WP No. 639

IIMB-WP No.
639/2021
WORKING PAPER NO: 639
Hybrid Agricultural Information Delivery System as a Means to Improve

Sustainability of Agriculture: Experimental Evidence from India
Gopal Naik
Professor
Economics & Social Sciences
Indian Institute of Management Bangalore
Bannerghatta Road, Bangalore – 560076
gopaln@iimb.ac.in
Aparna Krishna
Indian Institute of Technology (Indian School of Mines) Dhanbad
Dhanbad
aparna@iitism.ac.in
Year of Publication – April 2021
1|Page
IIMB-WP No. 639/2021
Hybrid Agricultural Information Delivery System as a Means to Improve Sustainability of

Agriculture: Experimental Evidence from India
Abstract
Extant research and policy on agriculture tend to treat financial sustainability and ecological sustainability
of agricultural operations as mutually exclusive goals. We show that sharing the current stock of
agricultural research with farmers through an innovative information delivery method can help in pursuing
these two goals simultaneously.
We used a mix of Information and Communication Technology (ICT) and agent-based information
delivery methods to provide information about all aspects of crop production with a focus on nutrient
management and plant protection to paddy (rice) and cotton farmers in a field experiment setup. The
method of information delivery enabled us to provide customized and detailed information. Our study
finds that two years of intervention led to an increase in paddy yields by 18 percent and cotton yields by
85 percent. The enhanced yields were achieved along with reduction in the use of inorganic fertilizers and
crop protection chemicals. At end of two years, intervention led to optimal, balanced usage of inorganic
fertilizers by farmers and an overall reduction in the quantity of fertilizers used. Intervention also led to
reduction in expenses related to crop protection chemicals owing to their proper usage. Yield enhancement
and cost reduction contributed to an increase in net returns by 78 percent and 221 percent for paddy and
cotton crops, respectively. Along with improving the financial viability of farming, proper use of
chemicals adds to the ecological sustainability of agriculture.
The results of this study are important because they show that improved delivery of agricultural
information can be used to attain goals beyond yield enhancement and that significant results can be
achieved in a relatively short span of time. The results also have important implications for the
management of agricultural information delivery initiatives and the continuity of such initiatives.
Keywords: Agricultural information delivery, Agricultural sustainability, Information and communication

technology, Randomized control trial
2|Page
1. Introduction
A set of practices developed around irrigation, seeds with higher yield potential, fertilizers, and crop
protection chemicals heralded the green revolution in India. However, due to their inefficient use of the
inputs and farm management practices, the productivity achieved by farmers has not reached full potential,
in addition to degrading land and water resources. On average, actual productivity realized by farmers in
India is 68 percent and 53 percent for paddy (rice) and cotton crops, respectively (Aggarwal, et al., 2008).
Moreover, 20 percent of the cultivable land in India is degraded due to incorrect use of fertilizers or
irrigation (GoI, 2016). Water use efficiency in agriculture in India is one of the lowest in the world (Shah,
2016) and 66 percent of states (the second rung of the Indian federal structure) have contaminated
groundwater due to leaching of harmful compounds from agricultural operations (Mali, Sanyal, & Bhatt,
2015).
Increasing productivity and protecting and enhancing natural resources are two of the five principles for
balancing the social, economic, and environmental sustainability of agriculture (FAO, 2018). In this study,
we explore whether providing information to farmers about efficient practices related to the use of
agricultural inputs could help the country move closer to these goals.
While the importance of delivering research-based information to farmers for improving the use of
agricultural inputs is well recognized, providing this information in the context of developing countries
has been a difficult task. Farmers might fall anywhere in the spectrum between subsistence agriculture
and commercial agriculture, and they might possess different capacities for information absorption (Babu,
Glendenning, Asenso-Okyere, & Govindarajan, 2012). In addition, inadequate regulatory mechanisms
and the proliferation of brands in the agricultural inputs market have made the choice of agricultural inputs
a complex decision for farmers. Any attempt to address these issues needs to meet two criteria, namely
customization of messages according to the farmers’ needs and communication of messages in a manner
that is easy for farmers to understand and adopt.
The two prevalent methods of communicating recommended practices to farmers fall short of one of the
two criteria. The traditional method of using extension agents—agents trained by state agricultural
departments, who contact farmers and provide them information—is suitable for customizing information
according to the farmers’ needs. However, this method has generally been useful only in cases where a
specific set of messages had to be delivered to farmers (Davis, 2008). Modern methods of information
delivery using information and communications technology (ICT) can be useful for disseminating vast
amounts of information in an easy-to-understand multimedia format to a large number of farmers, but
3|Page
these methods are not suitable for providing farmer-specific advice (Umadikar, Sangeetha, Kalpana,
Soundarapandian, & Prashant, 2014).
This study addresses the question of whether a mix of these two methods can help farmers to use key
agricultural inputs more effectively and measures the magnitude of the benefits in terms of financial and
ecological impact. Despite a growing understanding that neither of the methods is sufficient on its own,
recent studies have been limited to the evaluation of either of the traditional services, i.e., extension agents
visiting farmers and sharing information with them (Cerdan-Infantes, Maffioli, & Ubfal, 2008; Owens,
Hoddinott, & Kinsley, 2003; Sheng, Gray, & Mullen, 2010; Romani, 2003) or ICT-based methods (Aker,
2011; Fafchamps & Minten, 2012; Cole & Fernando, 2016). An exception to this trend is a study by
Gandhi et al., (2009) where the researchers studied the impact of digital videos along with mediation by
agents. However, the study limited its impact evaluation to the impact of intervention on changes in
agricultural practices and did not present quantifiable results on impact of information delivery methods
on agricultural outcomes such as yield, cost reduction, and returns.
In this study, we provided information to farmers about the effective use of agricultural inputs through
project-appointed extension agents whose agricultural knowledge was augmented through modules of
information on best practices and a special pest-diagnosis mobile application (app) that were loaded on to
an electronic tablet. The extension agent part of the intervention helped in customizing the information
while the tablet part helped in communicating accurate information in an understandable way. In the area
that we selected for intervention, we observed that agriculture was practiced on a commercial scale.
Inorganic fertilizers and crop protection chemicals were used excessively on average in this area, and there
were significant gaps between potential crop output and the output being achieved by farmers. We used a
randomized control trial design to establish the effects of the intervention on agricultural outcomes.
The intervention led to highly significant increases in yields, namely, 18 and 85 percent for paddy and
cotton crops, respectively, within a period of two years. The significant increases in crop yield were
achieved along with a statistically significant reduction in the expenses incurred by farmers for inorganic
fertilizers and crop protection chemicals and a reduction in the quantity of fertilizers used by farmers.
Within two years, the farmers moved closer to the recommended balanced combination of fertilizer use.
This study makes multiple contributions to the extant literature. Firstly, the magnitude of the results
achieved through our intervention is quite unprecedented and shows that agricultural information delivery
systems can help states to quickly achieve multiple goals linked to agriculture, especially those linked to
financial and ecological sustainability. Secondly, the present study is one of the few studies on the
4|Page
evaluation of agricultural information delivery methods that uses a widely accepted methodology to
establish causal impact of intervention on crop outcomes. Randomization at both the village level and
farmer level helped us in controlling the influence of farmer characteristics and agricultural endowments
of the area on the results of the study. Further, an emphasis on minimizing the potential sharing of
information between treatment farmers and control farmers helped us in accurately measuring the extent
of the impact of information delivery on agricultural outcomes. Thirdly, our study shows that a mix of
traditional and modern methods of agricultural information delivery can be effective, which opens doors
for the use of more innovative methods in the information delivery space.
The rest of the paper is structured as follows. In the following section, we briefly discuss the context of
poorly balanced agricultural practices and insufficient agricultural information provisioning in India
within which this study is nested, followed by details of the information delivery project from which we
draw our data. Subsequently, we discuss the identification strategy and the specifications of the
econometric and data envelopment analysis methods that we used for data analysis. Then, we discuss the
results obtained from the data analysis, and conclude the paper with the policy implications of our findings.
2. Contextual Background
2.1. Imbalance in agricultural practices in India
There are several dimensions of sub-optimal agricultural practices in India. The seed replacement rate,
i.e., percentage of crop area that is sown with certified or quality seeds is much below the prescribed
norms (Shreedhar, Gupta, Pullabhotla, Ganesh-Kumar, & Gulati, 2012). Another problem is the inefficient
use of crop protection chemicals that are unable to arrest crop losses. Nearly 25 percent of the potential
crop production in India is reported to be lost to pests, weeds, and diseases (FICCI, 2016).
The most significant sub-optimal practices are related to the use of inorganic fertilizers. While more than
three-fourths of the total cultivated area is treated with inorganic fertilizers (MoAFD, 2016), the three
major nutrients namely nitrogen, phosphorus, and potassium are not applied in sync with crop and land
requirements (Pavithra & Chand, 2015), and micro-nutrients are underused, on average (Shukla, 2010).
The problem of sub-optimal use of major nutrients is amplified by other factors such as a decline in the
use of manure (DoF, 2014) and the abandonment of traditional methods of land regeneration such as
leaving land fallow and returning crop residues to the land (Croppenstedt, Demeke, & Meschi, 2003).
Poorly balanced use of fertilizers has significant financial and ecological consequences. The expenditure
incurred toward inorganic fertilizers forms an average of 24 percent of the total expenses incurred in crop
5|Page
production (NSSO, 2013). Excessive, poorly balanced, or insufficient use of fertilizers prevents farmers
from reaping returns from their investment on fertilizers. The yield obtained by farmers per kilogram of
fertilizer used has steadily declined (Planning Commission, 2010) over time, and between 1950–55 and
2007–08, while fertilizer usage increased by 322 times, cereal production increased by only 5 times
(Prasad, 2009). Generally, when crop response to fertilizers declines, in the absence of any diagnosis of
the problem by experts, farmers tend to apply even more fertilizers. This leads to additional costs without
a commensurate increase in returns from crops (Mishra, 2007). Excess and poorly balanced use of
fertilizers might also result in more instances of pest or disease attacks and an increase in the problem of
weeds (Patil, Huggar, & Reddy, 2013).
Poorly balanced use of fertilizers has environmental consequences as well. Reduced use of organic sources
of plant nutrition and poorly balanced use of inorganic nutrients result in land degradation, i.e.,
deterioration of the physical, chemical, and biological characteristics of soil (Yedla & Peddi, 2009), which
in turn leads to the lower ability of the soil to utilize nutrients. About 20 percent of the cultivable area in
India has degraded due to soil alkalinity, sodicity, soil acidity, or soil salinity, all of which are result of
indiscriminate use of inorganic fertilizers (GoI, 2016). When land degrades, additional nutrients are not
taken up by crops, and these nutrients seep into the soil, leading to adverse environmental impacts such
as pollution of groundwater (Tilman, Cassman, Matson, Naylor, & Polasky, 2002), release of harmful
gases into the atmosphere (Prasad, 2009), and reduction in biodiversity (Horrigan, Lawrence, & Walker,
2002).
Imbalances in agricultural practices, especially those related to the use of fertilizers and crop protection
chemicals, also lead to yield gaps due to which the yield obtained by farmers is lower than the potential
yield that could be obtained if an ideal set of agricultural practices were followed. In India, the yield gap
for several crops is 50 percent or more, indicating that farmers obtain less than half of the possible yield
from the crop that they grow (Singh, 2012).
2.2. Role of agricultural information delivery methods

Yield gaps and problems in the use of agricultural inputs have been attributed to poor knowledge of crop
management techniques, among other factors (Alene & Manyong, 2006). The knowledge passed down
generations might not be sufficient in an agricultural setting where technology is rapidly evolving (Welch,
2001). While there is a need for continuous sharing of research-based information with farmers, the current
levels of access to agricultural information are quite low in India. Only 41 percent of Indian farmers have
6|Page
access to any source of information (NSSO, 2013), and only 6 percent have access to the public
agricultural extension system. In India, information is supposed to be delivered through state-appointed
extension agents, state-sponsored television and radio telecasts on agriculture, and articles in newspapers,
and also through centers from where farmers can access information (Gupta & Shinde, 2013).
Agricultural extension in India is probably best known for its contribution in bringing about the green
revolution. During the period starting in the 1960s and leading up to the late 1980s, agricultural extension
was aggressively used to enhance agricultural productivity and to expand food stocks of cereal crops
(Swanson, 2006). The Training and Visit (T&V) model of extension, which was launched in the late
1970s, was the most significant agricultural extension-related initiative of this period. Under the T&V
model, farmers were told about best practices for various crops by extension workers (Feder & Slade,
1993), who were in turn trained by subject matter specialists and were regularly monitored by supervisors
(Anderson & Feder, 2004). The T&V initiative lead to a higher level of contact between extension agents
and farmers, increased the farmers’ awareness of newer agricultural techniques, and increased agricultural
productivity (Feder & Slade, 1993).
The T&V systems were dismantled in the 1990s. Since then, there has been low provisioning of
agricultural information by public agencies, while the agricultural inputs market has become complex with
the proliferations of brands. Low provisioning of agricultural information by the public sector is mainly
linked to the structural adjustment and liberalization of the economy, which greatly reduced the funds at
the disposal of the government for various developmental initiatives (Rivera W. M., 2001). Low
provisioning of agricultural information is also related to a greater assurance of food security due to which
agricultural research and extension now receive less attention from the state (Balasubramanian, 2014).
Moreover, agricultural extension, like agriculture in general, is influenced by several factors and
infrastructural variables, which makes it difficult to isolate the impact of extension from other
environmental factors within the context of which extension operates (Birkhaeuser, Evenson, & Feder,
1991). That is, the impact of extension by itself is very difficult to trace (Anderson & Feder, 2004), which
pushes it behind other factors such as irrigation, whose benefits are easier to establish, for purposes of
budget allocation.
The agent-based system for information delivery has petered out in India, and only 6 percent of Indian
farmers have access to state-appointed extension agents (NSSO 70 th round), mostly due to budgetary cuts.
Moreover, there are other problems with the model, such as the system’s top-down approach and the
limited skills and knowledge of the agents.
7|Page
Several forms of information and communications technology (ICT) have emerged as alternatives to
agent-based agricultural information delivery. Some of the common ICT formats used to deliver
information are SMS-based and/or voice-based services available through mobile phones for
communicating agriculture-related information to farmers, and web-based question and answer (Q&A)
forums and web portals that act as information repositories (Umadikar, Sangeetha, Kalpana,
Soundarapandian, & Prashant, 2014). Web-based services can be accessed through tele-centers,
information kiosks, village knowledge centers, and multipurpose community centers (Mittal & Mehar,
2014). The main application of ICT in agriculture, at least in India, has been to provide farmers with
information about prices and other market-related information (Mittal & Mehar, 2014), although other
initiatives such as Kisan (farmer) call centers are meant to cater to all the information needs of farmers
(Ferroni & Zhou, 2012).
Mixed results have been reported about the performance of the various agriculture-related ICT initiatives
in India. Most applications are unable to provide farmer-specific advice, are not interactive, are mostly
query-based, and might not be suitable for farmers with limited literacy and limited understanding of the
use of internet or ICT applications (Umadikar, Sangeetha, Kalpana, Soundarapandian, & Prashant, 2014).
Access to the facilities required for such applications might also be limited. Farmers with larger land
holdings benefitted from these initiatives more than farmers with smaller landholdings (Ferroni & Zhou,
2012), and areas with better infrastructure have witnessed more progress in the use of ICT in agricultural
information delivery (Mittal & Mehar, 2014). Similar to the older systems, information is provided in a
top-down manner, with few options for farmers to seek further clarifications or to give feedback (Chapman
& Slaymaker, 2002). Some studies found that the content provided through ICT initiatives is of poor
quality (Heeks, 2002) and/or is not relevant (Roman & Colle, 2006).
The deficiencies in the various information delivery mechanisms are in large part due to the platform
designs, which do not consider the different needs of farmers based on where the farmer is located on the
spectrum that ranges from subsistence agriculture to commercial agriculture; the designs also do not
account for the differences in the information absorption ability of the farmers (Babu, Glendenning,
Asenso-Okyere, & Govindarajan, 2012).
There is little understanding of how groups with different levels of resources and skills or individuals of
different genders, age, and occupation absorb and use information (Richardson, 2005) and (Roman &
Colle, 2006), or what attributes of information—in terms of relevance, accuracy, affordability,
trustworthiness of the source, etc.—are considered necessary by farmers (Roman & Colle, 2006). This
8|Page
argument is supported by data from the NSSO 70th round, which shows that access to information does
not always translate into the adoption of the information. The three main reasons for not adopting
information are the lack of financial resources to put the information into practice, non-availability of the
suggested inputs and physical resources, and lack of technical follow-up related to the advice received
from the source of information (NSSO, 2013).
We hypothesize that through a mix of agent-based agricultural information delivery and ICT tools, it is
possible to design a system of agricultural information delivery that provides timely, reliable, accurate,
and relevant information in a way that can be easily understood and used by farmers.
3. Intervention
3.1. Details on the Intervention
To examine the impact of agricultural information delivery method on the sustainability of agricultural
operations, we designed a field intervention named Dynamic Agricultural Tablet-based Extension
Services (DATES).1 The intervention involved delivering agricultural information to farmers by project-
appointed extension agents who were equipped with electronic tablets. We used modules of agricultural
best practices, which had been developed and tested by a local agricultural university, 2 to provide
information to the farmers.3 Some of the information modules used for paddy crop are given in appendix
3. The intervention started in 2013 and concluded in 2015.
While information was provided about multiple aspects of crop production, namely, crop rotation, plant
variety, irrigation and drainage, weather forecasts, and weed control, the focus was on nutrient
management and plant protection. For pest management, the farmers were given information about the
types of pesticides to use for various pest infestations, the ideal time and procedure for applying the
chemicals, and organic alternatives to the chemicals. Similarly, for nutrient management, information
included organic alternatives to inorganic fertilizers 4 and effective use of inorganic fertilizers. The farmers
1
DATES was a joint effort of Indian Institute of Management Bangalore and the University of Glasgow. The initiative was
funded by the Economic and Social Research Council and Indian Institute of Management Bangalore.
2
The name of the local agricultural university: University of Agricultural Sciences, Raichur.
3
Information related to inorganic fertilizers and crop protection chemicals was found to be insufficient and a little dated;
therefore, we updated this information.
4
Organic alternatives are farmyard manure and green manure. Green manure is obtained by growing certain crops (mostly
field plants from the leguminous family) in the field and then incorporating the crops into the soil by ploughing after
sufficient growth.
9|Page
were encouraged to match the application of inorganic fertilizers to crop requirements, which mostly
involved applying fertilizers that were dense in certain nutrients at specific crop stages. 5
The information was delivered to farmers by agricultural graduates from the area designated for the study,
who were appointed under the project as extension agents. The extension agents were equipped with
electronic tablets on which we loaded the best practices modules and a special mobile application (app)
for pest diagnosis and pesticide recommendation, named Electronic Solutions against Agricultural Pests
(e-SAP). The main screen of the app contains photographs of crops infested with common pests. When a
user clicks on a photograph, the photograph gives way to details about the pest and suggested pesticide
details. The information contained in the e-SAP app was in the local language (Kannada) and was in
written, audio, and visual form. Printed copies of the suggested fertilizers and pesticides were provided to
enable farmers to purchase and apply the recommended chemicals accurately.
Armed with tablets, the extension agents paid regular visits to the treatment farmers, mostly in their
agricultural fields but occasionally at the farmers’ homes. Each agent provided information to 50 farmers.
In situations where the farmers faced pest-related problems, the agents would first try to diagnose the
problem with the help of the information available in the tablet and through the e-SAP app 6, and then
suggest remedial actions to the farmers 7.
If an agent was unable to diagnose the problem, he would take three photographs of the affected crop parts
and field conditions, and then submit the photographs to the online server. A scientist at the back-end
would diagnose the problem and upload suggested actions to the server, which the agent would then
communicate to the farmers. The names of crop protection chemicals were given as printouts to the
farmers by the extension agents. For other types of information, the agents relied on their own knowledge
and on the information modules in their tablets. For certain suggestions, such as those related to spacing
between crops or seed treatments, the agents demonstrated the process to the farmers and farm laborers.
The agents visited the farmers on a bi-monthly basis; over a period of two years, each treatment farmer
was visited 10 to 12 times by the extension agents based on the crop requirements 8. The visits were
planned to coincide with key stages in crop production, namely, sowing, flowering, and harvesting. The
5
Crop stages relevant for fertilizer application are sowing, crop establishment, flowering, and grain setting.
6
Appendix 3 contains representational photographs from e-SAP mobile application
7
Appendix 4 contains some photographs of interaction of extension agents with farmers
8
Appendix 5 contains a proforma of weekwise roster that was drawn for every extension agent throughout the intervention
10 | P a g e
farmers were also encouraged to call up the extension agents if they required any advice or additional
information.9
3.2. Mechanism for Impact of DATES intervention

Under DATES initiative, farmers were provided information on best practices for their crops, given their
unique economic and agricultural conditions. The aim was to bridge the gap between actual yield, i.e. the
yield realized by the farmers at start of the intervention and best practice yield, i.e. the maximum potential
yield of a given crop variety. This gap is known as extension gap. A typology of various possible crop
yields and corresponding gaps is given in figure 1.
Extension Gap Research Gap Science Gap
Best Practice Research Potential Science Potential

Actual Yield
Yields Yield Yield
Realised on Realised using best Realised by Realised by

average farmer’s available applying new research
fields technology research breakthroughs
Source: Evenson (2000)

Figure 1: Role of Agricultural Extension in Increasing Yields
Information provided under DATES fell in categories of quantity of variable inputs, embodied technical
change and disembodied technical change. These categories are situated under framework of total
agricultural output growth, given in figure 2.
Embodied technical change refers to use of products that have been modified and enhanced, such as seeds
with higher yield potential or better resistance to diseases. Disembodied technical change refers to change
in application of variable inputs and farm management practices. DATES encouraged farmers to use better
products, especially better seeds, fertilizers (organic and inorganic) and plant protection chemicals.
Farmers were also provided information on right product, right quantity, right method and right timing of
9
Several farmers reached out to the extension agents via mobile phones. However, an exact record of how many farmers
called and the number of times each farmer called an extension agent was not maintained.
11 | P a g e
use of variable inputs. The suggestions together led to increase in total factor productivity, defined as
output realized per unit of input.
Total Agricultural Output Growth
Short-term Shock Total Agricultural Input Change Total Factor Productivity Growth
Weather Input Fixed Input Quality Change in Quality and

Pests and diseases Quantity Application of Variable Inputs
Quality of Quality of Quality of Application

Labor Land variable of variable
inputs inputs
Education Irrigation Embodied Disembodied

Farmer health Conservation Technical Technical
Cropping pattern Change Change
Public R&D Extension

Private R&D Infrastructure
Source: (Wang, Heisey, Schimmelpfenning, & Ball, 2015)
Figure 2: DATES Intervention in terms of Total Agricultural Output Growth Model
Other components of total agricultural productivity growth, namely quality of fixed inputs and random
fluctuations like weather were not influenced by DATES intervention but were balanced between
treatment and control groups due to randomised control trial design.
4. Experimental Design
4.1. Randomization and sample characteristics
Since we were interested in preventing information sharing between treatment farmers and control farmers
as far as possible, we decided to stratify the farmers according to their gram panchayats (GP).10 We
assigned an entire GP as either treatment or control. The GPs were divided into two groups through a
10
For the purpose of this study, a gram panchayat (GP) is a cluster of villages. Across India, a gram panchayat is formed by
clustering three to five villages, on average. However, the number of villages under a GP can go up to 15 in certain
geographies. Formally, GPs are the base level of the rural local self-governance system in India, and representatives elected
at the GP level represent members at higher levels of local self-governance.
12 | P a g e
lottery system. To further increase the geographical distance between the groups of treatment and control
farmers, randomization was done to ensure that none of the control and treatment GPs were neighbors.
When a treatment GP was picked through lottery, all the GPs neighboring the selected treatment GP would
be removed from the lottery. The map of Siruguppa taluk given in Appendix 1 shows that the treatment
and control GPs were non-neighboring. Of the total 27 GPs in Siruguppa, six GPs were selected for the
treatment group, and six were selected for the control group. From each of the 12 GPs, we randomly
selected 50 farmers11 who met all the criteria in the screening survey.
The sample for this study consists of paddy (rice) and cotton 12 farmers whose land records existed in an
online portal for land records13. The farmers were selected from the Siruguppa taluk of Bellary district in
the state of Karnataka. We created a list of all the farmers in the selected treatment and control GPs, and
then conducted a screening survey and identified farmers from the shortlist who maintained residence in
the village, had grown at least one of the major crops in the previous year, and were planning to grow at
least one of the major crops in the upcoming agricultural season.
Based on power calculation (details in Appendix 2), we decided to conduct the study with 300 treatment
farmers and 300 control farmers.
5. Data and Empirical Strategy

5.1. Data
We used data collected at three different points during the intervention for our analysis. The first set of
data (the baseline) was collected at the start of the intervention for the year before the intervention. The
second set of data (the midline) was collected at end of the first year of intervention, and the third set of
data (the endline) was collected at end of two years of intervention14. At all three points, we collected
detailed crop production-related information and used that to construct variables for analysis. We analyzed
the impact of the intervention on yield (quantity of crop harvested per unit of cultivated area), revenue
11
Fifty farmers are about one-sixth of the total number of farmer-cultivators in a GP in the area of our study. Source:
https://data.gov.in/catalog/villagetown-wise-primary-census-abstract-2011-karnataka
12
Both rice and cotton are prominent crops in India and are grown on 31 percent and 8 percent respectively of country’s net
sown area (DACFW, 2017). Rice is an important food crop and staple food of almost 60 percent of the population of the
country (NFSM, 2016), and likewise cotton is an important fibre and cash crop (NFSM, 2018). Both crops play important role
in country’s agricultural economy but lag in terms of productivity when compared with rest of the world. Average productivity
of paddy and cotton realized by Indian farmers is 36 percent and 54 percent respectively of productivity realized by the most
productive nation (NFSM, 2018) , (NFSM, 2016).
13
Link to the portal: http://landrecords.karnataka.gov.in/service0/RTCHome.aspx
14
All three surveys, namely baseline, midline and endline had taken place after harvest of crops grown last agricultural
season of the year
13 | P a g e
(money from crop sales and imputed value of produce consumed at home), cost of production (actual and
imputed15 value of human labor, bullock and machine power, and agricultural inputs), net returns
(difference between revenue and cost), and the use of major nutrients (nitrogen, phosphorous, and
potassium).
5.2. Matching of observables

We wanted to understand the impact of the intervention on the major crops of the area. Hence, we
restricted the analysis to only those farmers who had grown either one of both of the two main crops
(paddy and cotton) at baseline and in either one or both of the other time periods (midline and endline).
50 out of 600 households (32 from the treatment group and 18 from the control group) did not meet this
criterion, and hence were not considered for analysis. For the remaining 550 households, we matched the
baseline observables of the treatment and control groups on 13 parameters. Table 1 presents the summary
statistics of the key variables categorized by the experimental groups.
Table 1: Comparison of Observables at Baseline
Panel A: General Observables
(1) (2)
(1) Vs (2)
Control Treatment
29.35 28.86 0.49
Age of family (Years)
(0.49) (0.46) (0.68)
4.10 4.96 -0.86***
Education of family (Years)
(0.18) (0.21) (0.27)
22.38 21.56 0.81
Experience in crop production (Years)
(0.69) (0.73) (1.01)
8.74 10.17 -1.42
Land owned at Baseline (Acre)
(0.60) (0.68) (0.91)
44.46 42.28 2.18*
Age of farmer (Years)
(0.82) (0.75) (1.11)
4.64 5.82 -1.18***
Education of farmer (Years)
(0.28) (0.30) (0.41)
0.99 .92 0.06***
Visits by extension agent at baseline
(0.01) (0.02) (0.02)
0.10 0.24 -0.15***
Visits to agricultural service centre at baseline
(0.02) (0.03) (0.03)
0.73 0.74 0.00
Faced income shortage
(0.03) (0.03) (0.04)
0.01 0.01 0.00
Female farmers
(0.01) (0.01) (0.01)
15
The imputed values were calculated by obtaining the number of home-owned units used for a certain process and multiplying
that with the per-unit rental rate. Example, if two units of machinery were used for a task and one of them was owned by the
family, then the imputed cost of machinery for that process was one machinery multiplied by the number of hours for which it
was used multiplied by the hourly rental rate.
14 | P a g e
Panel B: Caste Categories

(1) (2)
(1) Vs (2)
Control Treatment
0.30 0.44 -015***
General Caste
(.03) (0.03) (0.04)
0.08 0.09 0.00
Scheduled Caste
(0.02) (0.02) (0.02)
0.10 0.14***
Scheduled Tribe 0.24
(0.02) (0.03)
(0.03)
0.38 0.37 0.01
Other Backward Caste
(0.03) (0.03) (0.04)
Panel C: Land Ownership Categories
(1) (2)
(1) Vs (2)
Control Treatment
0.19 0.16 0.03
Marginal
(0.02) (0.02) (0.03)
0.24 0.21 0.04
Small
(0.03) (0.02) (0.04)
0.27 0.27 0
Semi-Medium
(.03) (0.03) (0.04)
0.22 0.27 -0.05
Medium
(0.02) (0.03) (0.04)
0.07 0.09 -0.02
Large
(0.02) (0.02) (0.02)
Panel D: Asset Index Categories
(1) (2)
(1) Vs (2)
Control Treatment
0.23 0.18 0.05
Asset Index 1
(0.03) (0.02) (0.03)
0.20 0.17 0.03
Asset Index 2
(0.02) (0.02) (0.03)
0.18 0.22 -0.04
Asset Index 3
(0.02) (0.03) (0.03)
0.17 0.22 -0.05
Asset index 4
(0.02) (0.03) (0.03)
0.21 0.20 0.01
Asset index 5
(0.02) (0.02) (0.03)
N 282 268
Columns (1) and (2) report sample means with standard errors in paraentheses. Column (3) reports difference
between the two experimental groups. Panel B reports the proportion of farmers from both groups in various caste
categories. Caste is a hereditary social stratification in India which is based on historic occupation of a group of
people. Belonging to certain caste has implications on status on social hierarchy. Panel C reports proportion of
farmers from both groups in various land ownership categories. Farmers are classified as marginal if they own less
that one hectare land, as small if they own between one to two hectares, as semi-medium if they own between two
to four hectares, as medium if they own between four to ten hectares and as large if they own more than ten hectares
of land. Cut-off limits for categorization by land ownership have been taken from
http://raitamitra.kar.nic.in/landholdings.html. Panel D reports categorization of farmers by asset index categories.
To calculate asset index categories, we ran principal component analysis on ownership of eleven types of
agricultural equipment and eleven types of household consumer durables. We then predicted scores based on PCA
15 | P a g e
results and using the scores divided farmers into five categories. Farmers with least predicted score were included
in asset index 1 and those with the highest score were included in asset index 5. The analysis has been done only
farmers who grew either or both of the main crops analysed in the study, namely cotton and paddy, at baseline.
The average age of the farmers was 43 years. In this group, the average number of years of schooling was
less than 6 years, and the average number of years of crop production experience was around 22 years. Of
the thirteen variables, difference between treatment and control groups was significant at the 1 percent
level for five variables. These five variables are average education of members of farmer’s family, years
of education of farmer, proportion of farmers visited by extension agents in previous year, proportion of
farmers who visited agricultural service centre in previous year and caste composition of farmers.
Similar summary statistics were compiled for cotton and paddy farmers and are given in appendix 7 and
8 respectively. We control for variables on which treatment and control farmers differ at baseline in
regression analysis of respective crops.
5.3. Empirical strategy

For our econometric analysis, we used a three-period difference-in-differences (DID) technique. The DID
specification is given in equation 1. Equation 1 is based on specification for decomposing partial factor
productivity provided by Evenson (2000), which allows us to consider agricultural extension as the main
independent variable. The DID specification controls for differences between the control group and
treatment group at baseline (Kusuma, et al., 2017) and assumes changes in values of the dependent
variable for the control group as a time effect (Athey & Imbens, 2006). Thus, this technique separates out
the effect of the intervention on the treated group from the effect of time and group-specific characteristics
(Puhani, 2008). Thus, β1 and β2, which are the coefficients of the interaction terms, capture the effect of
the treatment on the treated group for the midline and endline, respectively, relative to the baseline. We
cluster standard errors at the first level of randomization, i.e., at the GP level.
(1) 𝑌 = 𝑎 + 𝛽 𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡 ∗ 𝑀𝑖𝑑𝑙𝑖𝑛𝑒 + 𝛽 𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡 ∗ 𝐸𝑛𝑑𝑙𝑖𝑛𝑒 + 𝛽 𝑀𝑖𝑑𝑙𝑖𝑛𝑒 + 𝛽 𝐸𝑛𝑑𝑙𝑖𝑛𝑒 +
𝛽 𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡 + ∑ 𝑋 + ε
where,
Yit: Value of the outcome variable for farmer i at the baseline, midline, or endline
Treatmenti*Midline: Interaction between Midline and the treatment dummy, which takes the value 1 if the
farmer from the treatment group grew the crop in the midline
16 | P a g e
Treatmenti*Endline: Interaction between Endline and the treatment dummy, which takes the value 1 if the
farmer from the treatment group grew the crop in the endline
Midline: Time dummy that takes the value 1 if the farmer grew the crop in the midline
Endline: Time dummy that takes the value 1 if the farmer grew the crop in the endline
Treatmenti: Dummy for the treatment status of the farmer, which takes the value 1 if the farmer is a
treatment farmer
Xi: Variables on which treatment and control farmers differ at baseline
As a robustness check, we also estimate p-value of wild bootstrapping of standard errors, with 100
repetitions.
To determine whether the farmers became more efficient in their agricultural operations due to the
intervention, we computed and compared technical efficiency scores for the farmers growing the two
crops. The technical efficiency scores were computed using the data envelopment analysis (DEA) method.
We analyzed the data using the DEA method to understand whether the intervention had an impact on the
efficiency of the farming enterprise. The measure of change in efficiency in the DEA method is better
than that used in the DID of returns for three reasons: (a) we can include the element of land, which could
not be included in the regression analysis due to lack of data about land rent; (b) we could remove price
from the equation to obtain the direct impact of the intervention on the output obtained from various
inputs; and (c) unlike the regression equation for crop returns, where every input received equal weight,
in this case, the inputs received weight proportionate to their importance to crop cultivation.
The DEA method uses linear programming techniques to construct a production frontier of various
combinations of inputs and outputs (Coelli, Rahman, & Thirtle, 2005). The most efficient farms lie on the
frontier and are given a technical efficiency score of 1. Farmers that are not efficient lie below this frontier,
and their efficiency scores are computed in terms of their distance from the frontier (Ji & Lee, 2010).
Because the DEA method is non-parametric, we did not have to assume or specify the functional form of
the frontier or the distributional form of the error term (Coelli, 1995), or provide weights for the various
inputs and outputs (Cooper, Seiford, & Tone, 2000).
We used variable returns to scale and output oriented DEA model for analysis the data. Model
specification is given in equation 2.
(2) 𝑀𝑎𝑥 Φ s.t.
Φ𝑦 , ≤ ∑ 𝑧 𝑦 , Ɐm
∑ 𝑧𝑥, ≤ 𝑥, ɸ Ɐn
17 | P a g e
∑ 𝑧 =1
where, ɸ is a scalar outcome showing how much the production of each crop can increase by using the
inputs (both fixed and variable) in a technically efficient configuration. yj,m is the amount of output m by
firm j, xj,n is the amount of input n used by firm j, and zj are weighting factors. The restriction
∑ 𝑧 = 1 allows for variable returns to scale.
For DEA analysis we used crop yields as output and consumables, labor, machine and animal power, and
size of land in which the crop was cultivated as input.
6. Results
6.1 Impact of intervention on yields, revenues, costs and returns
The crop-wise details of the number of farmers who had grown the crops, the average area cultivated by
them, and the percentage of farmers who had access to irrigation are given in Table 2.
Table 2: Descriptive Statistics
Comtrol Farmers Treatment Farmers
Panel A: Number of farmers
Baseline Midline Endline Baseline Midline Endline
Paddy 178 173 147 199 197 146
Cotton 146 140 128 119 112 111
Panel B: Average area under cultivation (acre)

Paddy 7 8 9 8 9 10
Cotton 6 8 10 6 8 9
Panel C: Access to Irrigation (% of farmers)
Paddy 100 100 98 99 99 100
Cotton 28 37 31 19 38 22
Panel A reports the number of farmers who had grown the respective crop at the respective survey points. Panel B
reports the average area in acre cultivated of respective crop at the respective survey points and panel C reports the
percentage of farmers who had access to the irrigation for the respective crops. Farmers who did not grew the
respective crops at baseline have been excluded from analysis.
Given that paddy is the dominant crop of the region, at all the three time periods, the number of farmers
growing paddy was higher than the number growing cotton. Further, the farmers devoted more land to
18 | P a g e
paddy crop, on average. Most paddy farmers had assured access to irrigation in all the three time periods,
while less than 40 percent of the cotton farmers had access to irrigation.
The average values and standard deviations of yields, revenues, costs, and net returns for paddy and cotton
crop at the three time periods are reported in Table 3.
Table 3: Average Values of Yields, Revenues, Costs and Returns
Control Farmers Treatment Farmers
Panel A: Yields (Quintal/Acre)
Paddy 25 26 27 24 26 31
(7) (6) (4) (7) (5) (3)
Cotton 7 9 9 6 11 14
(4) (4.) (1) (4) (4) (2)
Panel B: Revenues (Rs./Acre)
Paddy 35830 39515 40268 35695 41566 50752
(13301) (10397) (7042) (12819) (9264) (5714)
Cotton 28610 42172 34941 22861 50820 58471

(16508) (19235) (5681) (15765) (20183) (11397)
Panel C: Costs (Rs./Acre)
Paddy 19396 20538 23960 20931 23011 19753
(4854) (4221) (3566) (5364) (5559) (2697)
Cotton 23589 25192 24704 23539 28803 23147

(6944) (7207) (4267) (7132) (8143) (4556)
Panel D: Returns (Rs./Acre)
Paddy 16434 18976 16308 16016 18555 30999
(14012) (11612) (7197) (13907) (11393) (6457)
Cotton 6021 16979 10237 -678 22017 35324

(15992) (17292) (6579) (15680) (20515) (10774)
The table reports average values of yields, revenue, cost and returns at the three survey points as reported by
farmers in the study. Standard deviations are given in parentheses. Farmers who did not grew the respective crops
at baseline have been excluded from analysis.
In terms of average yields, at the baseline, the paddy farmers from both the treatment group and control
group had similar average yields. However, the control farmers growing cotton had higher yields on
average compared to the treatment farmers at the baseline. At the midline, the average yields were similar
19 | P a g e
for the treatment and control farmers for paddy but were higher for the treatment farmers for cotton. At
the endline, the treatment farmers were getting higher yields than the control farmers for both crops.
Comparison of values of yields, revenues, costs and net returns obtained by treatment and control farmers
at baseline is given in Table 4.
Table 4: Comparing values at baseline: Yields, Revenues, Costs and Returns
Panel A: Yields (Quintal/Acre)
(1) (2)
(1) Vs (2)
Control Treatment
Paddy 25 24 1
(1) (0) (1)
Cotton 7 6 2***
(0) (0) (0)
Panel B: Revenues (Rupees/Acre)
(1) (2)
(1) Vs (2)
Control Treatment
Paddy 35830 36947 -1117
(997) (891) (1333)
Cotton 28610 22861 5749***

(1366) (1445) (1998)
Panel C: Cost (Rupees/Acre)
(1) (2)
(1) Vs (2)
Control Treatment
Paddy 19396 20931 1535***
(364) (380) (529)
Cotton 22589 23539 -950

(575) (654) (868)
Panel D: Returns (Rupees/Acre)
(1) (2)
(1) Vs (2)
Control Treatment
Paddy 16434 16016 418
(1050) (986) (1440)
Cotton 6021 -678 6699***

(1324) (1437) (1958)
The table reports comparison of yields, revenues, cost and returns observed at baseline for the two crops between
treatment and control group. Columns (1) and (2) report sample means with standard errors in paraentheses. Column
(3) reports difference between the two experimental groups. Farmers who did not grew the respective crops at
baseline have been excluded from analysis.
Between the two groups, there were statistically significant differences at baseline in yields, revenues and
returns for cotton crop and cost for paddy crop. At baseline, cotton farmers in control group realized higher
yields, which also resulted in higher revenues and returns as compared to the treatment group. On the
20 | P a g e
other hand, control farmers growing paddy incurred significantly lesser cost per acre for their crops at
baseline as compared to treatment farmers.
Table 5: Treatment effect on crop yields, revenue, cost and returns
Panel A: Paddy Crop

(1) (2) (3) (4)
Yield Revenue Cost Returns
Treatment*Midline 0.950 1273.7 864.5 249.5
(0.857) (1997.8) (1380.8) (2338.6)
[.31] [.57] [.56] [.92]
Treatment*Endline 5.036*** 9263.3*** -5715.8*** 14901.8***

(0.908) (1828.2) (870.3) (2103.0)
[.01] [0] [0] [0]
Mean of dep. Var 26.22 40367.5 21251.9 19122.4
Observations 1040 1040 1040 1040
R-Square 0.178 0.210 0.134 0.199
Adjusted R-square 0.165 0.197 0.121 0.187
Panel B: Cotton Crop
(1) (2) (3) (4)
Treatment*Midline 3.503** 14400.2** 2700.4 11550.1
(1.204) (5385.6) (2648.9) (6733.6)
[.01] [.01] [.44] [.14]
Treatment*Endline 7.174*** 29289.4*** -2451.4 31761.4***

(1.450) (6254.6) (1519.9) (5569.4)
[.01] [0] [.08] [.03]
Mean of dep. var 9.161 38913.6 24568.4 14365.2
Observations 756 756 756 756
R-Square 0.400 0.399 0.103 0.391
*
p < 0.05, ** p < 0.01, *** p < 0.001
Clustered standard errors in parenthesis. Wild bootstrap p-value in square brackets.
Panel A reports results of regressions of per acre yield, revenue, cost and returns for paddy crop on treatment and time
interaction terms in a difference-in-difference setup. Panel B reports the same results for cotton crop. Yield is measured in
terms of quintals per acre and revenue, cost and returns are measured in terms of rupees per acre. Values of all variables were
winsorized at the 99th percentile. The standard errors were clustered at level of Gram Panchayat (first level of randomization).
The interaction coefficients from the difference-in-differences (DiD) regression for the four dependent
variables are reported in Table 5.
In the DiD analysis16, no significant change in the yields of paddy crop was observed at the midline. By
the time DATES was operational, several farming-related decisions relating to paddy crop had been taken
16
Full results from DiD analysis are given in appendix
21 | P a g e
by the farmers, due to which the intervention did not have a significant impact on the crop at the midline.
However, a statistically significant increase in yields was observed for the treatment farmers for paddy
crop at the endline compared to the baseline. The intervention led to increase in yields of paddy crop by
19.2 percent.
For cotton crop, significant changes in crop yields were observed at both midline and endline for treatment
farmers. Cotton yields for the treatment farmers increased by 38 percent at the midline and 78 percent at
the endline, as compared to baseline.
There was greater scope for improvement of the yields for cotton compared to paddy, which also explains
the difference in the magnitude of impact of the intervention. The potential yields for the prominent
varieties of the two crops are given in Table 6. At the baseline paddy farmers in treatment group were
realizing 85 percent of the potential yield. In contrast, the cotton farmers in the treatment group were
realizing only 41 percent of the potential yields at baseline.
Table 6. Potential Yields of Prominent Varieties of Crops grown by Farmers
Average Yield Recorded at Baseline
Crop Variety Potential Yield
Treatment farmers Control farmers
Paddy Sona Masuri 28 23.8 25.9
Cotton Ajith Jadoo 13.9 5.7 7.2
Note: Potential yields were compiled by project staff using information from various sources. Unit for yield is quintals per acre
Cotton is a more difficult crop to manage compared to paddy due to the longer crop cycle and susceptibility
to various pests. The DATES intervention led to farmers achieving substantial part of the yield potential
of their crops. Since the yield gap was higher for cotton, we see a higher impact of the intervention on the
cotton crop compared to that of paddy.
The DiD coefficients for revenue were similar to those obtained for yields in terms of magnitude and
significance. Statistically significant results for revenue were obtained for paddy at the endline (23
percent) and for cotton at the midline and the endline (37 percent and 75 percent, respectively).
In terms of average costs, at the baseline, the per-acre costs incurred by the treatment farmers for paddy
and cotton were marginally higher than those incurred by the farmers in the control group. A similar
difference between the groups continued at the midline for both crops. However, at the endline, the
average costs incurred by the treatment farmers were lower than the costs incurred by their control
counterparts.
22 | P a g e
The DiD coefficient for costs was positive but not statistically significant at the midline for the treatment
farmers for paddy crop. However, the coefficients were negative and statistically significant for paddy at
the endline. The DATES intervention led to a reduction in the cost of cultivation by almost 26 percent for
paddy and 10 percent for cotton for the treatment farmers at the end of two years of intervention. Results
for paddy are significant at one percent and those for cotton are significant at eight percent level 17.
Table 7: Treatment effect on major components of cost
Panel A: Paddy Crop

(1) (2) (3) (4) (5) (6)
Plowing Seeds Transplanting Irrigation Weeding Harvesting
Treatment*Midline 48.40 725.9** -115.7 -1041.0*** 99.65 429.4
(123.1) (308.0) (90.32) (193.8) (266.6) (285.0)
[.57] [.01] [.33] [.01] [.73] [.2]
Treatment*Endline -349*** 6.075 -280.5*** -114.5 -1195*** 65.31

(104.3) (338.4) (77.04) (217.9) (139.1) (283.5)
[.02] [.97] [0] [.66] [0] [.8]
Mean of dep. var 1127.7 1648.7 2064.6 1767.8 1863.2 2605.1
Observations 1040 1040 1040 1040 1040 1040
R-Square 0.262 0.0788 0.485 0.171 0.112 0.129
Adjusted R- 0.250 0.0644 0.477 0.158 0.0978 0.115
square
(1) (2) (3) (4) (5) (6)
Plowing Seeds Sowing Intercultivation Weeding Harvesting
Treatment*Midline 298.3** 33.55 107.7 63.27 553.0 -842.9
(99.08) (211.3) (222.8) (279.9) (927.0) (502.4)
[.09] [.88] [.62] [.86] [.56] [.17]
Treatment*Endline 208.6 -246.3 95.80 -681.2** -54.94 201.7

(125.1) (142.7) (141.3) (227.0) (431.4) (767.6)
[.26] [.19] [.58] [.06] [.88] [.85]
Mean of dep. var 1036.5 2415.3 697.0 1957.9 2433.9 3927.2
Observations 756 756 756 756 756 756
R-Square 0.157 0.0645 0.0284 0.0621 0.0589 0.0967
Adjusted R- 0.140 0.0456 0.00866 0.0431 0.0398 0.0784
square
*
p < 0.05, ** p < 0.01, *** p < 0.001
Panel A reports results of regressions of major components of cost of production on treatment and time interaction terms in a
difference-in-difference setup. Panel B reports the same results for cotton crop. Values for all dependent variables are in terms
of rupees per acre. Values of all variables were winsorized at the 99 th percentile. The standard errors were clustered at level of
Gram Panchayat (first level of randomization).
17
Costs reduction for cotton crop are significant according to p-values compiled through wild bootstrapping of standard
errors
23 | P a g e
To understand the components of cost on which the project had an impact, we ran DiD regressions for the
main components of cost. The results are presented in Table 7.
For the paddy farmers, at the midline, the project led to a significant increase in the cost of seeds and a
significant reduction in the cost of irrigation. The farmers were advised to use ‘younger’ paddy saplings,
which were more expensive. This accounts for the increase in seed related expenses. Also, the farmers
adopted the advice on better water management, which might have led to them using lesser water for
irrigation at midline as compared to baseline, thus reducing the cost of labour needed for irrigation. At the
endline, the treatment farmers spent significantly lesser on plowing and weeding for paddy crop than they
did at baseline. For cotton crop, the regression coefficients are statistically significant and negative for
interculture. We explain reduced cost of weeding and interculture in later part of the section.
In terms of average net returns, at the baseline, the treatment farmers growing cotton received negative
returns on average, implying that when the imputed value of family labor was included in the calculation
of costs, the costs exceeded the revenue. In the case of both paddy and cotton, while the average returns
were higher for the treatment farmers at the midline compared to those for the control farmers, the results
of the DiD analysis showed that there was no statistically significant increase in the returns obtained by
the treatment farmers at the midline because of the intervention.
At the endline, for both the crops, the treatment farmers obtained higher returns than the control farmers
did. This finding is supported by the DiD coefficients. The project led to an increase in returns for the
paddy farmers by 78 percent and for the cotton farmers by 221 percent at the end of two years of
intervention. The substantial increase in returns is a result of the combined effect of the increase in
revenues and the decrease in costs for the treatment farmers at the endline.
6.2 Impact of intervention on efficiency of agricultural operations

The results of the analysis of the changes in yields, costs, and returns were validated through a comparison
of the technical efficiency scores obtained by the treatment farmers and control farmers in the three time
periods. As mentioned in section 5.3, the technical efficiency scores reflect the relative positioning of the
farmers with regard to the frontier of efficient transformation of inputs to outputs. The higher the technical
efficiency score, the closer a farmer is to the frontier, and the more efficient the farmer is in transforming
inputs to outputs.
Table 8 presents the percentage of the treatment and control farmers in various categories according to
their technical efficiency scores for the two crops in the three time periods.
24 | P a g e
Table 8: Technical efficiency score

Treatment Farmers Control Farmers
Panel A: Paddy Crop
TE
Score Baseline Midline Endline Baseline Midline Endline
Range
<.7 62.94 67.00 4.11 62.29 63.00 35.37
.7 - .8 16.24 13.71 20.55 18.28 14.45 35.37
.8 -.9 8.62 6.10 39.00 5.14 6.93 19.05
.9 – 1 12.18 13.20 36.30 14.28 15.61 10.20
TE
Score Baseline Midline Endline Baseline Midline Endline
Range
<.7 77.97 73.21 32.43 75.86 72.14 91.41
.7 - .8 5.93 7.14 37.83 6.89 7.14 0
.8 -.9 5.08 6.25 11.71 2.76 5.00 2.34
.9 – 1 11.02 13.39 18.02 14.48 15.71 6.25
Panel A reports percentage of farmers in respective categories of technical efficiency scores for paddy crop. While
calculating technical efficiency scores, crop yields were used as output and cultivated area (in acre) and expense on
consumables (seeds, fertilizers etc.), human labour (actual and imputed) and bullock and machine power (actual and imputed)
as inputs. Panel B reports the results for cotton crop.
Between the baseline and the endline, there is a clear movement of the treatment farmers from the lower
technical efficiency score ranges (lower than 0.7) to the higher ranges (0.8 to 1). For instance, 63 percent
of the treatment farmers growing paddy had technical efficiency scores lower than 0.7, and 12 percent had
technical efficiency scores between 0.9 and 1 at the baseline. At the endline, only 4 percent of the treatment
farmers had technical efficiency scores lower than 0.7, and 36 percent had technical efficiency scores
between 0.9 and 1.
In case of cotton, percentage of treatment farmers in 0.9 – 1 score category increased from 11 percent at
baseline to 18 percent at endline and percentage of treatment farmers in the lower score category (lower
than 0.7) reduced to 32 percent in the endline, compared to close to 78 percent in the baseline and the
midline. It is important to remember that the technical efficiency scores are relative in nature. Hence, any
worsening in the scores obtained by the control farmers, while supported by the earlier data about
declining revenues and increasing costs, is partly due to the treatment farmers becoming more efficient in
their operations as compared to control farmers and moving toward the efficiency frontier.
6.3 Impact of intervention on use of inorganic fertilizers and plant protection chemicals
Average value and standard deviation of expense incurred by treatment and control farmers on fertilizers
and insecticides are presented in table 9 and comparison of expenses incurred at baseline by the two groups
25 | P a g e
of farmers is presented in table 10. For both paddy and cotton, expense incurred by control farmers at
baseline on fertilizers and insecticides was significantly lower than expense incurred by treatment farmers.
Table 9: Average Values of Expenses on Fertilizers and Insecticides
Control Farmers Treatment Farmers
Panel A: Expense on Fertilizers (Rupees/Acre)
Baseline Midline Endne Baseline Midline Endline
Paddy 4817 4536 7138 5768 5990 4587
(2260) (1680) (1704) (2582) (2801) (1336)
Cotton 3641 6702 7363 4210 8645 6356

(2197) (3031) (1867) (2103) (3528) (1783)
Panel B: Expense on Insecticides (Rupees/Acre)
Paddy 1023 1065 2433 1234 1364 2114
(768) (513) (1154) (817) (1027) (974)
Cotton 1167 1666 2854 1637 2190 1991

(864) (949) (943) (1289) (1126) (607)
The table reports average quantities and standard deviation (in brackets) of expense incurred on fertilizers and
insecticides at the three time periods by treatment and control farmers. Farmers who did not grew the respective
crops at baseline have been excluded from analysis.
Table 10: Comparing values at baseline: Expense on fertilizers and insecticides

Panel A: Expense on Fertilizers (Rupees/Acre)
(1) (2)
(1) Vs (2)
Control Treatment
Paddy 4817 5768 950***
(169) (183) (251)
Cotton 3641 4210 -568**

(182) (193) (266)
Panel B: Expense on Insecticides (Rupees/Acre)
(1) (2)
(1) Vs (2)
Control Treatment
Paddy 1023 1234 -212**
(58) (58) (82)
Cotton 1167 1637 -470***

(72) (118) (133)
The table reports comparison of expense incurred on fertilizers and insecticides at baseline for the two crops by
treatment and control group. Columns (1) and (2) report sample means with standard errors in paraentheses. Column
(3) reports difference between the two experimental groups. Farmers who did not grew the respective crops at
baseline have been excluded from analysis.
26 | P a g e
On average, control farmers incurred higher expense on fertilizers and insecticides for both paddy and
cotton crop at midline and endline, as compared to baseline. On the other hand, expense incurred by
treatment farmers on the two products increased in midline as compared to baseline but reduced again in
midline for both paddy and cotton crop.
Table 11: Treatment effect on expense on fertilizers and insecticides
Panel A: Paddy Crop
(1) (2)
Fertilizers Insecticides
Treatment*Midline 464.7 63.38
(562.2) (181.1)
[.44] [.69]
Treatment*Endline -3464.4*** -531.9*

(390.2) (244.3)
[0] [.04]
Mean of dep. Var 5458.5 1476.7
Observations 1040 1040
R-Square 0.157 0.295
Adjusted R-square 0.144 0.284
(1) (2)
(901.3) (308.9)
[.19] [.92]
Treatment*Endline -1600.7** -1310.1***

(649.6) (198.8)
[.03] [.01]
R-Square 0.343 0.258
*
p < 0.05, ** p < 0.01, *** p < 0.001
Panel A reports results of regressions of expenses incurred on fertilizers and insecticides on treatment and time interaction
terms in a difference-in-difference setup. Panel B reports the same results for cotton crop. Values for all dependent variables
are in terms of rupees per acre. Values of all variables were winsorized at the 99th percentile. The standard errors were clustered
at level of Gram Panchayat (first level of randomization).
The DiD analysis (the coefficients for the interaction terms are presented in Table 11) of the expenses
incurred for plant protection chemicals shows that due to the intervention, at the endline, the treatment
27 | P a g e
farmers incurred 69 percent and 36 percent less expense for plant protection chemicals for cotton and
paddy crops, respectively.
Similar results are observed for fertilizers. The intervention led to a reduction in expenses incurred for
fertilizers by 63 percent for paddy and by 26 percent for cotton. Both results are statistically significant.
Rationalizing fertilizer usage was a big part of the intervention. Data collected at the baseline showed that
the farmers used fertilizers excessively. Figure 3 shows that, on average, the treatment farmers applied
100 kg of nitrogen per acre of cultivated area, while as per recommendation by local agricultural university
the crop required only 40 kg per acre.
Paddy
120
100
100
83
80 Treatment
58 53
60 49 Control
40 35
40 Recommended
20 20
20
0
Nitrogen Phosphorus Potassium
Cotton
80 73 73
70
60
50 44 Treatment
38
40 32 29 Control
30 21
20 16 16 Recommended
10
0
Nitrogen Phosphorus Potassium
Note: Graphs show average figures for kilograms of nutrients applied by treatment and control farmers per acre of cultivated
area. Data on recommended fertilizer quantity was obtained from a local agricultural university
Figure 3: Quantity of nutrients applied by farmers at baseline
28 | P a g e
Comparison between quantity of macro-nutrients used by treatment and control farmers at baseline is
presented in table 12. In almost all instances, nutrients applied by treatment farmers for paddy and cotton
crop were more than the nutrients applied by control farmers.
Table 12: Comparing values at baseline: Macro-nutrient use
Panel A: Nitrogen (Kilogram/Acre)

(1) (2)
(1) Vs (2)
Control Treatment
Paddy 83 100 -17***
(3) (3) (5)
Cotton 73 73 0
(4) (4) (5)
Panel B: Phosphorous (Kilogram/Acre)
(1) (2)
(1) Vs (2)
Control Treatment
Paddy 48 58 -9***
(2) (2) (3)
Cotton 38 44 -7**
(2) (2) (3)
Panel C: Potash (Kilogram/Acre)
(1) (2)
(1) Vs (2)
Control Treatment
Paddy 35 49 -13***
(2) (2) (3)
Cotton 21 29 -7***
(2) (2) (3)
The table reports comparison of macro-nutrient quantities used at baseline for the two crops by treatment and control
group. Columns (1) and (2) report sample means with standard errors in paraentheses. Column (3) reports difference
between the two experimental groups. Farmers who did not grew the respective crops at baseline have been excluded
from analysis.
The DiD analysis of the impact of the intervention on nutrient use (reported in Table 13) showed that the
intervention led to a significant reduction in the quantity of nitrogen and phosphorous used by the farmers
for both the crops. The treatment farmers reduced nitrogen use by 75 percent and 27 percent for paddy
and cotton crops, respectively. They also reduced phosphorous use by 73 percent and 40 percent for paddy
and cotton, respectively. The significant reduction in nutrient use was observed only at endline.
The reduction in the quantity of nutrient application and expenses related to fertilizers did not come at the
cost of the net returns from farming, as the regression results show that the treatment farmers gained
significantly in terms of additional net returns. One reason for this could be that though there was a
29 | P a g e
reduction in the quantity of nutrients applied, the nutrient absorption increased due to the application of
the right products in the right quantities and at the right time.
Table 13: Treatment effect on macro-nutrient application
Panel A: Paddy Crop

(1) (2) (3)
Nitrogen Phosphorous Potash
Treatment*Midline -3.218 -2.616 -0.631
(11.87) (4.939) (5.268)
[.87] [.67] [.95]
Treatment*Endline -71.66*** -37.31*** -16.53**

(8.158) (3.331) (6.427)
[0] [0] [.01]
Mean of dep. Var 93.11 51.25 44.45
Observations 1040 1040 1040
R-Square 0.189 0.109 0.0900
Adjusted R-square 0.176 0.0953 0.0758
(1) (2) (3)
Treatment*Midline 17.49** 10.06 6.875
(7.037) (11.76) (10.62)
[.07] [.42] [.64]
Treatment*Endline -29.74*** -22.80** 5.062

(7.342) (7.455) (6.253)
[0] [0] [.49]
Mean of dep. Var 109.2 56.59 45.15
R-Square 0.309 0.211 0.316
*
p < 0.05, ** p < 0.01, *** p < 0.001
Panel A reports results of regressions of quantity of macro-nutrients on treatment and time interaction terms in a difference-in-
difference setup. Panel B reports the same results for cotton crop. Values of all dependent variables are in terms of kilograms
per acre. Values of all variables were winsorized at the 99th percentile. The standard errors were clustered at level of Gram
Panchayat (first level of randomization).
The changes in nutrient management also explain the increase in labor outlay for fertilizer application and
reduced expenses for interculture operations and weeding. It was suggested to the farmers that fertilizers
should be applied in multiple doses during the crop lifecycle, which led to an increase in the cost of
applying fertilizers. The reduced use of fertilizers led to reduced growth of weeds, which in turn reduced
the labor required for weeding.
30 | P a g e
6.4 Cost of the intervention

Various components of cost of DATES intervention are given in table 14. In the DATES project, providing
customized information for two years cost Rs. 12,000 per farmer. This does not include the expenses
incurred for conducting research and for collecting data from the treatment and control farmers.
Table 14: Cost of running DATES project
Cost/unit Total
Particulars Qty./No. Total budget (Rs.)
(Rs.) Periods
I Salaries (Monthly Expenses)
a Extension Agents 3 15,000 24 1080000

b Field level supervisor 1 50,000 24 1200000
c Research Assistant (office level) 1 18,000 24 432000
II Travel and Other Allowances (Monthly Expenses)
a Extension Agents 3 5000 24 360000

b Field level supervisor 1 6000 24 144000
III Yearly Expenses
a Operational cost 1 20,000 2 40000

b Training and HRD 1 25,000 2 50000
c eSAP licenses 3 25,000 2 150000
IV Non-Recurring Expenses*
a Hand-held devices 3 9,000 1 27000

B Thermal Printer 3 9,500 1 28500
Total Expense for 300 Treatment Farmers 3511500
Expense per Treatment Farmer 11705
* the hand-held devices and thermal printer would have life expectancy of three years
The table reports total and per farmer cost of running the DATES intervention in the area of intervention. This does not include
cost of research and data collection. Costs and budget have been reported in rupees.
Returns to the farmers due to the intervention outweigh the the costs by a large amount and make
information delivery an effective initiative that the public sector can easily undertake. 18
18
Further research is required to explore whether the farmers would be willing to pay part of the cost of the intervention. The
farmers in the area of our study continued to call the graduates who served as extension agents even after the intervention was
completed to seek advice, which is a good indicator that they found the service extremely useful.
31 | P a g e
7. Conclusion and Policy Implications

While agricultural information delivery is not an actively used tool for the enhancement of agricultural
productivity in India, we show that it holds immense potential for enhancing agricultural productivity in
an ecologically and financially sustainable way. In this study, we presented results from an agricultural
information delivery project named DATES that used a mix of traditional extension methods and modern
information and communications technology to provide information about agricultural best practices to
300 farmers who were growing paddy and cotton.
The results achieved through the DATES intervention show that given the high gaps in yield of the various
crops grown in India, it is possible to achieve significant increases in yield via the proper delivery of
information related to agricultural practices. The intervention also shows that by providing information
about better farm management practices, it is possible to improve the technical efficiency of farming
enterprises. Additionally, the increase in productivity does not come with higher costs to farmers nor is it
at the cost of the environment. Information about the process of applying agricultural inputs can
simultaneously increase the effectiveness of the products and rationalize the quantity of inputs required,
thereby reducing the costs of production and reducing the various adverse environmental impacts that
these products may have. The findings of this study support the work of Pagani, Sawyer, & Mallarino
(2013) on the optimal nutrient rates for crops and show that a reduction in fertilizer quantity need not lead
to a drop in yields in areas where intensive agriculture is practiced. The results are also significant
considering that studies like Fishman et al., (2017) did not find any impact of providing agricultural
information on fertilizer use.
The DATES intervention showed that a substantial increase in productivity and profitability is possible
by providing information about best practices to farmers. However, our findings are limited to irrigated
areas, to crops for which large investments are made by farmers in terms of inputs, and to crops that are
highly susceptible to pest attacks. Similar interventions should be tried out for other agro-ecological zones,
other crops and on a larger scale to understand the robustness of the project outcome.
32 | P a g e
References
Aggarwal, P., Hebbar, K., Venugopa, l. M., Rani, S., Bala, A., Biswal, A., & Wani, S. (2008). Quantification of
yield gaps in rain-fed Rice, Wheat, Cotton and Mustard in India (Global theme on agrosystems Report
no. 43). International Crops Research Institute for the Semi-Arid Tropics.
Aker, J. C. (2011). Dial "A" for Agriculture: A review of information and communication technologies for
agricultural development in developing countries. Agriucltural Economics (Volume 42, Issue 6), 631-647.
Alene, A. D., & Manyong, V. (2006). Farmer‐to‐farmer technology diffusion and yield variation among adopters:
the case of improved cowpea in northern Nigeria. Agricultural Economics, Volume 35, Issue 2, 203-211.
Anderson, J. R., & Feder, G. (2004). Agricultural Extension: Good Intentions and Hard Realities. The World Bank
Reserach Observer (Volume 19, No. 1), 41-60.
Athey, S., & Imbens, G. (2006). Identification and inferences in nonlinear difference-in-difference models.
Econometrica, Vol. 74, No. 2, 431-497.
Babu, S. C., Glendenning, C. J., Asenso-Okyere, K., & Govindarajan, S. K. (2012). Farmers' information needs
and search behaviours: Case study in Tamil Nadu, India. Washington: International Food Policy
Research Institute.
Balasubramanian, R. (2014). Agricultural Reserach and Extension in India: Reflections on the Reality and a
Roadmap for Renaissance. Indian Journal of Agricultural Economics (Volume 69, No. 1), 14-43.
Birkhaeuser, D., Evenson, R., & Feder, G. (1991). The Economic Impact of Agricultural Extension: A Review.
Economic Development and Cultural Change (Volume 39, Number 3).
Cerdan-Infantes, P., Maffioli, A., & Ubfal, D. (2008). The impact of agricultural extension services: The case of
grape production in Argentina . Inter-American Development Bank. Office of evaluation and oversight.
Working paper: OVE/WP-05/08.
Chapman, R., & Slaymaker, T. (2002). ICTs and Rural Development: Review of the literature, current
interventions and opportunities for action. Overseas Development Institute (Working Paper 192).
Coelli, T. (1995). Recent developments in frontier modelling and efficiency measurement. Australian Journal of
Agricultural Economics. Vol. 39, No. 3, 219-245.
Coelli, T., Rahman, S., & Thirtle, C. (2005). Technical, allocative, cost and scale efficiencies in Bangladesh rice
cultivation: A non-parametric approach. Journal of Agricultural Economics (Volume 53, Issue 3), 607-
626.
Cole, S. A., & Fernando, A. N. (2016). ‘Mobile’izing Agricultural Advice: Technology Adoption, Diffusion and
Sustainability. Harvard Business School (Working Paper: 13-047).
Cooper, W. W., Seiford, L. M., & Tone, K. (2000). Data envelopment analysis: A comprehensive text with
models, applications, references and DEA-solver software. Kluwer Academic Publishers.
Croppenstedt, A., Demeke, M., & Meschi, M. M. (2003). Technology Adoption in the Presence of Constraints:
The Case of Fertilizer Demand in Ethiopia. Review of Development Economics, 7(1), 58-70.
33 | P a g e
Davis, K. E. (2008). Extension in Sub-Saharan Africa: Overview and Assessment of Past and Current models, and
Future Prospects. Journal of International Agricultural and Extension Education (Volume 15, Number 3),
15-28.
DoF. (2014). Indian Fertilizer Scenario 2014. Retrieved from http://fert.nic.in:

http://fert.nic.in/sites/default/files/Indian%20Fertilizer%20SCENARIO-2014_0.pdf
Evenson, R. (2000). Economic impacts of agricultural research and extension. In B. L. Gardner, & G. C. Rausser,
Handbook of Agricultural Economics (Volume 1, Part 1) (pp. 573-628). Elsevier.
Fafchamps, M., & Minten, B. (2012). Impact of SMS-based agricultural information on Indian farmers. The
World Bank Economic Review (Volume 26, Issue 3), 383-414.
FAO. (2018). Transforming Food and Agriculture to Achieve the SDGs: 20 interconnected actions to guide
decision-makers. Rome: Food and Agriculture Organization of the United Nations .
Feder, G., & Slade, R. (1993). Institutional Reform in India: The Case of Agricultural Extension. In K. Hoff, A.
Braverman, & J. Stiglitz, The Economics of Rural Organisation: Theory, Practice and Policy (pp. 530-
542). Oxford University press.
Ferroni, M., & Zhou, Y. (2012). Achievements and Challenges in Agricultural Extension in India. Global Journal
of Emerging Market Economies (Volume 4, Issue 3), 319-346.
FICCI. (2016, July). Next Generation Indian Agriculture - Role of Crop Protection Solutions: A Report on Indian
Agrochemical Industry. Retrieved from http://ficci.in: http://ficci.in/spdocument/20744/Agrochemicals-
Knowledge-report-2016.pdf
Fishman, R., Kishore, A., Rothler, Y., Ward, P. S., Jhan, S., & Singh, R. (2017). Can Information Help Reduce
Imbalanced Application of Fertilizers in India? Experimental Evidence from Bihar. New Delhi: South
Asia Office. Environment and Production Technology Division. International Food Polict Research
Institute.
Gandhi, R., Veeraraghavan, R., Toyama, K., & Ramprasad, V. (2009). Digital green: Participatory Video and
Mediated Instruction for Agricultural Extension. Information technology and International Development
(Volume 5, Number 1), 1–15.
GoI. (2016). State of Indian Agriculture (2015-16). New Delhi: Government of India. Ministry of Agriculture and
Farmers Welfare. Department of Agriculture, Cooperation & Farmers Welfare. Directorate of Economics
& Statistics.
Gupta, H., & Shinde, S. (2013). Agricultural Extension in India. International Journal of Management and Social
Sciences Research.
Heeks, R. (2002). i-Development not e-Development: Special Issue on ICTs and Development. Journal of
International Development, 14(1) , 1-12.
Horrigan, L., Lawrence, R. S., & Walker, P. (2002). How sustainable agriculture can address the environmental
and human health harms of industrial agriculture. Environmental Health Perspectives, 110(5), 445–456.
Ji, Y.-b., & Lee, C. (2010). Data envelopment analysis. The Stata Journal, 267-280.
34 | P a g e
Kusuma, D., Thabrany, H., Hidayat, B., Mcconne, M., Berman, P., & Cohen, J. (2017). New evidence on the
impact of large-scale conditional cash transfers on child vaccination rates: The case of a clustered-
randomized trial in Indonesia. World Development, Vol. 98, 497-505.
Mali, S. S., Sanyal, S. K., & Bhatt, B. P. (2015). Water Pollution and Agriculture. In H. Pathak, B. P. Bhatt, & S.
Gupta, State of Indian Agriculture - Water (pp. 39-47). National Academy of Agricultural Sciences.
Mishra, S. (2007). Risks, farmers' suicides and agrarian crisis in India: Is there a way out? Mumbai: Indira
Gandhi Institute of Development Research (WP-2007-014).
Mittal, S., & Mehar, M. (2014). Socio-Economic Impact of the Mobile Phone based Agricultural Extension. In S.
R, Mobile phones for agricultural extension (pp. 195-224). New Delhi: New India Publishing Agency.
MoAFD. (2016). All India Report on Input Survey 2011-12. New Delhi: Government of India (Ministry of
Agriculture & Farmers Welfare) .
NSSO. (2013). Some Aspects of Farming in India (NSS 70th Round). Retrieved from mospi.nic.in:
http://mospi.nic.in/sites/default/files/publication_reports/NSS_Report_573_16feb16.pdf
Owens, T., Hoddinott, J., & Kinsley, B. (2003). The impact of agricultural extension on farm production in
resettlement areas of Zimbabwe. Economic Development and Cultural Change.
Pagani, A., Sawyer, J. E., & Mallarino, A. P. (2013). Site-Specific Nutrient Management: For Nutrient
Management Planning To Improve Crop Production, Environmental Quality, and Economic Returns .
Extension and Outreach Publications.
Patil, S., Huggar, L., & Reddy, B. (2013). Cost and returns of plant proctection in paddy under different nitrogen
scenario in Tungabhadra project command. Karnataka Journal of Agricultural Sciences (Volume 26, No.
2), 229-232.
Pavithra, S., & Chand, R. (2015). Fertiliser Use and Imbalance in India. Economic & Political Weekly, Vol. 50,
Issue No. 44, 31 Oct, 2015, 98-104.
Planning Commission. (2010). Report of the Working Group on Fertilizer Industry for the Twelfth Plan (2012-13
to 2016-17). Government of India, Ministry of Chemicals & Fertilizers, Department of Fertilizers.
Prasad, R. (2009). Efficient fertilizer use: The key to food security and better environment. Journal of Tropical
Agriculture (Volume 47, No. 1), 1-17.
Puhani, P. A. (2008). The treatment effect, the cross difference, and the interaction term in nonlinear "difference-
in-differences" models. Economic Letters (Volume 115, Issue 1), 85-87.
Richardson, D. (2005). How can agricultural extension best harness ICTs to improve rural livelihoods in
developing countried. In E. Gelb, & A. Offer, ICT in Agriculture: Perspectives of Technological
Innovation. Food and Agricultural Organization of United Nations.
Rivera, W. M. (2001). Agricultural and rural extension worldwide: Options for institutional reforms in the
developing countries. Rome: Food and Agricultural Organisation of the United Nations.
Roman, R., & Colle, R. (2006). Content creation for ICT development projects: Integrating normative approaches
and community demand. Information Technology for Development (Volume 10), 85-94.
35 | P a g e
Romani, M. (2003). The impact of extension services in times of crisis: Cote d'Ivoire (1997-2000). CSAE
WPS/2003-07.
Shah, M. (2016). Urban Water Systems in India: A Way Forward (Working Paper 323). New Delhi : Indian
Council for Research on International Economic Relations.
Sheng, Y., Gray, E. M., & Mullen, J. D. (2010). Public investment in R&D and extension and productivity in
Australian Broadacre Agriculture. Sydney: 10th Economic Measurement Group Workshop.
Shreedhar, G., Gupta, N., Pullabhotla, H., Ganesh-Kumar, A., & Gulati, A. (2012). A review of input and output
policies for cereals production in India. IFPRI (Discussion Paper 01159).
Shukla, A. (2010). All India Coordinated Research Project of Micro and Secondary Nutrients and Pollutant
Elements in Soils and Plants. Retrieved from http://www.iiss.nic.in :
http://www.iiss.nic.in/downloads/pcm_materisl.pdf
Singh, M. (2012). Yield gap analysis in rice, wheat and pulses in India. Annals of Agricultural Research 33 (4).
Swanson, B. E. (2006). Extension Strategies for Poverty Alleviation: Lessons from China and India. Journal of
Agricultural Eductaion and Extension (Volume 12, Issue 4), 285-299.
Tilman, D., Cassman, K., Matson, P., Naylor, R., & Polasky, S. (2002). Agricultural sustainability and intensive
production practices. Nature (Volume 418), 671-677.
Umadikar, J., Sangeetha, U., Kalpana, M., Soundarapandian, M., & Prashant, S. (2014). mASK: A functioning
personalized ICT-based agricultural advisory system: Implementation, Impact and New Potential.
Chennai, India: 2014 IEEE Region 10 Humanitarian Technology Confernence.
Welch, F. (2001). Education in Production. Journal of Political Economy, Vol. 78, No. 1, 35-59.
Yedla, S., & Peddi, S. (2009). Agriculture and Environment. In S. Ray, Handbook of Agricutlure in India (pp. 57-
104). New Delhi: Oxford University Press.
36 | P a g e
Appendix 1
Map of treatment and control Gram Panchayats in DATES
Source: Map obtained from Karnataka Remote Sensing Applications Centre, Bangalore
Treatment GPs have been filled in dark blue and control gram panchayats have been filled in light blue.
37 | P a g e
Appendix 2
Power Calculation to Determine Number of Farmers to be Studied
Decision on number of farmers to be studied was made on basis of power size calculation. A RCT should have
sufficient statistical power to detect differences between treatment and control groups. For sample size calculation
a standard formula is used. The assumption behind this formula are: (i) one control & one treatment group of same
size; and (ii) standard deviation of the variable of interest is constant across the groups. Sample size in each group
is given by
2
 z  zP 
n  2s  C
2

  
where, s denotes pulled standard deviation of both comparison groups, z is standard normal variate, Z C and ZP are
the values for desired significance level and statistical power respectively, and Δ is the minimum expected
difference between means in two groups (or, effect size). We chose 80% power and 95% significance level for our
analysis. For sample size calculation we used the crop cutting experiment data maintained by Agriculture Insurance
Company of India (AIC) available at hobli level19.
Sample size calculation using pilot AIC data
No. of
Avg. yield (q/ac) Std. dev. (q/ac) Sample size (n)
obs.
Paddy (irrigated) 33 16.22 3.57 77
Sunflower (irrigated) 28 4.14 0.99 90
Sunflower (rainfed) 31 1.39 0.66 353
Source: AIC data and own calculations
Based on the above sample size computations, the team decided to have 300 farmers each in control and treatment
group in both project sites. Thus, for each GP, 50 farmers were surveyed.
19
Hobli refers to cluster of gram panchayats (generally 3-4 gram panchayats)
38 | P a g e
Appendix 3: e-SAP Mobile Application
39 | P a g e
Appendix 4: Using e-SAP to provide suggestion to farmers
40 | P a g e
Appendix 5: Information Module on Paddy

PADDY GROWTH STAGES
Transplanted Rice
Direct Seeded Rice
a) Vegetative (germination to panicle initiation) b) Reproductive (panicle initiation to flowering) c)

Ripening (flowering to mature gran)
41 | P a g e
CROP CALENDAR
CROP VARIETIES
BP 520 CSR 22 Gangavathi Sona Jaya
Varieties Situation Sowing Duration Characteristics

Rainfed Irrigated time (Days)
Jaya * * June 140-150 1. Long and bold seed 2. Resistant to
October 120-145 blast disease
BPT-5204 - * June 140-150 Long slender seed
CSR-22 - * June 130-135 Long slender seed
Gangavathi - * June 130-135 Medium slender seed
Siri-1253 - * June 135-140 Medium Small seed
IR-64 - * January 125-130 Long small seed
ES-18 - * January 120-125 Medium slender seed
Telhumsa - * January 120-125 Long small seed
Sujata - * January 130-135 Long small seed
42 | P a g e
SEED QUALITY TEST
Salt Water Treatment

Steps: a) Pour water into container b) Add salt or urea to increase specific density of water c) Keep on
adding salt or urea until egg can float on the surface d) Soak paddy seeds in salt water for fifteen
minutes and remove the chaffy seeds that float on water e) Collect the treated seeds of paddy from the
bottom of the bucket f) Wash it with clean water two times and dry it in shade
Rag Doll Test

a) Soak the cloth in clean water and spread them out on a flat surface b) From the seed sample, count out exactly
100 grains for each rag and distribute the grains evenly around the cloth (ten rows of ten grains facilitates
counting c) Carefully roll each rag around a separate stick, leaving the seeds undisturbed inside d) Fasten the rags
to the sticks with string and store the finished Rag Dolls in a warm moist place for five (5) days. e) Moisten the
cloth several times every day (this is very important; if the Rag Dolls are allowed to dry out, the seeds will die.) f)
After five (5) days, unroll the rags and count the number of seeds with roots. If each Rag Doll contains exactly
100 seeds, the number of sprouted seeds will equal the germination rate of the sample (e.g. if 85 seeds out of 100
43 | P a g e
sprouted, the germination rate equals 85%) g) Average out the germination rates indicated by the 3-5 separate
Rag Dolls to derive a more reliable overall germination rate.
SEED TREATMENT
Bio-agents seed treatment for paddy

Bio-agent Rate Purpose
Trichoderma harzianum/ 5 - 10 g/kg of Control Pythium seed rot and
T. viride/ T. virens seed damping-off and Bacterial shealth
blight
Pseudomonas fluorescens 5 - 10 g/kg of Control Pythium seed rot and
seed damping-off and Bacterial shealth
blight
Azospirillum 1g/kg of seed N fixation by rice seedlings
(mix with primed wet seed just before
sowing)
Chemical Seed Treatment

Method Chemical Rate Purpose
44 | P a g e
Wet seed Carbendazim 2 gram per litre of water Protection to

treatment or for 1 kg of seeds the seedlings
Tricyclozole (Soak the mixture for 10 up to 40 days
hrs and drain excess from blast
water) disease
Dry seed Captan or 4 gram per 1 kg of seed
treatment Thiram (Mix the chemical with
the seed 24 hours before
sowing)
Seedling Treatment
Steps: a) Prepare slurry by mixing Azospirillum @ 1 kg in 40 litres of water and dip the root portion of rice
seedlings in this bacterial suspension for 15-30 minutes and then transplant the seedlings in the field b) Dip
seedling root in chlorpyrifos @2 ml/litre of water (Control Yellow stem borer)
45 | P a g e
Seed treatment to break seed dormancy

Method: Soak the seeds in hot water(4 5°C) for 72 hours Or Soak seeds in KNo3 1.5 per cent or 50 mg
Gibberellic acid in one litre of water for 10 hours
Purpose: Breaks the seed dormancy in IR-64 and Gangavathi Sona varieties
Next steps: After the treatment dry the seeds in shade and continue with other seed treatment
methods( Care should be taken while using KNo3 )
NURSERY
Wet bed nursery
46 | P a g e
Steps: a) Plough and harrow the soil twice to obtain a fine till b) Puddling and levelling of nursery area
c) Construct drainage canals for proper water removal d) Nursery area: 300 m 2 (e) Seed bed size: 7 –
7.5 m x 1.2 – 1.5 m x 10 cm (f) Number of seed beds/ha: 75 (g) FYM or Compost: 250 kg (h) Urea: 2.17
kg (i) DAP: 0.868 kg (j) MOP: 0.835 kg (k) Top dressing 0.65 – 1.30 kg Urea six days before transplanting
(l) Seed rate: 62 kg/ha. Broadcast the pre-germinated seed. The application rate: 50-70 gm/m 2 (m)
Irrigation: when seedlins are 1inch height allow to stand a thin layer of water (n) Transplanting: 20-25
days onwards
LAND PREPARATION
Wet land preparation

Steps: a) Plough immediately after the previous harvest b) Plough the field using disc or mouldboard plough
preferably 6-8 weeks before planting with maximum depth of 10 cm and secondary ploughing 2-3 weeks before
planting c) Later, puddle the field with 5-10 cm of standing water 2-3 times d) Incorporate 5-7 tonne of FYM
or compost or 1 tonne of poultry manure per hectare or incorporate 10 tonne of green leaf manure three weeks
before transplanting. e) The fertilizers recommended for basal application have to be applied before the last
puddling and incorporated f) After fertilization avoid moving of water from one field to another g) Later,
proper levelling has to be made before transplanting the seedlings h) Apply mud paste to the side and top of the
bund to a thickness of 2.5 cm with a spade and plaster it using the flat surface of the spade.
47 | P a g e
Dry land Preparation (for Direct Sowing Rice)

Steps: a) Whenever possible give one or two summer ploughings to minimize weed growth b) Dry plough to
get fine tilth taking advantage of rains and soil moisture availability c) Perfect land leveling for efficient weed
and water management d) Ensure deep tillage and fine tilth at the time of final land preparation for sowing e)
Provide shallow trenches (15 cm width) at an interval of 3m all along the field to facilitate draining excess
water at the early growth stage.
PADDY PLANTING
Drill Sowing
Details: a) In irrigation command areas, before 30-35 days of receiving water from the canal after one
rain or even before drill sowing can be done. In case where irrigation facility is availing watering can
be done soon after sowing b) Tractor driven drill sowing can be done c) At the time of sowing after
proper land preparation basal dosage of fertilizer can be applied along with sowing d) Row to row
spacing to be maintained is 20 cm or 25 cm f) Time of Planting: May 3rd week to July 2nd week.
48 | P a g e
Machine Transplanting
There are two types of machines for transplanting paddy.
A) Paddy Transplanter – Walk behind and operate: i) Can transplant 4 rows of 30 cm apart at a time
ii) Has petrol engine which consumers 0.9 to 1 litre of petrol per hour iii) Can transplant 2.5 to 3.1 acre
for 8 hours.
B) Paddy Transplanter – Sit and operate: i) Can transplant 7.5 to 8.8 acres in 8 hours ii) 8 rows with 23
cm apart can be transplanted at a time iii) Has diesel engine which consumers 0.75 litre/hour.
Benefit of machine transplanting: 1. Cage wheel has been adopted for operating in puddled land 2.
Can be used for weeding in plots where transplanting is done by machine 3. Can obtain more that
18% higher yield and reduce cost by 30%.
Manual Transplanting
49 | P a g e
MANURE AND FERTILIZERS
Stages of Paddy at which fertilizer needs to be applied
Incorporation of green manure into land

Steps for green manure (transplanted paddy): Green manure crops : Sesbania aculeata, Crotalaria juncea
Seed rate : 25-30 kg/ha Planting method : Broadcast or line sowing with 45 cm row spacing Irrigation : One
pre-sowing irrigation followed by 1-2 irrigation in between Time of sowing : Last week of April to first
fortnight of May Time of incorporation: Crop to be turned down around 55-60 days after sowing.
TheTrichoderma (10 g/l) should be sprayed on turned Sesbania/Crotalaria crop in the field before one week of
puddling and transplanting rice crop.
Advantage: The 60-day-old crop can contribute approximately 100 kg N/ha, 25-30 kg P/ha and 75 kg K/ha.
Steps for green manure (direct seeded rice): a) Sow sunhemp green manure seeds @10kg/ha mixed with paddy
seeds b) Carry out hodta operation (Planking) in standing water after 40 DAS for in situ incorporation of
50 | P a g e
sunhemp in the soil (OR) Ex situ incorporation of green leaf manuring of Eupatorium/parthenium/cassia and
other weeds green material @ 5 t/ha in between the two paddy rows by carrying out hodta operation.
Advantage: Provides only 50% nutrients to maintain good yield
Chemical Fertilizers
Recommended nutrient quantity: 150:75:75 kg NPK / ha
Time of application
At the time of sowing or Basal dose 75:75:37.5 NPK/ha
transplanting
After 25-30 days 1st Top dressing 37.5 :0:0 kg NPK/ha
After 50-55 days or panicle 2nd Top dressing 37.5 :0:37.5 kg NPK/ha
initiation
Combination of fertilizers
Type of fertilizer Total qty Basal dose Top Dressing Remarks
(kg / ha) (kg/ha) I II
UreaRock 217 109 54 54
Phosphate 175 175 - -
MOP 80 40 - -
Urea 174 66 54 54
DAP 109 109 - -
MOP 80 40 - 40
Urea 109 - 55 54 Addition of MOP at the rate of
NPK 19-19-19 263 263 - - 40 kg/ha as 2nd top dressing is
Urea 109 - 55 54 advisable
NPK 15-15-15 333 333 - -
Application suggestion: 24 hours before the top dressing drain out the field and 24 hours after top
dressing irrigate the field
51 | P a g e
IRRIGATION
Irrigation in transplanted paddy:

Short Duration Variety Medium Duration Variety Long Duration Variety
Days No. of Water Days No. of Water Days No. of Water
irrigation level (cm) irrigation level (cm) irrigation level (cm)
125 5-7 2-3 1-30 5-7 2-3 1-35 6-8 2-3
25 - Thin film 30 - Thin film 35 - Thin film
of water of water of water
28 - Lift 33 - Lift 38 - Lift
irrigation irrigation irrigation
29-50 6 2-5 34-65 6-8 2-5 39-90 12-15 2-5
51-70 5-6 2-5 66-95 8-10 2-5 96-125 7-9 2-5
71-105 5-6 2-5 96-125 6-8 2-5 126- 5-6 2-5
150
Irrigation in Dry Seeded Rice
a) For drill sown rice care should be taken to drain out excess rain water during first 10-15 DAS and
the water level in the field should not be more than 2.5cm height during tillering stage. b)In drill sown
rice carry out hodta operation (Planking) in standing water at 40 DAS, and impound sufficient rain
water through the crop growth period.
52 | P a g e
WEED MANAGEMENT
Take up hand weeding at 20 and 40 days after transplanting and at 20 and 40 days after sowing in
transplanted and drill sown paddy respectively
Herbicide Quantity (ha) Time of application Remarks

2,4-D Sodium salt 80% 2.5 kg 3-4 weeks after transplanting Care should be taken to
avoid herbicidal drift to
nearby fields
Propanil 35 EC 7.5 lt Weeds are at 1 or 2 leaf stage Do not mix any chemical
with these herbicides
Butachlor 5% G 30 kg Broadcast the granules at 5 -7 -
days after transplanting
2,4 – Dethylester 5% G 15 kg Spray 5-7 days after -
transplanting
Anilogard 1.5 lt --do-- -
Thiobencarb 40 lt Within 5 days after -
(Benthiocarb) 50EC transplanting
Pendimethalin 30 EC 3.25 lt Within 3-5 days after -
transplanting
Oxadiazon 25 EC 1 lt Within 3-5 days after -
transplanting
53 | P a g e
Appendix 4: Interaction of Extension Agents with Farmers
Collection of soil sample
Soil samples for testing bagged and labeled
54 | P a g e
Inspection of paddy after sowing
Helping farmer install yellow stick trap (used for attracting insects)
55 | P a g e
Interaction of farmers with scientists from local agricultural university
Interaction of farmers with scientists from local agricultural university
56 | P a g e
Appendix 6: Proforma of Soil Test Result

SOIL TEST BASED FERTILIZER RECOMMENDATION-SIRUGUPPA TALLUK
Name of the Farmer: raghavendra s/o chanal siddlingappa HHID: 20601 Village: sirigeri Gram
Panchayat: sirigeri
SOIL TEST RESULTS:
pH: 6.5 NITROGEN: MEDIUMPHOSPHOROUS: MEDIUM POTASH: MEDIUM
FERTILIZER RECOMMENDATIONS
RICE: (BLANK RECOMMENDATION : 100:50:50; SOIL TEST BESED RECOMMENDATION: 100: 50: 50
kg/Ha)
OPTION1: UREA: 2200000000003 Kg +SUPER PHOSPHATE: 312.5 Kg + MURATE OF POTASH: 50 Kg
APPLY ALL SUPERPHOSPHATE AS BASAL, AND UREA AND POTASH MAY BE APPLIED 25%
BASAL, 25% AFTER 1ST WEEDING, 25 AFTER 2ND WEEDING AND 25% AFTER FLOWERING.
OPTION2: (17:17:17 COMPLEX BESED RECOMMENDATION)
17-17-17 COMPLEX: 295 Kg +UREA: 1100000000001 Kg + SUPER PHOSPHATE: 0 Kg + POTASH: 0 Kg
APPLY ALL THE 17-17-17 COMPLEX AND SUPER PHOSPHATE AS BASAL. UREA AND POTASH MAY
BE APPLIED 25% BASAL, 25% AFTER 1ST WEEDING, 25 AFTER 2ND WEEDING AND 25% AFTER
FLOWERING.
OPTION3: (15:15:15 COMPLEX BASED RECOMMENDATION)
15:15:15 COMPLEX: 335 Kg +UREA: 1100000000001 Kg + SUPER PHOSPHATE: 0 Kg + POTASH:0 Kg
APPLY ALL THE 15:15:15 COMPLEX AND SUPER PHOSPHATE AS BASAL. UREA AND POTASH MAY
BE APPLIED 25% BASAL, 25% AFTER 1ST WEEDING, 25 AFTER 2ND WEEDING AND 25% AFTER
FLOWERING.
COTTON: (BLANK RECOMMENDATION: 80:40:40; SOIL TEST BASED =80 : 40 : 40 Kg/Ha)
OPTION1: UREA: 176 + SUPER PHOSPHATE: 250 + 4068. APPLY ALL SUPERPHOSPHATE AS BASAL,
AND UREA AND POTASH MAY BE APPLIED 25% BASAL, 25% AFTER 1ST WEEDING, 25 AFTER 2ND
WEEDING AND 25% AFTER ONE MONTH.
OPTION2: 17-17-17 COMPLEX=236 Kg + UREA=88 Kg + SUPER PHOSPHATE=0 Kg + 0 Kg
OPTION3: 15-15-15 COMPLEX=268 Kg + UREA=88 Kg + SUPER PHOSPHATE=0 Kg + 0 Kg
APPLY ALL THE 17:17:17 OR 15:15:15 COMPLEX AND SUPER PHOSPHATE AS BASAL. UREA AND
POTASH MAY BE APPLIED 25% BASAL, 25% AFTER 1ST WEEDING, 25 AFTER 2ND WEEDING AND
25% AFTER FLOWERING.
Appendix 7: Roster for Farmer Visit
Agent # 1: RAM
Week: 1
Sl.
Day / GP No. Village Farmer Father Mobile
MON 1 Ibrahimpura
Bagewadi 2
3
4
5
6
7
8
TUE 1 Ibrahimpura
Bagewadi 2
3
4
57 | P a g e
5
6
7
8
9
WED 1 Ibrahimpura
Bagewadi 2
3
4 Bagewadi
5
6
7
8
THU 1 Bagewadi
Bagewadi 2
3
4
5
6
7
8
FRI 1 Bagewadi
Bagewadi 2
3
4
5
6
7
8
Appendix 8: Comparison of Observables at Baseline (Paddy)

0 1
Control Treatment (1) vs. (2)
Avg age of all family members (in years) 28.67 28.96 -0.29
(0.57) (0.53) (0.77)
Avg years of schooling of all family members (in yers) 4.25 5.20 -0.9***
(0.22) (0.25) (0.34)
Experience of farmer in crop cultivation (in years) 21.24 21.37 -0.13
(0.82) (0.81) (1.15)
Land owned by farmer (in acre) 9.09 10.39 -1.31
(0.84) (0.80) (1.16)
Age of farmer (in years) 42.43 41.81 0.62
(0.97) (0.83) (1.27)
Years of schooling of farmer (in years) 4.98 6.25 -1.27**
(0.37) (0.35) (0.50)
Visits to agri extension centre in previous year 0.10 0.21 -0.1***
58 | P a g e
(0.02) (0.03) (0.04)

Visits by agri extension agent in previous year 0.98 0.90 0.08***
(0.01) (0.02) (0.02)
Whether farmer faced income shortage in previous year 0.68 0.71 -0.03
(0.04) (0.03) (0.05)
Whether or not farmer is female 0.01 0.01 0.00
(0.01) (0.01) (0.01)
Panel B: Land Ownership Categories
0 1
Marginal Farmers 0.21 0.15 0.07*
(0.03) (0.03) (0.04)
Small Farmers 0.21 0.21 0.01
(0.03) (0.03) (0.04)
Semi-Medium Farmers 0.27 0.28 -0.01
(0.03) (0.03) (0.05)
Medium Farmers 0.22 0.27 -0.05
(0.03) (0.03) (0.04)
Large Farmers 0.08 0.10 -0.02
(0.02) (0.02) (0.03)
Panel C: Asset Index Categories
0 1
Asset Index 1 0.20 0.18 0.01
(0.03) (0.03) (0.04)
Asset Index 2 0.18 0.15 0.03
(0.03) (0.03) (0.04)
Asset Index 3 0.19 0.23 -0.04
(0.03) (0.03) (0.04)
Asset Index 4 0.19 0.23 -0.04
(0.03) (0.03) (0.04)
Asset Index 5 0.24 0.21 0.03
(0.03) (0.03) (0.04)
Panel D: Caste Categories
0 1
General Caste 0.29 0.49 -0.2***
(0.03) (0.04) (0.05)
Scheduled Caste 0.08 0.08 0.00
(0.02) (0.02) (0.03)
Scheduled Tribe 0.23 0.09 0.14***
(0.03) (0.02) (0.04)
Other Backward Caste 0.40 0.34 0.06
(0.04) (0.03) (0.05)
N 178 198 376
"Columns (1) and (2) report sample means with standard errors in parentheses. Column (3) reports the mean
difference between the two experimental groups."
Appendix 9: Comparison of Observables at Baseline (Cotton)
59 | P a g e

0 1
Avg age of all family members (in years) 30.10 27.97 2.13**
(0.72) (0.67) (1.00)
Avg years of schooling of all family members (in yers) 4.22 4.61 -0.40
(0.25) (0.30) (0.39)
Experience of farmer in crop cultivation (in years) 23.56 21.58 1.98
(1.02) (1.13) (1.52)
Land owned by farmer (in acre) 10.17 11.30 -1.12
(0.98) (1.07) (1.46)
Age of farmer (in years) 46.00 42.08 3.92**
(1.19) (1.15) (1.68)
Years of schooling of farmer (in years) 4.62 5.10 -0.48
(0.36) (0.42) (0.56)
Visits to agri extension centre in previous year 0.09 0.27 -0.18***
(0.02) (0.04) (0.05)
Visits by agri extension agent in previous year 0.99 0.97 0.01
(0.01) (0.01) (0.02)
Whether farmer faced income shortage in previous year 0.79 0.79 0.00
(0.03) (0.04) (0.05)
Whether or not farmer is female 0.00 0.00 0.00
(0.00) (0.00) (0.00)
Panel B: Land Ownership Categories
0 1
Marginal 0.12 0.13 -0.01
(0.03) (0.03) (0.04)
Small 0.23 0.18 0.05
(0.04) (0.04) (0.05)
Semi-Medium 0.32 0.26 0.06
(0.04) (0.04) (0.06)
Medium 0.23 0.32 -0.09
(0.04) (0.04) (0.05)
Large 0.10 0.11 -0.01
(0.02) (0.03) (0.04)
Panel C: Asset Index Categories
0 1
Asset Index 1 0.25 0.16 0.09*
(0.04) (0.03) (0.05)
Asset Index 2 0.21 0.21 0.00
(0.03) (0.04) (0.05)
Asset Index 3 0.15 0.19 -0.04
(0.03) (0.04) (0.05)
Asset Index 4 0.19 0.21 -0.02
(0.03) (0.04) (0.05)
Asset Index 5 0.20 0.23 -0.03
(0.03) (0.04) (0.05)
60 | P a g e
Panel D: Caste Categories

0 1
General Caste 0.33 0.37 -0.04
(0.04) (0.04) (0.06)
Scheduled Caste 0.08 0.08 -0.01
(0.02) (0.03) (0.03)
Scheduled Tribe 0.23 0.11 0.12***
(0.04) (0.03) (0.05)
Other Backward Caste 0.36 0.44 -0.07
(0.04) (0.05) (0.06)
N 146 119 265
"Columns (1) and (2) report sample means with standard errors in parentheses. Column (3) reports the mean
difference between the two experimental groups."
Appendix 10: Treatment effect on crop yields, revenue, cost and returns (Paddy Crop)
(1) (2) (3) (4)

Treatment*Midline 0.950 1273.7 864.5 249.5
(0.857) (1997.8) (138.8) (2338.6)
[.31] [.57] [.56] [.92]
Treatment*Endline 5.036*** 9263.3*** -5715.8*** 14901.8***

(0.908) (1828.2) (870.3) (2103.0)
[.01] [0] [0] [0]
Midline 1.267* 3559.1* 1153.4 2422.9

(0.664) (1853.8) (648.4) (1568.5)
Endline 2.635*** 4663.2** 4540.2*** 152.3

(0.767) (1580.2) (517.1) (1486.3)
Treatment -0.786 639.2 1674.6 -986.3

(1.405) (2041.5) (1132.8) (2397.6)
Years of schooling (Family) -0.0156 -21.62 6.142 -14.79

(0.0962) (168.5) (82.88) (206.2)
Years of schooling (Farmer) 0.177** 303.6** -7.286 305.7**

(0.0670) (106.0) (52.14) (137.2)
Visit to extension centre 0.102 -160.2 490.9 -632.2

(0.508) (1089.8) (480.7) (1157.3)
Visit by extension agent 0.113 -2278.4 1150.9 -3512.8

(0.881) (1629.4) (1042.6) (2036.7)
Marginal farmer 0 0 0 0
(.) (.) (.) (.)
61 | P a g e
Small farmer 0.724 502.8 -827.5 1282.9

(0.832) (1269.0) (535.8) (1408.1)
Semi-medium farmer -0.923 -1756.2 50.97 -1786.4

(0.848) (1305.3) (368.3) (1346.2)
Medium farmer -0.224 79.31 -520.9 608.9

(0.649) (1021.0) (380.4) (1275.1)
Large farmer -0.0196 81.57 -1972.6*** 1963.7

(0.723) (1228.2) (442.7) (1474.3)
General Caste -0.396 181.9 -567.2 655.6

(0.880) (1503.7) (844.5) (1911.4)
Scheduled Caste 0 0 0 0
(.) (.) (.) (.)
Scheduled Tribe 0.370 476.5 -402.3 801.7

(0.539) (801.1) (766.7) (1126.7)
Other Backward Caste 0.448 1069.5 -718.2 1786.5

(0.886) (1533.9) (845.3) (1902.2)
Constant 23.60*** 36279.2*** 19193.9*** 17194.7***

(2.089) (3860.1) (1434.7) (3596.7)
Mean of dep. Var 26.22 40367.5 21251.9 19122.4
Observations 1040 1040 1040 1040
R-Square 0.178 0.210 0.134 0.199
Standard errors in parentheses
All values were winsorized at the 99th percentile
*
p < 0.10, ** p < 0.05, *** p < 0.01
Appendix 11: Treatment effect on major components of cost (Paddy Crop)
(1) (2) (3) (4) (5) (6)

Plowing Seeds Transplanting Irrigation Weeding Harvesting
(labour)
Treatmen*Midline 48.40 725.9** -115.7 -1041.0*** 99.65 429.4
(123.1) (308.0) (90.32) (193.8) (266.6) (285.0)
[.57] [.01] [.33] [.01] [.73] [.2]
Treatment*Endline -349*** 6.075 -280.5*** -114.5 -1194*** 65.31

(104.3) (338.4) (77.04) (217.9) (139.1) (283.5)
[.02] [.97] [0] [.66] [0] [.8]
Midline -462*** -527.6** 335.7*** 192.0* 59.37 282.8

(109.3) (210.6) (72.93) (101.5) (175.1) (215.6)
62 | P a g e
Endline -351*** -395.3** 828.2*** -1223.9*** 465.0*** 528.6*

(60.44) (178.3) (56.02) (163.9) (109.0) (255.4)
Treatment 29.06 -207.6 274.6** 135.2 44.38 17.12

(75.03) (454.5) (94.96) (246.9) (145.4) (216.9)
Years of schooling 0.620 -19.71 -2.550 -26.28 11.58 -6.458

(Family)
(6.488) (21.78) (2.682) (22.78) (18.62) (13.96)
Years of schooling -4.006 -6.757 2.971 -4.630 -7.298 16.21

(Farmer)
(3.572) (20.53) (2.052) (16.60) (10.69) (10.03)
Visit to extension centre -7.465 -118.5 68.05** 129.7 -169.7 25.09

(59.84) (79.55) (25.44) (77.32) (122.0) (112.8)
Visit by extension agent 126.9* -320.7* 50.88 216.5 156.7 40.02

(65.97) (175.7) (35.55) (189.8) (111.6) (141.4)
Marginal farmer 0 0 0 0 0 0
(.) (.) (.) (.) (.) (.)
Small farmer -27.06 -432.6** -6.066 -265.2** 118.2 -88.05

(43.02) (173.3) (49.79) (112.4) (144.7) (94.29)
Semi-medium farmer - -118.8 59.88* -83.01 182.2 -38.73

128.0***
(15.26) (99.76) (31.47) (171.6) (119.2) (88.53)
Medium farmer -215.*** -241.3* 59.33** -232.7 371.2*** -188.8**

(56.42) (115.5) (25.86) (132.3) (94.41) (72.26)
Large farmer -477*** -624.3*** 52.80 -60.58 376.1** -348.8***

(62.87) (119.5) (34.12) (228.7) (133.5) (95.23)
General Caste -84.36 224.1 26.30 -5.619 -139.5 -60.13

(96.7) (230.1) (35.11) (153.4) (106.5) (82.17)
Scheduled Caste 0 0 0 0 0 0
(.) (.) (.) (.) (.) (.)
Scheduled Tribe -98.56 -7.868 52.66 42.36 56.21 -100.6

(71.99) (196.2) (38.57) (165.8) (90.82) (97.26)
Other Backward Caste -82.43 47.75 -24.37 -124.1 -91.59 -96.61

(93.30) (179.5) (43.62) (167.2) (112.2) (98.78)
Constant 1533*** 2500.2*** 1522.8*** 2298.3*** 1584.3*** 2335.0***

(120.2) (349.8) (89.59) (332.0) (174.6) (292.7)
Mean of dep. var 1127.7 1648.7 2064.6 1767.8 1863.2 2605.1
63 | P a g e
Observations 1040 1040 1040 1040 1040 1040

R-Square 0.262 0.0788 0.485 0.171 0.112 0.129
Adjusted R-square 0.250 0.0644 0.477 0.158 0.0978 0.115
*
p < 0.10, ** p < 0.05, *** p < 0.01
Appendix 12: Treatment effect on expense on fertilizers and insecticides (Paddy Crop)
(1) (2)
(562.2) (181.1)
[.44] [.69]
Treatment*Endline -3464.4*** -531.9*

(390.2) (244.3)
[0] [.04]
Midline -280.9 39.58

(424.8) (69.83)
Endline 2313.7*** 1412.3***

(249.0) (219.2)
Treatment 971.0* 236.4*

(445.8) (116.7)
Years of schooling (Family) 19.47 -5.862

(28.32) (14.16)
Years of schooling (Farmer) 10.11 8.234

(28.35) (11.10)
Visit to extension centre -47.30 43.19

(157.4) (95.50)
Visit by extension agent 899.3** 22.64

(328.8) (128.8)
Small farmer 33.32 -94.00

(150.4) (92.61)
Semi-medium farmer -64.95 -151.8

(124.3) (100.2)
64 | P a g e
Medium farmer -0.953 -188.9

(229.2) (112.3)
Large farmer -72.15 -401.6**

(223.1) (138.3)
General Caste -20.04 -144.2

(467.3) (90.28)
Scheduled Caste 0 0
(.) (.)
Scheduled Tribe -160.1 -32.45

(336.9) (139.0)
Other Backward Caste -31.86 -154.7*

(447.5) (83.61)
Constant 3877.4*** 1220.4***

(553.6) (202.1)
R-Square 0.157 0.295
*
p < 0.10, ** p < 0.05, *** p < 0.01
Appendix 13: Treatment effect on macro-nutrient application (Paddy Crop)

(1) (2) (3)
Treatment*Midline -3.218 -2.616 -0.631
(11.87) (4.939) (5.268)
[.87] [.67] [.95]
Treatment*Endline -71.66*** -37.31*** -16.53**

(8.158) (3.331) (6.427)
[0] [0] [.01]
Midline -4.290 -2.522 -0.200

(10.15) (3.585) (2.489)
Endline 46.81*** 17.68*** 17.60***

(4.801) (1.663) (5.596)
65 | P a g e
Treatment 17.30** 8.323 13.74**

(6.185) (5.184) (5.761)
Years of schooling (Family) 0.516 0.0299 0.596

(0.431) (0.365) (0.359)
Years of schooling (Farmer) -0.132 0.0672 -0.100

(0.364) (0.319) (0.253)
Visit to extension centre -2.159 -0.890 1.464

(2.019) (1.332) (2.376)
Visit by extension agent 6.582* 2.082 9.850**

(3.095) (2.804) (4.280)
Marginal farmer 0 0 0
(.) (.) (.)
Small farmer -1.540 1.651 0.820

(2.155) (2.034) (2.156)
Semi-medium farmer -0.362 -1.273 0.446

(2.684) (1.495) (1.737)
Medium farmer -2.932 1.329 -0.238

(3.819) (2.473) (2.669)
Large farmer -1.913 3.281 -0.926

(5.243) (2.510) (3.105)
General Caste 0.834 3.037 -2.993

(5.905) (4.899) (4.765)
Scheduled Caste 0 0 0
(.) (.) (.)
Scheduled Tribe -4.296 -0.395 -1.811

(4.250) (3.195) (4.559)
Other Backward Caste 2.712 0.984 -1.057

(4.023) (4.433) (4.991)
Constant 76.20*** 44.14*** 24.81***

(7.845) (5.959) (7.174)
Mean of dep. var 93.11 51.25 44.45
R-Square 0.189 0.109 0.0900
*
p < 0.10, ** p < 0.05, *** p < 0.01
66 | P a g e
Appendix 14: Treatment effect on crop yields, revenue, cost and returns (Cotton Crop)
(1) (2) (3) (4)
Treatment*Midline 3.503** 14400.2** 2700.4 11550.1
(1.204) (5385.6) (2648.9) (6733.6)
[.01] [.01] [.44] [.14]
Treatment*Endline 7.174*** 29289.4*** -2451.4 31761.4***

(1.450) (6254.6) (1519.9) (5569.4)
[.01] [0] [.08] [.03]
Midline 1.958** 13332.5*** 2588.2* 10871.6***

(0.865) (3322.4) (1170.3) (3003.0)
Endline 1.275 6269.5 2057.6*** 4226.3

(1.239) (5292.2) (424.5) (5036.7)
Treatment -1.678 -6260.8 1066.8 -7301.1

(1.140) (4781.1) (1579.6) (4145.5)
Age of family members(avg) 0.0324** 155.6** 15.11 143.5**

(0.0133) (66.51) (17.31) (56.17)
Age of farmer -0.00741 -42.19 15.99 -61.12

(0.00905) (40.31) (14.77) (41.64)
Visit to extension centre -0.351 -941.1 -393.1 -487.3

(0.330) (1385.2) (901.8) (1476.9)
Asset Index 1 0 0 0 0
(.) (.) (.) (.)
Asset Index 2 0.575* 2245.9 1733.8*** 522.3

(0.313) (1546.5) (448.5) (1330.9)
Asset Index 3 1.076* 4682.2* 1131.0 3531.0**

(0.522) (2204.2) (1165.9) (1452.3)
Asset Index4 0.267 1219.9 1572.9 -256.3

(0.441) (2124.8) (1095.0) (1232.2)
Asset Index 5 1.279** 5620.3** 163.4 5539.7**

(0.490) (2191.7) (1099.1) (1854.6)
General Caste 0.287 1826.2 -1421.7 3128.6

(0.549) (2493.4) (950.8) (1782.5)
Scheduled Caste 0 0 0 0
(.) (.) (.) (.)
Scheduled Tribe -0.721 -2315.4 -1042.9 -1333.3
67 | P a g e
(0.433) (2017.1) (835.9) (1937.2)
Other Backward Caste 0.212 1249.7 -1271.5 2501.3

(0.655) (2895.4) (994.0) (2456.2)
Constant 6.042*** 22913.8*** 21715.3*** 1234.1

(1.459) (6250.1) (1402.4) (5986.7)
Mean of dep. var 9.161 38913.6 24568.4 14365.2
Observations 756 756 756 756
R-Square 0.400 0.399 0.103 0.391
*
p < 0.10, ** p < 0.05, *** p < 0.01
Appendix 15: Treatment effect on major components of cost (Cotton Crop)
(1) (2) (3) (4) (5) (6)

Plowing Seeds Sowing Intercultivation Weeding Harvesting
Treatment*Midline 298.3** 33.55 107.7 63.27 553.0 -842.9
(99.08) (211.3) (222.8) (279.9) (927.0) (502.4)
[.09] [.88] [.62] [.86] [.56] [.17]
Treatment*Endline 208.6 -246.3 95.80 -681.2** -54.94 201.7

(125.1) (142.7) (141.3) (227.0) (431.4) (767.6)
[.26] [.19] [.58] [.06] [.88] [.85]
Midline -618.8*** -169.7 -50.52 -4.381 -883.0 1563.3***

(56.72) (170.4) (143.0) (109.2) (708.9) (370.5)
Endline -486.7*** -67.03 42.69 -105.4 -673.6** -668.8*

(108.8) (89.94) (85.14) (129.0) (270.4) (333.5)
Treatment -25.00 -149.4 -120.0 176.7 -76.58 150.4

(85.11) (126.6) (96.75) (192.2) (405.1) (629.9)
Age of family members(avg) 1.895 7.894** 0.0582 1.335 -1.686 -13.34

(1.735) (2.506) (1.905) (5.607) (6.109) (9.873)
Age of farmer 0.567 -1.461 1.317 0.483 3.849 9.690

(2.053) (1.595) (0.955) (1.936) (4.064) (7.261)
Visit to extension centre 61.74 40.27 57.18 -141.8* -99.97 -500.6*

(60.00) (95.18) (55.04) (70.56) (118.8) (248.1)
Asset Index 1 0 0 0 0 0 0
(.) (.) (.) (.) (.) (.)
Asset Index 2 -30.43 91.25 7.063 84.77 46.36 547.0

(71.76) (100.6) (43.38) (148.7) (198.2) (312.4)
68 | P a g e
Asset Index 3 51.40 172.6* -17.53 221.7 -184.5 257.6

(81.68) (81.84) (51.77) (151.6) (236.6) (289.6)
Asset Index4 -50.07 245.0* 8.599 124.0 24.66 333.5

(78.02) (112.5) (51.54) (144.8) (188.8) (331.6)
Asset Index 5 -205.0** 68.58 -72.83 -137.8 28.18 271.1

(79.36) (84.28) (51.71) (109.8) (206.0) (427.9)
General Caste -43.44 -68.41 -68.27 -254.6 -118.8 260.3

(76.79) (74.94) (69.95) (192.5) (241.6) (426.1)
Scheduled Tribe 8.108 145.2 -28.67 183.9 -116.1 -212.6

(57.64) (113.3) (35.18) (154.5) (261.9) (456.7)
Other Backward Caste -2.832 93.97 -17.27 -31.23 18.66 -217.7

(67.31) (84.10) (43.81) (221.2) (254.6) (322.3)
Constant 1308.1*** 2266.4*** 705.4*** 1984.4*** 2862.2*** 3440.2***

(86.85) (177.7) (70.57) (203.7) (623.0) (646.5)
Mean of dep. var 1036.5 2415.3 697.0 1957.9 2433.9 3927.2
Observations 756 756 756 756 756 756
R-Square 0.157 0.0645 0.0284 0.0621 0.0589 0.0967
Adjusted R-square 0.140 0.0456 0.0086 0.0431 0.0398 0.0784
*
p < 0.10, ** p < 0.05, *** p < 0.01
Appendix 16: Treatment effect on expense on fertilizers and insecticides (Cotton Crop)
(1) (2)
(901.3) (308.9)
[.19] [.92]
Treatment*Endline -1600.7** -1310.1***

(649.6) (198.8)
[.03] [.01]
Midline 3058.0*** 499.4**

(400.0) (157.0)
Endline 3711.0*** 1678.1***

(425.4) (22.42)
Treatment 522.2 493.7**

(487.3) (192.6)
69 | P a g e
Age of family members(avg) 12.59 3.981

(10.35) (3.416)
Age of farmer 1.008 -1.591

(7.375) (1.849)
Visit to extension centre 142.5 -91.58

(414.1) (107.2)
Asset Index 1 0 0
(.) (.)
Asset Index 2 633.1** 62.20

(258.7) (119.9)
Asset Index 3 538.5 8.612

(329.2) (151.9)
Asset Index4 592.6* -70.43

(290.1) (148.6)
Asset Index 5 286.4 -10.29

(426.2) (203.1)
General Caste -229.2 -303.4***

(295.6) (74.57)
Scheduled Caste 0 0
(.) (.)
Scheduled Tribe -256.3 -179.2**

(221.9) (66.57)
Other Backward Caste -211.6 370.6***

(286.2) (103.9)
Constant 3030.6*** 1407.3***

(589.0) (190.4)
R-Square 0.343 0.258
*
p < 0.10, ** p < 0.05, *** p < 0.01
70 | P a g e
Appendix 17: Treatment effect on macro-nutrient application (Cotton Crop)
(1) (2) (3)

Treatment*Midline 17.49** 10.06 6.875
(7.037) (11.76) (10.62)
[.07] [.42] [.64]
Treatment*Endline -29.74*** -22.80** 5.062

(7.342) (7.455) (6.253)
[0] [0] [.49]
Midline 47.27*** 28.97*** 25.30***

(5.880) (3.261) (7.175)
Endline 69.26*** 25.74*** 33.09***

(1.877) (6.534) (3.448)
Treatment 0.159 6.016 6.849

(8.153) (4.755) (4.831)
Age of family members(avg) 0.271 0.277*** -0.0234

(0.191) (0.0750) (0.126)
Age of farmer -0.119 -0.0240 -0.0606

(0.160) (0.0851) (0.0638)
Visit to extension centre -7.959 1.446 0.803

(7.626) (4.771) (3.487)
Asset Index 1 0 0 0
(.) (.) (.)
Asset Index 2 5.296 6.572 1.876

(6.965) (4.031) (1.291)
Asset Index 3 15.78** 5.559 1.587

(5.231) (4.403) (2.416)
Asset Index4 11.04* 7.288** 2.690

(5.828) (3.077) (3.363)
Asset Index 5 7.758 0.538 1.457

(7.918) (5.075) (2.145)
General Caste -10.10 -3.488 1.583

(7.936) (4.386) (2.085)
Scheduled Caste 0 0 0
(.) (.) (.)
71 | P a g e
Scheduled Tribe -9.011 -6.057* 3.161

(7.141) (3.289) (3.047)
Other Backward Caste -7.222 -4.812 2.648

(6.389) (3.812) (2.051)
Constant 72.23*** 30.88*** 20.94**

(12.29) (6.517) (6.486)
Mean of dep. var 109.2 56.59 45.15
R-Square 0.309 0.211 0.316
*
p < 0.10, ** p < 0.05, *** p < 0.01
72 | P a g e

WP No. 639

Uploaded by

Copyright:

Available Formats

WP No. 639

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

WP No. 639

Uploaded by

Copyright:

Available Formats

IIMB-WP No.

WORKING PAPER NO: 639

Hybrid Agricultural Information Delivery System as a Means to Improve

Year of Publication – April 2021

Hybrid Agricultural Information Delivery System as a Means to Improve Sustainability of

Keywords: Agricultural information delivery, Agricultural sustainability, Information and communication

2.2. Role of agricultural information delivery methods

3.2. Mechanism for Impact of DATES intervention

Extension Gap Research Gap Science Gap

Best Practice Research Potential Science Potential

Realised on Realised using best Realised by Realised by

Source: Evenson (2000)

Total Agricultural Output Growth

Weather Input Fixed Input Quality Change in Quality and

Quality of Quality of Quality of Application

Education Irrigation Embodied Disembodied

Public R&D Extension

Source: (Wang, Heisey, Schimmelpfenning, & Ball, 2015)

Figure 2: DATES Intervention in terms of Total Agricultural Output Growth Model

5. Data and Empirical Strategy

5.2. Matching of observables

Panel B: Caste Categories

5.3. Empirical strategy

Panel B: Average area under cultivation (acre)

Cotton 28610 42172 34941 22861 50820 58471

Cotton 23589 25192 24704 23539 28803 23147

Cotton 6021 16979 10237 -678 22017 35324

Cotton 28610 22861 5749***

Cotton 22589 23539 -950

Cotton 6021 -678 6699***

Panel A: Paddy Crop

Treatment*Endline 5.036*** 9263.3*** -5715.8*** 14901.8***

Treatment*Endline 7.174*** 29289.4*** -2451.4 31761.4***

Panel A: Paddy Crop

Treatment*Endline -349*** 6.075 -280.5*** -114.5 -1195*** 65.31

Treatment*Endline 208.6 -246.3 95.80 -681.2** -54.94 201.7

6.2 Impact of intervention on efficiency of agricultural operations

Table 8: Technical efficiency score

Cotton 3641 6702 7363 4210 8645 6356

Cotton 1167 1666 2854 1637 2190 1991

Table 10: Comparing values at baseline: Expense on fertilizers and insecticides

Cotton 3641 4210 -568**

Cotton 1167 1637 -470***

Treatment*Endline -3464.4*** -531.9*

Treatment*Endline -1600.7** -1310.1***

Figure 3: Quantity of nutrients applied by farmers at baseline

Panel A: Nitrogen (Kilogram/Acre)

Panel A: Paddy Crop

Treatment*Endline -71.66*** -37.31*** -16.53**

Treatment*Endline -29.74*** -22.80** 5.062

6.4 Cost of the intervention

a Extension Agents 3 15,000 24 1080000

II Travel and Other Allowances (Monthly Expenses)

a Extension Agents 3 5000 24 360000

III Yearly Expenses

a Operational cost 1 20,000 2 40000

a Hand-held devices 3 9,000 1 27000

7. Conclusion and Policy Implications

DoF. (2014). Indian Fertilizer Scenario 2014. Retrieved from http://fert.nic.in:

Map of treatment and control Gram Panchayats in DATES

TreatmentEndline 5.036 9263.3* -5715.8* 14901.8*

TreatmentEndline 7.174 29289.4* -2451.4 31761.4***

TreatmentEndline -349 6.075 -280.5* -114.5 -1195*** 65.31

TreatmentEndline -1600.7 -1310.1

TreatmentEndline -71.66 -37.31* -16.53**

TreatmentEndline -29.74 -22.80 5.062

TreatmentEndline 5.036 9263.3* -5715.8* 14901.8*

Endline 2.635* 4663.2 4540.2*** 152.3

Years of schooling (Farmer) 0.177 303.6 -7.286 305.7**

Constant 23.60* 36279.2* 19193.9* 17194.7*

TreatmentEndline -349 6.075 -280.5* -114.5 -1194*** 65.31

Midline -462* -527.6 335.7*** 192.0* 59.37 282.8

Endline -351* -395.3 828.2* -1223.9* 465.0*** 528.6*

Small farmer -27.06 -432.6 -6.066 -265.2 118.2 -88.05

Medium farmer -215.*** -241.3* 59.33 -232.7 371.2* -188.8**

Large farmer -477* -624.3* 52.80 -60.58 376.1 -348.8*