Abstract

Using a comprehensive data set collected through field survey of 950 farmers across Pakistan, the current study evaluates water-management practices and their impact on food security and poverty. The results show that rural households mainly adopted four water-management practices (bund making, deep plowing, the adoption of stress-tolerant varieties, and irrigation supplements) and that the wealth, education, and gender of the farmer (male) positively influences the adoption of improved water-management practices. The propensity score matching approach shows that the adoption of improved water-management practices improves wheat and rice yields, household income and food security levels, and reduces poverty levels. The food security levels of households adopting improved water-management practices are higher: in the range of 3–12%. Higher wheat yields are in the range of 26.8–70.4 kg/acre and higher rice yields are in the range of 48.4–85.2 kg/acre. Higher household income levels are in the range of rupees 2,573–4,926 and the lower poverty levels are in the range of 2–7%. Hence, agricultural policy should promote improved water-management practices among rural households.

Introduction

Depleting water resources, increasing water demand in agriculture, and climate change are major challenges to food security and poverty reduction globally and specifically in South Asia (Hijioka et al., 2014). With the increase in population, the demand for water is ever increasing. The global demand for agricultural water is expected to increase by 19% by 2050 (UN-Water, 2013). In addition to the increasing demand for water, wastage of water in agriculture due to inefficient water-management practices is a major problem. In the rice–wheat system of the Indo-Gangetic Plains, about 10–25% of irrigation water is lost due to poor water management and uneven fields (Kahlown et al., 2000). The food security challenges arising from water scarcity and wastage can be tackled by bringing about efficient changes in water management (Mueller et al., 2012). Water shortage in agriculture lowers yield and income, and makes the household more food insecure and vulnerable to poverty (Rahut et al., 2016). Hence, sound water-management practices are crucial for sustainable development, food security, and poverty reduction across the globe.

Water availability for agricultural production is essential for ensuring food security (Brown & Funk, 2008). Severe water scarcity and poor on-farm water-management practices in Northern Ethiopia, are found to be responsible for the decline in crop yields and soil salinization in the study area (Yohannes et al., 2017). The use of water-resource conservation and energy-efficient technology in agriculture is needed for increasing agricultural productivity (Ambast et al., 2006; Hanjra & Qureshi, 2010). Rockström et al. (2010) argued that major water investments in agriculture are required in those regions where yield gaps are large, not due to lack of water, but rather due to inefficient management of water, soils, and crops. Therefore, the proper water management in agriculture and investment in water conservation technologies in agriculture are extremely valuable. The adoption of improved water-management practices will influence soil quality and crop yield through improvement in soil moisture. Water-management techniques have a substantial impact on soil moisture and thermal conditions, and ultimately on the growth rate and development of maize (Yi et al., 2010). In West Africa, bund-feeding has significantly increased rice yield across sites by about 40%, and 25% less weed biomass was observed in bunded farms compared to open plots (Becker & Johnson, 2001).

Against the backdrop of water scarcity and competing demands for water, efficiency in water management is inevitable in agriculture for ensuring food security to the growing population. Studies have found that the use of recycled water and water-saving technologies such as drip irrigation have a positive effect on water use efficiency (Gadanakis et al., 2015). To take advantage of advances in agricultural technology, a combination of drought-tolerant varieties, redeemable water irrigation methods, and proper agronomic practices for soil manipulation is needed to improve the economic gains of water productivity (Ali & Talukder, 2008). Data of five consecutive agricultural years reveal that sub-soiling with mulch resulted in the highest precipitation-use efficiency, precipitation storage efficiency, and crop yield, whereas the no-till with mulch performed slightly less than sub-soiling with mulch. However, conventional tillage control produced intermediate yields (Jin et al., 2007).

Pakistan is among the world's 36 most water-stressed countries (Iqbal, 2015) and the situation is worsening due to the increasing population and the rising demand for water from the farm and non-farm sectors. Pakistan has the world's most extensive irrigation system (Iqbal, 2015); however, proper management of irrigation water is a serious issue. In Pakistan, canal water irrigation is underpriced and highly subsidized, recovering only one-fourth of its annual operating and maintenance costs. The per capita annual water availability in Pakistan has dropped, fundamentally due to population growth and increased water use by the non-farm sector, from 5,600 cubic meters in 1947 to the current level of 1,017 cubic meters (IMF, 2015). Alarmingly, it is projected to decline further under the changing climate scenarios and the current infrastructure and institutional conditions (Iqbal, 2015).

In Pakistan, the demand for water in the agricultural sector is increasing. The demand is projected to reach 274 million acres-feet (MAF) by 2025, but the supply is stagnant and even decreasing at 191 MAF and the demand–supply gap is approximately 83 MAF (IMF, 2015). The concerns over the widening gap of supply and demand are compounded by certain characteristics of Pakistan's geography, climate, and hydrological cycle. Pakistan mostly depends on a single source, the Indus system and its tributaries, for most of its water supply for crop irrigation. Compared to other countries, the water storage capacity is low in Pakistan. For example, Pakistan can only save water for 30 days due to its high demand, compared to 1,000 days in Egypt and 220 days in India. Nonetheless, due to Pakistan's arid and semi-arid climatic conditions, about 90% of the area in Pakistan is irrigated through canal irrigation which is approximately 56,073 kilometers and supplemented through groundwater (Iqbal, 2015). A recent study in Sri Lanka has shown that irrigation water access has a positive effect on income through livelihood choices (Senaratna Sellamuttu et al., 2014). Similarly, in a study, Ahmad (1993) found that improved water-management practices contribute to the increase in net return at the farm household level. In South Asia, the rainfall variability has serious implications on rural livelihoods (Palanisami et al., 2015).

The Global Hunger Index at the global level has declined from 35.2 in 1992 to 21.8 in 2017 and the hunger index in South Asia has declined dramatically from 46.3 in 1992 to 30.9 in 2017, which is considered as serious (Von Grebmer et al., 2017). Although the hunger index in Pakistan has declined from 42.7% to 32.6%, it occupies 106th position, which is one above Afghanistan (see Table 1). Therefore, ensuring food security should be a priority for Pakistan and worthy of investigation.

Table 1.

Global Hunger Index in South Asia.

Rank (2017) 1992 2000 2008 2017 
 World 35.2 29.9 25.7 21.8 
 South Asia 46.3 38.2 34.9 30.9 
72 Nepal 42.5 36.8 28.9 22.0 
84 Sri Lanka 31.6 26.8 24.2 25.5 
88 Bangladesh 53.6 37.6 32.2 26.5 
100 India 46.2 38.2 35.6 31.4 
106 Pakistan 42.7 38.2 34.7 32.6 
107 Afghanistan 50.2 52.7 37.9 33.3 
Rank (2017) 1992 2000 2008 2017 
 World 35.2 29.9 25.7 21.8 
 South Asia 46.3 38.2 34.9 30.9 
72 Nepal 42.5 36.8 28.9 22.0 
84 Sri Lanka 31.6 26.8 24.2 25.5 
88 Bangladesh 53.6 37.6 32.2 26.5 
100 India 46.2 38.2 35.6 31.4 
106 Pakistan 42.7 38.2 34.7 32.6 
107 Afghanistan 50.2 52.7 37.9 33.3 

Scale: ≤9.9 low; 10.0–19.9 moderate; 20.0–34.9 serious; 35.0–49.9 alarming; ≥50.0 extremely alarming.

The poverty headcount ratio was 64.3% in 2001, which declined to 50.4% in 2005, and 29.5% in 2013 (World Bank, 2018). Hence, developmental policies in Pakistan should focus on poverty reduction, and increasing agricultural productivity could play an important role as a significant number of the poor live in rural areas and are dependent on agriculture for their livelihood. In 2014–16, the prevalence of undernourishment in the total population of Pakistan was 19.9% and the absolute number of the undernourished population stands at 37.6 million (FAO IFAD UNICEF WFP & WHO, 2017), which is alarming and calls for a policy to strengthen food and nutritional security in Pakistan. The wasting in children (under 5 years of age) and stunting in children (under 5 years of age) was 10.5% and 45% in 2016 (FAO IFAD UNICEF WFP & WHO, 2017). Therefore, it is particularly important to investigate policy and practices that enhance yield and food security and reduce poverty. Water is of critical importance for food security, but is becoming scarce over time due to competing demands, declining groundwater levels, variability in rainfall, drought, and climate change.

The existing literature has mostly shown the positive impact of improved water management on water saving and farm productivity (Jonish, 1991; Rickman, 2002; Sakurai, 2002; Bekele & Drake, 2003; Ren et al., 2003; Bekele, 2005; Sidibé, 2005; Hagos & Holden, 2006; Mallappa & Radder, 2012). However, in Pakistan, the research on improved agricultural water-management practices and their impact on the well-being of rural farm household welfare is scanty (Erenstein, 2009). Hence, the current study focuses on improved agricultural water management and provides several novel contributions. First, it is among the few studies that focus on agricultural water management in Pakistan. Second, it covers the four major provinces of Pakistan (Punjab, Sindh, Khyber Pakhtunkhwa (KPK), Balochistan). Third, it employs a propensity score matching (PSM) approach that has rarely been used to evaluate the impact of water-management practices. Fourth, it uses a multivariate probit and Poisson regression to assess the factors determining the adoption of improved water-management technology. Finally, the current study findings provide an important policy recommendation.

The rest of the paper is organized as follows: econometric methods are described in the next section, followed by the presentation of materials and the general description of data. The following section presents empirical results, and finally conclusions and policy recommendations are given.

Methodology

This paper employs a multivariate probit model to estimate the determinants of the factors influencing the choice of improved agricultural water-management practices and a Poisson regression model to assess the determinants of the number of water-management practices adopted by a farm household. The impact of these approaches is estimated by employing the PSM approach.

Multivariate probit and poisson regression model

Multivariate probit

In our analysis, we categorized the commonly used improved agricultural water-management practices by farm households into four mutually inclusive categories: (i) bund making; (ii) deep plowing; (iii) supplementing irrigation water; and (iv) use of stress and drought-tolerant varieties. As the four discrete dependent variables are mutually inclusive, which means one farm household could adopt more than one type of agricultural water-management practice, the multivariate probit model is best suited for the estimation of the determinants of adopting improved agricultural water-management practices.

Poisson regression

As a farm household may adopt more than one improved agricultural water-management practice, we used the Poisson regression to analyze the determinants of the number of improved agricultural water-management practices adopted by rural farm households. The dependent variable is the number of improved agricultural water-management practices which ranges from 0 to 4 and the explanatory variables are demographic and farm-level characteristics.

Propensity score matching

PSM creates the condition of a randomized experiment, and, the estimated treatment effect for the treated groups is of principal significance when using PSM. This effect may be given as:  
formula
(1)
where τ is the average treatment effect for the treated (ATT), represents the value of the outcome for adopters of improved water-management techniques, and is the value of the same variable for non-adopters. The main problem is that we do not observe . Although the difference can be estimated, it may, potentially, be a biased estimation.
According to Rosenbaum & Rubin (1983), PSM can be defined as the conditional probability that a farmer adopts the improved water-management practices, given pre-adoption characteristics. To make it a randomized experiment, PSM uses the conditional independence assumption (CIA), which implies that once observable characteristics like Z are controlled for, the adoption of water-management practices are random and uncorrelated with the outcome variables1. In this case, the PSM can be expressed as:  
formula
(2)
where I = {0,1} is the indicator for adoption and Z is the vector of pre-adoption features. The conditional distribution of Z, given p(Z), is alike in both groups of adopters and non-adopters.

The main advantage of the PSM is that it does not need any functional form assumption, unlike the parametric methods. However, the main drawback of the approach is the strong assumption of unconfoundedness. Smith & Todd (2005) point out that there might be systematic variances among outcomes of farmers who embraced improved water-management practices and those who did not adopt improved water-management practices. However, Jalan & Ravallion (2003) pointed out that the assumption is no more restrictive than those of an instrumental variable approach employed in cross-sectional data analysis. Michalopoulos et al. (2004) compared different non-experimental methods in the absence of random assignment and found that PSM provides the most robust estimates. With PSM, the average treatment affect for the treated (ATT) is of most significance.

Average treatment effects

The average treatment effect for the treated (ATT) is appraised after estimating the propensity scores:  
formula
(3)

Many methods are available to match adopters with non-adopters of similar propensity scores like nearest-neighbor matching (NNM), kernel matching, caliper/radius matching, stratification matching, and mahalanobis metric matching. In practice, the choice of matching method often appears to make little difference (Smith & Todd, 2005). In analysis with a small sample size, the choice of matching approach can be vital (Heckman et al., 1997). It is advantageous to use more than one matching approach because the performance of alternative matching approaches varies case-by-case and depends mainly on the data structure at hand (Zhao, 2003). In the current paper, NNM and kernel-based matching (KBM) are employed. After matching, the quality of matching is tested by employing a number of econometric tests like median absolute bias before and after matching, the value of pre- and post-matching, and the joint significance of covariates before and after matching.

Data and description of variables

The current data were collected from the four major provinces of Pakistan, i.e., Punjab, Sindh, KPK, and Balochistan. A detailed questionnaire was developed and the questionnaire was pilot tested prior to the implementation of the survey. In total, data were collected from 950 farmers across Pakistan; about 350 farmers were interviewed from the Punjab province, 250 each from Sindh and KPK provinces, and about 100 farmers were interviewed from Balochistan. The survey for the collection of the data was managed by a team of well-trained enumerators in 2014. The questionnaire included information on a number of household and farm-level characteristics and particularly water-management practices adopted by the farmers (see Questionnaire in Appendix A, available with the online version of this paper).

The description of variables used in the study is presented in Table 2. The results show that a large number of farm households adopt improved agricultural water-management practices: about 62% of farmers adopted the bund-making technology to save irrigation water and 51% of the farmers have adopted the deep plowing technology. About 48% of the farmers have supplemented tube well water with canal water, but only 15% of the farmers have adopted stress-tolerant varieties to cope with water scarcity.

Table 2.

Data and description of variables.

Variable Description Mean Std. Dev. 
Province 
 Punjab 1 if the farmer is from Punjab province, 0 otherwise 0.38 0.48 
 Sindh 1 if the farmer is from Sindh province, 0 otherwise 0.26 0.44 
 KPK 1 if the farmer is from KPK province, 0 otherwise 0.26 0.44 
 Balochistan 1 if the farmer is from Balochistan province, 0 otherwise 0.11 0.31 
Water-management practice 
 Bund making 1 if the farmers have adopted bund-making practices, 0 otherwise 0.62 0.13 
 Deep plowing 1 if the farmers have adopted deep plowing practices, 0 otherwise 0.51 0.18 
 Supplementing irrigation 1 if the farmers practiced using water from more than one source, 0 otherwise 0.48 0.22 
 Stress-tolerant varieties 1 if the farmers have adopted stress-tolerant varieties, 0 otherwise 0.15 0.24 
Demographic 
 Age of farmer Age of the farmer in years 43.47 12.65 
 Male farmer (dummy) 1 if the farmer is male, 0 female farmer 0.93 0.24 
 Joint family system (dummy) 1 if the farmer is living in joint family system, 0 otherwise 0.56 0.49 
Human capital 
 Education Education of the farmer in years 7.91 5.54 
 Experience Experience of the farmer in years 21.56 14.29 
Land assets 
 Land owned Landholding of the farmer in number of acres 22.45 76.79 
 Landowner (dummy) 1 if the farmer is the owner of the land, 0 otherwise 0.81 0.38 
 Irrigated land (dummy) 1 if the area is irrigated, 0 rain-fed area 0.91 0.28 
 Good soil quality (dummy) 1 if the soil is of good quality, 0 otherwise 0.63 0.48 
Access to facilities 
 Electricity available (dummy) 1 if the village has electricity, 0 otherwise 0.94 0.23 
 Implement repairs (dummy) 1 if the village has implement repair shop, 0 otherwise 0.23 0.42 
 Agriculture extension (dummy) 1 if the village has access to agricultural extension, 0 otherwise 0.62 0.29 
 NGOs 1 if the village has NGOs, 0 otherwise 0.10 0.30 
Farm assets 
 Tractor owner (dummy) 1 if the household owns a tractor, 0 otherwise 0.32 0.46 
 Tube well owner (dummy) 1 if the household owns a tube well, 0 otherwise 0.35 0.47 
 Canal owner (dummy) 1 if the household uses canal water, 0 otherwise 0.74 0.31 
 Moldbold (MB) plow owner (dummy) 1 if the household has an MB plow, 0 otherwise 0.21 0.40 
 Seed drill owner (dummy) 1 if the household has a seed drill, 0 otherwise 0.12 0.32 
 Reaper owner (dummy) 1 if the household has a reaper, 0 otherwise 0.04 0.21 
Household durable assets 
 Car owner (dummy) 1 if the household has a car, 0 otherwise 0.20 0.40 
 Refrigerator owner (dummy) 1 if the household has a refrigerator, 0 otherwise 0.70 0.45 
 Air conditioner owner (dummy) 1 if the household has air conditioning, 0 otherwise 0.14 0.34 
 Room cooler owner (dummy) 1 if the household has a room cooler, 0 otherwise 0.27 0.44 
 Iron owner (dummy) 1 if the household has an iron, 0 otherwise 0.93 0.24 
 Television owner (dummy) 1 if the household owns a TV, 0 otherwise 0.79 0.78 
Outcome variable 
 Wheat yield Wheat yield in kilograms per hectare 3,161 239 
 Rice yield Rice yield in kilograms per hectares 2,675 561 
 Maize yield Maize yield in kilograms per hectare 4,293 783 
 Food security 1 if the household is food secure, 0 otherwise 0.69 0.34 
 Povertya Headcount index 1 if the household is poor, 0 otherwise 0.23 0.11 
 Income Per month household income in rupees 14,589 2,560 
Variable Description Mean Std. Dev. 
Province 
 Punjab 1 if the farmer is from Punjab province, 0 otherwise 0.38 0.48 
 Sindh 1 if the farmer is from Sindh province, 0 otherwise 0.26 0.44 
 KPK 1 if the farmer is from KPK province, 0 otherwise 0.26 0.44 
 Balochistan 1 if the farmer is from Balochistan province, 0 otherwise 0.11 0.31 
Water-management practice 
 Bund making 1 if the farmers have adopted bund-making practices, 0 otherwise 0.62 0.13 
 Deep plowing 1 if the farmers have adopted deep plowing practices, 0 otherwise 0.51 0.18 
 Supplementing irrigation 1 if the farmers practiced using water from more than one source, 0 otherwise 0.48 0.22 
 Stress-tolerant varieties 1 if the farmers have adopted stress-tolerant varieties, 0 otherwise 0.15 0.24 
Demographic 
 Age of farmer Age of the farmer in years 43.47 12.65 
 Male farmer (dummy) 1 if the farmer is male, 0 female farmer 0.93 0.24 
 Joint family system (dummy) 1 if the farmer is living in joint family system, 0 otherwise 0.56 0.49 
Human capital 
 Education Education of the farmer in years 7.91 5.54 
 Experience Experience of the farmer in years 21.56 14.29 
Land assets 
 Land owned Landholding of the farmer in number of acres 22.45 76.79 
 Landowner (dummy) 1 if the farmer is the owner of the land, 0 otherwise 0.81 0.38 
 Irrigated land (dummy) 1 if the area is irrigated, 0 rain-fed area 0.91 0.28 
 Good soil quality (dummy) 1 if the soil is of good quality, 0 otherwise 0.63 0.48 
Access to facilities 
 Electricity available (dummy) 1 if the village has electricity, 0 otherwise 0.94 0.23 
 Implement repairs (dummy) 1 if the village has implement repair shop, 0 otherwise 0.23 0.42 
 Agriculture extension (dummy) 1 if the village has access to agricultural extension, 0 otherwise 0.62 0.29 
 NGOs 1 if the village has NGOs, 0 otherwise 0.10 0.30 
Farm assets 
 Tractor owner (dummy) 1 if the household owns a tractor, 0 otherwise 0.32 0.46 
 Tube well owner (dummy) 1 if the household owns a tube well, 0 otherwise 0.35 0.47 
 Canal owner (dummy) 1 if the household uses canal water, 0 otherwise 0.74 0.31 
 Moldbold (MB) plow owner (dummy) 1 if the household has an MB plow, 0 otherwise 0.21 0.40 
 Seed drill owner (dummy) 1 if the household has a seed drill, 0 otherwise 0.12 0.32 
 Reaper owner (dummy) 1 if the household has a reaper, 0 otherwise 0.04 0.21 
Household durable assets 
 Car owner (dummy) 1 if the household has a car, 0 otherwise 0.20 0.40 
 Refrigerator owner (dummy) 1 if the household has a refrigerator, 0 otherwise 0.70 0.45 
 Air conditioner owner (dummy) 1 if the household has air conditioning, 0 otherwise 0.14 0.34 
 Room cooler owner (dummy) 1 if the household has a room cooler, 0 otherwise 0.27 0.44 
 Iron owner (dummy) 1 if the household has an iron, 0 otherwise 0.93 0.24 
 Television owner (dummy) 1 if the household owns a TV, 0 otherwise 0.79 0.78 
Outcome variable 
 Wheat yield Wheat yield in kilograms per hectare 3,161 239 
 Rice yield Rice yield in kilograms per hectares 2,675 561 
 Maize yield Maize yield in kilograms per hectare 4,293 783 
 Food security 1 if the household is food secure, 0 otherwise 0.69 0.34 
 Povertya Headcount index 1 if the household is poor, 0 otherwise 0.23 0.11 
 Income Per month household income in rupees 14,589 2,560 

a$US1 per person per day is used as the poverty line.

Approximately 43.5 years was the mean age of the farmers and about 93% of the farmers were male and only 7% were female. Against the belief that rural Pakistani households are predominantly joint family systems, the results show that only 56% of the surveyed households lived in the joint family system and about 44% lived in nuclear families. The average years of schooling of farmers stood at 7.91 years and the average years of experience of the farmer was 21.5 years.

The average land holding per household was about 22 acres and 81% of the farmers were landowners. Most of the agricultural area, i.e., 91%, was irrigated and 9% was rain-fed. Only about 63% of the farmers reported that the soil was of good quality; 37% had poor quality soil.

Access to facilities and infrastructure is crucial for technology adoption; hence, we carefully evaluated the access to facilities and infrastructure. The result shows that the vast majority of the villages (94%) have access to electricity and only 62% of the farm households have access to an agricultural extension, which is critical for scaling technology and improving farm productivity. Only 23% of the villages have implement repair shops and merely 10% of the households have membership in a non-governmental organization (NGO).

Household assets indicate family wealth and the household's purchasing power, and play a vital role in technology adoption. Analysis of the information on the farm-level assets shows that about 32% of the households own tractors, 35% of the households own tube wells, 21% of the households own moldbold (MB) ploughs, only 2% of the households own seed drills, 4% of the households own reapers, and about 74% of the households have access to canal water.

The data on the durable household assets show that about 20% of the households own cars, 70% of the households own refrigerators, 14% of the households own air conditioners (AC), 27% of the households own room coolers, 93% of the households own irons, and 79% of the households own televisions. Figure 1 presents the different irrigation sources adopted by the households: canals constitute 62% of the sources of irrigation, tube wells constitute 28%, 4% comes from wells, and 6% from other sources.

Fig. 1.

Water sources (percentage share).

Fig. 1.

Water sources (percentage share).

Empirical analysis

Determinants of the water-saving management practices: multivariate probit analysis

The surveyed farm households commonly practice four different water-saving practices: bund making, deep plowing, supplementing irrigation, and stress-tolerant varieties. Since most of the farm households adopt these improved agricultural water-management practices jointly, the multivariate probit model is estimated and the results are presented in Table 3. The cross-equation correlations for bund making, deep plowing, irrigation water supplement, and adoption of stress-tolerant varieties are positive and highly significant, indicating that these equations need to be estimated jointly.

Table 3.

Improved water-management practices (multivariate probit model).

Variable Bund making Deep plow Irrigation supplement Stress-tolerant varieties 
Demographic 
 Age of farmer −0.01***(−2.46) −0.02**(−2.75) −0.01*(−1.66) −0.03***(−2.74) 
 Male farmer (dummy) 0.02***(3.15) 0.02**(2.27) 0.01**(1.98) 0.01***(3.10) 
 Joint family (dummy) 0.02(1.25) 0.01(1.37) 0.02 (1.46) 0.01(1.21) 
Human capital 
 Education 0.02***(2.74) 0.01***(3.02) 0.03***(2.81) 0.01***(2.42) 
Land assets 
 Land size 0.03***(3.19) 0.02**(2.03) 0.02***(2.63) 0.01***(3.52) 
 Land owner (dummy) 0.01*(1.80) 0.02**(2.11) 0.03***(2.65) 0.02**(2.36) 
 Irrigation land (dummy) −0.02*(−1.70) −0.01***(−2.83) 0.01***(2.51) 0.01*(1.77) 
 Good soil quality (dummy) 0.02**(2.31) 0.01(1.56) 0.02***(2.97) 0.03*(1.72) 
Access to facilities 
 Electricity 0.02***(2.68) 0.01**(2.17) 0.01***(2.76) 0.02**(2.37) 
 Implement repairs −0.01**(−2.15) −0.02**(−2.30) −0.02***(−2.83) −0.01***(−2.94) 
 OFWM 0.02**(2.14) 0.01***(2.57) 0.02***(3.26) 0.01**(2.13) 
 Agriculture extension dummy 0.14***(2.73) 0.12**(2.06) 0.07(1.43) 0.11***(2.53) 
 NGOs (dummy) 0.01***(3.17) 0.02**(1.98) 0.01***(3.04) 0.01**(2.18) 
Farm assets ownership 
 Tractor owner (dummy) 0.02***(2.92) 0.02***(3.24) 0.01***(2.85) 0.01***(2.56) 
 Tube well owner (dummy) 0.01***(3.16) 0.03***(2.75) 0.02***(3.19) 0.01***(2.44) 
 Canal owner (dummy) 0.01***(2.75) 0.03***(3.06) 0.01***(2.81) 0.02***(2.77) 
 MB plow owner (dummy) −0.01***(−2.11) 0.02***(2.83) −0.03***(−1.92) −0.01***(−2.50) 
Household durable assets 
 Car owner (dummy) 0.03***(2.63) 0.01***(2.51) 0.01**(2.10) 0.02*(1.83) 
 Refrigerator owner (dummy) 0.01**(2.31) 0.02***(3.10) 0.02***(3.24) 0.03***(2.79) 
 Room cooler owner (dummy) 0.01***(2.62) 0.02**(2.30) 0.01***(2.51) 0.02*(1.69) 
 Television owner (dummy) 0.02**(2.15) 0.01***(2.74) 0.03***(2.87) 0.01***(3.11) 
Location (base category is Balochistan) 
 Punjab 0.01**(2.27) 0.02***(2.81) 0.03***(3.16) 0.02**(2.25) 
 Sindh 0.02**(2.30) 0.01***(2.72) 0.02**(2.13) 0.01***(2.71) 
 KPK −0.01**(−2.16) −0.02**(−2.18) −0.03***(−2.85) 0.02*(1.72) 
 Constant 0.02***(2.74) 0.01**(2.25) −0.03***(2.84) 0.01***(2.65) 
 Number of observations 950 950 950 950 
 Cross equation correlations 0.23***(2.84) 0.65***(2.60) 0.55***(3.18) 0.75***(3.42) 
 Cross equation correlations 0.76***(3.16) 0.41***(3.10)   
 0.19    
 LR- 123.41    
 Prob >  0.000    
Variable Bund making Deep plow Irrigation supplement Stress-tolerant varieties 
Demographic 
 Age of farmer −0.01***(−2.46) −0.02**(−2.75) −0.01*(−1.66) −0.03***(−2.74) 
 Male farmer (dummy) 0.02***(3.15) 0.02**(2.27) 0.01**(1.98) 0.01***(3.10) 
 Joint family (dummy) 0.02(1.25) 0.01(1.37) 0.02 (1.46) 0.01(1.21) 
Human capital 
 Education 0.02***(2.74) 0.01***(3.02) 0.03***(2.81) 0.01***(2.42) 
Land assets 
 Land size 0.03***(3.19) 0.02**(2.03) 0.02***(2.63) 0.01***(3.52) 
 Land owner (dummy) 0.01*(1.80) 0.02**(2.11) 0.03***(2.65) 0.02**(2.36) 
 Irrigation land (dummy) −0.02*(−1.70) −0.01***(−2.83) 0.01***(2.51) 0.01*(1.77) 
 Good soil quality (dummy) 0.02**(2.31) 0.01(1.56) 0.02***(2.97) 0.03*(1.72) 
Access to facilities 
 Electricity 0.02***(2.68) 0.01**(2.17) 0.01***(2.76) 0.02**(2.37) 
 Implement repairs −0.01**(−2.15) −0.02**(−2.30) −0.02***(−2.83) −0.01***(−2.94) 
 OFWM 0.02**(2.14) 0.01***(2.57) 0.02***(3.26) 0.01**(2.13) 
 Agriculture extension dummy 0.14***(2.73) 0.12**(2.06) 0.07(1.43) 0.11***(2.53) 
 NGOs (dummy) 0.01***(3.17) 0.02**(1.98) 0.01***(3.04) 0.01**(2.18) 
Farm assets ownership 
 Tractor owner (dummy) 0.02***(2.92) 0.02***(3.24) 0.01***(2.85) 0.01***(2.56) 
 Tube well owner (dummy) 0.01***(3.16) 0.03***(2.75) 0.02***(3.19) 0.01***(2.44) 
 Canal owner (dummy) 0.01***(2.75) 0.03***(3.06) 0.01***(2.81) 0.02***(2.77) 
 MB plow owner (dummy) −0.01***(−2.11) 0.02***(2.83) −0.03***(−1.92) −0.01***(−2.50) 
Household durable assets 
 Car owner (dummy) 0.03***(2.63) 0.01***(2.51) 0.01**(2.10) 0.02*(1.83) 
 Refrigerator owner (dummy) 0.01**(2.31) 0.02***(3.10) 0.02***(3.24) 0.03***(2.79) 
 Room cooler owner (dummy) 0.01***(2.62) 0.02**(2.30) 0.01***(2.51) 0.02*(1.69) 
 Television owner (dummy) 0.02**(2.15) 0.01***(2.74) 0.03***(2.87) 0.01***(3.11) 
Location (base category is Balochistan) 
 Punjab 0.01**(2.27) 0.02***(2.81) 0.03***(3.16) 0.02**(2.25) 
 Sindh 0.02**(2.30) 0.01***(2.72) 0.02**(2.13) 0.01***(2.71) 
 KPK −0.01**(−2.16) −0.02**(−2.18) −0.03***(−2.85) 0.02*(1.72) 
 Constant 0.02***(2.74) 0.01**(2.25) −0.03***(2.84) 0.01***(2.65) 
 Number of observations 950 950 950 950 
 Cross equation correlations 0.23***(2.84) 0.65***(2.60) 0.55***(3.18) 0.75***(3.42) 
 Cross equation correlations 0.76***(3.16) 0.41***(3.10)   
 0.19    
 LR- 123.41    
 Prob >  0.000    

Note: The results are significant at ***,**,* the 1%, 5%, and 10% levels, respectively. t-value in parentheses.

The age of the farmer is negative and significant for all four improved water-management practices, indicating that older farmers are less likely to adopt the improved techniques. This may be due to the fact that older households are inclined to continue with the traditional technology. The coefficient of the gender of the farmer dummy (1 for male and 0 for female), is positive and highly significant for all four practices, indicating that male farmers are more likely to adopt these water-management practices compared to female farmers. Female farmers seem to be disadvantaged when it comes to technology adoption because of the lack of a network and the necessary skills and resources to implement and manage improved practices. Although the coefficient of the joint family dummy is positive, it is non-significant.

The coefficient years of schooling of the farmer is positive and highly significant for all four improved practices, signifying that educated farmers have a greater likelihood of adopting these water-management practices compared to uneducated farmers. Educated farmers are aware of the practices and their impact on yield and knowledge.

As land is the most important asset of farm households, the results show that the amount of land owned is positive and highly significant, indicating that farm households with more land holdings have a higher probability of adopting improved practices. The coefficient of the landowner dummy (1 for owner and 0 for tenant) is positive and significant, demonstrating that the owners are more likely to adopt these practices compared to tenants. The dummy for the good soil quality was positive and significant for bund making, supplementing irrigation, and stress-tolerant varieties, while it was positive and non-significant for deep plowing. The coefficient of irrigated land dummy variable (1 for irrigated areas and 0 for rain-fed) was negative and significant for bund making and deep plow, indicating that farmers in rain-fed areas do not practice bund making and deep plow methods of water management. The results for the irrigation supplement and stress-tolerant varieties are positive and significant, indicating that farmers in irrigated areas mostly adopt irrigation supplement and stress-tolerant varieties as water-management practices.

The ownership of the farm and durable household assets indicate the wealth status of the households, which is an important driver for adoption of farm technology. The coefficient of the tractor, tube well, and canal ownership are positive and significant at the 1% level, indicating the importance of wealth on the adoption of all four water-management technologies. The coefficient of the MB ownership dummy was negative and significant. The coefficient of the ownership of the durable household assets such as a car, refrigerator, room cooler, and television are positive and significant, indicating the crucial role of wealth in adopting water-management technologies at the farm level. We also summarized the variables for the treatment (adopter) and control group (non-adopter) and the results are summarized in Appendix B (available with the online version of this paper), which confirms the findings of the multivariate probit.

Determinants of the number of water-management practices adopted by the farmer: Poisson regression

The Poisson regression is employed to assess the determinants of the number of water-management practices adopted by the farm household and the results are presented in Table 4. The dependent variable is the number of water-management practices adopted by the farmers and, based on the literature review, several independent variables are included in the model.

Table 4.

Number of irrigation water-management practices adopted by farmers (Poisson estimates).

Variable Coefficient t-values  
Demographic 
 Age of farmer −0.01*** 2.52  
 Male farmer (dummy) 0.02 1.34  
 Joint family system (dummy) 0.02* 1.74  
Human capital 
 Education 0.02*** 3.12  
Land asset 
 Land owned 0.02*** 2.55  
 Land owner (dummy) 0.03** 1.98  
 Irrigated land (dummy) −0.03*** −3.07  
 Good soil quality (dummy) 0.02 0.82  
Access to facilities 
 Access to electricity (dummy) 0.01 1.34  
 Implement repairs (dummy) −0.03 −0.69  
 OFWM (dummy) 0.02 1.45  
 NGOs (dummy) 0.01*** 2.56  
 Agriculture extension (dummy)  0.06*** 2.77 
Farm assets 
 Tractor owner (dummy) 0.02*** 2.80  
 Tube well owner (dummy) 0.03*** 2.55  
 Canal owner (dummy) 0.02*** 2.69  
 MB plow owner (dummy) 0.01 1.32  
Durable household assets 
 Car owner (dummy) 0.02*** 2.55  
 Refrigerator owner (dummy) 0.03** 2.16  
 Room cooler owner (dummy) 0.02 1.24  
 Television owner (dummy) 0.04** 2.05  
Location (base category is Balochistan) 
 Punjab 0.02*** 3.11  
 Sindh 0.02** 2.18  
 KPK 0.01*** 3.14  
 Constant 0.02*** 2.47  
 Number of observations 950 950  
 0.19   
 LR- 123.41   
 Prob >  0.000   
Variable Coefficient t-values  
Demographic 
 Age of farmer −0.01*** 2.52  
 Male farmer (dummy) 0.02 1.34  
 Joint family system (dummy) 0.02* 1.74  
Human capital 
 Education 0.02*** 3.12  
Land asset 
 Land owned 0.02*** 2.55  
 Land owner (dummy) 0.03** 1.98  
 Irrigated land (dummy) −0.03*** −3.07  
 Good soil quality (dummy) 0.02 0.82  
Access to facilities 
 Access to electricity (dummy) 0.01 1.34  
 Implement repairs (dummy) −0.03 −0.69  
 OFWM (dummy) 0.02 1.45  
 NGOs (dummy) 0.01*** 2.56  
 Agriculture extension (dummy)  0.06*** 2.77 
Farm assets 
 Tractor owner (dummy) 0.02*** 2.80  
 Tube well owner (dummy) 0.03*** 2.55  
 Canal owner (dummy) 0.02*** 2.69  
 MB plow owner (dummy) 0.01 1.32  
Durable household assets 
 Car owner (dummy) 0.02*** 2.55  
 Refrigerator owner (dummy) 0.03** 2.16  
 Room cooler owner (dummy) 0.02 1.24  
 Television owner (dummy) 0.04** 2.05  
Location (base category is Balochistan) 
 Punjab 0.02*** 3.11  
 Sindh 0.02** 2.18  
 KPK 0.01*** 3.14  
 Constant 0.02*** 2.47  
 Number of observations 950 950  
 0.19   
 LR- 123.41   
 Prob >  0.000   

Note: The results are significant at ***, **, * the 1%, 5%, and 10% levels, respectively.

The coefficient of the age of the farmer is negative and significant, indicating that young farmers are more likely to adopt more water-management practices as younger farmers ought to have the awareness and skills needed for implementing and managing the improved technology. The coefficient of the gender of farmer (male dummy) is positive but non-significant. The coefficient of the joint family system dummy is positive and significant, indicating those farmers living in joint family systems adopt more improved water-management practices, as a larger labor force is available for implementation and management of these practices. The years of schooling of the farmer is positive and significant, demonstrating that educated farmers are more likely to adopt more water-management practices.

The result for the landholding coefficient is positive and significant, indicating that farmers with large land assets have a higher probability of adopting more water-management practices. The irrigated land dummy, i.e., 1 for the irrigated areas and 0 for the rain-fed, is negative and significant, indicating that farmers with rain-fed areas have adopted fewer improved water-management practices compared to farmers with irrigated land. The dummy land ownership is positive and significant, indicating that farmers who own land have adopted more water-management practices.

The access to electricity, implement repair, and OFWM are not significant. The coefficient of NGOs dummy is positive and significant, highlighting the importance of awareness through NGOs in the adoption of improved water-management practices.

The ownership of farm assets dummy such as a tractor, tube well, and canal is positive and significant at the 1% level of significance, indicating the role of wealth in the adoption of improved water-management practices. The coefficient of the household durable asset ownership such as television, refrigerator, and car is positive and highly significant at the 1% level of significance, signifying the importance of wealth on the number of water-saving practices adopted by farmers. The is highly significant at the 1% level of significance, indicating the robustness of the variables included in the model.

Impact of irrigation water-management practices: propensity score matching approach analysis

The impact was estimated by employing the PSM approach2. The results are presented in Table 5. This PSM approach was employed to correct for the potential sample selection biasedness that might arise due to systematic differences between the adopters of irrigation water-management practices and the non-adopters of these practices. PSM creates the condition of a randomized experiment. Two different matching algorithms, i.e., NNM3 and KBM4 were employed to correct for the potential sample selection biasedness5. According to Sianesi (2004), it is always better to employ more than one matching6 algorithm; for this reason, in the current analysis, NNM and KBM matching algorithms were employed. For each irrigation water-management practice, i.e., bund making, deep plowing, supplementing irrigation, and adoption of stress-tolerant varieties, the impact was separately estimated. The impact was estimated on wheat and rice yields, household income, food security, and poverty levels. A dummy variable was used to measure the impact on food security: if a household is food secure, it is 1 and 0 if the household is food insecure7. The impact of water-management practices on wheat and rice crop yields was estimated in kilograms8 and the impact on the household income was estimated in terms of Pakistani rupees and the poverty level in terms of the headcount index.

Table 5.

Impact of the irrigation water-management practices.

Practice Matching algorithm Outcome ATT t-value Critical level of hidden bias Number of treated Number of control 
Bund making NNM Food security 0.09*** 2.77 1.25–1.30 256 397 
Wheat yield 1.28** 2.08 1.45–1.50 271 469 
Rice yield 1.43*** 3.15 1.10–1.15 384 425 
Income 3,346*** 2.54 1.25–1.30 264 352 
Poverty −0.04*** −3.08 1.35–1.40 273 251 
KBM Food security 0.14*** 2.63 1.55–1.60 253 385 
Wheat yield 1.45* 1.92 1.25–1.30 241 437 
Rice yield 1.62*** 2.51 1.85–1.90 232 369 
Income 4,237*** 3.16 1.35–1.40 182 265 
Poverty −0.06** −2.59 1.20–1.25 181 169 
Deep plowing NNM Food security 0.03 1.45 – 243 262 
Wheat yield 0.67* 1.88 1.30–1.35 230 274 
Rice yield 1.34** 2.55 1.25–1.30 215 265 
Income 2,573 1.28 – 231 265 
Poverty −0.03 −1.27 – 243 287 
KBM Food security 0.04 0.84 – 235 264 
Wheat yield 0.91 0.73 – 114 255 
Rice yield 1.21* 1.68 1.45–1.50 245 358 
Income 3,145 1.49 – 227 264 
Poverty −0.02 −1.37 – 243 285 
Irrigation supplement NNM Food security 0.05*** 2.94 1.15–1.20 252 296 
Wheat yield 1.26* 1.74 1.35–1.40 240 263 
Rice yield 2.13*** 2.70 1.25–1.30 285 358 
Income 3,752*** 2.56 1.35–1.40 271 396 
Poverty −0.07* −1.72 1.15–1.20 249 340 
KBM Food security 0.06** 2.14 1.35–1.40 264 312 
Wheat yield 1.23*** 3.87 1.30–1.35 236 319 
Rice yield 1.57* 1.82 1.25–1.30 144 328 
Income 4,528*** 3.16 1.35–1.40 242 315 
Poverty −0.04** −2.18 1.40–1.45 257 310 
Stress-tolerant varieties NNM Food security 0.13** 2.11 1.25–1.30 267 335 
Wheat yield 1.76** 2.03 1.05–1.10 254 285 
Rice yield 2.13*** 3.18 1.20–1.25 265 318 
Income 3,852* 1.96 1.35–1.40 237 315 
Poverty −0.05* 1.82 1.25–1.30 255 314 
KBM Food security 0.12*** 3.16 1.40–1.45 241 283 
Wheat yield 1.26** 2.44 1.35–1.40 312 354 
Rice yield 1.82*** 3.57 1.25–1.30 273 373 
Income 4,926*** 2.58 1.15–1.20 235 340 
Poverty −0.06*** 2.73 1.35–1.40 251 285 
Practice Matching algorithm Outcome ATT t-value Critical level of hidden bias Number of treated Number of control 
Bund making NNM Food security 0.09*** 2.77 1.25–1.30 256 397 
Wheat yield 1.28** 2.08 1.45–1.50 271 469 
Rice yield 1.43*** 3.15 1.10–1.15 384 425 
Income 3,346*** 2.54 1.25–1.30 264 352 
Poverty −0.04*** −3.08 1.35–1.40 273 251 
KBM Food security 0.14*** 2.63 1.55–1.60 253 385 
Wheat yield 1.45* 1.92 1.25–1.30 241 437 
Rice yield 1.62*** 2.51 1.85–1.90 232 369 
Income 4,237*** 3.16 1.35–1.40 182 265 
Poverty −0.06** −2.59 1.20–1.25 181 169 
Deep plowing NNM Food security 0.03 1.45 – 243 262 
Wheat yield 0.67* 1.88 1.30–1.35 230 274 
Rice yield 1.34** 2.55 1.25–1.30 215 265 
Income 2,573 1.28 – 231 265 
Poverty −0.03 −1.27 – 243 287 
KBM Food security 0.04 0.84 – 235 264 
Wheat yield 0.91 0.73 – 114 255 
Rice yield 1.21* 1.68 1.45–1.50 245 358 
Income 3,145 1.49 – 227 264 
Poverty −0.02 −1.37 – 243 285 
Irrigation supplement NNM Food security 0.05*** 2.94 1.15–1.20 252 296 
Wheat yield 1.26* 1.74 1.35–1.40 240 263 
Rice yield 2.13*** 2.70 1.25–1.30 285 358 
Income 3,752*** 2.56 1.35–1.40 271 396 
Poverty −0.07* −1.72 1.15–1.20 249 340 
KBM Food security 0.06** 2.14 1.35–1.40 264 312 
Wheat yield 1.23*** 3.87 1.30–1.35 236 319 
Rice yield 1.57* 1.82 1.25–1.30 144 328 
Income 4,528*** 3.16 1.35–1.40 242 315 
Poverty −0.04** −2.18 1.40–1.45 257 310 
Stress-tolerant varieties NNM Food security 0.13** 2.11 1.25–1.30 267 335 
Wheat yield 1.76** 2.03 1.05–1.10 254 285 
Rice yield 2.13*** 3.18 1.20–1.25 265 318 
Income 3,852* 1.96 1.35–1.40 237 315 
Poverty −0.05* 1.82 1.25–1.30 255 314 
KBM Food security 0.12*** 3.16 1.40–1.45 241 283 
Wheat yield 1.26** 2.44 1.35–1.40 312 354 
Rice yield 1.82*** 3.57 1.25–1.30 273 373 
Income 4,926*** 2.58 1.15–1.20 235 340 
Poverty −0.06*** 2.73 1.35–1.40 251 285 

*, **, *** indicates that the results are significant at 1, 5 and 10 percent levels, respectively.

Impact of the bund-making practice

Bund making is a traditional practice and farmers have been using it for centuries9. The impact of bund making was estimated on food security, rice and wheat yields, household income, and poverty levels. The results of the average treatment affect for the treated (ATT) for food security was positive and significant both for NNM and KBM, indicating that the households who adopted the bund-making technology have higher food security levels compared to similar households who have not adopted the bund-making technology10. The ATT results for the rice and wheat yields are positive and significant, indicating that farm households who adopted the bund-making technology have higher wheat yields (in the range of 51–58 kg per acre) and higher rice yields (in the range of 57–65 kg per acre). The ATT results for household income indicated that household income levels are higher (in the range of rupees 3,346–4,237) for those households that have adopted the bund-making technology compared to households that have not adopted it11. The ATT results for poverty are negative and significant, indicating that household poverty levels are less (in the range of 4–6%) for the households that have adopted the bund-making technology.

Impact of deep plowing

Deep plowing is a common agricultural water-management practice especially in the rain-fed areas; this helps to conserve soil moisture. The farmers in the rain-fed areas commonly practice deep-plowing technology to conserve moisture. The impact of deep plowing was estimated on food security, rice and wheat yields, household income, and poverty levels. The ATT results for food security are positive and non-significant. The ATT results for the wheat and rice yields are positive and significant, indicating that households practicing deep plowing have higher rice and wheat yields compared to households which do not practice deep plowing. The ATT results for household income are positive although non-significant for NNM and KBM. The ATT results for poverty are negative although non-significant12. The empirical results indicate that deep plowing has quite promising results, but it can be more effective if managed on time (before the start of crop season) especially in the rain-fed areas.

Impact of irrigation supplement

The majority of farmers supplement irrigation water from more than one source. The most popular sources are wells, tube wells, and canals13. In Pakistan, about 80% of the area is irrigated through canal irrigation systems14, but over the years the availability of canal water has declined considerably, forcing farmers to supplement with other sources; this is used as a water-management practice throughout Pakistan. The recent energy crisis has multiplied farmers' problems. The ATT results for the irrigation supplement are positive and significant for household food security levels and the food security levels are higher (in the range of 5–6%) in the case of NNM and KBM. The ATT results for the wheat and rice yields are positive and significant, indicating that households have higher yield levels (in the range of 49–51 kg per acre for wheat and 85–102 kg per acre for rice) compared to households not supplementing their crops with irrigation water. The ATT results for household income indicate that household income levels are higher: in the range of Pakistani rupees 3,752–4,528 per acre. The ATT results for poverty levels indicate that poverty levels are lower: in the range of 4–7% for those households with supplementary irrigation sources.

Impact of stress-tolerant varieties

The adoption of the less-water-demanding varieties is another option for managing irrigation water. The impact of stress-tolerant varieties was mostly estimated on household food security, crop yields, household income, and poverty levels. The ATT results indicated that household food security levels are higher (in the range of 12–13%) compared to households that have not adopted stress-tolerant varieties. The wheat yields are higher, in the range of 50–70 kg per acre and rice yields are higher, in the range of 127–142 kg per acre. The ATT results for household income indicated that income is higher in the range of Pakistani rupees 3,852–4,926. The household poverty levels are lower in the range of 5–6% for the households having adopted stress-tolerant varieties.

The critical levels of hidden bias are reported in Table 6. The critical level of hidden bias indicates the level up to which the adopters and non-adopters differ in their odds of adoption due to unobservable factors15. The PSM results are in line with the previous studies (e.g., Ali & Sharif, 2011). The number of treated and the number of controls are reported in Table 6.

Table 6.

Indicators of covariates balancing (before and after matching).

Practice Matching algorithm Outcome Median absolute bias before matching Median absolute bias after matching Percentage bias reduction Value of R-square before matching Value of R-square after matching Joint significance of covariates before matching Joint significance of covariates after matching 
Bund making NNM Food security 23.46 5.72 75.62 0.342 0.003 0.002 0.274 
Wheat yield 21.40 4.38 79.53 0.441 0.001 0.001 0.433 
Rice yield 22.57 5.13 77.27 0.382 0.002 0.003 0.452 
Income 20.75 5.67 72.67 0.345 0.003 0.002 0.321 
Poverty 22.93 6.25 72.74 0.293 0.002 0.001 0.275 
KBM Food security 19.21 4.61 76.00 0.314 0.001 0.003 0.358 
Wheat yield 21.68 5.26 75.74 0.320 0.002 0.002 0.234 
Rice yield 22.59 6.35 71.89 0.253 0.001 0.003 0.352 
Income 23.85 5.83 75.56 0.274 0.003 0.002 0.116 
Poverty 20.64 5.40 73.84 0.286 0.002 0.001 0.253 
Deep plowing NNM Food security 23.85 4.18 82.47 0.263 0.001 0.002 0.236 
Wheat yield 20.64 5.26 74.52 0.381 0.002 0.003 0.372 
Rice yield 23.74 4.51 81.00 0.240 0.001 0.001 0.250 
Income 22.83 5.19 77.27 0.261 0.003 0.002 0.244 
Poverty 21.42 4.37 79.60 0.335 0.002 0.001 0.285 
KBM Food security 24.51 5.10 79.19 0.269 0.004 0.001 0.243 
Wheat yield 23.49 4.36 81.44 0.350 0.003 0.002 0.215 
Rice yield 22.16 5.29 76.13 0.264 0.002 0.001 0.372 
Income 24.57 6.20 74.77 0.172 0.001 0.003 0.247 
Poverty 23.97 5.13 78.60 0.287 0.000 0.002 0.311 
Irrigation supplement NNM Food security 21.45 4.52 78.93 0.234 0.003 0.002 0.254 
Wheat yield 22.36 5.16 76.92 0.231 0.002 0.003 0.314 
Rice yield 24.51 4.72 80.74 0.252 0.001 0.001 0.243 
Income 21.36 5.27 75.33 0.243 0.003 0.002 0.272 
Poverty 19.42 4.70 75.80 0.315 0.002 0.003 0.253 
KBM Food security 18.35 5.23 71.50 0.262 0.001 0.002 0.238 
Wheat yield 20.45 5.10 75.06 0.274 0.003 0.001 0.241 
Rice yield 23.42 6.23 73.40 0.263 0.002 0.000 0.237 
Income 22.31 5.23 76.56 0.250 0.001 0.003 0.391 
Poverty 20.16 5.30 73.71 0.331 0.003 0.002 0.236 
Stress-tolerant varieties NNM Food security 22.34 5.13 77.04 0.415 0.002 0.001 0.352 
Wheat yield 21.52 4.43 79.41 0.438 0.001 0.002 0.344 
Rice yield 22.56 5.36 76.24 0.516 0.003 0.003 0.452 
Income 18.23 4.62 74.66 0.482 0.002 0.002 0.382 
Poverty 19.45 5.26 72.96 0.462 0.003 0.002 0.511 
KBM Food security 20.21 4.72 76.65 0.563 0.002 0.003 0.673 
Wheat yield 21.63 5.24 75.77 0.455 0.001 0.002 0.418 
Rice yield 24.21 4.69 80.63 0.382 0.002 0.000 0.462 
Income 23.73 5.12 78.42 0.336 0.003 0.001 0.456 
Poverty 22.12 4.36 80.29 0.274 0.002 0.001 0.472 
Practice Matching algorithm Outcome Median absolute bias before matching Median absolute bias after matching Percentage bias reduction Value of R-square before matching Value of R-square after matching Joint significance of covariates before matching Joint significance of covariates after matching 
Bund making NNM Food security 23.46 5.72 75.62 0.342 0.003 0.002 0.274 
Wheat yield 21.40 4.38 79.53 0.441 0.001 0.001 0.433 
Rice yield 22.57 5.13 77.27 0.382 0.002 0.003 0.452 
Income 20.75 5.67 72.67 0.345 0.003 0.002 0.321 
Poverty 22.93 6.25 72.74 0.293 0.002 0.001 0.275 
KBM Food security 19.21 4.61 76.00 0.314 0.001 0.003 0.358 
Wheat yield 21.68 5.26 75.74 0.320 0.002 0.002 0.234 
Rice yield 22.59 6.35 71.89 0.253 0.001 0.003 0.352 
Income 23.85 5.83 75.56 0.274 0.003 0.002 0.116 
Poverty 20.64 5.40 73.84 0.286 0.002 0.001 0.253 
Deep plowing NNM Food security 23.85 4.18 82.47 0.263 0.001 0.002 0.236 
Wheat yield 20.64 5.26 74.52 0.381 0.002 0.003 0.372 
Rice yield 23.74 4.51 81.00 0.240 0.001 0.001 0.250 
Income 22.83 5.19 77.27 0.261 0.003 0.002 0.244 
Poverty 21.42 4.37 79.60 0.335 0.002 0.001 0.285 
KBM Food security 24.51 5.10 79.19 0.269 0.004 0.001 0.243 
Wheat yield 23.49 4.36 81.44 0.350 0.003 0.002 0.215 
Rice yield 22.16 5.29 76.13 0.264 0.002 0.001 0.372 
Income 24.57 6.20 74.77 0.172 0.001 0.003 0.247 
Poverty 23.97 5.13 78.60 0.287 0.000 0.002 0.311 
Irrigation supplement NNM Food security 21.45 4.52 78.93 0.234 0.003 0.002 0.254 
Wheat yield 22.36 5.16 76.92 0.231 0.002 0.003 0.314 
Rice yield 24.51 4.72 80.74 0.252 0.001 0.001 0.243 
Income 21.36 5.27 75.33 0.243 0.003 0.002 0.272 
Poverty 19.42 4.70 75.80 0.315 0.002 0.003 0.253 
KBM Food security 18.35 5.23 71.50 0.262 0.001 0.002 0.238 
Wheat yield 20.45 5.10 75.06 0.274 0.003 0.001 0.241 
Rice yield 23.42 6.23 73.40 0.263 0.002 0.000 0.237 
Income 22.31 5.23 76.56 0.250 0.001 0.003 0.391 
Poverty 20.16 5.30 73.71 0.331 0.003 0.002 0.236 
Stress-tolerant varieties NNM Food security 22.34 5.13 77.04 0.415 0.002 0.001 0.352 
Wheat yield 21.52 4.43 79.41 0.438 0.001 0.002 0.344 
Rice yield 22.56 5.36 76.24 0.516 0.003 0.003 0.452 
Income 18.23 4.62 74.66 0.482 0.002 0.002 0.382 
Poverty 19.45 5.26 72.96 0.462 0.003 0.002 0.511 
KBM Food security 20.21 4.72 76.65 0.563 0.002 0.003 0.673 
Wheat yield 21.63 5.24 75.77 0.455 0.001 0.002 0.418 
Rice yield 24.21 4.69 80.63 0.382 0.002 0.000 0.462 
Income 23.73 5.12 78.42 0.336 0.003 0.001 0.456 
Poverty 22.12 4.36 80.29 0.274 0.002 0.001 0.472 

It is essential to check the matching quality after carrying out the PSM. The median absolute bias before matching is high and is relatively low after matching. The percent drop in the bias is in the range of 70–80% which indicates that, after matching, the adopters of water-management practices and non-adopters are quite similar to each other. The value of before matching is quite high and is quite low after matching. The value of before and after matching indicates that after matching both the adopters and non-adopters are quite similar to each other. The p-value of joint significance of covariates should always be accepted before matching and should always be rejected after matching, indicating that after matching both the groups are quite similar to each other. The indicators of covariates balancing are also presented in Figure 2.

Fig. 2.

PSM estimates.

Fig. 2.

PSM estimates.

Conclusion

The present paper uses the rich primary data set collected in 2014 from four provinces of Pakistan, i.e., Sindh, Punjab, KPK, and Balochistan. The analysis shows that farmers commonly use four main types of improved agricultural water-management practices, namely bund making, deep plowing, supplementary irrigation from different sources, and the adoption of stress-tolerant varieties. The empirical analysis shows that education, land assets, farm assets, and durable household assets positively influence the adoption of these water-management practices.

The PSM estimates indicated the positive impact of the adoption of improved water-management practices on wheat and rice yields, food security, income levels, and lower poverty levels compared to the farm households who have not adopted these water-management practices. Among the different water-management practices, bund making, supplementary irrigation water, and the adoption of stress-tolerant varieties have a higher positive impact compared to the practice of deep plowing; however, deep plowing is especially helpful in the rain-fed areas compared to the irrigated areas.

Empirical findings of the paper indicate that the agricultural water-management policy should invest substantially in providing useful knowledge to the farmers on better management of scarce irrigation water for higher benefits. Importantly, the general education of the farmers in developing countries must be ensured; however, for most of the farmers, going back to school is not a feasible option. Hence, knowledge on improving irrigation management can be provided by organizing farmers' field days, demonstrations and strengthening agricultural extension services by the national government. Last, but not least, the practice of improved irrigation management can generate positive externality to the society by conserving increasingly depleting irrigation water across the globe. Thus, legislators and policymakers should prioritize training on irrigation water management in developing countries, not only to ensure the improved livelihood of farmers, but also to ensure water availability for the next generation.

1

Although it is a strong assumption.

2

The STATA statistical software was used for carrying out the propensity score matching analysis.

3

Nearest-neighbor matches with the closest similar nearest neighbor only.

4

Kernel-based matching takes the weighted average of all the non-participants and then matches with a similar participant.

5

In the case of propensity score matching the most important parameter of interest is the ATT (average treatment affect for the treated), i.e., the difference in the outcome of the similar adopters and non-adopters of the water-saving technology.

6

By employing more than one matching algorithm the results across different groups can be compared.

7

If the household was food secure then it was treated as 1 and if the household was food insecure then it was treated as 0.

8

One maund is equal to 40 kg.

9

Although bund making is a centuries-old technology it is still very effective.

10

Indicating that, although bund making is a simple technology, it has a huge impact on ensuring rural household food security levels.

11

One US$ is equal to 107 Pakistani rupees, hence 3,000–4,000 rupees indicates 30–40 $US.

12

The deep plowing is an effective practice to conserve moisture especially when drought conditions prevail.

13

There also exists other sources of irrigation like wells, streams, Karazs, etc.

14

This proportion is one of the highest in the world.

15

The critical levels of hidden bias for the insignificant variables is not significant hence it is not reported.

References

References
Ahmad
S.
, (
1993
).
Viability of Agriculture Resource Base: A Critical Appraisal. Agricultural Strategies in the 1990s: Issues and Options
.
Pakistan Association of Agricultural Social Scientists
,
Islamabad
, pp.
449
466
.
Ali
M.
&
Talukder
M.
, (
2008
).
Increasing water productivity in crop production – a synthesis
.
Agricultural Water Management
95
(
11
),
1201
1213
.
Ambast
S.
,
Tyagi
N.
&
Raul
S.
, (
2006
).
Management of declining groundwater in the Trans Indo-Gangetic Plain (India): some options
.
Agricultural Water Management
82
(
3
),
279
296
.
Becker
M.
&
Johnson
D.
, (
2001
).
Improved water control and crop management effects on lowland rice productivity in West Africa
.
Nutrient Cycling in Agroecosystems
59
(
2
),
119
127
.
Brown
M. E.
&
Funk
C. C.
, (
2008
).
Food security under climate change
.
Science
319
(
5863
),
580
581
.
FAO IFAD UNICEF WFP & WHO
(
2017
).
The State of Food Security and Nutrition in the World 2017. Building Resilience for Peace and Food Security
.
FAO
,
Rome
,
Italy
.
Gadanakis
Y.
,
Bennett
R.
,
Park
J.
&
Areal
F. J.
, (
2015
).
Improving productivity and water use efficiency: a case study of farms in England
.
Agricultural Water Management
160
,
22
32
.
Hanjra
M. A.
&
Qureshi
M. E.
, (
2010
).
Global water crisis and future food security in an era of climate change
.
Food Policy
35
(
5
),
365
377
.
Heckman
J. J.
,
Ichimura
H.
&
Todd
P. E.
, (
1997
).
Matching as an econometric evaluation estimator: evidence from evaluating a job training programme
.
The Review of Economic Studies
64
(
4
),
605
654
.
Hijioka
Y.
,
Lin
E.
,
Pereira
J. J.
,
Corlett
R. T.
,
Cui
X.
,
Insarov
G. E.
,
Lasco
R. D.
,
Lindgren
E.
,
Surjan
A.
, (
2014
).
Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part B: Regional Aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change
(
Barros
V. R.
,
Field
C. B.
,
Dokken
D. J.
,
Mastrandrea
M. D.
,
Mach
K. J.
,
Bilir
T. E.
,
Chatterjee
M.
,
Ebi
K. L.
,
Estrada
Y. O.
,
Genova
R. C.
,
Girma
B.
,
Kissel
E. S.
,
Levy
A. N.
,
MacCracken
S.
,
Mastrandrea
P. R.
&
White
L. L.
eds).
IPCC, Cambridge
,
UK
and New York, NY, USA
, pp.
1327
1370
.
IMF
(
2015
).
Issues in Managing Water Challenges and Policy Instruments: Regional Perspectives and Case Studies
.
Technical Companion Note
.
International Monetary Fund
,
Washington, DC
,
USA
.
Iqbal
A.
, (
2015
).
Water Scarcity May Threaten National Economy: IMF
.
DAWN
,
Karachi
,
Pakistan
.
Jalan
J.
&
Ravallion
M.
, (
2003
).
Estimating the benefit incidence of an antipoverty program by propensity-score matching
.
Journal of Business & Economic Statistics
21
(
1
),
19
30
.
Jin
K.
,
Cornelis
W. M.
,
Schiettecatte
W.
,
Lu
J.
,
Yao
Y.
,
Wu
H.
,
Gabriels
D.
,
De Neve
S.
,
Cai
D.
&
Jin
J.
, (
2007
).
Effects of different management practices on the soil–water balance and crop yield for improved dryland farming in the Chinese Loess Plateau
.
Soil and Tillage Research
96
(
1
),
131
144
.
Jonish
J. E.
, (
1991
).
Benefits and costs of laser land leveling in Egypt
.
Advances in Desert and Arid Land Technology and Development
.
Kahlown
M.
,
Raoof
A.
&
Hanif
M.
, (
2000
).
Rice yield as affected by plant densities
.
Mona Reclamation Experimental Project
. (
238
).
Mallappa
M.
&
Radder
G.
, (
2012
).
Evaluation of different proportion of levelling in Zingg conservation terraces on soil moisture conservation, crop growth and yield of green gram
.
Karnataka Journal of Agricultural Sciences
6
(
3
),
37
43
.
Michalopoulos
C.
,
Bloom
H. S.
&
Hill
C. J.
, (
2004
).
Can propensity-score methods match the findings from a random assignment evaluation of mandatory welfare-to-work programs?
The Review of Economics and Statistics
86
(
1
),
156
179
.
Mueller
N. D.
,
Gerber
J. S.
,
Johnston
M.
,
Ray
D. K.
,
Ramankutty
N.
&
Foley
J. A.
, (
2012
).
Closing yield gaps through nutrient and water management
.
Nature
490
(
7419
),
254
257
.
Palanisami
K.
,
Kakumanu
K.
,
Ranganathan
C.
,
Haileslassie
A.
&
Wani
S.
, (
2015
).
Mapping of Household Vulnerability and Identification of Adaptation Strategies in Dryland Systems of South Asia
.
International Crops Research Institute for the Semi-Arid Tropics (ICRISAT)
,
Patancheru 502 324, Telangana
,
India
.
Rahut
D. B.
,
Ali
A.
,
Imtiaz
M.
,
Mottaleb
K. A.
&
Erenstein
O.
, (
2016
).
Impact of irrigation water scarcity on rural household food security and income in Pakistan
.
Water Science and Technology: Water Supply
16
(
3
),
675
683
.
Ren
W.
,
Hu
Z.
,
Cui
H.
,
Yang
C.
,
Liu
Y.
,
Wang
Y.
,
Zhang
Z.
&
Li
B.
, (
2003
).
Effect of laser-controlled land leveling and baby rice seedling direct planting on saving water
.
Transactions of the Chinese Society of Agricultural Engineering
19
(
3
),
72
75
.
Rickman
J.
, (
2002
).
Manual for Laser Land Leveling
.
Rice-Wheat Consortium Technical Bulletin Series 5
,
New Delhi
, p.
24
.
Rockström
J.
,
Karlberg
L.
,
Wani
S. P.
,
Barron
J.
,
Hatibu
N.
,
Oweis
T.
,
Bruggeman
A.
,
Farahani
J.
&
Qiang
Z.
, (
2010
).
Managing water in rainfed agriculture – The need for a paradigm shift
.
Agricultural Water Management
97
(
4
),
543
550
.
Sakurai
T.
, (
2002
).
Adoption of Modern Rice Variety and Water Control Technology in West Africa's Rainfed Lowland
.
Workshop on the Green Revolution in Asia and its Transferability to Africa
,
Tokyo
,
Japan
.
Senaratna Sellamuttu
S.
,
Aida
T.
,
Kasahara
R.
,
Sawada
Y.
&
Wijerathna
D.
, (
2014
).
How access to irrigation influences poverty and livelihoods: a case study from Sri Lanka
.
Journal of Development Studies
50
(
5
),
748
768
.
Sianesi
B.
, (
2004
).
An evaluation of the Swedish system of active labor market programs in the 1990s
.
The Review of Economics and Statistics
86
(
1
),
133
155
.
Smith
J. A.
&
Todd
P. E.
, (
2005
).
Does matching overcome LaLonde's critique of nonexperimental estimators?
Journal of Econometrics
125
(
1
),
305
353
.
UN-Water
(
2013
).
Water for Food Factsheet
.
UN Water
,
Geneva
.
Von Grebmer
K.
,
Bernstein
J.
,
Brown
T.
,
Prasai
N.
,
Yohannes
Y.
,
Towey
O.
,
Foley
C.
,
Patterson
F.
,
Sonntag
A.
,
Zimmermann
S.-M.
&
Hossain
N.
, (
2017
).
2017 Global Hunger Index: The Inequalities of Hunger
.
International Food Policy Research Institute (IFPRI)
,
Washington, DC
.
World Bank
(
2018
).
World Development Indicators
.
World Bank, Washington, DC
.
Yi
L.
,
Shenjiao
Y.
,
Shiqing
L.
,
Xinping
C.
&
Fang
C.
, (
2010
).
Growth and development of maize (Zea mays L.) in response to different field water management practices: resource capture and use efficiency
.
Agricultural and Forest Meteorology
150
(
4
),
606
613
.
Yohannes
D. F.
,
Ritsema
C.
,
Solomon
H.
,
Froebrich
J.
&
van Dam
J.
, (
2017
).
Irrigation water management: Farmers’ practices, perceptions and adaptations at Gumselassa irrigation scheme, North Ethiopia
.
Agricultural Water Management
191
,
16
28
.
Zhao
Z.
, (
2003
).
Data Issues of Using Matching Methods to Estimate Treatment Effects: an Illustration with NSW Data set
.
China Center for Economic Research
,
Peking
,
China
.

Supplementary data