Comparing the impacts of different irrigation systems on the livelihoods of women and youth: evidence from clustered data in Ghana

ABSTRACT

In Sub-Saharan Africa, empirical evidence has shown that irrigation can have positive impacts on agricultural production and farm incomes. This study uses a rigorous propensity score matching analysis applied to clustered data from two state-led and two farmer-led irrigation examples in Ghana to quantitatively evaluate the intersectional impacts of different types of irrigation on multiple dimensions of farmers’ livelihoods. The results of our study indicate that although farmer-led irrigation enhances farmer incomes, this does not necessarily translate into poverty alleviation and prosperity. Furthermore, impacts on young men, young women and adult women are of a different nature.

KEYWORDS:

• Farmer-led irrigation
• impact evaluation
• gender
• youth
• Ghana
• Sub-Saharan Africa

Introduction

The links between irrigation and poverty have been explored by a vast literature that converges towards a global conclusion that the development of irrigation constitutes an effective means of fighting poverty (Hussain & Hanjra, 2004; Lipton et al., 2003). In Sub-Saharan Africa after the benefits of large-scale irrigation projects started being questioned in the late 1980s, small-scale irrigation was proposed as a viable alternative. Investments in small-scale water management technologies have been shown to generate positive productivity and economic impacts including increased net farm incomes, improved land and labour productivity, and important indirect economic benefits (Adeoti et al., 2007; Burney & Naylor, 2012; Dillon, 2011; Kumar et al., 2005; Shah et al., 2000).

In the past decade, questions of irrigation scale have been replaced with actions focused on the process through which irrigation unfolds. In this context, farmer-led irrigation can be defined as a process where farmers assume a driving role in improving their water use for agriculture through changes in knowledge production, technology use, investment patterns and market linkages, and the governance of land and water (Woodhouse et al., 2016). Several studies have outlined the prospected potential of the farmer-led irrigation to improve income, poverty alleviation and employment (Giordano & de Fraiture, 2014), enhance nutrition (Domènech, 2015; Passarelli et al., 2018) and generate greater resilience to seasonal weather variability and climatic shocks (Zorom et al., 2013). However, important research gaps remain. First, most of the research concentrates on the coverage and official documentation of this practice and its benefits for agricultural production and incomes, leaving a gap on the actual impacts it has on farmers’ livelihoods (Osewe et al., 2020). Second, although previous research reported that women and youth are especially vulnerable to climate change because of insecure land rights, the capital they possess, their late investment in irrigation, the sources of irrigation they make use of, and the social relationships they are part of and that structure water governance (Fischer et al., 2022), there has been very limited research on the differentiated effects of irrigation development on women and youth (Dossou-Yovo et al., 2022). Third, most of the existing studies exploring the links between irrigation use and livelihoods, either argue without empirical support or fail to account for selectivity bias.

This paper seeks to examine:

• How the processes through which irrigated agriculture comes about affect the impact of irrigation on the multiple dimensions of livelihoods such as incomes, food security, education, health status, etc. (Do all types of irrigation lead to poverty alleviation and prosperity?)

• Whether women and youth benefit differently from irrigation and if these gender and generational differences depend on the role taken by these groups in improving their water use for agriculture. (Does irrigation alleviate or deepen inequalities across gender and generational gaps?)

There are at least three contributions of this study to the literature on rural irrigation development. First, it supplements existing evidence on the impact of irrigation on economic and livelihood outcomes in an agricultural setting. Our study substantially furthers the evidence base by assessing the impacts of farmer-led irrigation development, alongside public large-scale irrigation investments, on a broad range of livelihood dimensions. Second, to our knowledge, this is the first study that explores how different types of irrigation affect women and youth differently. Understanding the impacts of irrigation on young peoples’ livelihoods is especially important, as approximately 60% of Africa’s population are under 35 years old, and most of this youth are unemployed and living in rural areas (Castañeda et al., 2018). Finally, our study also contributes to the growing need for rigorously conducted impact evaluations of irrigation development. Our study relates to a growing number of papers that adopt non-experimental approaches to quantify the impacts of rural infrastructure development. Specifically, our study adds to the small but growing number of counterfactual-based impact evaluations of irrigation projects (Dillon, 2011; Gebregziabher et al., 2009; Hirko et al., 2018; Zeweld et al., 2015) by adapting the evaluation methodology to clustered data and rigorously assessing impacts on a broad range of livelihood dimensions.

We evaluate these questions through a quantitative analysis of comparative case studies in Ghana. Agriculture in Ghana is predominantly practiced on smallholder, family-operated farms, which produce about 80% of total agricultural output. Despite considerable potential for development and the emphasis placed on irrigation development in many government plans, less than 2% of the total cultivable area in Ghana is irrigated. Less than a third of the estimated total irrigated land in Ghana lies within 22 well-known public schemes, and not enough is known of the location, development, and management of the farmer-led irrigation schemes that account for the remaining two-thirds of total irrigated land (Namara, Horowitz, et al., 2011). The performance and productivity of existing irrigation schemes, particularly those that were publicly developed, are generally low and little is known about their impacts on the multiple dimensions of livelihoods and on women’s empowerment.

We consider two examples of farmer-led and two examples of state-led irrigation developments. The farmer-led irrigation initiatives analysed are petrol pump riverside irrigation along the White Volta River in the Bawku West district and low-intensity shallow groundwater irrigation in the vicinity of small unused reservoirs in the Nabdam district. The two examples of state-led interventions analysed are the Tono (Kassena Nankana district) and Bontanga (Kumbungu district) irrigation schemes.We analyse data collected through a semi-structured farmer questionnaire of irrigators and non-irrigators through the application of propensity score matching to clustered data. We compare impacts on multiple dimensions of poverty and livelihoods and empowerment of farmer-led irrigation developments with state-led interventions.

Our results contribute to the literature on irrigation impacts in Sub-Saharan Africa and inform investments and policymaking along three dimensions. First, we find that land ownership and social capital are the main factors that affect access to irrigation. Second, the effect of access to irrigation varies considerably by the dimension of farmers’ livelihoods analysed, the type of irrigation development and the gender/age of the irrigator. Finally, our results indicate that taking into consideration the cluster structure of the data collected in our estimation of the propensity scores and treatment effects reveals important insights concealed within pooled (and biased) estimates of treatment effects.

The article is organized as follows. The next section presents the conceptual framework and the methodology used to estimate irrigation impacts and describes the data collection process. The third section presents the empirical results, and the final section concludes.

Methods

Conceptual framework

Our conceptual framework examines the role of irrigation development on multiple aspects of the livelihoods of men, women and youth. The livelihoods aspects that this article considers are foundationally based on the agricultural water and micro-pathways proposed by Hussain and Hanja (2004) (Figure 1). In this article, we focus on direct linkages, which operate via localized and household-level effects. Indirect linkages operate via aggregate or national level impacts and refer mainly to spillover or multiplier effects, which, although relevant, fall outside the scope of this paper.

Figure 1. Irrigation–livelihoods pathways.

Source: Hussain & Hanjra (2004).

Investments in irrigation may help increase farmers’ production by raising crop yields through greater and more stable supply of water for agriculture (Hussain & Hanjra, 2003, 2004). Irrigation also gives farmers the confidence to invest in productivity-enhancing inputs that complement irrigated water, such as improved seeds, fertilizer and pesticides (Evenson & Gollin, 2003), thus supporting intensification and diversification (Binswanger & Braun, 1991; Huang et al., 2006; Smith, 2004). In addition, irrigation may also benefit farmers through increased demand for labour resulting in higher wages (Narayanamoorthy & Deshpande, 2003; Von Braun, 1995), reduced outmigration (Fishman & Li, 2022), and lead to crucial changes on all four food security dimensions – food availability, access, utilization and stability (Domènech, 2015).

Impact evaluation of agricultural investments

Broadly, the approaches to impact evaluation can be divided between experimental and non-experimental approaches. In an experimental approach, data are generated through a properly implemented random experimental design, so the expectations of the treatment and comparison groups are equal because the groups are composed of randomly allocated members, ensuring that the distribution of observable and unobservable characteristics of the groups are equivalent in a statistical sense. However, reducing selection bias by random assignment of treatment is not possible for ex-post studies. In this case, non-experimental approaches apply econometric methods to create a control group that represents a reasonable counterfactual to the treatment group by requiring identification assumptions with non-experimental data.

There are a number of methods available for estimating treatment effects while addressing selection bias using non-experimental research designs. The most important factors in deciding which of the methods to use are the nature of data and the type of agricultural investment being investigated. The difference-in-differences estimator requires panel data to compare the mean changes between treatment and control groups over two periods. Although this estimator does not completely control for selection bias (if initial conditions that affect the outcome are correlated with the selection criteria for agricultural investment programmes or self-selection), this method has been widely used to estimate irrigation impacts. For instance, it was used to evaluate the impact of access to irrigation on poverty, production and nutrient intakes in Mali (Dillon, 2008), the impact of rural roads and irrigation on household welfare in Vietnam (Nguyen et al., 2017), the impacts of large irrigation dams on household consumption in Nigeria (Takeshima et al., 2016) and the impacts of a set of irrigation rehabilitation projects in Peru (Del Carpio et al., 2011), among others. The Heckman correction (Heckman et al., 1997) and the instrumental variable approaches can be used together with the difference-in-differences estimator to control for unobservable differences between treated and control groups. However, panel data and valid instruments are rarely available for impact evaluation.

An alternative non-experimental method for estimating the effects of an agricultural investment while dealing with selection bias is propensity score matching. This method uses propensity scores to match households with similar observable characteristics, varying only the treatment. Because it does not require the assumptions of functional and distributional form or exogeneity of covariates, propensity score matching has gained popularity in recent years. Propensity score matching estimates the effects of a treatment, in this case irrigation, by comparing how given outcomes of interest differ between two seemingly similar groups, with the only difference being the treatment administered. Rather than attempting to match on all values of the variables, cases can be compared on the basis of propensity scores (the probability that the farmer irrigates). Propensity score matching is based on the assumption that the only source of selection bias is the set of observed variables (conditional independence assumption). If pre-treatment data are available, this strong assumption can be relaxed by implementing a difference-in-differences matching estimator (Andersen et al., 2015).

A number of studies have estimated irrigation impacts using propensity score matching (Dillon, 2008, 2011; Osewe et al., 2020). This method, however, has not yet been applied to compare the impact of different irrigation development processes on the livelihoods of female versus male irrigators and adult versus young irrigators. Furthermore, to the best of our knowledge, existing propensity score matching applications on agricultural investments or projects have failed to consider the cluster structure of observational data. If unobserved cluster-level variability is correlated with the outcome and the treatment assignment mechanism, this will lead to biased estimates of the treatment effect.

Estimation strategy

We use propensity score matching to evaluate the impact of irrigation access, by type of irrigation development and gender/age, on different dimensions of farmers’ livelihoods (economic, health, food and physical security and empowerment). For a detailed explanation of the econometric method used, please refer to Appendix 1 in the online supplemental data. The causal effect of irrigation on the variables of interest was estimated in two stages: in the first stage, the propensity scores were estimated using the logit models of Eqs. [1–3] in Appendix 1; in the second stage, irrigators and non-irrigators were matched by their propensity scores using a nearest-neighbour matching algorithm.¹ Table 1 outlines the different models we estimate by taking clustered data into account in different ways. We first estimate the overall effect of irrigation using the whole sample. As a benchmark, in model 1 we estimate average treatment effects without considering clustering in either the propensity score estimation or the matching procedure (as per a standard propensity score matching estimation). In models 2 and 3, we account for clustering in our matching procedure and in models 4 and 5 we account for clustering in our propensity score estimation model. Second, we estimate the effect of irrigation separately by cluster: first for each type of irrigation development process and second for each gender/age group. In this set of estimations, the complete matching procedure is implemented separately for each group. This is analogous to insisting on a perfect match (in terms of irrigation type first and gender/age second) and then carrying out propensity score matching.² In models 6–8, matching is done separately for each irrigation type. In models 7 and 8, we add dummy variables for gender/age and community in the propensity score estimation, but it is still possible that, for instance, adult women in community A with irrigation type A are matched with young men in community B with irrigation type A, because the gender/age and community dummies are only a subset of all available variables. Finally, in models 9–11, matching is done separately for each gender/age group.

Table 1. Average treatment effect of irrigation: alternative propensity score matching model specifications.

Overall and sub-cluster analysis	Model number	Propensity score estimation	Matching
Overall effect of irrigation	1	Single level	Pooled
	2	Single level	Within type of irrigation
	3	Single level	Within gender/age group
	4	Fixed effects by type of irrigation, community and gender/age group	Pooled
	5	Random effects by type of irrigation, community and gender/age group	Pooled
Effect of irrigation by scale/type of irrigation system	6	Single level	Within type of irrigation
	7	Fixed effects by community and gender/age group	Within type of irrigation
	8	Random effects by community and gender/age group	Within type of irrigation
Effect of irrigation by gender/age of irrigator	9	Single level	Within gender/age group
	10	Fixed effects by community and type of irrigation	Within gender/age group
	11	Random effects by community and type of irrigation	Within gender/age group

The most common method for estimating propensity scores is logistic regression, and when we have clustered data it is possible to include a cluster-level fixed effects or random effects to account for any unmeasured cluster-level covariates. The main difference between these models is that random effects assumes a normal distribution, which can lead to more precise estimates if correct. However, it may be biased if individual-level confounders are correlated with cluster-level confounders. Conversely, fixed effects does not make distributional assumptions about cluster-level effects. In most data sets (like this case study) it may be expected for cluster-level covariates to be correlated with individual-level covariates, which means the fixed effects may be preferred to random effects. However, fixed effects may not perform well in data sets with a large number of small clusters due to the large number of parameters included in the model, and thus random effects may be preferred. As our clusters are relatively small, at around 150–250 observations, we decide to present estimates from both fixed effects and random effects propensity score models. For all estimations, we have now graphed the propensity score of treatment-group observations versus control-group observations before matching and redone the same graph after matching. In all models, our results suggest that the balance improves significantly after matching (Appendix 2 in the online supplemental data).

The observable covariates considered, i.e., factors that were likely to affect the probability of irrigation access, were selected based on previous irrigation impact studies in developing countries (Dillon, 2008, 2011; Domenech & Ringler, 2013). Respondents’ characteristics that could influence the probability of irrigating included age, civil status, education, origin, farming experience, access to credit, ownership of irrigable land and membership to social groups. In addition, the gender of the household head might influence the decision to irrigate.

More educated and experienced farmers and those that participate in various kinds of social groups are more likely to be knowledgeable on solutions to water management problems and proactive in investing in irrigation. Farmers who own irrigable land are more likely to irrigate than those who have to rent land in the dry season. Access to savings increases the probability of receiving credit to finance irrigation investments. Civil status, age and gender may have a positive or negative effect on access to irrigation.

Data

This study used one household survey (questionnaire included in Appendix 3 in the online supplemental data), for the control and the treatment. Irrigators were matched with non-irrigators to form a representative sample of 864 farmers covering the major irrigated zones in the White Volta Basin in Ghana. The White Volta Basin has a total catchment area of 105,000 km², with 49,200 km² in Ghana and the rest in Togo and Burkina Faso. The climate is semi-arid with annual rainfall on the basin ranging from 1010 mm/year in the north to 1140 mm/year in the south. The predominant land use is arable agriculture and widespread grazing of large numbers of cattle and other livestock (Ampim et al., 2021). Irrigation farming in the study area is mainly practised during the dry season (October–April). In addition to the government developed irrigation systems, there are other irrigation development processes led by farmers and groups of farmers scattered across the basin. As a result, the basin has witnessed a spectacular rise of irrigated agriculture since the early 2000s, which seems to have been triggered by a strong and growing demand for vegetables, notably tomatoes from the urban centres of southern Ghana (Ofosu et al., 2010).

Four types of irrigation practiced in the basin were considered: large-scale irrigation (schemes larger than 500 ha); medium-scale irrigation (schemes between 50 and 500 ha); small-scale groundwater irrigation (schemes smaller than 50 ha where the source of irrigation water is groundwater) and small-scale riverine irrigation (schemes smaller than 50 ha where the source of irrigation water is a river). The survey used a stratified multi-stage (cluster) sampling design, with the four major irrigation districts as the strata (Kassena-Nankana, Bawku West, Nabdam and Kumbungu). In each of these districts, irrigating farmers specialize in one of the four irrigation types considered: large-scale in Kassena-Nankana (Tono, the only operational large-scale scheme on the White Volta Basin in Ghana); medium-scale in Kumbungu (Bontanga, the only operational medium-scale scheme on the White Volta Basin in Ghana); small-scale groundwater based in Nabdam and small-scale riverine in Bawku West (Figure 2). Crops produced under rain-fed are mostly grains such as maize, rice, millet, sorghum, groundnut and soybean, whereas irrigated crops are mainly vegetables including tomato, onion, pepper, okra and leafy vegetables.

Figure 2. Map of the study location in Northern Ghana.

Source: Authors’ own production.

The sampling design consisted of sublocations (communities, Ghana’s smallest administrative units) as primary sampling units, and farmers as secondary sampling units. The number of primary sampling units in each strata and the number of secondary sampling units in each primary sampling unit were determined to achieve minimum sample requirements to detect a certain minimum effect size for each outcome variable at 5% significance (probability of type I error) and 80% power (20% probability of type II error) with a balanced sample of treated and untreated units (White & Raitzer, 2017).³ The required number of sublocations was established at two per strata (eight in total). A list of all irrigating sublocations in each of the four districts was established, with the number of farmers in each obtained from district statistics. For each district, the required number of sublocations (two) was selected with probability proportionate to size, to maintain a fixed number of farmers per sublocation and a self-weighted sample. We use the definition of youth proposed by the African Union’s Africa Youth Charter, which defines youth as peoples between the ages of 15 and 35 years (African Union Commission, 2006). For each sublocation, 36 farmers were selected from each gender/age group (men from 35 years old, women from 35 years old, and men and women less than 35 years old) by random sampling. Thus, the total sample included 108 farmers per sublocation (Table 2).

Table 2. Survey sample.

District	Type of irrigation system	Communities	Irrigators			Non-irrigators			Total per community	Total per district
			Men	Women	Youth	Men	Women	Youth
Nabdam	Small-scale groundwater	Zanlerigu	18	18	18	18	18	18	108	216
		Zanlerigu Gia	18	18	18	18	18	18	108
Bawku West	Small-scale riverine	Googo	18	18	18	18	18	18	108	216
		Yarigu	18	18	18	18	18	18	108
Kumbungu	Medium-scale (Bontanga)	Wuba	18	18	18	18	18	18	108	216
		Kpalsogu	18	18	18	18	18	18	108
Kassena-Nankana	Large-scale (Tono)	Bonia	18	18	18	18	18	18	108	216
		Korania	18	18	18	18	18	18	108
Totals		8	144	144	144	144	144	144	864	864

Both irrigators and non-irrigators were interviewed with the same pretested questionnaire. Respondent characteristics included age, gender, literacy, years of formal education and years of rain-fed and irrigated farming experience. Farmers were asked to estimate their rain-fed and irrigated production, how much they consumed and how much they sold. Economic security indicators included total rain-fed land cultivated, irrigated farm and non-farm income, income diversification (number of income sources), savings (whether the farmer kept cash or in-kind savings) and value and diversity of farm and non-farm assets.

Concerning food security, respondents were first asked to estimate the number of months in which their household did not have adequate food to feed all the members, to give the months of inadequate household food provisioning score (Bilinsky & Swindale, 2005). The second measure of food insecurity is the household dietary diversity score, given by the number of different foods eaten from different food groups.

On health security, farmers were asked whether any member of the household suffered from any sickness and whether they had to stop the usual activities because of sickness or injury in the previous 3 months, travel times to nearest health facility, and whether they were National Health Service subscribers. Similarly, on physical security, farmers were asked about the building materials of their dwelling, and its storage, energy, lighting, sanitation and water facilities. Health and physical security indexes were created giving equal weight to factors within each category.

Concerning empowerment, farmers were asked questions on a range of areas related to production, resources, income, leadership and time. To test for attributable changes in empowerment of irrigators, we used the Women’s Empowerment in Agriculture Index, which was pioneered by researchers of the International Food Policy Research Institute (IFPRI), USAID, and the Oxford Poverty and Human Development Initiative (OPHI) to evaluate outcomes of interventions on women’s empowerment (USAID, IFPRI and OPHI, 2012). To the best of our knowledge, the Women’s Empowerment in Agriculture Index has not yet been used to evaluate the impact of irrigation on farmers’ empowerment.

Finally, an aggregate livelihoods index for each farmer was constructed by averaging all the four groups of livelihood security (economic, food, health and physical) and empowerment with an equal weight. Usually, irrigation projects and investments channelled to poor farmers are based on income data. This study attempts to develop an index that captures comprehensively all livelihoods’ aspects of the rural poor that can be readily used by decision- and policy-makers to assess irrigators’ specific needs and maximize intervention impacts.

Results and discussion

Propensity scores

In the first step of the propensity score matching, the logit models of Eqs. [1–3] were estimated in order to analyse the factors that affect irrigation access, and to calculate the propensity to irrigate for each farmer. The results show that significant characteristics that increase the probability of irrigation include being married, land ownership and being a member of social groups (Table 3). Although this model does not include all observable household characteristics from the data set that could possibly influence irrigation, the specification chosen satisfies the balancing property.

Table 3. Estimation of the logit models on the propensity to irrigate.

Variable name	Single level	Fixed Effects	Random Effects
Female household head	Omitted	0.36 (0.34)	0.45 (0.33)
Age	0.07 (0.04)*	0.05 (0.05)	0.05 (0.04)
Age squared	−0.00 (0.00)**	−0.00 (0.00)	−0.00 (0.00)
Single	0.22 (0.33)	0.56 (0.28)*	0.42 (0.24)*
Education – Primary	0.09 (0.23)	−0.07 (0.26)	−0.02 (0.25)
Education – Secondary/Tertiary	−0.38 (0.32)	−0.39 (0.34)	−0.34 (0.33)
Origin	0.30 (0.22)	0.11 (0.25)	0.19 (0.24)
Credit	0.21 (0.17)	0.25 (0.18)	0.25 (0.18)
Experience	−0.00 (0.01)	−0.01 (0.01)	−0.01 (0.01)
Ownership of irrigated land	2.69 (0.21)***	2.96 (0.22)***	2.92 (0.22)***
Membership to social groups	0.36 (0.36)***	0.46 (0.23)**	0.38 (0.21)*
Number of observations	848	848	848

Notes: Constant is included. *Significant at the 10% level; **significant at the 5% level; ***significant at the 1% level.

Overall impact of irrigation

Table 4 reports the results of the different models from Table 1 used to estimate the overall effect of irrigation. The results indicate significant effects when all the approaches are calculated. However, more effects are significant when the clustered nature of the data is considered in the estimation of the propensity scores. While irrigation reduces income diversification, the level of non-farm income increase ranges from 215 to 261 Ghana Cedis (GHS). In addition, the increase in the value of assets varies from 412 to 553 GHS),⁴ depending on the approach used. Irrigation increases food availability by approximately half a month and the health index is significant and positive at the 5% and 10% levels. The effect of irrigation on the composite livelihoods index is positive and significant in 4 of the 5 approaches used. Rain-fed farm size, savings, assets diversity, dietary diversity, physical security and empowerment are not statistically significant.

Table 4. Overall impact of irrigation.

Model number	1	2	3	4	5
Propensity Score Estimation	Single level	Single level	Single level	Fixed effects Irrigation and gender/age	Random effects Irrigation and gender/age
Matching	Pooled data set	Within irrigation system	Within gender/age	Pooled data set	Pooled data set
Effect on outcome variables
1. Economic security
Income – diversification (GHS)	−0.18 (0.09)*	0.06 (0.1)	−0.04 (0.11)	−0.13 (0.08)*	−0.19 (0.08)***
Income – irrigation (GHS)	1297.45 (144.46)***	1293.46 (150.98)***	1325.56 (145.35)***	1324.93 (146.66)***	1310.96 (147.75)***
Non-farm income (GHS)	181.83 (128.62)	260.93 (115.22)**	−225.69 (233.43)	78.25 (144.33)	214.60 (123.91)*
Farm size (rain-fed) (ha)	−0.28 (0.51)	0.58 (0.38)	0.24 (0.35)	−0.28 (0.37)	−0.32 (0.40)
Savings	0.02 (0.067)	0.07 (0.09)	0.00 (0.06)	−0.01 (0.05)	−0.02 (0.05)
Assets value (GHS)	483.57 (200.94)**	514.74 (235.75)**	553.41 (265.72)**	512.23 (174.67)***	411.84 (186.56)**
Assets diversity (number of asset types)	0.68 (0.52)	0.79 (0.47)*	0.75 (0.47)	0.31 (0.46)	0.66 (0.45)
2. Food security
Food availability (months of inadequate food provisioning)	−0.47 (0.27)*	−0.05 (0.2)	−0.34 (0.24)	−0.48 (0.24)**	−0.52 (0.23)**
Household dietary diversity (number of food groups)	0.32 (0.28)	0.24 (0.31)	0.38 (0.23)	0.28 (0.23)	0.15 (0.23)
3. Health security
Health index	0.05 (0.03)	0.02 (0.03)	−0.03 (0.03)	0.05 (0.03)*	0.05 (0.03)**
4. Physical security
Physical security index	0.00 (0.03)	0.04 (0.03)	−0.01 (0.02)	−0.01 (0.02)	−0.02 (0.02)
5. Empowerment
Empowerment index	0.00 (0.00)	−0.01 (0.01)	0.01 (0.01)	−0.01 (0.01)	−0.01 (0.01)
6. Multidimensional
Livelihoods index	0.02 (0.01)**	0.02 (0.01)**	0.01 (0.01)	0.02 (0.01)***	0.02 (0.01)***

Note: All standard errors in brackets are bootstrapped with 500 repetitions. *Significant at the 10% level; **significant at the 5% level; ***significant at the 1% level.

Irrigation impacts by type of system

Table 5 reports the results of the different models used to estimate the impact of irrigation by type of system. As with the estimation of the overall effect of irrigation, more effects are significant when the gender/age clusters are considered in the estimation of the Propensity Score. Irrigation in the Tono public-led system has a positive effect on nonfarm income (increase ranges from 428 to 479 GHS), household dietary diversity and the composite livelihood index. Farmer-led river pumping is the only other system that also increases the livelihood index, and has many other positive effects on non-farm income, assets value (increases of over 1000 GHS) and diversity, food availability, empowerment and physical security. Irrigation in the state-led Bontanga scheme has mixed results: it increases income diversity, farm size and assets value but decreases empowerment. Finally, the farmer-led groundwater-based irrigation is the system that fares the worst, with negative effects on farm size, physical security and dietary diversity. This may be due to a variety of factors, including siltation of an upstream dam, tired soils, low use of fertilizers, low yields of seasonal wells and the diversity and type of crops grown (low-profit and perishable vegetables). It also supports existing evidence that a lack of crop diversity can have a negative effect on dietary diversity and nutrition (Sekabira & Nalunga, 2020).

Table 5. Impacts of different types of irrigation systems.

Approach	6				7				8
Propensity Score Estimation	Single level				Fixed effects (gender/age)				Random effects (gender/age)
Matching	Within irrigation type				Within irrigation type				Within irrigation type
Irrigation system	Tono	Bontanga	Farmer-led groundwater	Farmer-led riverine	Tono	Bontanga	Farmer-led groundwater	Farmer-led riverine	Tono	Bontanga	Farmer-led groundwater	Farmer-led riverine
Effect on outcome variables
1. Economic security
Income – diversification (GHS)	0.12 (0.21)	0.43 (0.28)	−0.28 (0.22)	−0.04 (0.11)	−0.01 (0.15)	0.63 (0.29)**	−0.21 (0.20)	−0.12 (0.10)	−0.05 (0.14)	0.63 (0.24)***	−0.31 (0.20)	−0.15 (0.10)
Income – irrigation (GHS)	1377.79 (276.69)***	1773.61 (605.41)***	586.98 (57.53)***	1461.21 (194.10)***	1375.94 (254.23) ***	1782.61 (455.16)***	586.98 (56.96)***	1485.21 (194.71)***	1414.83 (241.60)***	1782.18 (474.91)***	586.98 (60.62)***	1493.21 (195.61)***
Non-farm income (GHS)	433.59 (200.30)***	−138.31 (133.27)	−600.02 (385.36)	692.70 (305.34)	428.47 (149.26)***	−13.31 (204.82)	−353.06 (281.91)	891.20 (315.5)***	479.04 (135.22)***	71.69 (161.66)	−291.97 (275.92)	712.02 (296.61)
Farm size (rain-fed) (ha)	1.5 (0.77)*	1.25 (0.61) **	−1.06 (0.51)**	−0.13 (1.06)	1.03 (0.63)	1.44 (0.41)***	−0.44 (0.47)	0.04 (0.88)	1.11 (0.62)*	1.20 (0.38)***	−0.97 (0.47)**	−0.25 (1.06)
Savings	0.18 (0.21)	0.11 (0.25)	−0.06 (0.06)	0.06 (0.12)	0.22 (0.16)	0.29 (0.23)	0.00 (0.06)	0.02 (0.09)	0.16 (0.18)	0.26 (0.23)	−0.03 (0.06)	0.06 (0.11)
Assets value (GHS)	111.01 (543.08)	11,107.79 (296.71) ***	33.61 (122.98)	598.54 (619.65)	556.84 (435.59)	1103.79 (366.00)***	137.78 (104.63)	786.64 (439.95)*	655.76 (383.53)*	1095.93 (257.89)***	14.94 (108.60)	816.82 (429.82)*
Assets diversity (number of asset types)	0.04 (1.21)	2.16 (1.16)*	−0.13 (0.50)	1.24 (0.91)	0.98 (1.17)	1.76 (1.13)	−0.58 (0.44)	1.65 (0.64)***	1.24 (1.05)	1.62 (1.01)	−0.80 (0.41)*	1.57 (0.73)**
2. Food security
Food availability (months of inadequate food provisioning)	0.16 (0.33)	0.06 (0.06)	0.22 (0.44)	−0.69 (0.51)	−0.13 (0.36)	−0.01 (0.05)	0.22 (0.40)	−0.86 (0.38)**	−0.18 (0.35)	0.05 (0.04)	−0.03 (0.41)	−1.24 (0.50)**
Household dietary diversity (number of food groups)	1.25 (0.45)***	−0.47 (0.95)	−0.59 (0.36)	0.73 (0.57)	1.1 (0.31)***	0.20 (0.82)	−0.55 (0.35)	0.31 (0.38)	1.07 (0.33)***	0.32 (0.85)	−0.70 (0.34)**	0.46 (0.40)
3. Health security
Health index	0.09 (0.07)	0.00 (0.08)	−0.06 (0.06)	0.03 (0.06)	0.09 (0.06)	−0.04 (0.08)	−0.06 (0.05)	0.05 (0.04)	0.07 (0.05)	−0.06 (0.07)	−0.03 (0.06)	0.03 (0.04)
4. Physical security
Physical security index	0.14 (0.10)	−0.02 (0.02)	−0.02 (0.03)	0.06 (0.04)	0.07 (0.09)	−0.02 (0.01)	−0.03 (0.03)	0.00 (0.03)	0.10 (0.07)	−0.02 (0.01)	−0.07 (0.04)*	0.06 (0.03)**
5. Empowerment
Empowerment index	−0.00 (0,03)	−0.03 (0.02)	0.02 (0.03)	0.01 (0.02)	−0.02 (0.02)	−0.03 (0.02)**	0.00 (0.02)	−0.00 (0.01)	−0.02 (0.02)	−0.03 (0.02)	−0.01 (0.02)	0.01 (0.02)
6. Multi-dimensional
Livelihoods index	0.06 (0.02) ***	−0.00 (0.02)	−0.02 (0.02)	0.04 (0.02)**	0.05 (0.02)***	0.00 (0.01)	−0.02 (0.02)	0.02 (0.02)	0.06 (0.02)***	−0.00 (0.01)	−0.02 (0.02)	0.04 (0.02)**
N	216	209	218	218	216	209	218	218	216	209	218	218

Note: All standard errors in brackets are bootstrapped with 500 repetitions. *Significant at the 10% level; **significant at the 5% level; ***significant at the 1% level.

As in other locations with similar conditions as Nabdam, where smallholders’ livelihoods are more static, and issues of connectivity, market access and unbalanced economic growth hinder livelihood expansion and value chain access, farmer-led irrigation struggles to contribute to improvements in farmers’ livelihoods (Higgins et al., 2021). Our findings are in line with the few existing studies that analyse the impact of irrigation on outcomes other than yields. Without sufficient gains from specialization, irrigation may constrain income growth and increase vulnerability to crop-specific shocks, thus reducing resilience and potentially hindering livelihood choices and worsening livelihood outcomes (Lin, 2011; Makate et al., 2016).

Impacts of irrigation by gender/age group

Table 6 reports the results of the different approaches used to estimate the effect of irrigation by gender/age group using one-to-one matching. Adult men present the highest increases in non-farm income (430 to 494 GHS) and assets value (707 to 1224 GHS), almost the same increase in irrigation income as young irrigators (around 1600–1700 GHS) and the livelihoods index increases when using all approaches.

Table 6. Irrigation impacts by gender/age group.

Approach	9				10				11
Propensity Score Estimation	Single level				Fixed effects (scale/type)				Random effects (scale/type)
Matching	Within gender/age				Within gender/age				Within gender/age
Gender/age group	Adult Male	Adult Female	Youth Male	Youth Female	Adult Male	Adult Female	Youth Male	Youth Female	Adult Male		Adult Female	Youth Male	Youth Female
Effect on outcome variables
1. Economic security
Income – diversification	−0.07 (0.12)	0.18 (0.20)	−0.62 (0.33)*	0.44 (0.37)	−0.16 (0.12)	−0.04 (0.15)	−0.16 (0.35)	0.56 (0.30)*	−0.12 (0.12)	−0.12 (0.16)		−0.12 (0.33)	0.54 (0.28)*
Income – irrigation	1666.22 (395.98)***	783.40 (80.19)***	1747.01 (287.11)***	1143.39 (187.17)***	1620.42 (391.14)***	780.15 (74.43)***	1747.01 (263.05)***	1143.39 (154.17)***	1657.82 (405.26)***	779.83 (76.81)***		1747.01 (278.36)***	1143.39 (160.61)***
Non-farm income	328.14 (277.58)	84.13 (101.09)	−1700.59 (1087.18)	73.08 (134.75)	430.33 (243.03)*	47.25 (87.14)	−1691.69 (983.18)*	119.12 (66.61)*	494.44 (228.80)**	42.77 (89.61)		−1693.58 (925.96)*	133.37 (58.24)**
Farm size (rain-fed)	0.32 (0.68)	0.42 (0.71)	−0.23 (0.79)	0.97 (0.85)	0.49 (0.59)	−0.07 (0.58)	0.68 (0.63)	0.76 (0.67)	0.32 (0.62)	−0.00 (0.59)		0.47 (0.68)	1.11 (0.72)
Savings	0.02 (0.11)	0.00 (0.11)	−0.07 (0.07)	0.07 (0.23)	0.06 (0.10)	−0.04 (0.08)	−0.09 (0.06)	0.02 (0.23)	0.01 (0.10)	−0.05 (0.08)		−0.10 (0.04)**	−0.03 (0.21)
Assets value	1224.01 (314.77)***	183.20 (291.76)	563.35 (423.34)	391.67 (158.03)**	897.47 (374.40)**	−124.72 (265.60)	543.54 (302.71)*	300.35 (144.26)**	707.0(366.76)*	−59.70 (298.44)		569.71 (327.79)*	341.89 (143.54)**
Assets diversity	1.45 (0.64)**	−0.29 (0.83)	1.55 (0.74)**	0.86 (0.87)	0.76 (0.54)	−0.19 (0.78)	0.35 (0.68)	0.39 (0.71)	0.65 (0.52)	−0.09 (0.82)		0.80 (0.67)	0.58 (0.72)
2. Food security
Food availability	−0.24 (0.39)	0.08 (0.47)	−1.45 (0.74)**	0.25 (0.18)	0.05 (0.37)	−0.26 (0.32)	−0.11 (0.67)	0.25 (0.15)*	−0.21 (0.37)	−0.57 (0.34)*		−0.08 (0.67)	0.27 (0.15)*
Household dietary diversity	0.93 (0.37)**	−0.11 (0.54)	1.05 (0.50)*	−0.80 (0.67)	0.50 (0.30)*	−0.38 (0.42)	0.82 (0.41)**	−0.93 (0.56)*	0.45 (0.31)	−0.32 (0.45)		0.74 (0.38)**	−1.07 (0.64)*
3. Health security
Health index	0.01 (0.06)	0.04 (0.05)	−0.15 (0.09)*	−0.07 (0.08)	0.05 (0.05)	0.05 (0.05)	−0.01 (0.08)	−0.03 (0.08)	0.05 (0.05)	0.07 (0.04)		−0.03 (0.08)	−0.03 (0.08)
4. Physical security
Physical security index	0.00 (0.05)	0.01 (0.03)	−0.12 (0.06)**	0.06 (0.04)*	−0.00 (0.05)	−0.01 (0.03)	−0.05 (0.05)	0.10 (0.03)***	0.00 (0.04)	−0.01 (0.03)		−0.07 (0.05)	0.09 (0.03)***
5. Empowerment
Empowerment index	0.01 (0.02)	−0.02 (0.02)	0.07 (0.03)**	−0.02 (0.03)	−0.01 (0.01)	−0.02 (0.01)	0.03 (0.02)	−0.02 (0.02)	−0.00 (0.01)	−0.03 (0.01)**		0.04 (0.03)	−0.02 (0.02)
6. Multi-dimensional
Livelihoods index	0.05 (0.02)**	0.01 (0.01)	−0.01 (0.02)	0.03 (0.02)	0.04 (0.02)**	0.01 (0.01)	−0.01 (0.02)	0.04 (0.02)**	0.04 (0.02)**	0.02 (0.01)		−0.01 (0.02)	0.04 (0.02)**
N	262	275	164	157	262	275	164	157	262	275		164	157

Note: All standard errors in brackets are bootstrapped with 500 repetitions. *Significant at the 10% level; **significant at the 5% level; ***significant at the 1% level.

Our findings do not support previous evidence that farmer-led irrigation is largely benefitting young farmers (Colenbrander & van Koppen, 2013; Namara et al., 2014), as our non-experimental design allows us to compare increased irrigation income to forgone non-irrigation income (which is not possible to estimate in the absence of rigorous evaluation designs). On the contrary, while young men involved in irrigation see the value of their assets increase by around 400–500 GHS, their non-farm incomes are reduced by a striking 1700 GHS, probably because the proceedings from irrigation are much lower than the remittances from those young men who travel to the cities to work during the dry season. This finding also limits previous evidence suggesting the potential of irrigation to reduce outmigration, with previous studies in Northern Ghana suggesting that farmer-led development of shallow groundwater irrigation is associated with a reversal in rural–urban migration (Laube et al., 2012). Although it is true that youth are more likely to find employment in the informal farming sector to meet immediate basic needs, underemployment offers limited social protection and rights, low wages, poor job security, and limited future career development opportunities (Verbruggen et al., 2015). Addressing youth participation in irrigation requires a holistic approach on rural economies, service delivery and livelihoods broadly (Geza et al., 2021).

Young female irrigators are the other group (besides adult men) that has their non-farm income increase, but by a smaller amount (119–133 GHS), as well as the value of their assets (also by a smaller amount: 300–391 GHS). The composite livelihoods index also increases for young women irrigators but only when the scale/type of the irrigation system is considered in the estimation of propensity scores. Whereas young female irrigators’ physical security increases, both food availability and dietary diversity are negatively impacted . Previous studies in Northern Ghana failed to find any significant differences on the dietary diversity of rain-fed farmers and irrigated farmers. With irrigation improving male dietary diversity and reducing female dietary diversity, our findings on the impact on nutrition highlight the importance of exploring gender dynamics and women’s roles in irrigated agriculture as key determinants of food access and utilization across different members of the household. Irrigated crops are often cash crops, and cash crops are often men’s domain. If decisions regarding the crop are in male hands, including the sale and income from the sale, then intrahousehold food and nutrition outcomes might not improve (Quisumbing, 1995) and may even deepen inequalities in nutrition outcomes, changing farmers’ time use with negative trade-offs for women (Steiner-Asiedu et al., 2012).

Irrigation seems to have no significant effect on any livelihood aspect of adult women, which resonates with the extensive literature on women’s challenges to benefit from irrigation. Women are underrepresented in owners of irrigation equipment (Namara et al., 2014). Some women farmers use ‘jerri cans’ and buckets to water their crops, which is usually tedious in nature, requiring them to stay long hours to be able to water their entire farms. Women farmers are particularly disadvantaged by high up-front investment costs, absence of proper financing products, weak land rights, poor market integration and inadequate information services (Theis et al., 2018). At the same time, women are more time-constrained than men, and financial resources and local norms can make it difficult for women to hire labourers to support irrigation and other on-farm tasks (FAO, 2011). In some countries, such as Ghana, interest rates make loans unaffordable and would render irrigation investments unprofitable, and institutional barriers particularly restrict women from obtaining loans. In some cases, women may be forced to give male household members the microfinance loans they receive (Ganle et al., 2015). These intrahousehold power dynamics further reduce the ability of, and incentives for, women to invest in irrigated production.

In groundwater-based farmer-led irrigation, land access seemed to be a more critical factor limiting the engagement of women and youth. In Nabdam, the youth explained that not having access and control over land made it very difficult for them to make serious investments on the land they currently cultivate. In particular, women also confront cumbersome customary requirements that reduce their access to family land on which to farm (Theis et al., 2018). They are also faced with the difficult and expensive task of fencing their irrigable areas to protect their crops from animals.

Conclusions

A vast systematic empirical research body has measured the impacts of irrigation on poverty alleviation. However, the impacts of irrigation investments on broader measures of poverty and livelihoods have been seldom rigorously established, most likely because insufficient attention is often given to these goals during the design of irrigation interventions and policies. Poverty reduction and productivity gains are usually the most important drivers of irrigation programmes. Calls for incorporating nutritional, health, youth and gender considerations into the design of new irrigation programmes and policies are now more heightened than ever. Our results suggest that the impacts of irrigation on farmers’ livelihoods depend on the characteristics of irrigation development processes and contextual factors, and on the characteristics of the irrigators. The challenges facing young men, young women and adult women to act as catalysts for sustainable farmer-led irrigation development are of a different nature. Adult and young men irrigators reap the highest incomes from irrigation. However, the proceedings from irrigation of young men only slightly outweigh the remittances from those young men who travel to the cities to work during the dry season. Several constraints remain to make irrigation enterprises more profitable for young men considering the high opportunity costs they face if they stay behind during the dry season. Farmer-led irrigation appears to widen inequalities across gender gaps within the older generations. Whereas adult men realize very large and positive effects on irrigating and non-farm income, farmer-led irrigation seems to have no positive effect on any livelihood aspect of adult women, and the value of incomes from irrigation farming is not even half of those earned by their male counterparts. Irrigation income of young female irrigators is much higher than that of adult women, as rural–urban migration is higher among young men (compared to adult men), giving young women higher chances to access and farm family lands and make decisions over the sale of and incomes from their irrigated crop (but at the expense of higher workload and possibly less time for nutritious food choices).

Although several policies in Ghana seek to promote the use of small pumps and flexibility in collateral demand by informal and formal microfinancial institutions for farmer-led irrigation, technology and credit acquisition are still major challenges for the majority of women. These policies do not present measurable and achievable strategies and guidelines for mainstreaming gender in water management and support women’s participation. There is also an absence of clear by-laws in the districts concerning the involvement of women in irrigation, thus creating new inequalities or deepening existing ones.

A lack of basic facilities and infrastructures; the difficulty in accessing resources such as land, finance, and market information; and the opportunity cost of staying behind in the dry season instead of migrating to cities and earning alternative incomes limit the participation of youth in farmer-led irrigation development. Effectively increasing the participation and investment of youth in irrigation requires targeted investments to increase access to high-quality markets for inputs, outputs and credit. Shifts in livelihood activities and benefits from specialization cannot be supported through irrigation without market and value chain connectivity. Investments by and collaborations with state, private or civil society actors must be aligned with these farmer groups’ own investments and needs, in order to foster gender equality and engage and empower the youth.

Six main aspects must be considered to assess the multi-faceted impacts on farmers’ livelihoods when designing irrigation investments or policies: (1) multi-dimensional livelihood gains should be stated goals of irrigation programmes; (2) training programmes and awareness campaigns should accompany irrigation interventions to minimize negative trade-offs with non-economic benefits of irrigation; (3) multiple uses of irrigation water should be recognized in order to improve access to water supply and sanitation and ensure benefits across multiple livelihood dimensions; (4) women’s empowerment and women’s participation in irrigation programmes should be promoted; (5) programmes should promote irrigation as part of a dynamic youth’s livelihood portfolio; and (6) policy synergies between different sectors (agriculture, nutrition, health, water supply and sanitation, education) should be sought.

It is worth highlighting that focusing on irrigation alone is not the only answer to achieving more prosperous livelihoods. Our findings in relation to the opportunity cost of irrigation to youth underline the need to develop farmer-led irrigation support programmes as part of integrated livelihood strategies, where isolated ‘irrigation solutions’ may not be a desirable option. Policies and investments need to consider lock-in risks, for example, by confining youth into an irrigation avenue with loans or contracts, which may limit their options for alternative forms of livelihood diversification (Duker et al., 2023). It is imperative to consider how best to support agrarian development where farmer-led irrigation is only one of many forms to survive or prosper, often within a diversified livelihood portfolio.

We acknowledge a number of limitations in our study. First, our study focused on impacts at individual level and did not delve into the complexities of impacts unfolding across different household members. Second, the sampling strategy of the household survey used in this study did not allow us to estimate spillover or meso-level effects. Given the nature of irrigation investments, the presence of spillovers within communities is highly likely. Estimating spillover effects, when they exist, may help identify the additional indirect benefits of irrigation beyond irrigators themselves. This finding motivates future research into the mechanisms through which irrigation may generate additional impacts at community and regional levels. Finally, given the cross-sectional nature of this study and its inherent data limitations, we were not able to control directly for any time-varying unobservable characteristics that may drive the results, such as changing market conditions or agroclimatic factors. All in all, we conclude that more rigorous evaluations of the impact of irrigation interventions on livelihoods outcomes are needed. Developing such evidence will be important for the successful implementation of new irrigation projects and policies, especially in Sub-Saharan Africa, where the potential to expand irrigation is large and where recent projections indicate that growth conditions remain insufficient to reduce extreme poverty and boost shared prosperity in the medium to long term.

Future research is needed to further investigate: (i) the under-studied links between different types of irrigation development, women’s empowerment and women’s economic opportunities and their role in the community; (ii) the pathways through which water availability, access and use in different irrigation systems impact nutrition outcomes and the role of contextual and individual mediating factors; (iii) the impacts of specific patterns and characteristics of irrigation development processes on livelihood outcomes and poverty reduction, such as farmer-led institutional innovations within irrigation systems originally established by the state or other actors; (iv) how the use, fructus, management, and alienation rights held by different people (Theis et al., 2018) results in intrahousehold differences in control over irrigation technology, experience of its costs and benefits, and associated shifts in power. This evidence will help ensure that irrigation forms part of a holistic strategy that advances development objectives such as food and nutritional security, resilience, youth participation and women’s empowerment.

Acknowledgements

This work was part of a CGIAR Research Program project on Water, Land and Ecosystems (WLE) supported by CGIAR Fund Donors [V9-Reorienting agricultural water management] and part of the WLE’s funded Volta-Niger Focal Region Projects. Views expressed in this study are the sole opinion of the authors.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplementary data for this article can be accessed at https://doi.org/10.1080/02508060.2024.2330272.

Notes

1. As a first robustness check of the matched estimates, three other estimators (nearest-neighbour matching with 10 closest neighbours, kernel matching and local linear matching) were employed to produce point estimates. Estimates of the impacts of irrigation are robust to the choice of estimator. As a second robustness check, we performed the Rosenbaum Bounding sensitivity analysis on the computed outcome variables concerning deviation from the conditional independence assumption (Rosenbaum, 2002). We found that, in all models, inference for the impact of irrigation use on all aspects of livelihoods does not change, even though the irrigating and non-irrigating participants were allowed to differ in their odds of being treated up to gamma = 2 in terms of unobserved covariates (where gamma represents the log odds of differential due to unobserved factors where Wilcoxon significance level for each significant outcome variable is calculated). Generally, if gamma is greater than or equal to 2 and the p-value is significant, the results are insensitive to hidden bias (Hu & Hibel, 2014).

2. We do not estimate the effect of irrigation by community because the focus of this study is to compare differential irrigation impacts by type of irrigation and gender/age of the irrigator. We do account for community-level determinants of irrigation access by adding community dummy variables in the estimation of propensity scores.

3. The following formula was used to determine the minimum sample size required (n) to detect a minimum effect size (MES), given a chosen significance level (α) and level of power (1 – β), the standard error of the outcome variable (${\sigma }_y$) and the proportion of the sample in the treatment group (P): $\mathit{n}=\frac{{\left({\mathit{t}}_{\mathit{\alpha }/2}+t_{1-\mathit{\beta }}\right)}^2{\sigma }_y^2}{\mathit{ME}{S}^2\mathit{P}\left(1-\mathit{P}\right)}$ (White & Raitzer, 2017). Estimates of MES and (${\sigma }_y$) were taken from relevant irrigation impact and quantitative studies in the same region wherever possible (Acheampong et al., 2014; Domenech & Ringler, 2013; Hussain, 2007; Hussain & Hanjra, 2004; Namara et al., 2010; Namara, Awuni, et al. 2011; Namara, Horowitz, et al., 2011). The formula was applied separately to each outcome variable and sub-group (sex/age and irrigation scale/type groups) and adjusted for community clusters by multiplying the required sample size by a design effect: $\mathit{DE}=1+\left(\mathit{m}-1\right)\mathit{\rho }$, where m is the number of farmers per community in each sub-group, and ρ is the intracluster correlation coefficient, commonly assumed to be 0.2.

4. At the time of the survey, 1 USD = 4 Ghana cedi approximately.