Page 53 - AJWEP-22-4

Impact of cyclones on rice farming

3.2. Sampling and data technique of rice seed used (kg); X is the quantity of chemical

4

The sampling strategy targeted rice-producing farmers fertilizer (kg); X is the quantity ofpesticide (kg); D is the
5

who were severely affected by the cyclones. A total cyclone dummy variable (1 = cyclone-affected farmers,
of 400 cyclone-affected farmers were selected using a 0 = non-affected farmers); γ captures the marginal effect
simple random sampling technique—200 each from five of cyclone exposure on rice production; β , β ,…, β
5
0
1
unions in Shyamnagar and Koyra upazilas. Similarly, are the output elasticities of the respective inputs; and
400 farmers were randomly selected from the ten is the error term. The model assumes no perfect

unions in the control upazilas, Dumuria and Kalaroa. multicollinearity. Robust standard errors are applied
This selection ensured the inclusion of respondents to correct heteroskedasticity, and autocorrelation is
with relevant experience and knowledge of agricultural assumed to be absent.
practices and challenges in both cyclone-affected and
unaffected regions. 3.3.2. Event study with panel fixed effects regression
Data were collected through face-to-face, model
in-depth interviews using a pre-tested semi- An event study approach, using a fixed effects panel
structured questionnaire designed to capture relevant regression, was used to estimate the impact of cyclones
information systematically. The interviews were on rice production across the three periods: before,
further supplemented by focus group discussions. In during, and after the cyclone. The purpose of the model
addition, key informant interviews were conducted is to capture the dynamic effects of cyclones on rice
with local leaders, agricultural experts, and government production over time. The model, modified to include
officials. These interviews provided expert opinions and lagged variables, is expressed in Equation III:
contextual information, offering critical insights that Y = α + δ D + δ D + βX + ε (III)
influence farmers’ adaptation strategies. A case study it i 1 duringt 2 aftert it it
th
approach was also employed to provide an in-depth where Y is the rice output (maunds) for the i
it
analysis of the impacts and adaptive responses of the
most affected unions. All data collection activities were farmer at time t; αi is the farmer-specific fixed effect
(e.g., education, soil quality, market access); D
performed between March and August 2023. during
is the dummy variable = 1 during the cyclone, 0
otherwise; D is the dummy variable = 1 after the
3.3. Empirical methods after
3.3.1. Cobb-Douglas production function cyclone, 0 otherwise; D before is the base period (omitted);
The Cobb-Douglas production function is a widely X is the vector of control variables (e.g., land, labor,
it
used model in empirical research to analyze production seed, fertilizer, pesticide); is the idiosyncratic error
it
term (i.i.d). In order to adjust for heteroskedasticity and
efficiency. The general form of the function is expressed autocorrelation, clustered standard errors at the farmer
in Equation I.
level were used. 47
β
Y = A X⋅ 1i β 1 ⋅ X 2i 2 ⋅… . X⋅ ki β k e ⋅ ∈i (I) 3.3.3. Relative cyclone loss (RCL) model
i
where Y is the output of rice production for the i RCL measures the proportion of a household’s annual
th
i
farmer; X , X ., X represents input variables (e.g., land, income lost due to cyclone-induced damage. It provides
1i
2i
ki
labor, seeds, fertilizers); β , β ..., β are the output a normalized metric to compare impacts across
2
1
k
elasticities of the respective inputs; and ϵ is the error households of varying income levels, with higher
i
48
term. RCL values reflecting greater financial vulnerability.
To facilitate estimation, the function is transformed Equation IV shows the formula for estimating RCL.
into a linear logarithmic form, as shown in Equation II. Cyclone loss
RCL = (IV)
InY = β + β In X + β In X + β In X + β In X + β Annualaverage income
5
2i
0
3
1
1i
2
i
4i
4
3i
In X + γD + ε (II)
5i i I
To identify the determinants of RCL, a multiple
where Y is the output measured in maunds linear regression model was estimated (Equation V).
i
(1 maund = 40 kg) of the i rice farmers; X is the It normalizes the disaster loss, making it possible to
th
1
land size in acres (1 acre = 0.4047 hectares); X is the compare the impact across households of varying
2
human labor measured in man-days; X is the quantity economic scales.
3
Volume 22 Issue 4 (2025) 45 doi: 10.36922/AJWEP025100063