RES EARCH

◥ tension agents and enhance supply chain

REVIEW SUMMARY functionality. Realizing the potential of digital
agriculture will require an interdisciplinary
AGRICULTURE effort to develop and rigorously test a variety
of approaches, incorporating insights from
Realizing the potential of digital development: behavioral science, agriculture, economics,
and data science.
The case of agricultural advice
OUTLOOK: Multiple market failures associated
Raissa Fabregas, Michael Kremer, Frank Schilbach* with information markets limit the ability
of mobile phone–based extension systems
to reach socially efficient scale through purely
BACKGROUND: Sustainably raising agricultural cost of information transmission, benefits commercial financing. Because the marginal
productivity for the 2 billion people living likely exceed costs by an order of magnitude. costs of disseminating information are close
in smallholder farming households in the Even basic phones and inexpensive text and ◥ to zero, the optimal scale
developing world is critical for reducing world voice messages can influence farmer behavior. ON OUR WEBSITE
of such systems is very
poverty and meeting rising food demand in Smartphones with GPS systems create the Read the full article large. However, fixed sys-
the face of climate change. Nevertheless, most potential for larger gains through the trans- at http://dx.doi. tem development costs
smallholder farmers have no access to science- mission of more sophisticated media, such as org/10.1126/ still must be covered. Mul-
based agricultural advice. The widespread videos, and for locally customized informa- science.aay3038 tiple organizations have

Downloaded from http://science.sciencemag.org/ on December 17, 2019

..................................................
adoption of basic mobile phone technology tion on soil characteristics, weather, and pest introduced digital agri-
presents opportunities to improve upon exist- outbreaks, delivered at the appropriate time cultural extension systems with financial
ing in-person agricultural extension efforts during the agricultural season. models based on selling subscriptions to indi-
that are expensive and fraught with account- Messages could be customized on the basis vidual farmers, but such systems have been
ability problems. of farmer characteristics, such as education or able to reach only a small fraction of farmers
financial circumstances. Experimentation, ma- in the developing world. Farmer payments
ADVANCES: Meta-analyses suggest that the trans- chine learning, and two-way communication may be insufficient to cover the fixed costs,
mission of agricultural information through with and between farmers could facilitate im- because information is difficult to exclude
mobile technologies in sub-Saharan Africa and provements of information and other services from nonpurchasers and because it is chal-
India increased yields by 4% and the odds of over time. Advances from behavioral science lenging for farmers to verify the quality of the
adoption of recommended agrochemical inputs can improve information transmission and information. Existing evidence suggests sub-
by 22%. The delivery of market information can address behavioral barriers to the adoption stantial gaps between farmers’ willingness
have additional system-wide impacts, reducing of improved agricultural techniques. Mobile to pay for information and its social value.
price dispersion and lowering transaction phone–based systems could increase the pro- Advertising or agrochemical input sales could
costs. Given the low and rapidly declining ductivity and accountability of in-person ex- be used to finance information provision, but
this approach could incentivize providers to
distort information content in the absence of
strong reputational costs of misinformation
or appropriate regulation.
Public financing could cover fixed costs
and enable scale-up. Although agriculture
ministries often deliver messages in ways
that farmers find difficult to understand and
use, recent examples suggest that if feedback
mechanisms are in place, governments can
improve their services over time. Models that
incentivize farmers to share their experiences
create scope for customization and efficiency
gains as systems grow, because this data may
be used to improve recommendations for other
farmers. If successful, digital agricultural ad-
visory systems could supply a model for dig-
ital development more broadly.
▪
PHOTO: JAKE LYELL/ALAMY STOCK PHOTO

The list of author affiliations is available in the full article online.

*Corresponding author. Email: fschilb@mit.edu
Cite this article as R. Fabregas et al., Science 366,
eaay3038 (2019). DOI: 10.1126/science.aay3038

Mobile phones can benefit farmers in low- and middle-income countries by improving access to
agricultural advice and market price information. Mobile technologies, particularly smartphones, have the
potential to bring sophisticated science-based agricultural advice to smallholder farmers to improve TOMORROW’S EARTH
productivity, especially under rapidly changing economic and environmental conditions. However, market Read more articles online
failures likely preclude efficient scaling of valuable digital advice applications. at scim.ag/TomorrowsEarth

Fabregas et al., Science 366, 1328 (2019) 13 December 2019 1 of 1

RES EARCH

◥ Digital agriculture: Potential and challenges

REVIEW There is good reason to believe that emerging
digital technologies can improve the func-
AGRICULTURE tioning of agriculture markets at a very low cost
per farmer. Establishing initial mobile phone
Realizing the potential of digital development: coverage involves fixed costs, but the marginal
cost of phone communication in rural areas is
The case of agricultural advice close to zero because cell phone towers typ-
ically operate below capacity. Cellular phone
Raissa Fabregas1, Michael Kremer2, Frank Schilbach3* companies charge prices well above marginal
cost, but they are often highly regulated, and
The rapid spread of mobile phones creates potential for sustainably raising agricultural productivity for governments could negotiate access at prices
the 2 billion people living in smallholder farming households. Meta-analyses suggest that providing with lower markups.
agricultural information via digital technologies increased yields by 4% and the odds of adopting Mobile phones, particularly GPS-enabled
recommended inputs by 22%. Benefits likely exceed the cost of information transmission by an order of smartphones, facilitate the provision of tail-
magnitude. The spread of GPS-enabled smartphones could increase these benefits by enabling ored information. Recommendations for
customized information, thus incentivizing farmers to contribute information to the system. Well-known agrochemical inputs that address specific
distortions in markets for information limit the ability of such systems to reach the socially efficient soil conditions on the basis of digital maps
scale through purely commercial means. There is a clear role for public support for digital agricultural can improve yields while reducing environ-
extension, but messages designed by agricultural ministries are often difficult for farmers to understand mentally harmful and wasteful use (22–24).

Downloaded from http://science.sciencemag.org/ on December 17, 2019

and use. Realizing the potential of mobile communication systems requires feedback mechanisms to Messages can target specific areas with re-
enable rigorous testing and continuous improvement. ported pest outbreaks or be customized to
other local conditions such as market prices.

M
Farmers can tailor their investment decisions
obile phones have penetrated the de- enabled smartphones will create opportuni- to expected weather patterns and benefit
veloping world to a greater extent than ties for customization and two-way commu- from improvements in weather forecasting
most other technologies (Fig. 1). More nication, but an interdisciplinary effort will (25, 26). Customized information allows farm-
than three of four people in low- and be required to experiment with different ap- ers to choose language, dialect, or literacy
middle-income countries (LMICs) own proaches and rigorously measure impact. Dis- levels. Mobile technologies can also provide
a phone. Approximately one in three people tortions in the markets for information limit reminders and other nudges to address be-
have internet access, and access is expected the ability of systems to reach the socially ef- havioral biases (27).
to increase markedly as smartphone costs ficient scale through purely commercial means, Running these systems at scale allows for
decline (1). such that scaling programs beyond their cur- testing variations to establish the most effec-
The spread of phones presents opportu- rent modest levels will likely involve an active tive approaches (A/B testing) and feedback
nities for digital development by reducing public-sector role. loops to improve accuracy and effectiveness
information acquisition costs, allowing cus- of messages over time. Images taken from
tomization of information, and enabling mon- Traditional agricultural extension satellites can provide rich data about crop
itoring and accountability in public services Raising agricultural productivity is critical to growth and, when linked with Geographic
(1–3). Digital technologies have been deployed reducing poverty and satisfying the growing Information System (GIS) on plot boundaries,
in a range of sectors—including finance, edu- global food demand (10) in the face of envi- can improve measurements of productivity at
cation, health, and civic participation—to im- ronmental stress and climate change. Improved scale and allow for ongoing experimentation
prove development outcomes (1, 4, 5). access to agricultural information and target- (28, 29). Mobile phones facilitate two-way
The proliferation of phones may also carry ing of agricultural inputs can raise agricultural communication, whereby farmers can ask
risks, such as the potential to exacerbate vio- productivity and reduce negative environmen- questions and request information. Such
lent conflict (6), enable state surveillance and tal footprints (11, 12). platforms can also provide opportunities
propaganda (7), accelerate the spread of fake Nevertheless, most smallholder farmers lack for networking and information exchange
news via social media, or further widen in- access to science-based agricultural advice. Al- among farmers. Information from farmers
equality because of uneven access to digital though ~400,000 agricultural extension agents using the system can further improve future
technologies (8, 9). Finance and governance (13) are employed by governments in LMICs, recommendations for all users.
systems will affect the sustainability, scale, the ratio of farmers to extension workers ex- As smartphone use continues to expand,
equitable reach, and effective design and im- ceeds 1000 to 1 in many countries (14). Trans- farmers will increasingly have the means to
plementation of these systems. port budgets are often meager, and training, watch videos demonstrating new agricultural
We review the evidence on “digital agricul- management, and accountability of extension techniques or take pictures of pests affecting
ture.” With current technologies, impacts on workers are inadequate. In India, only 6% of their crops and either request automatic iden-
farming practices and yields are modest in farmers report having received any advice tification and recommendations or raise ques-
absolute terms but large relative to the cost from an extension agent in the past year, and tions with agronomists (30). Smartphones may
of information delivery. The spread of GPS- 70% of farmers distrust extension worker rec- also provide farmers access to interventions
ommendations (15). and apps that can enhance psychological well-
More generally, there is limited evidence of being (31). Increased aspirations, grit, and
1
Lyndon B. Johnson School of Public Affairs, University of extension services’ impact or cost effective- improved mental health may boost farmer
Texas at Austin, Austin, TX 78712, USA. 2Department of ness (13, 16). Extension workers have been income by increasing investment and facil-
Economics, Harvard University, Cambridge, MA 02138, USA.
3 found to favor their own social networks itating learning among farmers (32–34).
Department of Economics, Massachusetts Institute of
Technology, Cambridge, MA 02139, USA. (17) and neglect the most vulnerable farmers Mobile phones may create opportunities to
*Corresponding author. Email: fschilb@mit.edu (18, 19) and women (20, 21). complement and strengthen existing in-person

Fabregas et al., Science 366, eaay3038 (2019) 13 December 2019 1 of 9

RES EARCH | R E V I E W

Downloaded from http://science.sciencemag.org/ on December 17, 2019

Fig. 1. The spread of mobile phones in LMICs. (A) Growth of mobile phone subscriptions relative to other services in LMICs. “Improved sanitation” denotes the
percentage of people using basic sanitation services, at minimum [data from (1) and World Development Indicators: https://datacatalog.worldbank.org/dataset/
world-development-indicators]. (B) Mobile phone penetration, as determined by the percentage of adults who report owning any mobile phone and those who own a
smartphone [data from Pew Global Attitudes Survey, Spring 2017: www.pewresearch.org/global/dataset/spring-2017-survey-data/].

agricultural extension efforts. Many agricul- demonstration plots or conducting training Finally, digital agricultural services can im-
tural extension workers already have smart- sessions. Mobile phones could also be used prove the functioning of agricultural supply
phones and thus could download information to improve accountability among extension chains. For example, these services could make
on pests, flooding, or other problems arising in workers—for example, by allowing extension it easier for farmers to check and compare
their region, as well as information needed to workers and their supervisors to set goals and input or output prices, potentially lowering
respond to farmer queries. Automatic notifica- track performance, enabling automatic collection markups; notify farmers whether inputs are
tions can allow extension agents to alert farm- of feedback from farmers, or tracking wheth- in stock with particular dealers; and facilitate
ers in their region when they are visiting er extension agents actually visit farmers. coordination among farmers in an area and

Fabregas et al., Science 366, eaay3038 (2019) 13 December 2019 2 of 9

RES EARCH | R E V I E W

Downloaded from http://science.sciencemag.org/ on December 17, 2019

Fig. 2. Effects of text messages on acquisition of recommended inputs. Meta-analysis of the effects of text messages on following advice for purchasing
agricultural lime, as measured by administrative data. Kenya Agriculture and Livestock Research Organization (KALRO), Precision Agriculture for Development and
Innovations for Poverty Action (PAD/IPA), and One Acre Fund (OAF) implemented the programs. The effects are measured in odds ratios (OR). The OR is the
ratio of the odds of following recommendations in the treatment group divided by the odds in the control group. The odds refer to the probability of following the
advice over the probability of not following the advice. Weights are from a random effects analysis [data adapted from (47)].

with traders. Firms could use projections of annoyed by unwanted spam messages or feel planned adoption of recommended agricul-
harvest quality to inform lending decisions patronized by reminders and exhortations tural inputs and practices (44–46). Each of these
(35). Satellite-based yield assessment could be (36). Taken together, such issues could lead outcome variables has limitations. Knowledge
used to inform social insurance programs to reduced trust in the messaging system. may or may not translate into behavior change.
that provide support for farmers in response Realizing the potential of customization and Relying on self-reported data on the use of
to weather or pest disasters. two-way communication in LMICs carries par- inputs may lead to overestimation of impact.
However, despite the potential of digital agri- ticular challenges. Customization requires in- For example, farmers who receive messages
culture advisory services, reasons for skepticism formation about a farm’s location, which is advocating certain behavior may over-report
remain. Overcoming informational constraints difficult to collect remotely unless farmers this behavior because of social desirability bias.
may not result in substantially increased agri- have GPS-enabled smartphones, because in Indeed, a recent meta-analysis of four trials in
cultural productivity, given the existence of many countries there is a lack of precisely- Kenya found that the measured impact of
other barriers such as credit constraints, input defined physical addresses (37), area names mobile phone messages using self-reported
shortages at local markets, and missing insur- are often ambiguous, and user text entry is data exceeded the impacts based on admin-
ance markets and infrastructure (12). Even to error prone (38). Gathering agricultural data istrative data (47).
the extent that informational barriers are im- from farmers is challenging because response To alleviate such concerns, administrative
portant, mobile phone messages may not rates to phone surveys are typically low; farm- data on input purchases from agricultural
overcome them: Some farmers ignore messages, ers may be hesitant to provide accurate infor- suppliers or redemption of discount coupons
especially from unknown sources, because phone mation; and some information, such as exact were used to measure farmer behavior in six
spam is common in many LMICs. Some farmers yields, can be difficult to quantify. experimental evaluations of text messaging
are illiterate and have difficulty using voice programs that encouraged farmers in East
menus. Senders may design obscure and con- Impacts of digital agriculture: Africa to use locally appropriate inputs (47).
fusing messages or may provide messages Empirical evidence Figure 2 depicts the results from a meta-
designed to target objectives at odds with Earlier reviews of the impacts of digital agri- analysis of these studies, which found that the
farmer interests, such as messages aimed at cultural extension report mixed results and odds ratio for following the recommendation
increasing sales of inappropriate agricultural considerable context dependence (39–43). How- to purchase agricultural lime, an input to
inputs. Certain kinds of information may be ever, sufficient evidence is now emerging to reduce soil acidity, is 1.22 [95% confidence
too complicated to convey by text or voice; begin quantitatively assessing the farmer-level interval (CI): 1.13 to 1.31]. For context, the
effective communication may require pictures impact of digital agricultural extension by proportion of people acquiring recommended
or video. Smartphones are thus required to meta-analysis. inputs in each of the control groups ranged
receive these messages, but few smallholder from 0.03 to 0.32.
farmers currently have access to this technol- Impacts on individual farmers Some of the individual experiments had
ogy in the poorest countries. Finally, farmers Several experimental studies have found that statistically significant impacts and others
may begin to ignore reminders or nudges if mobile phone–based programs increase farm- did not. However, we cannot reject the hy-
they are repeated too often, or they may be er knowledge and self-reported adoption or pothesis that the effects were the same across

Fabregas et al., Science 366, eaay3038 (2019) 13 December 2019 3 of 9

RES EARCH | R E V I E W

Downloaded from http://science.sciencemag.org/ on December 17, 2019

Fig. 3. Effects of digital agriculture programs on yields. Meta-analysis of effects of digital technologies on crop yields drawn from six studies that report yields
[data from (46, 48–52)]. For the study by Udry (52), the outcome variable is harvest value. The upper portion of the figure shows the impact of mobile-based
programs delivered directly to farmers. The lower portion shows the impact of digital programs with an in-person component. Weights are from a random effects
analysis. ES, effect size as a percentage increase.

contexts and that the estimated effects dif- workers (51), and a program in Ghana de- advanced technology, such as video, and bet-
fered only because of sampling variation, livered by community extension workers who ter customization to local conditions. Video-
which suggests that we need to be cautious relied on a mobile software application (52). based interventions and a gamified app have
in claims regarding heterogeneous treatment Several statistical approaches indicate that also been found to improve knowledge and
effects and, in particular, in interpreting the digitally delivered advice to farmers increases farmers’ practices (30, 50, 53, 54).
sources of differences in estimated effects yields by ~4% (see supplementary materials). As noted, traditional in-person agricultural
across studies. Combining these estimates Notably, the impacts are not larger for services extension has been found to favor certain
with agricultural trial data on the impact on that include more costly in-person compo- groups. It seems likely that digital extension
yields of treating soil with lime suggests that nents. On average, the value of increased will also favor men, as well as richer, younger,
farm profits increased by one to two orders output greatly exceeds the marginal cost of and more educated farmers with better digital
of magnitude beyond the marginal cost of delivery via mobile phones, such that policy- access. However, current data are inconclusive,
sending the messages. Similar estimates were makers would invest in mobile-based programs and it is possible that biases will be less se-
found for fertilizer purchases (47). unless they are highly risk-averse. vere than with in-person extension. Cole and
Figure 3 reports on a complementary meta- Several factors suggest that the true returns Fernando (46) report suggestive evidence that
analysis measuring the impact of experimen- to investment in digital agricultural extension richer farmers were modestly more engaged
tally evaluated digital agricultural extension may be higher than these numbers suggest. in the service they studied and were more
interventions on farm yields or harvest value First, farmers who receive information via dig- likely to adopt recommended practices. In con-
(unfortunately we do not have sufficient data ital agricultural extension sometimes transmit trast, Fabregas et al. (47) found little evidence of
on farm costs to estimate impacts on profits). it to other farmers, thus creating additional heterogeneous impact in their meta-analysis,
This analysis encompasses four trials of mes- benefits (46, 47). Second, to the extent that although the underlying studies did not include
sages delivered purely through mobile phones: impacts vary across contexts and policy-makers farmers without phones.
two text message interventions with sugarcane have data to assess impact in their own context,
farmers in Kenya (48) and two season mea- there is value in testing such systems, assessing Market- or system-level impacts
sures for an interactive voice response (IVR) their effects, and adjusting policy accordingly. Beyond individual-level effects, digital tech-
intervention with cotton farmers in India (46). Unsuccessful programs can be abandoned and nologies can also affect farmers by altering
It also includes four studies with an in-person successful ones scaled up. Finally, impacts are entire markets or systems. In particular, im-
component: two video interventions with maize likely to improve over time as farmers learn to proved access to price information can enable
and rice farmers in Uganda implemented via use the systems, program operators improve farmers to sell their products in markets with
in-person visits (49, 50), a program providing message content and delivery through A/B higher prices and reduce price dispersion
customized information on rice cultivation to trials, and smartphone use spreads, enabling across markets. By reducing waste of perish-
Nigerian farmers offered through extension digital extension services to incorporate more able goods and the need for middlemen,

Fabregas et al., Science 366, eaay3038 (2019) 13 December 2019 4 of 9

RES EARCH | R E V I E W

Downloaded from http://science.sciencemag.org/ on December 17, 2019

P

Fig. 4. Farmer valuation curve and usage under a subscription model. Given the downward-sloping valuation (or demand) curve, P* is the profit-maximizing
price. Usage at this price is lower than the socially efficient level, giving rise to a deadweight loss. Firms would not be able to cover fixed costs if they set prices at the
distribution costs. Firms will invest in a system only if the development costs are less than the expected profits, whereas a hypothetical benevolent social planner
would invest as long as costs are less than the sum of profits, farmer surplus, and deadweight loss.

improved information access can thus increase ers are common, but by establishing a hotline level, phone surveys allow collection of high-
producer prices and lower average prices for for farmers to report problems, late delivery frequency survey data on agricultural produc-
consumers. Access to mobile phones allowed was reduced by 23% and nondelivery by 54%. tion at a much lower cost than traditional
Indian fishermen to compare prices while still The benefits spill over to neighbors because methods (68). Yet despite some successes in
at sea and then transport their catch to the mark- the company schedules deliveries to neighbor- eliciting information through phone-based
ets offering the highest prices, thus causing a ing farms at the same time. feedback tools (69), phone surveys are plagued
reduction in price dispersion across markets An important area for future work is ex- with low response rates and thus may be sub-
(55). Fewer fish were wasted, profits increased ploring whether digital agriculture can address ject to selection bias. If users are required to
by 8%, and consumer prices declined by 4%. supply chain problems, such as limited compe- provide information to access content, they
Studies in Uganda and Niger recorded sim- tition and high markups (65) and adulteration may prioritize speed over accuracy, degrading
ilar results for other crops (56–58); delivering and counterfeiting of inputs (66). Moreover, the quality of information.
price information for staple grains was also studying the distributional effects of different Systems that foster information exchange
found to cause positive effects in Ghana and interventions, arising from a combination of to facilitate participation and truthfulness
Peru (59, 60). In contrast, sending farmers the direct effects of receiving digital advice by using a “Netflix model” are one solution.
price and weather information through text itself and through positive or negative spill- Netflix shares recommendations for video
messages and phone calls did not affect av- overs of interventions, remains a fruitful area content with its users and tethers those rec-
erage prices for crops from farmers in India for investigation. ommendations to what users have liked in
or Colombia (61–63). These differences are the past. This procedure incentivizes users to
hypothesized to result from a combination Learning from farmers share information with the platform to im-
of factors, including differences in target pop- A key open empirical question is the extent prove the quality of its future recommenda-
ulations, crop varieties, the importance of in- to which mobile systems can gather valuable tions. This information is then used to benefit
formational constraints, message design, and information from farmers, which in turn can other users of the service by improving the
barriers to the effective use of information and be used to inform other farmers. In the United quality of Netflix’s recommendations to them.
communications technology (41). States, Farmers Business Network applies ma- A comparable model could potentially work
Digital technologies may also improve sup- chine learning to hundreds of thousands of for agriculture. Farmers could be convinced to
ply chains by helping farmers shop for ade- acre-years of data to provide high-quality yield supply information on what has recently worked
quate inputs or report inefficiencies or fraud. predictions for seed varieties (67). Mobile phone for them, because doing so would improve
Casaburi et al. (64) examined a contract-farm systems in LMICs could potentially be used to the advice the mobile-advisory service pro-
setting in which farmers sign contracts with a collect data to serve as inputs in machine learn- vides them in the future. Such a system would
sugar company in Kenya. The company pro- ing applications, learn from farmers’ experi- incentivize farmers to share their experiences,
vides agrochemical inputs to farmers and then ences with particular agricultural technologies, because sharing would enable them to receive
deducts the costs of the inputs from the and facilitate networking among farmers. better-tailored recommendations. The result-
amount it pays farmers for the sugarcane. De- However, gathering high-quality data from ing data could also be used to improve recom-
lays or failures in fertilizer deliveries to farm- farmers is challenging. At the most basic mendations for other, similar farmers.

Fabregas et al., Science 366, eaay3038 (2019) 13 December 2019 5 of 9

RES EARCH | R E V I E W

Financing and governance of digital the creation of the service and the conditions un- agricultural practices may differ from year
agriculture systems der which it will be privately profitable to do so. to year, it may be difficult for farmers to as-
Digital agricultural extension systems cur- More sophisticated forms of subscription sess the quality of advice, even after they pur-
rently reach only a small proportion of farmers models may ameliorate these distortions. To chase it.
(70). Here, we discuss barriers to commercial the extent that firms can charge different Weak regulation makes it difficult to trust
scaling, as well as problems with public scal- prices for information on the basis of farmers’ those selling information. Fraudulent operators
ing and emerging evidence on ways to ad- willingness to pay, these firms can serve more can set up firms and offer useless information.
dress them. farmers and increase their profits. “Freemium” Even firms with legitimate information would
models are a step in this direction because they have incentives to inflate the benefits of their
Barriers to scaling of subscription models give consumers a chance to learn about the information. Farmers may thus discount any
Many of the efforts to establish digital exten- quality of advice before they pay for it. claims and reduce their willingness to pay for
sion systems, such as iCow Global in Kenya or Nonrivalry does not imply that no knowledge- information. Markets for information can un-
RML Agtech in India, have sought to finance creation investments will be commercially ravel entirely, preventing any transactions.
themselves by selling subscriptions to farmers, viable. Indeed, a considerable share of all Sellers address this issue by investing in a
but these types of efforts have reached only a research and development investment is made reputation for trustworthiness, but this involves
small fraction of the potential market (70). by the private sector. If the only market failure some costs, and farmers may still retain doubts.
Economic theory suggests that three features associated with agricultural information mar- Beyond these distortions specific to agricul-
of markets for agricultural information— kets was nonrivalry, then a subscription model tural information markets, other factors make
nonrivalry, nonexcludability, and asymmetric might become viable once technological prog- selling any investment product in these mar-
information—make it difficult for pure sub- ress and the spread of smartphones and data kets difficult. Many developing-country farm-

Downloaded from http://science.sciencemag.org/ on December 17, 2019

scription models to reach as many farmers as plans sufficiently drove down the costs of in- ers have no readily available cash and may not
would be efficient from a social point of view. formation production and distribution. How- be able to borrow money either. In addition,
ever, markets for agricultural information a variety of behavioral factors, ranging from
Nonrivalry are subject to two additional distortions that present bias to loss aversion, may inhibit
Information differs from most other goods further reduce farmers’ willingness to pay for investment (27). Customer acquisition costs
(71, 72). Creation of information involves fixed information. and transaction costs in payments are also
costs—for example, collecting data from soil likely obstacles to successful scaling with
tests and weather stations or designing, test- Nonexcludability subscription models.
ing, and refining comprehensible and action- Agricultural information is also nonexclud- Together, these factors erode farmers’ wil-
able messages for farmers. However, unlike able or only partially excludable—i.e., once an lingness to pay for information and thus limit
most other goods, such as agricultural products, individual has access to the information, this the financial viability of digital agricultural
information is a nonrival commodity: Once it person can easily share it with others. In their extension efforts. Cole and Fernando (46)
has been created, it can be used by additional study of digital agricultural extension in India, estimate that their IVR service in India in-
people at a minimal marginal distribution cost, Cole and Fernando (46) found significant creased farmers’ incomes by more than the
with no cost to others. knowledge spillovers to farmers who had not costs of the service, but despite a high rate of
From a social point of view, it is efficient for received the services in the trial. A rich lite- farmer engagement, the average price a farmer
all who value the information at more than rature documents the flow of agricultural in- was willing to pay for a 9-month subscription
the distribution cost to have access to it. How- formation in rural communities (73–75). Sharing was only $2, whereas the cost of provision
ever, a firm using a pure subscription model of information not only directly reduces the was $7.
would need to charge a higher fee than the number of potential customers for digital agri-
price of distribution to cover the fixed cost of cultural extension services but may also re- Other commercial financing models
information creation. As a result, some farm- duce willingness to pay among those who do Beyond pure subscription models, other com-
ers for whom it would be efficient to obtain purchase, which could affect the financial via- mercial models may partially address some of
information would be excluded (Fig. 4). For bility of subscription-based services. the market failures.
the particular curve drawn, the majority of In a study of willingness to pay for local soil
farmers would value the information at more information in western Kenya, individual Contract farming
than the cost of distribution, but only about farmers were not willing to pay the full cost of Some agricultural products, such as sugarcane
one-third would be willing to pay the profit- local soil test results (76). However, the aggre- or dairy, require local processing, often fea-
maximizing price of a commercial firm using gate valuation of all farmers for a given test in turing a dominant local buyer. If the buyer
a pure subscription model. an area exceeded the cost of testing, poten- profits sufficiently from having a greater input
In addition, under nonrivalry, a potential tially making investment in this information supply, this person may be willing to pay to
provider operating via a subscription model worthwhile from a social standpoint. Farmers’ provide digital agricultural extension services
may not have the commercial incentives to willingness to pay for information was also for all farmers in the area. Because the dom-
create the service, even if it is socially efficient lower in a setting where they could ask others inant buyer would want all farmers in the
to do so. It is socially efficient to invest if the for the information, which suggests that the area to increase production, this would help
total area between the farmer valuation curve option of resale or free riding depressed any address the problems of nonrivalry and
and the cost of distributing information— willingness to pay. nonexcludability. Additionally, a buyer with
profits plus farmer surplus and deadweight professional staff may be less subject to
loss—exceeds the cost of information creation. Asymmetric information asymmetric information problems. The buyer
However, a private firm will invest only if the Buyers do not necessarily know the value of may personally operate a digital agricultural
profits from selling information are sufficient the information sold to them, and they may extension service or purchase these services
to cover the cost of information creation. Non- not trust sellers’ claims about its value (77). in bulk from another provider, thus reducing
rivalry leads to a gap between the conditions Because agricultural production is highly customer acquisition costs relative to the cost
under which it is socially efficient to invest in variable and the profitability of recommended of selling to individual farmers.

Fabregas et al., Science 366, eaay3038 (2019) 13 December 2019 6 of 9

RES EARCH | R E V I E W

This approach may be particularly effective to market failures, government solutions are warrant waiving rules against sending un-
when the buyer cares not only about the extra subject to government failures. Governments solicited information. As smartphones and
profits from greater input supply, but also are not known for nimble product develop- data plans spread, it will become cheaper to
about farmers. Dairy farmers, for example, ment or user-friendly technology interfaces, distribute content to farmers, so nonexclud-
often organize themselves into cooperatives and they lack the immediate customer feed- ability of information will be less of a barrier to
that jointly buy milk-processing equipment. back mechanism the market provides. Gov- information provision. But it is likely to become
In some cases, lenders also provide digital ernment agencies often provide agricultural more difficult to control the provision of mis-
agricultural advice. information that might be too technical or leading information, and hence asymmetric
detailed for communicating with the average information may become more of a problem.
Advertising and selling inputs farmer. For example, several Indian state gov- Customization could have great potential
Digital agricultural extension providers could ernments distribute personalized soil health benefits, but it also raises questions about how
also try to finance themselves by selling ad- cards, based on local soil tests, to farmers. These to protect users’ privacy. Governments must
vertising, selling own-brand inputs, or enter- cards are difficult for farmers to understand, decide whether and under what circumstances
ing strategic alliances with agricultural input and many farmers report never receiving them to share contact information for agricultural
providers. Incentives such as commissions can (15). Those who do receive the report cards extension agents or farmers.
lead to biased advice (78); nevertheless, some often distrust the content (79). Simplifying Regulatory issues arise even for messages
firms succeed in developing a reputation for the design, making the cards less technical, sent by government agencies. Messages from
providing objective advice. For example, Farm- and complementing them with information an agricultural ministry, for example, could
ers Business Network in the United States has delivered by mobile phones increased baseline crowd out equally important health messages.
grown rapidly with a financial model based on comprehension from 8% to at least 40% (15). Messages that misleadingly imply that socially

Downloaded from http://science.sciencemag.org/ on December 17, 2019

selling own-brand agricultural inputs, as well Similarly, a government-sponsored IVR help- desirable behavior (such as environmentally
as providing information. line in Africa required farmers to answer a favorable agricultural practices) has individual
As smartphones and data plans become series of registration questions before they benefits could reduce trust in all messages
more common and it becomes possible to could access content, preventing many from from the government. Too many messages
transmit more types of information at low reaching the agricultural content. Merely could annoy people and make them feel
cost, advertising and input sales may gen- postponing user registration until after the uncomfortable (82).
erate enough revenue to finance some pro- farmers received some useful information in-
vision of agricultural information, but markets creased the share of farmers getting to the Outlook
are still unlikely to deliver socially optimal content by approximately one-fifth, from The available evidence suggests that the ben-
outcomes. Information will probably be under- 52 to 63% (80). efits of providing digital extension far exceed
supplied on agricultural techniques that do not In these cases, governments responded to the costs but that subscription-based models
involve input purchases or that involve only evidence by adjusting their programs, rede- will not reach optimal scale. This disparity
purchases of inputs that have become commo- signing the soil health cards, and postponing creates a potential role for public financing,
dified and hence have low markups. False or registration requirements, raising the possi- which LMIC governments and aid donors
misleading information may be supplied on bility that the systematic incorporation of on- are increasingly supporting.
the merits of different brands or the suitabil- going surveys and A/B trials into government Delivering on the full promise of digital
ity of techniques that involve input purchase. programs may serve as a partial substitute for agriculture, including customization of in-
Although regulation could potentially address the lack of market feedback. formation provision, will require sustained
such issues, designing and implementing ap- However, governments, like private busi- cycles of iteration and testing. The develop-
propriate regulation is likely to be difficult. nesses, may distort information provision to ment of lessons that are viable and useful in
farmers. For example, governments may want multiple contexts will be essential to avoid
Public financing and government provision to increase production of certain commodities reinventing the wheel for each application.
The above examples of market failures provide to achieve export goals, but they may not suf- Because these lessons may constitute a global
a rationale for the public sector to fully or ficiently value the time and effort required for public good, multilateral institutions and
partially finance provision of digital agricul- farmers to adopt the corresponding agricul- global donors may wish to financially support
tural extension. A public-sector entity could tural practices. If government systems grow, digital information provision efforts by gov-
either operate a digital agricultural extension input sellers may start lobbying or bribing gov- ernments or private actors in exchange for
service itself, as the government of Ethiopia ernment officials to recommend their brands undertaking experimentation and making the
currently does, or contract with or provide fi- as opposed to others. results widely available. Equally, many of the
nancial support for private providers, as donors emerging lessons on provision of information
in international development are currently Regulation to farmers could also apply in other sectors,
doing. Governments could also use their Digital agricultural extension raises regulatory such as education or health.
position as regulators to encourage telecom- questions that require further research. Infor-
munication firms to provide such services or mation providers with financial interests in REFERENCES AND NOTES
to make capacity available to other digital selling certain products might send misleading 1. World Bank, “World Development Report 2016: Digital Dividends”
(World Bank, 2016). doi: 10.1596/978-1-4648-0671-1
agricultural extension providers. messages. One approach to tackling this chal- 2. J. C. Aker, J. E. Blumenstock, “The economic impacts of
Scaling through governments entails typi- lenge would be to mandate certain disclosures new technologies in Africa” in The Oxford Handbook of Africa
cally lower customer acquisition costs than of financial conflicts. However, evidence from and Economics: Volume 2: Policies and Practices, C. Monga,
J. Y. Lin, Eds. (Oxford Univ. Press, 2015), pp. 354–371.
faced by individual companies, because gov- the regulation of financial and medical pro-
3. J. C. Aker, I. M. Mbiti, Mobile Phones and Economic
ernments can leverage their relationships with ducts industries suggests that such mandates Development in Africa. J. Econ. Perspect. 24, 207–232 (2010).
telecommunications companies, the existing are not sufficient (81). Telecommunications doi: 10.1257/jep.24.3.207
agricultural extension apparatus, and regula- authorities typically have rules limiting phone 4. J. C. Aker, “Using digital technology for public service provision
in developing countries” in Digital Revolutions in Public Finance,
tory powers to draw farmers to their platforms. spam and will have to decide whether specific S. Gupta, M. Keen, A. Shah, G. Verdier, Eds. (International
However, just as market solutions are subject emergencies, such as severe pest outbreaks, Monetary Fund, 2017), pp. 201–224.

Fabregas et al., Science 366, eaay3038 (2019) 13 December 2019 7 of 9

RES EARCH | R E V I E W

5. T. Suri, W. Jack, The long-run poverty and gender impacts of 28. M. Burke, D. B. Lobell, Satellite-based assessment of yield 49. B. Van Campenhout, W. Walukano, F. Nattembo, L. Nazziwa-Nviiri,
mobile money. Science 354, 1288–1292 (2016). doi: 10.1126/ variation and its determinants in smallholder African systems. J. Blom, “The Role of Information in Agricultural Technology
science.aah5309; pmid: 27940873 Proc. Natl. Acad. Sci. U.S.A. 114, 2189–2194 (2017). Adoption: Experimental Evidence from Rice Farmers in Uganda”
6. J. H. Pierskalla, F. M. Hollenbach, Technology and Collective doi: 10.1073/pnas.1616919114; pmid: 28202728 (IFPRI Discussion Paper, IFPRI, 2017); https://core.ac.uk/
Action: The Effect of Cell Phone Coverage on Political 29. D. Donaldson, A. Storeygard, The View from Above: download/pdf/153429888.pdf.
Violence in Africa. Am. Polit. Sci. Rev. 107, 207–224 (2013). Applications of Satellite Data in Economics. J. Econ. Perspect. 50. B. Van Campenhout, D. J. Spielman, E. Lecoutere, “Information
doi: 10.1017/S0003055413000075 30, 171–198 (2016). doi: 10.1257/jep.30.4.171 and Communication Technologies (ICTs) to Provide
7. B. Qin, D. Strömberg, Y. Wu, Why Does China Allow Freer 30. K. Vasilaky, K. Toyama, T. Baul, M. Mangal, U. Bhattacharya, Agricultural Advice to Smallholder Farmers: Experimental
Social Media? Protests versus Surveillance and Propaganda. “Learning Digitally: Evaluating the Impact of Farmer Training Evidence from Uganda” (IFPRI Discussion Paper, IFPRI, 2018);
J. Econ. Perspect. 31, 117–140 (2017). doi: 10.1257/jep.31.1.117 via Mediated Videos” presented at the Northeast Universities www.ifpri.org/publication/information-and-communication-
8. J. James, Digital Divide Complacency: Misconceptions and Development Consortium Conference, Providence, RI, 7 and technologies-icts-provide-agricultural-advice-smallholder.
Dangers. Inf. Soc. 24, 54–61 (2008). doi: 10.1080/ 8 November 2015; https://editorialexpress.com/cgi-bin/ 51. A. Arounaa, J. D. Michlerb, W. G. Yergoa, K. Saitoa, “One Size
01972240701774790 conference/download.cgi?db_name=NEUDC2015&paper_id=519. Does Not Fit All: Experimental Evidence on the Digital Delivery
9. J. Blumenstock, N. Eagle, “Mobile divides: Gender, 31. M. Ridley, G. Rao, P. Vikram, F. Schilbach, “Poverty and Mental of Personalized Extension Advice in Nigeria” (2019);
socioeconomic status, and mobile phone use in Rwanda” in Illness: Causal Evidence” (2019); https://economics.mit.edu/ www.researchgate.net/publication/336927609_One_Size_
Proceedings of the 4th ACM/IEEE International Conference files/18694. Does_Not_Fit_All_Experimental_Evidence_on_the_Digital_
on Information and Communication Technologies and 32. S. Alan, T. Boneva, S. Ertac, Ever Failed, Try Again, Succeed Delivery_of_Personalized_Extension_Advice_in_Nigeria.
Development - ICTD ’10 (Association for Computing Machinery, Better: Results from a Randomized Educational Intervention on 52. C. Udry, “Information, Market Access and Risk: Addressing
2010), article no. 6; .doi: 10.1145/2369220.2369225 Grit. Q. J. Econ. 134, 1121–1162 (2019). doi: 10.1093/qje/ Constraints to Agricultural Transformation in Northern Ghana”
10. Food and Agriculture Organization of the United Nations (FAO), qjz006 (2019); https://sites.northwestern.edu/christopherudry/files/
“The economic lives of smallholder farmers: An analysis 33. E. La Ferrara, Aspirations, Social Norms, and Development. 2019/03/DIRTS-Final-report-1362c0g.pdf.
based on household data from nine countries” (FAO, 2015); J. Eur. Econ. Assoc. 10.1093/jeea/jvz057 (2019). doi: 10.1093/ 53. R. Gandhi, R. Veeraraghavan, K. Toyama, V. Ramprasad,
www.fao.org/3/a-i5251e.pdf. jeea/jvz057 “Digital Green: Participatory video for agricultural extension,” in
11. Z. Cui et al., Pursuing sustainable productivity with millions of 34. T. Bernard, S. Dercon, K. Orkin, A. S. Taffesse, “The future in 2007 International Conference on Information and
smallholder farmers. Nature 555, 363–366 (2018). mind: short and long-run impact of an aspirations intervention Communication Technologies and Development (IEEE, 2007).

Downloaded from http://science.sciencemag.org/ on December 17, 2019

doi: 10.1038/nature25785; pmid: 29513654 in rural Ethiopia,” Working paper (Univ. of Oxford, 2018); doi: 10.1109/ictd.2007.4937388
12. B. K. Jack, “Market inefficiencies and the adoption of http://ibread.org/bread/system/files/bread_wpapers/429.pdf. 54. E. Tjernström, T. J. Lybbert, R. Frattarola Hernández,
agricultural technologies in developing countries,” [White 35. G. Murthy, M. Fernandez-Vidal, X. Faz, R. Barreto, “Fintechs J. S. Correa, “Learning by (virtually) doing: experimentation
paper, Agricultural Technology Adoption Initiative, JPAL (MIT)/ and Financial Inclusion: Looking past the hype and exploring and belief updating in smallholder agriculture” (2019);
CEGA (Berkeley), 2011]; https://escholarship.org/content/ their potential” (Consultative Group to Assist the Poor, 2019); www.lafollette.wisc.edu/sites/tjernstrom/
qt6m25r19c/qt6m25r19c.pdf. www.cgap.org/sites/default/files/publications/ Learning_by_virtually_doing.pdf.
13. J. Anderson, G. Feder, in Agricultural Development: Farmers, 2019_05_Focus_Note_Fintech_and_Financial_Inclusion_1_0.pdf. 55. R. Jensen, The Digital Provide: Information (Technology),
Farm Production and Farm Markets, vol. 3 of Handbook of 36. M. T. Damgaard, C. Gravert, The hidden costs of nudging: Market Performance, and Welfare in the South Indian Fisheries
Agricultural Economics, R. Evenson, P. Pingali, Eds. (Elsevier, Experimental evidence from reminders in fundraising. J. Public Sector. Q. J. Econ. 122, 879–924 (2007). doi: 10.1162/
2007), pp. 2343–2378. Econ. 157, 15–26 (2018). doi: 10.1016/j.jpubeco.2017.11.005 qjec.122.3.879
14. K. Davis, B. Swanson, D. Amudavi, D. A. Mekonnen, A. Flohrs, 37. U. P. Union, “Addressing the World–An Address for Everyone” 56. J. C. Aker, Information from Markets Near and Far: Mobile
J. Riese, C. Lamb, E. Zerfu, “In-depth assessment of the (White paper, Universal Postal Union, 2012); www.upu.int/ Phones and Agricultural Markets in Niger. Am. Econ. J. Appl.
public agricultural extension system of Ethiopia and fileadmin/documentsFiles/activities/addressingAssistance/ Econ. 2, 46–59 (2010). doi: 10.1257/app.2.3.46
recommendations for improvement” [International Food Policy whitePaperAddressingTheWorldEn.pdf. 57. J. C. Aker, M. Fafchamps, Mobile Phone Coverage and
Research Institute (IFPRI) Discussion Paper, IFPRI, 2010]; 38. A. Erlich, D. F. Jung, J. D. Long, C. McIntosh, The double-edged Producer Markets: Evidence from West Africa. World Bank
www.ifpri.org/publication/depth-assessment-public-agricultural- sword of mobilizing citizens via mobile phone in developing Econ. Rev. 29, 262–292 (2015). doi: 10.1093/wber/lhu006
extension-system-ethiopia-and-recommendations. countries. Dev. Eng. 3, 34–46 (2018). doi: 10.1016/ 58. J. Svensson, D. Yanagizawa, Getting Prices Right: The Impact
15. S. Cole, G. Sharma, “The Promise and Challenges in j.deveng.2017.11.001 of the Market Information Service in Uganda. J. Eur. Econ.
Implementing ICT for Agriculture,” Working paper (2017); 39. J. C. Aker, Dial “A” for agriculture: A review of information and Assoc. 7, 435–445 (2009). doi: 10.1162/JEEA.2009.7.2-3.435
https://precisionag.org/uploads/cole-sharma-july1-2017.pdf. communication technologies for agricultural extension in 59. P. Courtois, J. Subervie, Farmer Bargaining Power and Market
16. M. Gautam, “Agricultural extension: The Kenya experience: developing countries. Agric. Econ. 42, 631–647 (2011). Information Services. Am. J. Agric. Econ. 97, 953–977 (2015).
An impact evaluation” (World Bank Publications, 2000). doi: 10.1111/j.1574-0862.2011.00545.x doi: 10.1093/ajae/aau051
17. O. Bandiera, R. Burgess, E. Deserranno, R. Morel, I. Rasul, 40. E. Nakasone, M. Torero, B. Minten, The power of information: 60. E. Nakasone, “The Role of Price Information in Agricultural
M. Sulaiman, “Social Ties and the Delivery of Development The ICT revolution in agricultural development. Annu. Rev. Markets: Experimental Evidence from Rural Peru,” Working
Programs,” Working paper (2018); https://drive.google.com/ Resour. Econ. 6, 533–550 (2014). doi: 10.1146/annurev- paper (2014); https://fc317c2d-a-62cb3a1a-s-sites.
file/d/1uESM-AfGaiY-7eZ7Zq1WBzyP1IKE4fuS/view. resource-100913-012714 googlegroups.com/site/eduardonakasone/Nakasone%20-%
18. J. R. Alwang, P. B. Siegel, “Rural poverty in Zambia: An analysis 41. J. C. Aker, I. Ghosh, J. Burrell, The promise (and pitfalls) of ICT 20Price%20Info%20and%20Ag%20Markets%20-%
of causes and policy recommendations” (World Bank, 1994). for agriculture initiatives. Agric. Econ. 47, 35–48 (2016). 20OCT2014.pdf?attachauth=ANoY7crbtOVyu8ZK0saI6Lj
19. B. Cunguara, K. Moder, Is Agricultural Extension Helping the doi: 10.1111/agec.12301 ZFX5YTjR1uGo1cHVxIjGqIIwCGBxMJ_IlNbLwB9TYQn6vfZSW
Poor? Evidence from Rural Mozambique. J. Afr. Econ. 20, 42. H. Baumüller, The Little We Know: An Exploratory Literature 7ol2aX2JLozIxPvAWqbjvGZDtpdAtqLTQTQyWHFL4OMZ22
562–595 (2011). doi: 10.1093/jae/ejr015 Review on the Utility of Mobile Phone-Enabled Services for 94iGS4VN2afdYzi-rbC5gUKq8ADwq0aKtPVCX0iAMynd
20. F. Kondylis, V. Mueller, G. Sheriff, S. Zhu, Do Female Smallholder Farmers. J. Int. Dev. 30, 134–154 (2018). 5Q3QK2oljTngQPoyLOYuNdJrQtyFxJV3pnkb9kLrEE8X6_oO2
Instructors Reduce Gender Bias in Diffusion of Sustainable doi: 10.1002/jid.3314 JBZJiF2snZppL-Edb5LwSVErpxNJJa4DARTJXcQPIsm
Land Management Techniques? Experimental Evidence From 43. U. Deichmann, A. Goyal, D. Mishra, Will digital technologies 7suHAaddvPwWGSiX75x7mE–gG9osAlVdVGg%3D%
Mozambique. World Dev. 78, 436–449 (2016). doi: 10.1016/ transform agriculture in developing countries? Agric. Econ. 47, 3D&attredirects=0&d=1.
j.worlddev.2015.10.036 21–33 (2016). doi: 10.1111/agec.12300 61. M. Fafchamps, B. Minten, Impact of SMS-Based Agricultural
21. K. A. Saito, H. Mekonnen, D. Spurling, “Raising the Productivity 44. C. Larochelle, J. Alwang, E. Travis, V. H. Barrera, Information on Indian Farmers. World Bank Econ. Rev. 26,
of Women Farmers in Sub-Saharan Africa” (Discussion paper, J. M. D. Andrade, Did You Really Get the Message? Using Text 383–414 (2012). doi: 10.1093/wber/lhr056
World Bank, 1994); .doi: 10.1596/0-8213-2749-6 Reminders to Stimulate Adoption of Agricultural Technologies. 62. A. Camacho, E. Conover, “The Impact of Receiving Price and
22. R. A. Viscarra Rossel et al., A global spectral library to J. Dev. Stud. 55, 548–564 (2019). doi: 10.1080/ Climate Information in the Agricultural Sector,” Working paper
characterize the world’s soil. Earth Sci. Rev. 155, 198–230 00220388.2017.1393522 no. IDB-WP-220 [Inter-American Development Bank (IDB)
(2016). doi: 10.1016/j.earscirev.2016.01.012 45. X. Fu, S. Akter, The Impact of Mobile Phone Technology on Working Paper Series, 2011]; doi: 10.2139/ssrn.1755057
23. D. Arrouays, P. Lagacherie, A. E. Hartemink, Digital soil Agricultural Extension Services Delivery: Evidence from India. 63. S. Mitra, D. Mookherjee, M. Torero, S. Visaria, Asymmetric
mapping across the globe. Geoderma Reg. 9, 1–4 (2017). J. Dev. Stud. 52, 1561–1576 (2016). doi: 10.1080/ Information and Middleman Margins: An Experiment with
doi: 10.1016/j.geodrs.2017.03.002 00220388.2016.1146700 Indian Potato Farmers. Rev. Econ. Stat. 100, 1–13 (2018).
24. P. A. Sanchez et al., Digital soil map of the world. Science 325,
46. S. A. Cole, A. N. Fernando, “‘Mobile’izing Agricultural Advice: doi: 10.1162/REST_a_00699
680–681 (2009). doi: 10.1126/science.1175084;
Technology Adoption, Diffusion and Sustainability,” Working 64. L. Casaburi, M. Kremer, R. Ramrattan, “Crony Capitalism,
pmid: 19661405
paper (2018); www.dropbox.com/s/0rewpp8oeost0o5/ Collective Action, and ICT: Evidence from Kenyan Contract
25. M. Rosenzweig, C. R. Udry, “Forecasting Profitability,” Working ao_final_v24_tables.pdf?dl=0. Farming,” Working paper (2019); http://econ.uzh.ch/dam/jcr:
paper 19334 [National Bureau of Economic Research (NBER) e2ffc4e5-ab32-4405-bfa4-70b0e962aa81/
47. R. Fabregas, M. Kremer, M. Lowes, R. On, G. Zane,
Working Paper Series, 2013]; .doi: 10.3386/w19334
“SMS-extension and Farmer Behavior: Lessons from hotline_paper_20191015_MERGED.pdf.
26. R. B. Alley, K. A. Emanuel, F. Zhang, Advances in weather Six RCTs in East Africa,” Working paper (2019); 65. R. J. Sexton, Imperfect Competition in Agricultural Markets and
prediction. Science 363, 342–344 (2019). doi: 10.1126/science. www.dropbox.com/s/guiog3ief0hnwl7/textfarmers.pdf?dl=0. the Role of Cooperatives: A Spatial Analysis. Am. J. Agric. Econ.
aav7274; pmid: 30679358 72, 709–720 (1990). doi: 10.2307/1243041
48. L. Casaburi, M. Kremer, S. Mullainathan, R. Ramrattan,
27. M. Kremer, G. Rao, F. Schilbach, “Behavioral development “Harnessing ICT to increase agricultural production: 66. T. Bold, K. C. Kaizzi, J. Svensson, D. Yanagizawa-Drott, Lemon
economics” in Handbook of Behavioral Economics - Evidence from Kenya,” Working paper (2019); Technologies and Adoption: Measurement, Theory and
Foundations and Applications 2, B. D. Bernheim, S. DellaVigna, www.econ.uzh.ch/dam/jcr:873845ce-de4d-4366-ba9a- Evidence from Agricultural Markets in Uganda. Q. J. Econ. 132,
D. Laibson, Eds. (North Holland, 2019), pp. 345–458. d60accda577d/SMS_paper_with_tables_20190923_merged.pdf. 1055–1100 (2017). doi: 10.1093/qje/qjx009

Fabregas et al., Science 366, eaay3038 (2019) 13 December 2019 8 of 9

RES EARCH | R E V I E W

67. S. Cole, T. L. He, “Farmers Business Network: Putting Farmers 75. L. Beaman, A. BenYishay, J. Magruder, A. M. Mobarak, “Can 81. G. Loewenstein, S. Sah, D. M. Cain, The unintended
First,” Harvard Business School Case 217-025 (2018); Network Theory-based Targeting Increase Technology consequences of conflict of interest disclosure.
www.hbs.edu/faculty/Pages/item.aspx?num=51649. Adoption?” Working paper 24912 (NBER Working Paper Series, JAMA 307, 669–670 (2012). doi: 10.1001/jama.2012.154;
68. B. Dillon, Using mobile phones to collect panel data in 2018); .doi: 10.3386/w24912 pmid: 22337676
developing countries. J. Int. Dev. 24, 518–527 (2012). 76. R. Fabregas, M. Kremer, J. Robinson, F. Schilbach, “The Value 82. H. Allcott, J. B. Kessler, The Welfare Effects of Nudges: A Case
doi: 10.1002/jid.1771 of Local Agricultural Information: Evidence from Kenya,” Study of Energy Use Social Comparisons. Am. Econ. J. Appl. Econ.
69. M. Jones, F. Kondylis, Does feedback matter? Evidence from Working paper (2019); www.dropbox.com/s/ 11, 236–276 (2019). doi: 10.1257/app.20170328
agricultural services. J. Dev. Econ. 131, 28–41 (2018). bumw8uj5dyhnt9n/valueinformation.pdf?dl=0.
doi: 10.1016/j.jdeveco.2017.10.013 77. G. A. Akerlof, The Market for “Lemons”: Quality Uncertainty
AC KNOWLED GME NTS
70. M. Tsan, S. Totapally, M. Hailu, B. K. Addom, “The digitalization and the Market Mechanism. Q. J. Econ. 84, 488 (1970).
of African agriculture report 2018-2019: Executive summary” doi: 10.2307/1879431 This paper reflects conversations and comments from many
(CTA/Dalberg Advisers, 2019); https://cgspace.cgiar.org/ people, including D. Bjorkegren, S. Cole, T. Harigaya, J. Li,
78. S. Anagol, S. Cole, S. Sarkar, Understanding the Advice of
bitstream/handle/10568/103198/Executive%20Summary% R. Maertens, R. Meager, G. Rao, W. Więcek, and G. Zane.
Commissions-Motivated Agents: Evidence from the Indian Life
20V4.5%20ONLINE.pdf. We are grateful to F. Mashwama and J. Que for assistance in
Insurance Market. Rev. Econ. Stat. 99, 1–15 (2017).
preparing this Review. Competing interests: R.F. and M.K.
71. K. J. Arrow, “Economic welfare and the allocation of resources doi: 10.1162/REST_a_00625
have received research grants from Precision Agriculture for
for invention” in The Rate and Direction of Inventive Activity: 79. R. Fishman, A. Kishore, Y. Rothler, P. S. Ward, S. Jha,
Development. M.K. serves on the board of directors of Precision
Economic and Social Factors (Princeton Univ. Press, 1962), R. K. P. Singh, “Can information help reduce imbalanced
Agriculture for Development. The authors declare no other
pp. 609–626. application of fertilizers in India?: Experimental evidence from
competing interests.
72. P. M. Romer, Increasing Returns and Long-Run Growth. J. Polit. Econ. Bihar” (IFPRI Discussion Paper, IFPRI, 2016); http://ebrary.
94, 1002–1037 (1986). doi: 10.1086/261420 ifpri.org/utils/getfile/collection/p15738coll2/id/130249/
73. A. D. Foster, M. R. Rosenzweig, Learning by Doing and Learning filename/130460.pdf. SUPPLEMENTARY MATERIALS
from Others: Human Capital and Technical Change in 80. Precision Agriculture for Development, “Mid-term science.sciencemag.org/content/366/6471/eaay3038/suppl/DC1
Agriculture. J. Polit. Econ. 103, 1176–1209 (1995). evaluation of IVR/SMS project: Findings and Supplementary Text
doi: 10.1086/601447 recommendations” (Ethiopian Agricultural Transformation References (83–86)
74. T. G. Conley, C. R. Udry, Learning about a new technology: Agency, 2018); https://precisionag.org/uploads/January-

Downloaded from http://science.sciencemag.org/ on December 17, 2019

Pineapple in Ghana. Am. Econ. Rev. 100, 35–69 (2010). ATA-PAD-Report-FINAL-FOR-VALIDATION-WORKSHOP-
doi: 10.1257/aer.100.1.35 compressed.pdf. 10.1126/science.aay3038

Fabregas et al., Science 366, eaay3038 (2019) 13 December 2019 9 of 9

Realizing the potential of digital development: The case of agricultural advice
Raissa Fabregas, Michael Kremer and Frank Schilbach

Science 366 (6471), eaay3038.

DOI: 10.1126/science.aay3038

Mobile farming advice

Mobile phones are almost universally available, and the costs of information transmission are low. They are used
by smallholder farmers in low-income countries, largely successfully, to optimize markets for their produce. Fabregas et
al. review the potential for boosting mobile phone use with smartphones to deliver not only market information but also

Downloaded from http://science.sciencemag.org/ on December 17, 2019

more sophisticated agricultural extension advice. GPS-linked smartphones could provide locally relevant weather and
pest information and video-based farming advice. But how to support the financial requirements of such extension
services is less obvious, given the unwieldiness of government agencies and the vested interests of commercial
suppliers.
Science, this issue p. eaay3038

ARTICLE TOOLS http://science.sciencemag.org/content/366/6471/eaay3038

SUPPLEMENTARY http://science.sciencemag.org/content/suppl/2019/12/11/366.6471.eaay3038.DC1
MATERIALS

RELATED http://science.sciencemag.org/content/sci/366/6471/eaax3100.full
CONTENT
http://science.sciencemag.org/content/sci/366/6471/eaaw9256.full
file:/content

REFERENCES This article cites 44 articles, 4 of which you can access for free
http://science.sciencemag.org/content/366/6471/eaay3038#BIBL

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Use of this article is subject to the Terms of Service

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement of
Science, 1200 New York Avenue NW, Washington, DC 20005. The title Science is a registered trademark of AAAS.
Copyright © 2019 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of
Science. No claim to original U.S. Government Works