This document discusses a study on agricultural production in polygynous households in Burkina Faso. The study finds that while altruism can encourage cooperation, it can also inhibit efficiency by reducing the credibility of punishments. Using data on plot yields, the study finds greater cooperation between co-wives than between wives and husbands in polygynous households, suggesting altruism makes punishment threats less credible. Various robustness checks considering alternative explanations and unobserved factors support the role of altruism in impacting cooperation and efficiency.
1. ALTRUISM, COOPERATION, AND
EFFICIENCY: AGRICULTURAL
PRODUCTION IN POLYGYNOUS
HOUSEHOLDS
Joyce Chen (Ohio State), Richard Akresh
(Illinois-Urbana), Charity Moore (Ohio State)
2. Efficiency in the Household
Households have many features that encourage
cooperation and the efficient allocation of resources
Altruism/shared public goods
Repeated interaction
Better information
And, yet, there is ample evidence of inefficiency, in
both consumption and production
Udry (1996), Goldstein and Udry (2008)
Duflo and Udry (2004), Dubois and Ligon (2010)
3. Inefficiency in the Household
Common culprits of inefficiency include
Imperfect monitoring
Limited commitment/enforcement
Social norms
Separate spheres
Perhaps altruistic preferences themselves can inhibit
cooperation and efficiency
Leadto higher utility in a non-cooperative equilibrium
Make threats of punishment less credible
4. Altruism and Exchange
Bernheim and Stark (1988) first suggested, in a
theory paper, that altruism can inhibit cooperation
by reducing the credibility of punishments
Empirical studies typically compare relationships
between family and non-family members, making it
difficult to distinguish altruism from information, etc.
We look at relationships with differing degrees of
altruism, within the same family
5. Outline
Context and Data
A Simple Game-Theoretic Model
Main Results
Robustness and Extensions
6. Context: Burkina Faso
Data are drawn from the 1984-85 ICRISAT Survey
Married Burkinabé women often have access to
private plots under their own control
But they also must provide labor on household
communal plots, usually at discretion of the head
Husbands typically provide staple foods and
contribute to medical expenses and school fees
7. Table 1. Average Yield and Plot Area
Monogamous
Household Wife of Other Other
Head Head Male Female
Yield (1000 FCFA) 126.29 49.15 142.93 124.82
(651.6) (267.0) (498.2) (434.7)
Area (Hectare) 0.748 0.075 0.318 0.069
(1.24) (0.13) (0.54) (0.12)
Observations 743 425 172 319
Polygynous
Household Wife of Other Other
Head Head Male Female
Yield (1000 FCFA) 85.47 59.50 145.51 71.57
(341.3) (208.4) (358.6) (250.6)
Area (Hectare) 0.756 0.099 0.385 0.074
(1.14) (0.14) (0.48) (0.10)
Observations 1156 1305 407 699
8. Context: Burkina Faso
Polygyny is quite common in our data (51%)
Mostof these households (56%) have just two wives
One-third have three wives, 11% have more
Although interaction between co-wives is often
characterized by conflict, there is also a fair amount
of cooperation for pragmatic goals
Labor-sharing
Exchange of goods and/or services
9. Altruism and Cooperation
Consider a household with three individuals – one
husband and two wives
Each player engages in agricultural production on
his/her own plot, using the same technology
But different plot characteristics mean the optimal
labor allocation differs across plots as well
10. Altruism and Cooperation
Husbands and wives care about each other’s
consumption of certain goods
“Separate spheres”– otherwise, production and
consumption decisions are separable
Consistent with anthropology of Burkinabé households
Co-wives have no altruistic linkage
Each individual can choose to coordinate production
(i.e., share labor) with other household members
11. The Basic Game
With altruism, there will be some exchange
between spouses, even without explicit cooperation,
but not between co-wives
The gains to cooperation can be higher for co-
wives, even if aggregate production is not
Lack of altruism allows co-wives greater scope for
punishment, while altruism makes the husband’s
punishment susceptible to renegotiation
12. Altruism and Efficiency
Results are sensitive to separate spheres assumption
Consistent with Burkinabé households
Not an uncommon formulation of altruism to have
preferences over a good that one does not directly
control (e.g., children’s education, utility of spouse)
Cooperation implies efficient allocation of inputs
across plots controlled by cooperating parties
Controllingfor land quality, crop choice and time-
varying shocks, plot yields should be equalized
13. Empirical Application
Estimate yields as a function of plot characteristics
(size, toposequence, soil type, location) and
cultivator characteristics
𝑄𝑄ℎ𝑡𝑡 𝑡𝑡 𝑡𝑡 = 𝑿𝑿ℎ𝑡𝑡 𝑡𝑡 𝑡𝑡 𝛽𝛽 + 𝛾𝛾 𝐺𝐺 𝐺𝐺ℎ𝑡𝑡 𝑡𝑡 𝑡𝑡 + 𝛾𝛾 𝑂𝑂𝑂𝑂 𝑂𝑂𝑂𝑂ℎ𝑡𝑡 𝑡𝑡 𝑡𝑡 + 𝛾𝛾 𝑂𝑂𝑂𝑂 𝑂𝑂𝑂𝑂ℎ𝑡𝑡 𝑡𝑡 𝑡𝑡 + 𝜆𝜆ℎ𝑡𝑡 𝑡𝑡 + 𝜀𝜀ℎ𝑡𝑡 𝑡𝑡 𝑡𝑡
where 𝛾𝛾 𝑘𝑘 = 𝛾𝛾 0 + (𝑃𝑃𝑃𝑃𝑃𝑃 𝑃𝑃ℎ𝑡𝑡 ∗ 𝛾𝛾 𝑘𝑘𝑃𝑃 ) for k = G, OM, OF
𝑘𝑘
Include household-crop-year fixed effects (λ)
Allow cultivator characteristics to vary with conjugal
status (monogamous/polygynous)
14. Empirical Application
If cooperation is greater among co-wives, should
observe smaller negative effect of gender in
polygynous households
If polygyny also provides husbands with more
opportunities for cooperation, then we should
observe a smaller gap between heads and other
male cultivators as well
𝛾𝛾 𝑂𝑂𝑂𝑂 ≥ 0
𝑃𝑃
15. Alternate Explanations
Household head can enforce cooperative
arrangements between players other than himself –
also implies smaller yield differences between other
𝛾𝛾 𝑂𝑂𝑂𝑂 < 𝛾𝛾 𝑂𝑂𝑂𝑂
cultivators than between husbands and wives
Strong preferences (lower costs) for cooperation
among women – also implies smaller differences
𝛾𝛾 𝑂𝑂𝑂𝑂 > 0
between females in polygynous households
𝑃𝑃
16. Table 2. Fixed Effects Estimates of the
Effect of Cultivator Characteristics on Plot Yield
Other
Men Only Cultivators Women Only
(I) (II) (III) (IV)
Gender (1=female) -202.21 *** -160.72 ***
(34.14) (54.01)
Other Male -97.18 ** -74.78 **
(39.38) (36.06)
Other Female -31.96 18.16
(31.39) (20.77)
Gender*Polygynous 168.94 *** 131.04 **
(40.09) (61.80)
Other Male*Poly 86.50 * 69.99 *
(45.82) (42.05)
Other Female*Poly 28.71 -18.87
(35.81) (23.23)
Observations 5230 2478 1597 2748
17. Main Results
Estimates consistent with altruism– co-wives more
likely to cooperate with each other than with
husband
Husbands do not do as well under polygyny, relative to
other male cultivators – evidence of transaction costs
No clear evidence for alternate explanations
Gap between other males and other females is not
significantly smaller
No significant gap between wives and other women
18. Table 3. Alternate Tests
Gender (1=female) -63.60 Gender (1=female) -45.46
(66.14) (50.68)
Gender*Other Female -132.29 * Gender*No. of Kids -23.28 **
(78.33) (10.27)
Gender*Poly 33.67 Gender*Poly 41.94
(75.20) (64.85)
Gender*Poly*OF 126.61 Gender*Poly*Kids 22.15 *
(88.86) (11.62)
Observations 3629 Observations 4701
If females have stronger preferences for
cooperation, addition of other female cultivator
should increase efficiency among women
Cooperation should be decreasing in the degree
of altruism (# of shared goods) between players
19. Robustness Checks
Fixed effects control for factors that are fixed
across individuals planting the same crop, in the
same year, within the same household
But they do not control for differences in crop choice
or preferences for cooperation across monogamous
and polygynous households
Main results include full set of interactions with
polygyny to allow for different technologies
20. Table 4. Robustness Checks, Fixed Effects Estimates
Polygynous Polygynous Household-
=2 Wives >2 Wives Year FEb
(I) (II) (V)
Gender (1=female) -155.14 *** -155.14 *** -81.52 ***
(40.11) (39.01) (23.50)
Other Male -56.35 -56.35 -29.70
(47.21) (45.91) (29.82)
Other Female -16.02 -16.02 0.97
(36.50) (35.49) (26.27)
Gender*Polygynous 136.33 ** 154.32 *** 76.40 ***
(53.59) (53.01) (28.36)
Other Male*Poly 72.42 45.40 29.62
(62.19) (62.09) (35.80)
Other Female*Poly 14.76 13.91 -8.50
(48.07) (44.86) (30.85)
Observations 3112 3142 5230
21. Unobserved Land Quality
Fallow decisions can provide some indication of
land quality
Omitting observed land quality measures can
reveal the direction of bias, if correlation between
observed and unobserved characteristics is known
Data for 1981-83 only included co-wives’ plots for
cotton, cereal and root crops
22. Table 5. Checks for Unobserved Land Quality
Years Since No Plot 1981-83
Fallow Chars. Only
(I) (II) (III)
Gender (1=female) -6.73 *** -125.67 *** -35.13 ***
(2.20) (31.15) (12.48)
Other Male -9.48 ** -8.52 -30.30 **
(3.99) (36.98) (12.58)
Other Female 2.73 -3.58 2.74
(2.14) (31.80) (15.60)
Gender*Polygynous 1.31 128.65 *** 1.66
(2.32) (35.90) (14.88)
Other Male*Poly 2.34 21.09 17.15
(4.03) (43.12) (15.25)
Other Female*Poly -3.21 6.01 -23.77
(2.33) (36.33) (18.53)
Observations 4356 5230 4198
23. Unobserved Heterogeneity
While greater efficiency among wives seems clear,
proximate cause is not yet clear
No evidence of greater cooperation among husbands
and wives, contract enforcement by head or greater
propensity for cooperation among women
But polygyny may require the head to contribute
more time to administrative tasks
Or, polygynous men may have a different level of
underlying productivity
24. Table 6. Fixed Effects Estimates of the Effect of
Cultivator Characteristcs on Plot Yield, by Household Structure
Vertical a Horizontal b
(I) (II)
Gender (1=female) -8.43 -516.33 ***
(21.02) (111.29)
Other Male -18.55 -237.79 **
(25.80) (109.52)
Other Female -22.94 -5.00
(20.73) (74.17)
Gender*Polygynous -9.68 518.79 ***
(26.50) (117.88)
Other Male*Poly 8.06 251.77 **
(34.52) (116.05)
Other Female*Poly 20.62 2.30
(27.05) (78.45)
Observations 2878 1823
25. Table 7. Household Fixed Effects Specification
Gender (1=female) -99.54 ***
(26.25)
Other Male -32.28 Implied Fixed Effect
(31.46) Switch to Poly -30.98
Other Female 7.914 (49.66)
(27.03) Always Polygynous -67.78 ***
Polygynous -82.87 (20.49)
(83.71) Total Hh Plot Area 7.790 ***
Gender*Poly 65.37 ** (2.752)
(30.21) Capital Intensity 30.21
Other Male*Poly 12.21 (37.21)
(37.27) Observations 120
Other Female*Poly -15.13
(31.65)
Observations 5230
26. Unobserved Heterogeneity
Cooperation is likely more difficult in horizontal
households, given varying family allegiances
Polygyny, by providing more opportunities for mutual
benefit, has a greater positive effect on efficiency
Not consistent with greater demands on time for
polygynous household heads
Differences between polygynous household heads
and other male cultivators seem to arise over time,
rather than being intrinsic
27. Dynamic Inefficiency
Greater cooperation may make investments in large
capital goods more cost-effective
Wealthier households are more likely to make such
investments, but wives have found to be a substitute
for investment in durable assets
Look at expenditure on large capital goods as a
percentage of total expenditure on farm inputs
Treat polygyny and household wealth (total area under
cultivation) as endogenous
28. Dynamic Inefficiency
Instruments include:
Ethnic group of the household – polygyny based in
ethno-cultural traditions, while farming practices are
similar across ethnic groups
Area of land inherited – land for permanent cultivation,
reflects relative position in lineage, unobserved skill
Also include village-year fixed effects to account
for aggregate shocks and differences in agro-
climatic zones
29. Table 8. Percentage of Input Purchases in Large Capital Goods, IV Estimates
First Stage Polygyny Total Area
Polygynous 0.592 *** Dagari-Djula 0.708 *** 0.841
(0.209) (0.190) (1.172)
Total Hh Plot Area 0.008 Bwa 0.201 4.140 ***
(0.017) (0.147) (0.909)
Observations 231 Other Ethnic Group 0.096 0.648
(0.195) (1.208)
Sargan Test 0.27 Inherited Area 0.004 0.260 ***
(p-value) (0.87) (0.008) (0.050)
30. Dynamic Inefficiency
Polygynous households invest more, proportionally,
in large capital equipment
Without cooperation, lumpy investments should lead to
greater divergence in yields
They seem to have more labor and more capital
Perhaps monogamous households use more chemicals
and/or improved seed varieties
Need to estimate production function to fully
understand the implications for income and growth
31. Conclusions
In polygynous households, co-wives cooperate more
with each other, at the expense of husbands
Difference between heads’ and wives’ yields is smaller,
as is the difference between heads and other males
Some evidence of head acting as an enforcement
mechanism, but only in certain contexts
Some evidence of greater cooperation among all
women, but relationship matters
32. Conclusions
Results do not seem to be explained by
Differing preferences for cooperation
Endogenous crop choice
Unobserved plot quality
Differing demands on the household head
Self-selection into polygyny
Although polygynous households are a very specific
case, there are many cases of 3+ player games in
which there are differing degrees of altruism