exogenous variables, an u as unerlying isturbance terms, the following two equation system is efine: 13

16 = γ s = γ u, u s s s + β X + β X s ~ BVN( σ s + u + u, σ s s, ρ) (16) Given our analysis above, the parameter of interest is the sign an significance of γ s in the istance equation. If estimation reveals it to be negative an significant, this result is consistent with the cooperative moel. If it is not significant, the result is consistent with the traitional moel. Finally, if we fin it to be positive an significant, this provies a result consistent with the conteste moel. Three issues emerge when attempting to estimate this system of equations. First, both epenent variables are by construction non-negative an censore at zero. Distance from town equals zero for 7% of observations in Chalbi an 3% in Dukana. In aition, no milk was sol for 7% of observations in Chalbi an 8% in Dukana. 11 Failure to take account of the censore nature of epenent variables results in inconsistent parameter estimates. As the equations are specifie as a system, the methoology use is full information maximum likelihoo estimation of a bivariate tobit system (Maala 1983). A secon issue arises ue to the longituinal nature of the ata. It is possible that there are unerlying househol specific characteristics that influence livestock transfer behavior. If not controlle for, the presence of such characteristics will lea parameter estimates to be inconsistent (Hsiao 1986). The response use to aress this issue was to inclue a time invariant househol specific effect by creating a matrix recoring the means of househol specific variables for all time perios observe. The unobserve househol specific effect is assume to be a function of these househol specific means. A thir issue is the possibility that milk prouction is enogenously etermine. In other wors, if milk prouction is a function of istance, an milk prouction increases the further one 14

17 is from town, we coul misinterpret a husban s ecision to move away from town in orer to increase milk prouction as evience supporting the conteste moel. We take two steps to aress this issue. First, we present the results of a fixe effects estimation of milk prouction. These results (see Table ) inicate that the istance a househol settles from town has no significant impact on milk prouction. Secon, we present results of estimations of equation (16) with an without milk prouction inclue as a regressor. Estimations are conucte separately for the Chalbi an the Dukana ata. As both the lan an the market associate with these two towns iffer, there may be parametric ifferences in the ecision processes. Table 3 presents results of simultaneous tobit estimation of milk sales an istance from town ecisions for the Chalbi sample an the Dukana sample. The results show clearly that the coefficient on milk sales in the istance estimation is positive an significant, thereby supporting the conteste moel of the househol. As expecte, the coefficient on istance in the milk sales estimations are negative an significant. As istance increases, milk sales ecrease. There is almost no ifference between the two specifications for each sample; the moel is robust to the inclusion or exclusion of milk prouction. In Dukana, milk prouction affects milk sales, but this is not the case in Chalbi. In neither sample oes milk prouction affect istance. The her size variables are not statistically significant in the estimation of either location or milk sales ecisions. Age of the olest members of the househol oes affect these ecisions. There is some evience of seasonality in both istance from town an milk sales base on the seasonal ummies. In Dukana, foo ai eliveries ecrease the istance from town an increase milk sales. Foo ai is usually elivere to the towns an thus they provie an incentive for people to locate closer to town. 15

18 Conclusion The results are consistent with a conteste moel of househol ecision-making. Men appear to be making ecisions about the istance from town in orer to limit women s ability to market milk. This result is consistent with the notion that men resist the ability of their wives to obtain inepenent sources of income. While women gain the benefits of milk marketing, men are reluctant to facilitate this. Is this contestation a goo thing or a ba thing for overall househol welfare? We o not have the ata to aequately aress this issue in this context. Since most stuies inicate that income in women s control is more likely than men s income to be spent on goos for chilren 1, it may be that chilren s welfare woul increase if women earne income from milk sales. On the other han, by selling milk, women are also reucing the amount of milk available to the househol. Thus, the impact on chilren is ambiguous. We leave as a topic for further stuy who is right in this case; husbans who argue milk marketing has a negative impact on househol utility or wives who argue it has a positive impact on househol utility. What we can say is that husbans an wives are in a process of ajusting to the new opportunity brought about by milk marketing in this area. Our evience suggests the most appropriate way to unerstan the process is one of contestation. Husbans are using their traitional right to ecie migration patterns to influence wives sales ecisions. Wives are asserting that their traitional right over milk management extens to this new setting. This fining suggests that the introuction of market opportunities for goos that are traitionally home consume may meet with resistance within the househol. 16

19 Table 1. Descriptive Statistics Variable Chalbi Mean Chalbi Stanar Deviation Dukana Mean Dukana Stanar Deviation Distance --base camp to town (hours walk) Milk Sales (liters per perio * shillings per liter) a Milk Prouction (liters per ay) Her size in TLU Her size in TLU / Ault Equivalent Househol Size in Ault Equivalents Percent at satellite camp (of labor force for Chalbi, of her for Dukana) Rainfall in mm over past six months Long Rains Dummy Short Rains Dummy Foo ai eliveries in tons per perio Age of olest male in househol Age of olest female in househol Number of Observations Number of Househols a Note that the price of milk was constant at 0 shillings per liter over the entire perio. 17

20 Table. Fixe Effects Estimation of Milk Prouction Function Chalbi Fixe Effects Dukana Fixe Effects Her Size in TLU *** (0.0137) *** (0.0403) Her Size in TLU (x 10 - ) (0.0044) (0.080) Distance in hours from town (0.0440) (0.089) Distance in hours from town (0.0013) * (0.0008) Percent of her at satellite camp 0.09 (0.795) (0.3081) Rainfall in past six months *** (0.0059) (0.0036) Rainfall in past six months (x 10 - ) (0.0036) (0.0019) Long rains ummy *** (0.7).3547 *** (0.141) Short rains ummy ** (0.04) *** (0.1309) Time tren (1993 long rains =1, 1997 n ry season =0) *** (0.0893) *** (0.0503) Perio *** (0.0045) *** (0.005) Constant *** (0.3907) *** (0.540) Her significance χ ( ) *** *** Distance significance χ ( ) R.38.8 Number of observations * inicates significance at the.10 level, ** at the.05 level, *** at the.01 level 19

21 Table 3. FIML Simultaneous Tobit Results for Distance from Town an Milk Sales Milk Sales Chalbi Distance from Town Milk Sales *** (0.451) - Distance from Town Milk Sales *** (0.486) - Dukana Distance from Town Milk Sales *** (0.6004) - Distance from Town Milk Sales.995 *** (0.7039) - Distance from Town *** (0.3518) *** (0.3644) *** (0.0334) *** (0.0310) Constant * (.7869) *** (4.3405) ** *** (1.9430) (4.01) ** *** (5.5419) (.1667) * (6.3416) ** (4.3670) Last perio istance *** (0.0443) *** (0.0448) *** (0.084) *** (0.080) - Her size per ault equivalent (1.9809) 0.44 (.1897) (1.058) (0.681) ( ) (4.797) ( ).975 (3.4564) Her size per ault equivalent (.6590) Foo ai eliveries (6.3676) (7.963) (3.3505) (3.5015) (.560) (3.4069) (3.083) (9.4313) *** (0.4693) (3.448) *** (0.994) (8.4033) ** (0.5639) (3.0587) *** (0.306) Rainfall in past six months (0.811) (0.780) (0.5985) (0.7377) (0.6770) (0.3731) (1.1648) (0.356) Long rain ummy Short rains ummy Age of olest male (0.5435) * (0.6351) ** (0.5789) (0.6855).1030 (5.7519) (0.553) *** (0.5130) (1.497) (0.676) 0.16 (0.6474) *** (0.745) *** (0.508) *** *** (0.6837) (0.5056) *** (0.3865) *** (1.0810).81 *** (0.4479) ** *** (0.6976) (0.4494) *** (0.493) - Age of olest *** male (0.7791) ** (0.8004) * (1.757) * (1.8445) - Age of olest female ** (1.656) ** (1.658) *** (0.1903) *** (0.07) Age of olest female ** (1.4843) ** (1.4736) *** (0.9086) * (1.005) Milk prouction in perio (0.60) (1.406) (.41) *** (0.7476) σ 5.13 *** (0.1881) *** (0.7715) *** *** (0.1891) (0.7934) *** (0.84).4949 *** (0.46) *** (0.313).913 *** (0.196) σ *** (.0918) *** (.1315) *** (1.8373) *** (1.9797) Her size joint sig. χ ( ) Age joint sig. χ ( ) Fixe effect joint sig. χ ( 4 ) Beta sig. χ * ** 6.34 ** 6.9 ** 6.61 ** *** *** 8.75 ** 7.8 ** 15.4 *** *** 16.5 *** *** 17.7 *** ** 14.7 *** 9.5 ** *** *** *** *** *** 8.47 *** Equation Joint sig. χ *** *** *** *** * inicates significance at the.10 level, ** at the.05 level, *** at the.01 levr 0

22 References: Bourguinon, F., M. Browning, P. Chiappori, an V. LaChene Intra Househol Allocation of Consumption: A Moel an Some Evience From French Data. Annales 'Economie et e Statistique 39: Dercon, S. an P. Krishnan In Sickness an in Health: Risk Sharing within Househols in Rural Ethiopia. Journal of Political Economy 108(4): Doss, C.R Intra-househol Resource Allocation in Ghana: The Impact of the Distribution of Asset Ownership within the Househol. In G.H. Peters an J. von Braun, es.. Foo security, iversification an resource management: Refocusing the role of agriculture? Proceeings of the Twenty-thir International Conference of Agricultural Economists hel at Sacramento, California, August Alershot, U.K. an Brookfiel, Vt.: Ashgate, pp Doss, C.R Is Risk Fully Poole within the Househol? Evience from Ghana. Economic Development an Cultural Change, : Fratkin, E. an E. Roth Drought an Economic Differentiation among Ariaal Pastoralists of Kenya. Human Ecology, 18: Hoinott, J. an L. Haa Does Female Income Share Influence Househol Expenitures? Evience from Cote D'Ivoire. Oxfor Bulletin of Economics an Statistics 57 (1): Hsiao, C Analysis of Panel Data. Cambrige, Cambrige University Press. Jones, C The mobilization of women's labor for cash crop prouction: A game theoretic approach. American Journal of Agricultural Economics: Laipo, P Looking beyon the farm for gener issues in FSRE. Journal for Farming Systems Research Extension (): Lunberg, S., R.A. Pollak an T.J. Wales Do Husbans an Wives Pool Their Resources? Evience from the U.K. Chil Benefit. Journal of Human Resources 3 (3): Maala, G.S Limite-Depenent an Qualitative Variables in Econometrics. Cambrige, Cambrige University Press Martin, M Design of a foo intake stuy in two Bambara viallages in the Segou Region of Mali with preliminary finings. Population, Health an Nutrition in the Sahel. A. Hill. Lonon, KPI Limite Phipps, S. A. an P. S. Burton What's Mine Is Yours?: The Influence of Male an Female Incomes on Patterns of Househol Expeniture. Economica 65 (60):

23 Quisumbing, A. an J. Maluccio Intrahousehol Allocation an Gener Relations: New Empirical Evience from Four Developing Countries. Foo Consumption an Nutrition Division Discussion Paper No. 84, International Foo Policy Research Institute, Washington, DC. Schwartz, H.J., S. Shaabani an D. Walther Es. (1991) Range Management Hanbook of Kenya. Nairobi, Republic of Kenya, Ministry of Livestock Development Thomas, D The Distribution of Income an Expeniture within the Househol. Annales 'Economie et e Statistique 9(January/March 1993) Thomas, D. an C. Chen Income shares an shares of income: Empirical tests of moels of househol resource allocations. Mimeo.Santa Monica, CA: RAND. Ury, C Gener, agricultural prouction, an the theory of the househol. Journal of Political Economy 104(5): von Braun, J., an P.J.R. Webb The impact of new crop technology on the agricultural ivision of labor in a West African setting. Economic Development an Cultural Change 37(3):

24 Notes 1 In this paper, we o not try to istinguish among the ifferent types of cooperative outcomes. Much of the intrahousehol literature has focuse on etermining which cooperative outcome results, base on bargaining power or other factors. Here we are intereste in whether or not the outcome is a cooperative one. This is the milk prouce for human consumption. Traitionally, half the uer of a milking animal is taken for human consumption an the other half is left for young stock to suckle. 3 The ault equivalent weighting scheme use in this stuy assigns a value of one to iniviuals of both sexes oler than 15, a value of.6 to iniviuals 6-14 years ol, a value of.3 to chilren ages -5, a value of.1 for chilren uner. 4 One livestock unit = 10 sheep or goats = 1 hea of cattle = 0.7 camels. This iffers slightly from the scheme in Schwartz et al. as they weigh 11 goats equal to one TLU. As the total number of sheep an goats is the variable recore in the ata set, the composite measure of smallstock is assigne a weight of 1 animal = 0.1 TLU. 5 The rainfall an foo ai recors were provie by the Catholic mission in North Horr an the AIC mission in Kalacha. 6 While the focus of this paper is the ynamic process of cultural aaptation to market evelopment, we evelop our argument through moels of ifferent states of this process as separate static moels rather than through use of a unifie ynamic moel. This keeps the moel as simple as possible while illuminating our main points. We leave as a future extention the connection of these ifferent phases in a unifie ynamic moel. 7 Within the cooperative moels, the outcomes woul iffer epening on the weights assigne to each person s utility. But changing the weights woul not affect the sign on the istance variable, which is our concern in this paper. 8 This allows us to focus attention on the static aspect of the istance an milk sales ecision. Livestock sales introuce a ynamic element to the moel that we chose to ignore at this stage in the interest of simplicity. 9 The price of milk was constant over the stuy perio at a price of 0 shillings per liter. 10 Wives avance the story that milk sales will enhance househol welfare overall, as they provie foo an clothing for themselves an the chilren with this income, leaving the husban to consume the milk-base iet that he expects. In this case, assume the chilren s welfare is subsume uner the wife s utility. Milk sales in this case expan the househol s buget constraint ue to the avantageous caloric terms of trae, thus expaning overall househol welfare, while not etracting from the husban s utility. Husbans, alternatively, argue that 3

25 when women gain control over income they will aversely impact the househol buget as they will spen the procees of milk sales on town base boyfriens thus epriving the househol of both milk an income from milk sales. In this case, assume the chilren s welfare is subsume uner the husban s utility. While the former story is perhaps more creible than the latter, it remains an area for further research. 11 Note that although most of the househols sol milk uring at least one of the perios of the survey, most of these househols also ha perios where they sol no milk. 1 See for example, Hoinott an Haa (1995), Doss (1999), Thomas (1993). 4