Fredriksson, Per G.; Gupta, Satyendra Kumar (2020): Irrigation and Culture: Gender Roles and Women’s Rights, GLO Discussion Paper, No. 681, Global Labor Organization (GLO), Essen. https://www.econstor.eu/bitstream/10419/225005/1/GLO-DP-0681.pdf
Abstract: This paper proposes that ancestral use of irrigation reduces contemporary female labor force participation and female property rights. We test this hypothesis using an exogenous measure of irrigation and data from the Afrobarometer, cross-country data, the European Social Survey, the American Community Survey, and the India Demographic and Household Survey. Our hypothesis receives considerable empirical support. We find negative associations between ancestral irrigation and actual female labor force participation, and attitudes to such participation, in contemporary African and Indian populations, 2nd generation European immigrants, 1.5 and 2nd generation US immigrants, and in cross-country data. Moreover, ancestral irrigation is negatively associated with attitudes to female property rights in Africa and with measures of such rights across countries. Our estimates are robust to a host of control variables and alternative specifications. We propose multiple potential partial mechanisms. First, in pre-modern societies the men captured technologies complementary to irrigation, raising their relative productivity. Fertility increased. This caused lower female participation in agriculture and subsistence activities, and the women worked closer to home. Next, due to the common pool nature of irrigation water, historically irrigation has involved more frequent warfare. This raised the social status of men and restricted women’s movement. These two mechanisms have produced cultural preferences against female participation in the formal labor market. Finally, irrigation produced both autocracy and a culture of collectivism. These are both associated with weaker female property rights.
JEL Codes: J16, J21, N50, O10, P14, Q15, Z13.
Keywords: Irrigation; agriculture; culture; gender; norms; labor force participation; property rights.
5. Possible Mechanisms
In this section, we discuss possible mechanisms for the effect of irrigation potential on
female labor force participation and female property rights, respectively.
5.1. Female participation on pre-industrial agriculture
Boserup (1970) argues that males have historically captured technologies complementary
to irrigation, which increased their productivity in irrigated agriculture. Women worked with
less advanced technology and were relatively less productive. Irrigation activities may also be
dangerous (including to accompanying children), and may involve confrontations with
neighbors over water allocation. These factors all contribute to male labor specialization in
irrigated agriculture. While women still worked in agriculture to some degree, their work
gravitated toward the homestead, e.g. specializing in processing cereals and child care,
spending their lives mainly indoors or in the courtyard with little contact with non-relatives
(Ember, 1983).20
The Standard Cross-Cultural Sample (SCCS), compiled by Murdock and White (1969), has
data from 186 separate pre-industrial societies.21 The Ethnographic Atlas by Murdock (1967)
has data on 1,267 pre-industrial societies from around the world. The Atlas contains little
information about the societies themselves, however. The measure of the female-relative-tomale participation in overall agricultural production takes values from 1 to 5, where relative
participation is coded as follows. 1: males exclusively; 2: males predominantly; 3: equally; 4:
females predominantly; and 5: females exclusively. The dependent variables. The measure of
the relative time and effort spent on subsistence activities by females varies from 1 to 3, where
1: men expend more; 2: men and women expend roughly equal; 3: women expend more.
While the SCCS and the Atlas contain information on the centroid of each society, the
measurement of the precise location may involve errors. Moreover, information about the land
area covered by these pre-industrial societies is missing. We use a buffer zone of 200 km
around the centroid in order to construct the measure of irrigation potential and control
variables (Alesina et al., 2013). The ethnographic controls include suitability of the local
environment for agriculture, the presence of large domesticated animals, the proportion of the
local environment that is tropical or subtropical, an index of settlement density, and an index
of political development.
The mean variable values are provided in the table heading of Table 6. While column (1)
uses data from the Ethnographic Atlas, the remaining columns use SCCS data. Columns (1)-
(4) use female overall participation in agriculture as outcome variable, column (5) studies the
overall relative female contribution to subsistence in time and effort, and columns (6)-(8)
provide estimates for crop tending, harvesting, and milking, respectively. All columns include
fixed effects for the century-of-observance and language, ethnographic controls including plow
use, suitability of the local environment for agriculture, the presence of large domesticated
animals, the proportion of the local environment that is tropical or subtropical, economic
complexity, and political development.
Column (1) irrigation potential has a negative and statistically significant influence on
relative female participation in agricultural activities overall. The effect of moving from zero
to complete irrigation potential equals -1.45, which is substantial given the 3.04 mean. The
effect is stronger in column (2), using SCCS data. Column (3) adds a dummy for whether
cereals are the main crops, which Ember (1983) and Hansen et al. (2015) suggest influences
female labor force participation. The coefficient on irrigation potential rises further. Column
(4) controls for societies with formal class stratification systems, where males may be less
likely to participate in agricultural field work. The two added controls are insignificant, while
irrigation potential remains significant. Column (5) suggests that the female relative
contribution to subsistence activities were lower in societies with greater irrigation potential.
This reflects a shift away from agricultural field activities towards greater domestic and child
rearing duties. Columns (6)-(7) provide evidence that irrigation is associated with a decline in
female participation in some important agricultural activities outside the home, while column
(8) suggests the opposite effect occurred to milking which occurs closer to home. Table C5 in
online Appendix C provides cross-country OLS evidence that irrigation is associated with
restrictions on women’s freedom of movement within a country, using data from Coppedge et
al. (2019).
Overall, the results are consistent with women staying closer to the homestead in irrigated
areas during the pre-industrial time period. This suggests one possible partial mechanism which
links irrigation to lower contemporary female labor force participation rates and related
attitudes. The division of labor appears to have persisted through intergenerational cultural
transmission. [Table 6]
5.2. Warfare
Irrigation water is frequently a common pool resource. Since water consumption by
upstream communities may affect consumption by downstream communities, especially in
times of drought, conflicts are likely to occur. Anecdotal evidence includes Iraq and the
Andes.22 Irrigation agriculture also led to a storable surplus and relatively more valuable land,
providing incentives for raids and external warfare by other groups (e.g., Ember, 1982; Ang
and Gupta, 2018). We hypothesize that irrigation societies had a greater incidence of violent
external conflict. This yielded a greater demand for men due to greater muscle strength and
aggressiveness, improving their social standing (Chagnon, 1988; Ramos-Toro, 2019). RamosToro (2019) provides evidence of a negative relationship between exposure to conflict and
contemporary female labor force participation.23 Warfare may also have been associated with
reduced female mobility, and stricter social norms and restrictions on women’s labor market
participation outside the home.
Table 7 provides some support for this hypothesis. Columns (1)-(3) utilize society level
data from the SCCS (Murdock and White, 1969), while columns (4)-(9) use historical district
level data from India. The dependent variable in column (1) is a measure of external warfare.
We focus on societies with agricultural activities. We recode the Murdock and White (1969)
external warfare measure as follows: it takes a value of 1 if external war is ‘frequent, occurring
at least yearly’ or ‘common, at least every five years’; it takes a value of zero if ‘occasional, at
least every generation’ or ‘rare or never’. The sample size declines to 54. This renders language
and century fixed effects and the inclusion of tropical climate infeasible in this analysis. The
results should thus be interpreted with these drawbacks in mind. The logit model in column (1)
suggests that irrigation potential is positively associated with external warfare. Columns (2)-
(3) present ordered logit regressions. Column (2) is the basic model from Table 1 with female
participation in overall agriculture as the outcome variable (but with a smaller sample size).
External warfare is negative and significant in column (3), while the irrigation potential point
estimate declines modestly (in absolute value) from -2.00 to -1.67. This provides some support
for the hypothesis that a history of external warfare a partial channel between irrigation and
female participation in agriculture in pre-industrial societies.
Columns (4)-(9) provide district level evidence from India in support of the proposed
warfare mechanism. In rural India, most households working outside the home are engaged in
agricultural sector work (Kapsos et al., 2014). We measure female employment outside the
home by the ratio of the population of female agricultural workers to the total female population
in year 2011. In column (4), war count is a measure of the total number to land battles over
years 610-1962 occurring within a distance of 50km from the district, geocoded using Jacques
(2007).24 In column (7), the period of observations for war count is instead restricted to 1001-
1867AD. Jacques (2007) records relatively few instances of wars before 1001 (data quality
may be an issue), and in 1858 the British Crown took over the administration of India from the
British East India Company. The wars for accession of the native states ceased after 1857.
Columns (4) and (7) present negative binominal regressions, while columns (5), (6), (8), and
(9) are generalized linear model regressions. Geographical controls (temperature, precipitation,
latitude, area, land quality, and nightlight luminosity) and language fixed effects (largest
language group within the district, reflecting cultural variation across districts) are included.
One benefit of using data on India is the low interstate migration rate, which maintains longterm cultural differences (Kone et al., 2018).
Column (4) suggests a positive association between a district’s irrigation potential and the
historical experience with wars. When war count is included in column (6), the (absolute value
of the) irrigation potential point estimate declines relative to column (5), from -0.137 to -0.115,
while war count is negative and significant. The takeaway from columns (7)-(9) is similar.
These findings provide some support for the hypothesis that a history of warfare is a possible
partial mechanism linking irrigation and female participation in agriculture in contemporary
India. [Table 7]
5.3. Autocracy
Next, we provide evidence that historical irrigation may affect contemporary female
property rights regimes via an autocracy channel. Underpinned by resource curse theory,
Bentzen et al. (2017) argue that historical irrigation agriculture raised the likelihood that a preindustrial society was ruled by an elite based on the control of a natural resource. This has
yielded lower levels of contemporary democracy. Autocracies have weaker property rights
(Gradstein, 2007), and Fish (2002) suggests that authoritarianism is associated with negative
outcomes for women.25
To study the determination of female property rights, we utilize the standard general
measures of democracy, Polity2. The male political majority (in both democracies and
autocracies) is likely to determine the extent of female property rights. Table 8 shows that
democracy works as channel for the effect of irrigation potential on contemporary female
property rights. Column (1) establishes that irrigation potential has a negative effect on
democracy, measured by Polity2. In column (2), irrigation potential has a negative effect on
female property rights as shown above, but this effect disappears when we include democracy
in column (3). This suggests that the effect of irrigation on female property rights works at least
partially through the democracy channel. In contrast, while democracy has a positive
association with male property rights in column (5), there is no evidence that democracy works
as a link between irrigation and male property rights. [Table 8]
Table C6 in the online appendix provides additional evidence that attitudes, actual female
political participation rates, and the freedom to discuss politics are negatively associated with
irrigation potential. The first column uses a statement from the Afrobarometer as outcome
variable, and the remaining three columns use variables from Coppedge et al. (2019). 26
5.4. Individualism vs. Collectivism
This section investigates the cultural dimension of individualism vs. collectivism as a
possible potential mechanism. Buggle (2020) finds that irrigation is negatively correlated with
individualism, because irrigation required constant collaboration which yielded collectivism.
Gorodnichenko and Roland (2017) argue that the degree of individualism influences societies’
institutional choices. In particular, they find that individualism has a positive association with
the level of protection against expropriation risk, a measure of property rights. Binder (2019)
reports that collectivism is correlated with a belief in traditional gender roles.
Table 9 explores whether individualism vs. collectivism may function as a partial channel
between irrigation potential and contemporary female property rights. We use Hofstede et al.’s
(2010) measure of individualism and Coppedge et al.’s (2019) measure of property rights.
Irrigation potential has a negative correlation with individualism in column (1) and with female
property rights in column (2). This effect declines moderately (in absolute value) when
individualism is entered in column (3), from -1.75 to -1.64. However, a similar pattern occurs
for male property rights in columns (4) and (5), from -1.67 to -1.57. It appears that the
individualism vs. collectivism dimension of culture may provide a partial link between
irrigation potential and property rights for both genders. Thus, we do not find strong evidence
that this link is stronger for female than for male property rights. To investigate this further,
Table C7 in the online appendix includes both individualism and democracy as potential
mechanisms. It appears that both mechanisms work for female property rights, but only
democracy constitutes a channel for male property rights. [Table 9]