Impact of climate-smart agriculture practices on multidimensional poverty among coastal farmers in Bangladesh

The study examines whether and how climate-smart agriculture technologies (CSAT) reduce multidimensional poverty among climate-vulnerable coastal farm households in Bangladesh and what factors drive CSAT adoption. The context is rising food security pressures amid climate change, with Bangladesh’s southwestern coastal zone highly exposed to cyclones, salinity intrusion, drought, and waterlogging that threaten paddy yields and livelihoods. Small and marginal farmers dominate agriculture, and coastal poverty exceeds national averages. Climate-smart agriculture (CSA) aims to increase productivity, enhance resilience, and reduce emissions. Despite evidence that CSA improves yields and food security, its relationship with multidimensional poverty in coastal contexts remains underexplored. The study addresses three questions: (i) Which factors influence CSAT adoption among coastal farm households? (ii) Do CSAT practices impact multidimensional poverty (MPI)? (iii) Which specific CSA technologies most effectively reduce MPI?

Prior research documents diverse CSA practices (e.g., conservation tillage, agroforestry, crop rotation/diversification, drip irrigation, precision farming) and identifies adoption drivers such as land size, prior climate shock experience, land fertility, market distance, farming experience, and information sources. Studies from Vietnam and Mali show socio-demographic factors (age, gender, education, experience) affect adoption. Evidence links CSA to higher yields and improved food security and suggests CSA can reduce agricultural emissions and carbon intensity. However, barriers include high input and maintenance costs, weak organizational capacity, asymmetric information, limited credit, input accessibility, water scarcity, market uncertainty, and insufficient extension services. The link between CSAT adoption and multidimensional poverty in hazard-prone coastal settings is limited in the literature, motivating this study’s focus on Bangladesh’s coastal farmers and technology-specific effects on MPI.

Study setting and sample: The study was conducted in Khulna district (southwestern coastal Bangladesh). Using multi-stage random sampling, three sub-districts (Batiaghata, Paikgachha, Dacope) were selected; from each, three Union Parishads (UPs) were randomly chosen. From each UP, 39 farm households were randomly sampled, yielding 351 households (exceeding the Cochran formula target n≈349). Inclusion criteria: ≥3 consecutive years of farming and ≥5 years of paddy farming; familiarity with conventional or modern farming. Data were collected via face-to-face structured interviews in May 2022 after a two-week pilot. Measures: CSAT adoption was captured across seven technologies: crop diversification (CD), agroforestry (AGRO), farming decisions based on weather forecasting (FDWF), on-farm diversification (OD), conservation tillage (CT), contingent crop planning (CCP), and rainwater harvesting/drip irrigation (RH/DI). Adoption responses were binary per technology; an aggregate CSAT adoption intensity classified households as insignificant (≤2), moderate (>2 and <5), or intensive (≥5) adopters. Moderate and intensive adopters were grouped as adopters for impact analysis. Crop vulnerability index: Constructed from seven climate-related indicators (drought, flood, storm/cyclone, pest attack, heavy/low rainfall, riverbank erosion, salinity) rated on a 1–5 Likert scale, aggregated into an index (mean ~0.662). Multidimensional Poverty Index (MPI): Following Alkire–Foster, three dimensions (education, health, standard of living) with nine indicators and pre-specified weights. Deprivation cut-offs were applied per indicator; a second cut-off identifies multidimensionally poor if weighted deprivations C_i ≥ k with k=3 (one-third of indicators). Continuous MPI scores were also analyzed. Analytical approach: To address endogeneity in CSAT adoption and selection bias, an Endogenous Switching Regression (ESR) model estimated via Full Information Maximum Likelihood (FIML) was employed. A selection equation modeled CSAT adoption; outcome equations modeled MPI for adopters and non-adopters. Counterfactual outcomes produced ATT and ATU. Robustness checks included maximum likelihood ESR with CSAT as an endogenous regressor on MPI, alternative MPI constructions (average deprivation score, equal weights), binary MPI (poor/non-poor), and technology-specific ESRs for each CSAT.

Descriptive statistics: Among 351 households, 83% are male-headed; mean household size ≈5.3; mean schooling of heads ≈5.9 years. Average monthly crop income ≈ BDT 5,217 (SD 2,333); total household income ≈ BDT 12,618. Mean paddy yield ≈ 6.37 t/ha; crop vulnerability index mean ≈ 0.662. Financial inclusion is limited: 47.6% have savings accounts; 54.7% access formal credit; 68.9% have loan burdens (mean BDT 24,681). Only 40.7% have access to extension services. CSAT adoption rates: 25.1% insignificant, 46.7% moderate, 28.2% intensive; moderate and intensive were merged as adopters. Determinants of CSAT adoption (ESR selection, FIML/ML): - Positive and significant: ln(crop income) (FIML: 0.587***; ML: 0.744***), access to extension services (FIML: 0.495***; ML: 0.537**), crop vulnerability index (FIML: 1.053**), paddy yield (FIML: 0.0675***; ML: 0.0879***), training on input management (FIML: 0.374**; ML: 0.480**). - Negative: access to formal credit (-0.260*, FIML selection). Other covariates generally insignificant. Impact of CSAT on MPI: - Counterfactual treatment effects (Table 4): ATT = -0.413*** (current adopters’ MPI 0.488 vs 0.901 if they had not adopted); ATU = -0.152*** (non-adopters’ MPI would fall from 0.588 to 0.436 if they adopted); overall treatment effect (adopters vs non-adopters) = -0.100***. - Alternative ESR (ML) confirms CSAT adoption reduces MPI by 0.183*** (Table 3). Athrho indicates endogeneity, supporting ESR use. Technology-specific impacts on MPI (Table 6): - Significant MPI reductions (percentage points): agroforestry (≈ -0.276***), on-farm diversification (≈ -0.239***), conservation tillage (≈ -0.328***), contingent crop planning (≈ -0.35***), rainwater harvesting/drip irrigation (≈ -0.35***). - Positive associations (higher MPI): crop diversification (+0.277***), farming decisions based on weather forecasts (+0.203***). Additional associations in MPI equations: Savings account ownership is consistently associated with lower MPI across models; higher ln(crop income) often reduces MPI. Robustness checks with alternative MPI constructions and binary MPI yield consistent results: adopters are significantly less likely to be multidimensionally poor.

Findings support the hypothesis that CSAT adoption significantly reduces multidimensional poverty among coastal Bangladeshi farm households. Adoption is most likely among households with higher crop income, greater crop vulnerability, better paddy yields, access to extension services, and training on input management, indicating both capacity and need-based drivers. Counterfactual analyses show large poverty reductions for adopters relative to non-adoption scenarios and meaningful potential gains for non-adopters if they adopt. Technology-wise results suggest targeting agroforestry, on-farm diversification, conservation tillage, contingent crop planning, and rainwater harvesting/drip irrigation yields substantial poverty reductions, likely via resilience and productivity channels. Conversely, crop diversification and weather-forecast-based decisions correlate with higher MPI, plausibly due to small landholdings limiting economies of scale in diversification and incomplete access to timely, reliable forecasts (smartphone/internet gaps), underscoring the need for context-appropriate guidance and information delivery. The results emphasize strengthening extension services and input-management training to scale CSAT adoption, and tailoring interventions (e.g., salt- or flood-tolerant seeds, water-saving irrigation) to local climate risks to maximize poverty impacts.

Climate-smart agriculture technologies provide a viable pathway to reduce multidimensional poverty among climate-exposed coastal farmers in Bangladesh. Using ESR with FIML, the study shows substantial reductions in MPI among adopters and meaningful prospective gains for non-adopters. Technology-specific evidence highlights agroforestry, on-farm diversification, conservation tillage, contingent crop planning, and rainwater harvesting/drip irrigation as particularly effective. Policy implications include scaling access to extension services, expanding training on input management, and deploying vulnerability-specific CSAT packages (e.g., tolerant seed varieties, water-smart irrigation) to bolster resilience and livelihoods. These strategies align with SDG1 and SDG2 and can help transition poor farm households toward sustained, climate-resilient livelihoods.

External validity is limited to the study area and cross-sectional design; results should be re-examined with larger samples across diverse geographies. Longitudinal and experimental designs would strengthen causal inference. Technology-specific anomalies (e.g., CD and FDWF positively associated with MPI) warrant deeper investigation, potentially addressing constraints like small landholding-induced scale effects and limited, timely access to forecast information. General constraints in data (e.g., reliance on self-reports) and context-specific factors (extension access, digital connectivity) may affect generalizability.