Skip to content

When Do Farmers Adopt and Respond to Early Warning Systems?

At first glance, one might presume the answer to this question to be an emphatic “always”. After all, these systems are designed to shield communities from severe environmental hazards, thereby boosting agricultural productivity, so it would seem that farmers would jump at the chance to use them. In reality, however, even when access isn’t a barrier, the use of Early Warning Systems can vary (Sharafi et al., 2021; Andersson et al., 2020). This raises some interesting questions: What drives some farmers to embrace EWSs? What holds others back? And how can research on climate risk communication help improve the adoption of these life-saving technologies?

In exploring these questions with existing literature, EPAR PhD student Nnenna Ogbonnaya-Orji conducted a Web of Science (WOS) search using the criteria “early warning system” AND farmer AND climate on studies published between January 2015 and March 2025. The majority of studies retrieved were focused on early warning systems pertaining to specific hazards, most notably droughts. Papers were excluded for one or more of the following reasons: either they did not address factors relevant to the uptake of early warning systems or other weather/climate information service; focused on advanced economies, or did not pertain to the early warning of climate-related events or disasters.

Why EWSs Matter to Farmers in Low- and Middle-Income Countries

Climate adaptation is increasingly critical for agriculture, especially rainfed and small-scale producer systems in countries where farming is not just an economic activity but a way of life, woven into social and cultural fabrics. In these regions, farming is often tied to deeply rooted beliefs and traditions. This creates a complex decision-making environment that makes it difficult to promote new information-based technologies. The high uncertainty of climate impacts introduces cognitive biases, which further complicates decisions to adopt new practices and technologies.

Though definitions of EWSs may vary, they all revolve around two main goals: detecting risks early and advising specific actions to reduce impact. Risk in this context is defined as the probability of an adverse event occurring, with its impact contingent on the frequency and intensity of exposure to the event and the capacity to mitigate or adapt to it (IPCC, 2022). Exposure and vulnerability depend on the scope and scale of the particular event, and are shaped by socioeconomic and political factors such as literacy levels, access to infrastructure, government policies and priorities, and trust in institutions – which in turn, influence how warning messages are communicated and the extent to which they prompt action. Therefore, to be effective, EWSs must be regularly updated and customized to fit the changing dynamics of local contexts. A key component of many EWSs are Seasonal Climate Forecasts (SCFs), which estimate rainfall and other important meteorological variables for periods ranging from months to entire seasons Despite increased availability, and proven gains to agricultural yield (e.g. see Lupiya et al., 2024), adoption among farmers is, at times, considered suboptimal (Andersson et al., 2020).

What drives EWS Adoption  by Farmers?

Though some progress has been made in recent times, not all countries have functional EWSs. As at 2023, 69 of 100 countries assessed under the Sendai Framework for Disaster Risk Reduction reported having “comprehensive coverage” of local or national dissemination mechanisms for early warning information (WMO, 2023). Notably, none of the 11 Least Developed Countries (LDCs) assessed were included in this number – highlighting  wide disparities in EWS access and significant knowledge gaps on the nature of this problem, especially as it pertains to developing countries. In some contexts where this technology is available, some evidence suggests that uptake by farmers is low, due in part, to a range of socio-cultural, demographic, economic, and institutional factors. For example, wealth and socioeconomic status reportedly play a big role in determining which farmers have access to these technologies, meaning that uptake can be limited to those with more resources (Nyoni et al., 2024). Adoption by farmers has also been found to be more likely with higher literacy rates and less likely with increasing age (Kolawole et al., 2018). Relative to other climate smart agricultural practices, information-based interventions such as EWSs are distinct in their primary reliance on effective communication rather than on physical tools or techniques. In Sub Saharan Africa, for example, low uptake of weather and climate services has been linked to limited comprehensibility of forecast information, inappropriate use of language and perceived incompatibility with indigenous practices (Nkiaka et al., 2019). Thus, to dig deeper into what drives EWS adoption, it helps to break down the communication process. Berlo’s (1960) SMCR model (where SMCR denotes the Source, Message, Channel, and Receiver), offers a useful framework for exploring these factors.

The Source of the Message Matters

EWSs are often managed and delivered through a top-down structure, starting from national or regional levels and moving down to local authorities, who then communicate with end-users (Andersson et al., 2020). This structure means that how farmers view the credibility of the government or scientific institutions often influences their trust in these systems. In many low-income countries, farmers rely on traditional sources of weather information, embedded in their cultural frameworks, rather than on scientific sources. For example, a study in North Central Namibia found that while about 50% of households accessed climate information, many still found it insufficient and continued to trust traditional knowledge instead (Gitonga et al., 2020). Promising paths to circumventing this challenge emphasize scientific interdisciplinarity and participatory approaches to the design and delivery of EWSs (Van Ginkel & Biradar, 2021; Walker, 2021; Hermans et al., 2022).

Trust in the Message Itself

It’s not just about who delivers the message, but also about the message’s content, quality, and clarity. Farmers need to feel confident that the information is reliable and relevant. For an EWS to build this confidence, it typically needs to deliver consistently accurate and precise forecasts—a tall order given the high uncertainties inherent in climate modeling. This is complicated as communicating these uncertainties without compromising message credibility is difficult. Furthermore, many low-income countries lack the resources needed for up-to-date methods, further limiting the system’s ability to generate reliable predictions.

Studies also show that how a message is framed can impact its reception in a given context. For instance, Calvel et al. (2020) found that small-scale farmers in the Mangochi and Salima Districts of Malawi adopted and responded to early warning messages when framed as advice pertaining to agricultural practices (what to do) as opposed to weather-related information (when to do it). In a study of coastal communities in Vietnam, Ngo et al. (2022) found that when messages were more concrete, and framed as gains, they led to a stronger risk awareness and intent to act on climate adaptation than abstract- or loss- framed messages. To be effective, they conclude, EWS messages should be clear, culturally relevant, and specific in the actions they ask farmers to take.

Reaching Farmers: The Importance of Communication Channels

Whether it’s traditional media, new digital platforms, or word-of-mouth, the medium of communication is critical to broadening EWS access and uptake. Thus, to be effective, EWS messages must be disseminated via communication channels that are tailored, contextually relevant and accessible. This can be challenging, as in many poor or remote regions, access to these channels remains limited and can vary along gender lines, making it difficult to reach all potential users.

An illustrative case is the flood devastation of rural communities in Delta state, Nigeria despite timely early warnings disseminated via mass media. In a qualitative study, Ebhuoma & Leonard (2021) found that fatigue after a long day of work prevented farmers who owned radios from tuning into local news, while poverty restricted access to television broadcasts, suggesting that more personalized communication e.g. via extension workers may  have been more impactful. Using intrahousehold survey data, Ngigi & Muange (2022) report gender variations in access to climate information services and preferences for specific communication channels. Whereas husbands were found to have significantly more access to early warning systems and advisory services on adaptation than their wives, wives were shown to have greater access to weather forecasts than their husbands. Additionally, husbands indicated a greater preference for obtaining climate information services from extension workers, print media, television and local leaders, while wives preferred to obtain such information from their social networks and/or the radio. Other channel-relevant  challenges cited in literature include delays in data exchange between agencies and the need for timely, multi-channel dissemination, all of which can affect the salience of EWSs in local communities.

Understanding the Receiver

For EWSs to be impactful, they need to engage with farmers’ perspectives on risk and adaptation. Some factors,  like innate cognitive biases, may be harder –if not infeasible – to address. But others, such as perceptions of self-efficacy and adaptive capacity, can be influenced through well-crafted messages. Research shows that with a higher material cost of adaptation (e.g. buying drought-resistant seeds or hiring equipment), discrepancies between plans and actions tend to widen (Sutcliffe et al., 2024). This suggests that when economic constraints prevent adaptation, communication strategies can sometimes help bridge the gap by fostering a sense of control. Farmers may be more likely to act if they feel capable of managing risks despite limited resources.

Bringing It All Together

The SMCR model focusing  on the Source, Message, Channel, and Receiver provides a useful framework for dissecting the complex and layered communication environment in which farmers operate. By addressing gaps in trust, message framing, communication channels, and farmers’ perceptions, researchers and policymakers can improve EWS adoption, ultimately helping farmers make more informed, adaptive decisions in a changing climate.

There is no simple answer to why some farmers use and respond to climate information while others do not. Risk perception, often influenced by cognitive biases, remains a central factor in these decisions, but Early Warning Systems (EWSs) can also be optimized to improve uptake. EWS deployment must balance environmental and economic benefits with social realities to be effective in different cultural and economic contexts (Sharafi et al., 2021). Practical barriers, such as limited forecast precision, comprehension challenges, weak infrastructure, and perceived self-efficacy, underscore the need for EWS designs that engage local communities in identifying social priorities (Andersson et al, 2020; Otieno et al., 2024).

Studies suggest the importance of integrating EWSs with indigenous knowledge systems to build trust and foster effective two-way engagement with end-users of warning messages (Fragaszy et al., 2020; Otieno et al.,2024; Funk et al., 2023). Although some studies have looked into ways that practitioners have tried to do this (e.g. see Walker, 2021 and Hermans et al,. 2022 for examples) some questions remain. For instance, how might warning messages be communicated in ways that do not undermine/patronize indigenous belief systems? How can discrepancies which sometimes arise between scientifically derived information and farmers perceptions of climate realities be reconciled?  (Solano-Hernandez et al., 2020).

Risk communication models that account for the behavioral nuances of adoption, such as consumer choice models with relaxed rationality assumptions or mental models that align messaging with farmer perspectives, could provide a stronger theoretical basis for adaptive decision-making.

Continuous EWS interventions face sustainability challenges, especially in changing policy landscapes. Research could address the effects of evolving political environments on EWS credibility, community buy-in, and funding stability. Institutionalizing EWSs, with stable support and adaptability to shifting needs, could enhance their long-term effectiveness.

Ultimately, addressing informational, attitudinal, and behavioral barriers through well-designed risk communication will be essential for supporting climate change adaptation. Future research that hones methods for effective risk messaging and clarifies potential trade-offs in EWS updates will enable decision-makers to allocate resources effectively and enhance farmers’ resilience in the face of growing climate challenges.

Blog written by Nnenna Ogbonnaya-Orji.

Gendered languages

Introduction

Language is an important cultural institution in human society. The World Bank (2019) estimates that 38 percent of world languages have grammatical gender, which can be defined as a system of word classification in which words are grouped into categories or “genders”. In the last decade, gendered language has attracted increasing attention from social scientists as a force shaping socio-economic inequalities and development outcomes. Evidence from recent research in economics, psychology and management has shown that the gender system of a language has causal impacts on wage inequality, female labor force participation, gender gap in math achievement, innovation output, gender norm attitudes, among others.

Source: World Atlas of Language Structure dataset (WALS)

Grammatical gender in language may or may not be linked to biological sex. In some languages, gender is based on an animacy (living/non-living) rather than a male/female distinction, such as in Swahili. Grammatically gendered languages vary in the specific location in syntax to which gender is attached (Kramer 2016). Some languages gender-mark nouns only, whereas in others gender is marked in various locations in the syntax.  Gender can manifest in (1) adverb modifiers: for example, “la voiture” the car, “le marteau” the hammer in French; (2) adjective modifiers: for example, “la nouvelle voiture” the new car, “lemarteau perdu” the lost hammer in French; (3) pronouns: For example, the sex-based pronominal gender system she/he/it in English; and/or (4) verbs: For example, “молодая медсестра пошла в больницу” – “The young nurse went to the hospital” in Russian. Here, the modifier and the verb are both conjugated to the female gender in agreement with the gender of the subject. Arabic is another language where verbs are gendered in accordance with the gender of the subject. Some languages whose noun gender system is sex-based do not indicate sex in pronouns, such as Tamil and Farsi; whereas some gender-mark pronouns only, not nouns, such as English.  

Operationalizing grammatical gender as an independent variable

What does genderedness of a language capture?

Explanations vary regarding the mechanism through which grammatical gender impacts decision-making and behavior. Proponents of one view posit that grammatical gender, like other language structures, arises from evolutionary pressures shaping patterns of reproduction and the division of labor in the history of a given society (Dahl 2004; Johansson 2005; Galor et al. 2018; Shoham 2019). Language structure, along this view, has been described as “institutional memory” (Hicks et al. 2015), “gender roles mirror” (Shoham 2019) and “a vehicle for cultural transmission” (Gay et al. 2013). Therefore, grammatical gender encodes historical institutions and acts as a proxy for their effects. Another view suggests that variations in language structures give rise to differences in individuals’ reasoning, learning, and other mental activities (Whorf 1956; Lucy 1997).

Exploring mechanisms is especially relevant to policy because determining what policy can change and cannot change requires elucidating the source of constraints or shifters of preference. It is difficult for policy to cause wholesale change in social and cultural norms. So if gendered language has no independent effect apart from mediating that of social and cultural norms, language reform policy has little leverage. However, if gendered language alone constitutes a source of psychological constraints or primers with impact on behavior or decision-making, language reform policy can be an easy-to-implement and low-cost way of attaining desired change in agents’ behavior or decision-making.

Language as a measure of culture

Language structures are part of culture, broadly understood. However, unlike other cultural features which evolve independently, language structures are highly correlated with other subsets of culture. Attributing autonomous causal power to gendered language faces difficulty given the presence of a rich set of confounders (Maviskayan 2015). For example, what historically caused the emergence of certain features of linguistic structures may directly shape the outcome variable, eg. historical migration and labor specialization patterns.

Scholars have used language structure as an instrumental variable for culture. For example, Licht et al. (2007) use the grammar of pronouns as an instrumental variable to study how countries that favor autonomy, egalitarianism, and mastery exhibit less corruption. However, this approach assumes that language structure has no autonomous, unmediated causal impact on individual propensity to corruption, which is contradicted by the linguistic relativity hypothesis. Studies, such as Gay et al. (2013), have shown that linguistic structures can have significant effects on socioeconomic outcomes after controlling for measures of culture.

Measurements

What constitutes a grammatically gendered and a grammatically genderless language is subject to contested definitions. Authors of the World Atlas of Language Structures (WALS) articulate four gender-related grammatical features of language, each of which can be represented in a dummy variable as follows (summarized in Gay et al. 2013):

Measurement strategy Definition Binary/continuous Adoption (# articles)
Gay I (2013) GII = SBI + NGI + GAI + GPI Continuous (0-4) 6
Gay II (2018) GII = SBI × (GPI + GAI + NGI) Continuous (0-3) 1
Jakiela & Ozier (2018) GG = SBI × GAI Binary 2
Stahlberg (2007) A grammatically gendered language is one where every noun is assigned feminine or masculine or neuter gender. It is distinguished from natural gender languages and genderless languages Binary 1
Givati & Troiano (2012) The number of cases of gender-differentiated pronouns Continuous 1
Mavisakalyan (2015) A language is either highly gendered, mildly gendered, or gender neutral Continuous (0-2) 1
Table 1. Measurement strategies

Other measurement strategies exist: in Jakiela & Ozier (2018) define grammatical gender by requiring formal gender assignment (GAI=1), combined with sex-based assignment (SBI=1) to create their metric (GAI × SBI); Stahlberg et al. (2007) classify languages into three categories: grammatical gender languages (nouns have gender), natural gender languages (no grammatical gender, but pronouns may reflect sex), and genderless languages (no gender in nouns or pronouns). Givati & Troiano (2012) quantify grammatical gender as the number of gender-differentiated pronouns, e.g. Spanish has four cases of gender-differentiated pronouns (third-person singular, first-person plural, second-person plural, and third-person plural). Lastly, Mavisakalyan (2015), building on Siewierska (2008)’s categorization of languages based on the intensity of gender distinctions in pronouns into six types, offers a revised framework with three types: highly gendered (Gender distinction in third-person and also the first- and/or the second-person singular pronouns), mildly gendered (Gender distinction in third-person singular pronouns only), and gender-neutral (No gender distinction in pronouns).

Table 2. Grammatical gender features in WALS

  0 1
Sex-Based Intensity (SBI)[1] languages without a sex-based gender system languages with a sex-based gender system
Number Gender Intensity (NGI) languages with three or more genders or with no gender distinctions languages with exactly two genders
Gender Assignment Intensity (GAI) languages with no gender assignment system or where the gender assignment system is only semantic[2] Languages where the gender assignment system is both semantic and formal[3]
Gender Pronouns Intensity (GPI) languages which do not distinguish gender in pronouns (or does so only in the third-person pronoun). languages with a gender distinction in third-person pronouns and also in the first and/or the second person
Table 2: Four gender-related grammatical features of language according to classification by authors of the World Atlas of Language Structure dataset (WALS)

Existing Evidence

A large body of work in psychology exploring the effect of gendered language on psychological processes and attitudinal and cognitive outcomes exist; however, the recent economics literature has increasingly focused on socioeconomic outcomes. Existing literature on the effect of gendered language predominantly find a negative effect of gendered language on gender equality, understood as women’s attainment, or the parity between men and women, of socioeconomic outcomes. These outcomes include the Global Gender Gap Index score (Prewitt-Freilino et al. 2012); education attainment (Davis & Reynolds 2018; Galor et al. 2020); labor force participation (Givati and Troiano 2012; Gay et al. 2013; Mavisakalyan 2015; Jakiela and Ozier 2018; Gay et al. 2018); political participation (Gay et al. 2013), access to land and credit (Gay et al. 2013); representation of women on corporate boards (Santacreu-Vasut et al. 2014; Pisera 2023); gender wage gap (Van der Velde et al. 2015; Shoham & Lee 2018); entrepreneurship activity (Hechavarría et al. 2018; Xie et al. 2021); women’s math performance (Kricheli-Katz & Regev 2021; Cohen et al. 2023).

Exceptions include Del Caprio & Fujiwara (2023), focusing on women’s application to online tech jobs, which finds no evidence of a significant effect of changing the gendered form of address in job ads except for fields where there is existing substantial representation of women; and Santacreu-Vasut et al. (2013), who find a positive relationship between gendered language and gender equality, although equity is legislated (gender quota in a parliament) hence possibly reflecting that a society with gendered language is more likely to require a quota for representation than a society able to move toward equality more naturally.


[1] SBI includes nomial as well as pronominal sex-based gender system. English, for example, is considered to have SBII=1 since it has sex-specific pronouns, although it is otherwise usually considered a grammatically genderless language.

[2] Semantic gender assignment is based on the inherent meaning or characteristics associated with nouns. Nouns are assigned to gender classes based on their semantic properties, such as biological sex, animacy, or natural gender. In this system, the gender assignment of a noun is often predictable based on the noun’s meaning, eg. “la grand-mère”, the grandmother; “le fils”, the son.

[3] Formal gender assignment, also known as grammatical or arbitrary gender assignment, is based on the form or shape of nouns rather than their inherent meaning. Nouns are assigned to gender classes based on grammatical rules and patterns within the language, which may not correspond to natural or semantic gender distinctions, eg. “la voiture”, the car; “le Pakistan”, Pakistan.


Glossary

Grammatical gender A system in which nouns are always categorized into classes, which are often but not always linked to biological sex
Syntax The order in which words are arranged in a sentence
Syntax agreement Agreement occurs when a word changes form depending on the other words in a sentence to which it relates.
Semantic Relating to meaning in language
Modifier An adjective or an adverb used to modify a noun
Lexical Relating to the words or vocabulary of a language
Lexical gender Semantic property carried in lexical units, such as “mother” and “son” in English which assign female or male quality to the referent
Referential gender A sub-category of lexical gender where referential lexical units, ie. pronouns have gendered semantic property, such as “he” and “she” in English
Social gender (also called natural gender) Socially imposed dichotomy of masculine and feminine traits associated with nouns, such as “trucks” (masculine) v. “lamp” (feminine), which often but do not always coincide with grammatical gender
Generic masculine (also called false generics) Male generic usage in cases of gender-indefinite reference. E.g., the singular generic masculine in English, “An American drinks his coffee black”; the plural generic masculine in Spanish, “Los olímpicos han vuelto” (The Olympians have returned).
Speaker-gender dependent expressions Expressions or linguistic structures which male or female speakers of the language ought to use exclusive to their gender, eg. politeness expressions reserved for women in Japanese; separate languages for men and women in the Ubang community in Nigeria