Valuation and Aspirations for Drip Irrigation in Punjab, Pakistan

Productive use of scarce water is increasingly critical for South Asia (e.g., Barker and Molle 2004; Bengali 2009; Rasul 2014), where massive public-funded irrigation systems consume on average more than 90% (World Bank 2015) of surface water supplies, whose future is increasingly uncertain as rainfall, glacier-melt, and snowmelt shift under climatic changes (Archer et al. 2010; Hewitt 2011; Hosterman et al. 2012; Immerzeel et al. 2009). Modern drip-irrigation technologies can not only improve water-use efficiency in currently irrigated areas [by targeted delivery to crop root zones and reduced conveyance, distribution, and evaporative losses (Narayanamoorthy 2008)], but can also further expand water and land productivity by enabling transitions to higher-value crops or transforming areas that are not otherwise irrigable (too distant or too high to be reached by surface waters).

In India, where drip irrigation is expanding rapidly, savings in water and electricity costs at the farm scale have been consistently found in analyses on sugar (Narayanamoorthy 2005), grapes, cotton, and bananas (Narayanamoorthy 2010; Suresh Kumar and Palanisami 2011). For example, drip-irrigation systems were reportedly installed on 55,000 ha in 1994 (Sivanappan 1994), 246,000 ha in 1998 (Narayanamoorthy 2004), and 650,000 ha by 2013 (Chandrakanth et al. 2013). Despite high installation costs, the cost benefit ratios of drip investments range from 1.7 to 2, depending on the crops and availability of subsidies.

Improved water productivity at the farm scale from drip systems can contribute to a range of policy objectives at the landscape scale. Appropriate policy instruments that either limit global water access or (via water-user fees or the potential for trade) make water more valuable elsewhere (Tiwari and Dinar 2000) can translate the efficiency gains from tools like drip system improvements into global water-use efficiency improvements. In the absence of such instruments, the higher water productivity can create a Jevons paradox of higher overall water consumption (Berbel and Mateos 2014; Dumont et al. 2013), but increasing overall water consumption can also often be a goal of policy. Where water (as opposed to land or other inputs) is physically or economically scarce and limiting, improved water productivity can bring nonirrigated land into production or shift planting toward higher-value cropping patterns (e.g., Batchelor et al. 2014). In short, even where tools like drip systems improve water productivity for the farmer, the landscape outcomes (e.g., reduced consumption, improved environmental flows, or increased production) will depend on the policy context into which they are placed.

For Pakistan, goals for water use in agriculture include both increased efficiency and productivity. Vision 2025, the development framework put forward by the Ministry of Planning, Development, and Reform, calls for investment “in proven methods and technologies to minimize wastage (e.g., in the agricultural sector)” and to “improve efficiency of usage in agriculture by 20 [percent]” in the coming decade (Planning Commission 2014). The Punjab Irrigated-Agriculture Productivity Improvement Project (PIPIP), begun in 2012 with support from the World Bank, has the explicit goal to “improve water productivity,” contributing to “increased agricultural production, employment and incomes” (Punjab Agriculture Department 2011a). Given these priorities, the observed farm-scale benefits and the rapid expansion of drip systems in climatically similar areas in India, the persistently low levels of adoption in Pakistan—where seasonal costs for pumping groundwater can measure in the hundreds of USD per acre (Bell et al. 2014)—are something of a puzzle.

The International Food Policy Research Institute’s (IFPRI) 2012 Rural Household Survey (covering 942 households across 76 sampling units that are representative of the provinces of Sindh, Punjab, and Khyber Pakhtunkhwa) (IFPRI and IDS 2015) did not encounter a single farmer using a drip-irrigation system, suggesting that adoption rates are not likely any greater than 1%–2% of rural households and raising the question of what factors could be constraining them. In other regions, constraints have included high up-front capital investment costs and risks associated with operation (e.g., due to poor quality equipment), as well as a lack of knowledge about the systems and crops they can be applied to (e.g., Dai et al. 2015). Structurally, it would be helpful to understand whether the lack of drip-irrigation systems in use is better explained by such upfront barriers preventing initial adoption, or by challenges in using and maintaining drip systems hindering their sustained use by early adopters. Diagnosing this would require data spanning farms with no, little, and moderate experience with drip systems. A key challenge to such analysis would be identifying such a diverse population in a landscape where adoption is known to be very low.

PIPIP is currently the leading example of an instrument that has had some modicum of success in enhancing adoption (Punjab Agriculture Department 2011a), and it may provide the critical frame to identify and examine drip users. PIPIP aims to install drip-irrigation systems on 48,563 ha (approximately 120,000 acres), with cost sharing to reduce the capital burden and incentivize adoption. Specifically, the government covers 60% of the cost of installation through a subsidy paid directly to select service providers (registered with the government to install drip-irrigation systems) who are bound to subsequently provide after-sale technical and maintenance support to farmers on an as-needed basis for up to 2 years. Eventually, PIPIP aims to trigger upscaling by the local private sector, thus improving access to technical support and reducing capital investment costs. Through December 2015, PIPIP had subsidized 8,900 ha (approximately 22,000 acres) of drip systems installation across Pakistan’s Punjab Province (World Bank 2016), although to date, there has been very little published in the scientific literature to examine how the factors shaping adoption in Pakistan compared with India, or other contexts where drip irrigation has provided productivity and efficiency gains in dry environments, such as Australia (Mushtaq et al. 2013) or China (Burnham et al. 2015; Dai et al. 2015). With beneficiaries of PIPIP located across a range of agroecological conditions, and with some of these beneficiaries now having several years of experience and new users continuing to join the program, PIPIP provides an ideal backdrop for a rigorous evaluation of the factors shaping perceptions and constraining adoption of drip-irrigation systems.

This study contributes to an underdeveloped literature on drip-irrigation adoption by introducing a discrete choice experiment (DCE) across adopters and nonadopters in four districts of Punjab Province. The DCE is designed to diagnose the structure of the adoption problem in Pakistan, and to assess farmers’ aspirations for cultivation that could be realized under the availability of drip systems, through the lens of the current PIPIP regime. Specifically, this study examines how participants with different levels of drip experience value various aspects of a hypothetical drip-irrigation system, described by the amount of farmland covered, the overall cost of installation, the subsidy provided, and periods of knowledge and maintenance support (mirroring the attributes of PIPIP). This study provides a baseline assessment of drip system valuation, against which future (longitudinal) studies can assess changes in valuation over time. An increased valuation by users with increased exposure to drip systems could indicate a knowledge gap because such a result could imply that nonusers are less aware of benefits that can accrue from the use of drip irrigation. In contrast, decreased valuation with experience could indicate a support or maintenance gap because such a result could reflect the difficulties in using drip systems becoming clearer with use.

To foreshadow the obtained results, it was broadly found that utility for drip irrigation is higher among those with at least some drip experience, potentially suggesting an upfront knowledge gap as an important barrier to drip adoption. Encouraging drip adoption through subsidies is likely to accelerate adoption of drip-irrigation systems, but doing so entails weighing the benefits of more widespread adoption of drip against the fiscal burden of distributing large subsidies. An additional key result is that stated aspirations for cropping under a drip system are better predictors of the utility of drip than are current practices. In other words, farming system aspirations under drip differ from those under current systems, and in some parts of Punjab, the promotion of drip could be expected to be truly transformative of the agricultural landscape.

DCEs are a form of stated choice experiment where preferences are elicited based on responses to hypothetical scenarios rather than observed decisions. Although these methods were originally developed to address questions in the marketing and transportation literature (e.g., Louviere and Hensher 1982; Louviere 1988, 1992), they have been increasingly adapted and applied to other sectors, such as environment, health, and agriculture (Bennett and Birol 2010). Across all these disparate applications, choice experiments are particularly useful for studying preferences when functional markets are sparse or nonexistent, or when it may be beneficial to consider implications of multidimensional policy changes (Louviere et al. 2000). In a DCE, individuals are presented a series of hypothetical choice scenarios, in which they must choose between bundles containing different traits (in this case, different characteristics of a drip-irrigation system), where each trait takes one of a number of prespecified levels (i.e., the expression of the trait). Through statistical analysis of participants’ choices given the alternatives available in each choice scenario, the researcher can estimate preferences for the incremental changes in the various attributes embodied in the alternatives.

In every choice task for which the respondent selected one of the hypothetical alternatives, they were asked to carefully consider whether they would change any of their current cropping patterns given the new high-efficiency input selected. They were not asked for numerical estimates of how much cropping area they would change, but rather simply asked whether they would plant more or less of any particular crops (or perhaps entirely different crops than were currently being planted). Assessing aspirations is a relatively new area of research (Kosec et al. 2012), with current approaches divided among those that assess respondents’ agreement with hypothetical conditions or maxims, such as “Each person is primarily responsible for his/her success or failure in life” (Bernard et al. 2012), and those that ask respondents directly about their desires for income, education, or their children’s futures (e.g., Beaman et al. 2012; Kosec et al. 2012). The used approach borrows more from this latter approach, although the focus is very narrowly on aspirations for cropping patterns rather than other aspects of rural livelihoods, and thus different measures of aspirations are not drawn together in any kind of composite index. The only interpretation among aspirations is to distinguish those that lie along the extensive margin (i.e., those who aspire only for more of some crop) versus those along the intensive margin (i.e., those who aspire for more of some crop with less of some other crop).

The data used in this study come from a household survey conducted among 475 households across four districts (Attock, Chakwal, Layyah, and Sahiwal) in Punjab Province, Pakistan (Fig. 1). These districts capture contrasting agroecological zones (FAO 2004) and contrasting agricultural systems. Chakwal and Attock are predominantly rainfed areas, with some parts of Attock relying on groundwater use; Layyah is considered sandy desert, with low rainfall and poor access to canal infrastructure; Sahiwal sits within the irrigated Indus floodplain (Table 2).

Several features of the agricultural sector in Punjab affect outcomes and interpretations in this analysis. Of total cropped area, 42% in Punjab is in wheat, but in the winter (rabi) season, it is by far the major crop grown, with over 90% of acreage (Punjab Agricultural Department 2016). This is also the season with the lowest flows in the canal system and the least rainfall. Therefore, the choice of how much acreage to plant, especially in wheat, and given uncertainty about water, is of concern to most farmers during rabi season. In 2018, for example, the government announced an expected reduction in canal water of 45% in nonperennial canals, which serve approximately one-third of the irrigated acreage (Hasan 2018). This concern is exacerbated in rainfed districts such as Chakwal and Attock.

The long-term availability of water, along with a proximity to population centers and transport systems, have also affected market access, and likely the respondent’s aspirations in each district. Although all districts have central markets, with buyers and wholesalers participating, these will not have sufficient volumes of higher-valued products in areas with less varied production. Therefore, with a small diversity in production, the development of sufficient opportunities for high-valued products is not easy, and the aspirations for change could be muted. In contrast, in Sahiwal, there is a well-developed potato marketing and storage system, with sales across the country and to exports, and a strong trade association and access to improved technologies, so farmers there have seen successes. Finally, a major wheat procurement program has existed for many years, and larger farmers, such as those in this program, can make use of it. Currently, the prices in this program exceed international ones by over 25%, so in effect, this creates a barrier to shifting to production of higher-valued commodities.

In designing the sampling frame, one of the authors’ primary interests was to capture a gradient of drip-irrigation experience and exposure. Because current rates of drip-irrigation adoption are very low, any sort of random sampling (e.g., simple, stratified, or clustered) was unlikely to provide enough observations to reasonably estimate preferences among current drip users with any degree of precision. To circumvent this issue, this study’s sampling design purposively oversampled existing drip users. First, rosters of farmers receiving a subsidy under PIPIP in the sample districts were obtained. Attempts were made to contact all individuals on these lists to secure participation in the survey, with modest success, obtaining interviews from 237 respondents across the four districts. Where the authors were successful in engaging current drip users, the users were asked to identify another farmer in their village who was very similar to them (in terms of cropping patterns, farm size, and socioeconomic status, for example), but who, crucially, was not a current drip-irrigation user. These nominated farmers were subsequently interviewed, doubling the number respondents to a total sample of 475 farmers.

In some cases, farmers who were identified as having received subsidies had not in fact followed through with using drip systems, either because they were incorrectly listed as an adopter or because they actually purchased a system and paid the net cost but do not currently use it. Thus, the resulting sample consists of 164 current users and 311 nonusers, among whom 296 never used drip and 14 that had, at some point, used drip, but were currently not using the technology at the time of the interview. Of the current users, most were relatively new adopters. Roughly 63% of current drip users had fewer than 3 years of experience with these systems, whereas only 3% had more than 5 years of experience. Although the sample is not representative (due to oversampling drip adopters and consisting of larger farmers), the sample is arguably reflective of farmers who are most relevant for PIPIP and other such programs promoting drip irrigation. Importantly, it is a single-period sample, and although the level of experience with drip as controlled for, endogeneity or selection bias owing to farmers’ original decision to adopt cannot be resolved.

Basic summary statistics for the households included in our sample are reported in Table 3. Detailed summary statistics, including basic household demographic and agricultural characteristics, experiences with (for current drip users) and perceptions of (for nonusers) drip-irrigation systems (including gross installation costs per acre, subsidy per acre, and support provided), and information on current cropping patterns are provided in Tables S1–S3.

The results are presented in five parts. First, the results from the estimation of the overall utility function via the mixed logit model introduced in Eq. (2) are presented. Second, the estimated marginal utilities on cost and subsidies, along with results of two additional embedded experiments, are drawn upon to identify the appropriate income measure for estimating willingness to pay (WTP). Third, this model structure is applied to estimate WTP for installation of a hypothetical drip-irrigation system, across nonusers and current drip users. Fourth, a descriptive analysis of participants’ reported aspirations for their land, should they adopt the drip systems presented in the choice scenarios, is presented. Fifth, the aspirations results are integrated with the modeled marginal utilities to examine key explanatory factors in participants’ valuation of drip systems using regression analysis.

The principal objective in carrying out this study was to establish a baseline valuation of drip irrigation in the study area by respondents with different socioeconomic characteristics and experience with drip systems, such that further data collection would engender diagnoses of factors constraining adoption through changes in valuation with experience. The baseline findings unambiguously show a higher valuation of the coverage in a program by users with some experience with drip irrigation (compared with nonusers), along with a higher sensitivity to the cost of the program and an increased valuation of the value of the subsidy. No difference is found in the valuation of maintenance support with increased experience, and a lower valuation of knowledge support is observed among those respondents with 1–2 years of experience with drip systems. Taken together, these results are consistent with the narrative that adoption is limited by gaps in knowledge regarding drip—as observed by Dai et al. (2015) in the case of China—as opposed to issues with maintenance or realizing benefits (which a lower valuation with experience would indicate). Again, this study’s single round of data collection cannot resolve whether this higher valuation among new users simply means that those who value drip will adopt it, and those who do not value it in turn do not adopt it. However, these initial findings do help to refute the possibility of support or maintenance gaps being constraints on adoption (which a lower valuation among new users would suggest), and they set a foundation for future data collection. If it is in fact the case that those who do (and will) value drip systems highly have already adopted, then the costs of employing drip systems in the future may be progressively more costly to encourage—perhaps an opposing outcome to one in which a spread of shared knowledge among peers leads instead to lower costs of encouragement with time. This is an important question that only a panel-level analysis of the future path of adoption can inform, for which this study provides the first of two (or more) critical pieces. The ongoing promotion of drip systems via PIPIP will hopefully provide just such an opportunity for panel analysis in the years to come.

A second finding is that although some characteristics shaping higher valuation of a drip subsidy program are intuitive, i.e., those facing constraints in access to water, capital, or electricity, there is a counterintuitive lower valuation of drip subsidies by farmers with larger landholdings. PIPIP has explicit adoption targets by land holding size (5,400 units on 3-acre holdings; 5,400 units on 5-acre holdings; 4,800 units on 10-acre holdings; and 1,920 units on 15-acre holdings), and this study’s results suggest that it may be more successful in reaching targets for the smaller holdings (or perhaps reaching them earlier).

Finally, this study’s results show that aspirations may provide better predictors of valuation (and consequently, the likelihood of adoption) of a drip program than using existing cropping patterns. More specifically, it is observed that aspirations for expanded production along the extensive margin (stating aspirations for more of what they are currently cultivating, specifically more cereals or fruit and nuts) predict higher valuation of coverage area and a greater sensitivity to the cost of the system. In contrast, it is observed that aspirations for increasing production along the intensive margin (stating aspirations to do less of what they are currently cultivating, typically wheat, alongside aspirations to grow more of other products, specifically chilies, fruit and nuts, or oilseeds) predict higher sensitivity to the subsidy. Therefore, increasing the subsidy on drip-irrigation systems could encourage more high-value crop production, such as citrus, as farmers transition out of staple crops such as wheat. In contrast, increasing the areal coverage of the program could do more to encourage expansion of crop cultivation into currently undercropped areas, albeit in ways that might add to cereals production.

It is an open question of whether such aspirations can be reached and whether there are sufficient inputs and supporting market growth to permit this expansion. Although a complex assessment, a recent study by Young et al. (2019), looking at water demand for agriculture and nonagricultural uses in Pakistan, found that as income grows, consumption progressively moves toward more fruits and vegetables. These are commodities that would expand with more drip systems in place, even with slowing fertility, and Young et al. (2019) reported that demand from population growth will rise for all products, including basic cereals. Thus, the respondents’ aspirations appear to fit some assessments of the longer-run market outlook.

The fact that most respondents expect to expand on the extensive margin fits likely opportunities. The irrigation system in Pakistan was originally developed to cover subsistence water requirements and typically aimed at a 70% cropping intensity (the number of crops per year per acre), so adding cropped acreage would appear to be possible. This figure is now closer to 100%, only possible using additional groundwater, which often costs up to 10 times that of canal water (Government of Punjab 2015). Drip use will economize on expensive tubewell water to make it less costly. Moreover, the Punjab Government is encouraging water storage facilities via subsidies to raise water availability, which if developed in combination with a drip system, will create improved economic outcomes. In this regard, the gross value of production rises exponentially with cropping intensity, implying another payoff from extending land via a drip system (Tahir and Habib 2000).

More importantly, the linkage from aspirations to valuation of drip subsidy programs suggests that the expansion of drip use in Pakistan could have a transformative effect on the range of production. Although aspirations for citrus and other fruits in rainfed Attock and Chakwal correspond closely with current crops, the dreams of expanding vegetable and melon cropping in Chakwal and Layyah, or chili farming and vegetable gardens in Sahiwal, do not map onto the current agricultural landscape.

There is a clear linkage from drip program design to the question of the role of wheat and the programs that focus on wheat self-sufficiency (e.g., Farooq and Iqbal 2000) in Pakistan. For years, the inefficiencies and expense of the wheat procurement program has been of concern for many observers; results here suggest that an appropriate balance of subsidy and expansion of coverage could aid in making a transition away from a dependence on wheat production more successful, with the potential to expand the range of high-value output. To this point, efforts to diversify agricultural production have paled compared with the emphasis on food security via production of wheat, but part of a pathway out of this focus is supported by the results here.

Certainly, these stated aspirations are as subject to hypothetical bias as the findings of any nonbinding experiment, and, in the case of Sahiwal, the lack of local experience with drip might lead one to view the broad range of aspirations as being uncalibrated, although the extent of the higher-valued agriculture in the district may help bring more reality into their aspirations. Additionally, the study used a cross-sectional (single time period) data set from a nonrepresentative sample, and so there clear limitations on the types of inferences that can be made. One cannot make any inference about how drip systems could actually change farm performance or structure, nor can one infer adoption rates that could follow particular incentive structures. These baseline data do, however, allow identification of how the best available sample of drip system adopters in Pakistan think about farming with drip compared with their nonadopting peers, providing some compelling evidence that successful promotion of drip irrigation via programs like PIPIP could have a transformational impact on Pakistan’s agricultural economy.

This study has provided some evidence on the potential for drip-irrigation systems to transform the agricultural landscape in Punjab, Pakistan. A discrete choice experiment was undertaken framed around the hypothetical subsidized purchase of a drip-irrigation system in four districts within Punjab, a region where an existing subsidy program is helping to augment currently low levels of adoption of the technology. The nonrepresentative sample of adopters and nonadopters in the study districts identified a clear increase in valuation of drip systems among those who had adopted drip in recent years, possibly suggesting that farmers may be unaware of its opportunities or the benefits that may accrue. A possible counternarrative is that those who value drip systems have already adopted, implying a costly path to further expansion and highlighting the importance of a follow-up (panel) evaluation. It was also observed that aspirations for cropping systems under drip were better predictors of the valuation of drip systems than were current cropping patterns, implying that a different agricultural landscape might reasonably emerge under improved adoption of drip.

Aspirations differed across the different agroecological zones and water regimes captured by the study, with specific ambitions, such as for enhanced fruit and nut cropping in the rainfed districts of Attock and Chakwal, occurring when greater levels of experience with drip were observed in the study sample. In addition, there was a wider range of envisioned crops under drip in canal-fed Sahiwal, where very little experience with drip was observed in the study sample, but where it is also suspected that conditions with regard to market access for higher-valued production are good in that district. Aspirations to substitute wheat for fruit and vegetables were associated with higher sensitivity to the subsidy level, whereas aspirations to expand wheat were associated with a higher sensitivity to the area covered by the drip system; together, these imply a degree of control over the extent of wheat production in the landscape via careful design of the drip subsidy program. Although the penetration of drip irrigation is not sufficient enough yet to draw inferences from a representative sample, these results suggest a number of ways through which drip irrigation may transform Pakistan’s agricultural landscape and for enhancing its adoption.

Tables S1–S8 are available online the ASCE Library (www.ascelibrary.org).

This work was undertaken as part of the CGIAR Research Program on Policies, Institutions, and Markets (PIM) and facilitated by the Pakistan Strategy Support Program (PSSP) of the International Food Policy Research Institute (IFPRI) with funding support from PIM and USAID. The authors thank Mr. Omar Majeed and Mr. Khurshid Arshad for their efforts in coordinating and managing data collection.