Introduction

Agricultural farm income is extremely variable, and household welfare and agricultural production can be affected by farm income (Key et al. 2017). In this context, extant literature states that Agriculture is the key economic sector of Bangladesh. Much of the country’s rural population are reliant on agriculture as a critical source of livelihood and employment. Many farm households in Bangladesh continue to earn low and negative net farm incomes and cannot stop farming due to financial difficulties and lack of alternative opportunities. Farmers have been experiencing increasingly adverse effects on agricultural production and livelihoods due to environmental stress (Kabir et al. 2017). However, many farm families have continued the occupation in Bangladesh; besides, they are also involved in other off farm income sources, since employment, through livestock and fisheries are the top choices for farm family members. Farmers of different areas in Bangladesh face diverse issues regarding agricultural production. The country is extremely vulnerable to severe climate events; floods, tidal waves, and droughts are frequent and damaging (Karim 1996; Yu et al. 2010; IFPRI 2013). Climate change is the recognized input factor that affects the agricultural production in different divisions of this country (Sikder and Xiaoying 2014). The relatively low-lying areas are more frequently submerged (Fafchamps 2003) during floods, which affects farm income. Bangladesh forms the largest delta of the world by the merging of three major rivers—the Ganges, Brahmaputra, and Meghna. Consequently, the country faces massive floods and heavy monsoon rains; in general, 55–60% of the country is inundated during extreme floods (BDP 2100 2018). Every year, from June to October, around 50% of the country’s low-lying lands remain inundated by flood (Banglapedia 2003). Recent evidence reveals that, as a consequence of climate change, the magnitude and frequency of mega floods are increasing (BDP 2100 2018). Mymensingh is the central division of Bangladesh, and it faces monsoon flood, flash flood in haor (haor lies below the northeastern Himalayan foothills and it experiences the highest rainfall in the world, which makes it a flashflood-prone basin) areas, and heavy rains, each year (Sarker et al. 2014). The southern, north-central, and north-western regions of Bangladesh face the most impactful climate change factors—coastal areas are particularly vulnerable to river floods and heat stress; the Mymensingh division experiences river flood almost every year (Ruane et al. 2013). This division has highest upper poverty rate in Bangladesh (Poverty and inequality in Bangladesh 2015), and its farmers were worst hit because they face the largest declines in rice production due to climate variability affects (Yu et al. 2010).

Although literature shows that this division’s farm income is vulnerable to climate change, there has been no attempt to study regional farm income variation (Alamgir et al. 2018), and limited number of studies have been carried out on the effects of climatic variability on farm income. In addition, appropriate policy, along with suitable adaptation technologies, are required to support farm families who are facing income variability. Thus, it is necessary to understand the income distribution and income inequality of farmers and composition of income sources. Moreover, studying the potential impact of climate changes on the earnings from agricultural sources is important because it threatens to escalate the magnitude of poverty among farm families. Thus, the principal objectives of this paper are to assess the contributions of income sources to the incomes of the households and to determine the distributional implications of income earned from different sources for four districts of Mymensingh division. This paper also examines the income inequality and similarities of income sources among the farmers and projects the potential poverty from the perspective of income distribution under climate change impact.

This paper is organized as follows. It presents a literature review in section two. This part reviews the literature on climate change and agriculture. The third part contains study area and survey data, wherein the authors discuss the study sites and sources of rural farm household data in Mymensingh division. Section four address the empirical methods used in this study. The fifth part is results and discussions section, where empirical research with relevant data is classified into four districts of Mymensingh division; then, it compares the results of per capita income differences, income inequality, similar income sources group, cost and income disparity of crop production, and poverty of rural farm households. The final section includes conclusions and policy implications.

Literature review of climate change impact on agriculture nexus

Climate change is a serious risk factor for agricultural production, and this varies for crop yields in different regions. The literature on climate change and agriculture is outlined concisely in this section. Climate change’s effects on agriculture is unevenly distributed across the world (IPCC 2014). Agriculture is highly sensitive to climate change; however, climate changes’ impact is widely understood to be key factor that increases the vulnerability of agricultural production (Hossain et al. 2013; Ruane et al. 2013; Yu et al. 2010; IFPRI 2013). It could affect agricultural productivity, in terms of quantity and quality of crops. Moreover, peoples from many parts of the world are already suffering from food insecurity due to climate change, and it is predicted that global crop production will decline by 2–6% per decade (Smith et al. 2014). The varying latitude and irrigation practices cause opposite effects in regions; while in some regions climate change impact increases the crop yield because of these factors, in others it decreases yield (Popova and Kercheva 2005). To ensure sustainable development and reduce the vulnerabilities of the rural poor, adaptation is necessary in face of constant climate change (Munasinghe 2010; African Development Bank et al. 2002). Agricultural productivity is climate sensitive and the fluctuations in temperatures and rainfall patterns adversely affect the food grain productivity (Kumar and Sharma 2013). Climatic conditions, including temperature and precipitation, may affect the agricultural production and food prices (Hertel et al. 2010; Ahmed et al. 2009; Ahmed et al. 2011), and climatic variability may worsen the poverty in less developed countries (Leichenko and Silva 2014). Many of the previous studies has indicated the significance of climate change impact on agriculture and suitable adaptation strategies by farmers (Masutomi et al. 2009; Yu et al. 2010; IFPRI 2013; Alauddin and Sarker 2014; Brulle et al. 2012; and Rashid et al. 2014). Besides, some studies also project the climate change impacts on crop agriculture and food production (UNDP 2009; Porter et al. 2014; Kobayashi and Furuya 2011; Salam et al. 2016; Alamgir et al. 2018).

Bangladesh is a predominately an agrarian country where extreme weather-related disasters like flood and drought is common, and the magnitude and frequency of these disasters are increased by the variation in climatic conditions (Parry et al. 2007; IPCC 2012; IFPRI 2013). The Asian Development Bank (ADB) and International Food Policy Research Institute (IFPRI) studied the climate change impact on the agriculture of some Asian and Pacific countries, including Bangladesh, and showed that these countries are most vulnerable to climate change (ADB and IFPRI 2009). Agriculture is vulnerable in Bangladesh due to the unfavorable climate condition; however, the agriculture sector involves 70% of labor forces and accounts for around 35% of the total GDP contributions in Bangladesh’s economy (Basak et al. 2009, 2010). The sector faces increasing food demands due to considerable pressure by the population and reduced agricultural land (Ahmed and Ryosuke 2000; Israt et al. 2016). Additionally, climate change is a potential threat to food grain production, particularly rice production (Basak et al. 2009, 2010). The International Food Policy Research Institute report (IFPRI 2012) noted that rice is a staple food and the most important crop in Bangladesh’s agriculture, covering 73.94% of the country’s gross cropped area (Banglapedia 2003). Rice production is an important indicator of Bangladesh’s economy and is essential for the social and political stability of the country (Nath 2015). During the post-independence period, Bangladesh faced severe food shortage; however, the country focused on food production, increasing rice production with new varieties of rice and implementing irrigation technology as well. After the Liberation War and until now, some research institutes, specifically the Bangladesh Rice Research Institute (BRRI), engaged in adaptive research and the development of new rice varieties that suit the agroecological condition of Bangladesh, in partnership with the International Rice Research Institute (IRRI) (Israt et al. 2016). According to the crop calendar of Bangladesh, there are three rice seasons: the aus (former rainy season rice), aman (wet season rice), and boro (dry season irrigated rice). Aus rice is planted in March–April and harvested during June–July; aman rice is planted in June–July and harvested during November–December, and boro rice is planted in December–January and harvested during mid-April–June (BBS 2018a). Since the inception of BRRI in 1973, a number of smart rice varieties of aus, aman, and boro under the name BR, and later BRRI dhan, have been developed and released for different environments and seasons in Bangladesh, which certainly helped in increasing rice production (Kabir et al. 2015). More than 1000 traditional rice cultivars, including modern high-yielding varieties developed by BRRI-69 cultivars, Bangladesh Agricultural University (BAU)-2 cultivars, and Bangladesh Institute of Nuclear Agriculture (BINA)-17 cultivars, are now being grown in Bangladesh (Hossain and Jaim 2009). In the latest 2017–18 season, the total rice production in Bangladesh was around 36.28 Million Tons (MT), which was thrice that of the 1971–72 period, when the production was only 11 Million Tons (BBS 2018b). Although rice production reached a secure level in Bangladesh, increasing climate variability poses a great challenge for future rice production and for ensuring food security in Bangladesh (Brolley 2015; Hossain and Silva 2013). Rimi et al. (2009) studied the recent trend of climate change in Bangladesh, and predicted the future scenarios by investigating the impacts of climate change on rice production. This study mentioned that under calcareous soil condition, boro rice production was reduced from 3.47 to 48.64 percentage points. Sarker et al. (2012) examined the relationship between yield of rice crops and climate variables for Bangladesh and confirmed that the boro rice yield has been adversely affected. The Intergovernmental Panel on Climate Change (IPCC 2007) estimates rice production’s decline in Bangladesh due to the changing climate change’s impact, such as changing rainfall pattern, flooding, drought, and salinity. Islam et al. (2011) checked the climate variation on farming system in Bangladesh and concluded that climate variability will result in significant yield reduction for the agriculture sector, in the future. Yu et al. (2010) estimated the climate change risk and food security in Bangladesh and mentioned their studies that rice is the dominant crop and is extremely vulnerable to climate change and extreme climate events, such as floods and droughts; besides, this study also projected that an average of 7.4% overall rice production in Bangladesh will reduce every year, over the period 2005–2050. The high dependency on weather causes the farm income to be unstable; the total annual income of a farm household depends on farm and non-farm income (Chaudhry and Wimer 2016). Unpredictable weather can easily damage crop production, which creates vulnerability and uncertainty in farm income (Key et al. 2017).

From the above review, it should be noted that most of the studies on climate change impacts on agricultural production have been on a national level. In addition, in-depth empirical research on farm income, inequality, poverty, and climate change impact has been deficient in Mymensingh division. Therefore, to consider various adaptation strategies, it is necessary to study the farm income and climate change vulnerability in root level farming in Bangladesh. Moreover, an inclusive study of farm income and climate change impacts on farm production in the districts of Mymensingh division is extremely significant. Thus, the current study intends to bridge the literature gap by describing the root level farm household income scenarios with the application of different empirical methods, projecting the farm income and future poverty, if extreme climatic events occur and affect farm production, thereby identifying the vulnerable districts, and suggesting appropriate coping strategies.

Description of the study sites and survey data

Outline of the study sites

The north-middle part of Bangladesh, called the Mymensingh division, was chosen for the study, and it consists of four districts named Jamalpur, Mymensingh, Netrokona, and Sherpur that represent the major agriculture production areas of the country (Fig. 1). This division was formed in 2015, from districts formerly comprising the northern part of Dhaka Division. It is the eighth administrative division of Bangladesh comprising a total area of 10,668.8 square kilometers and a population of 10,990,913, as of the latest 2011 census (BBS 2018a). It is located between 24°15′ and 25°12′ north latitudes and 90°04′ and 90°49′ east longitudes and is predominantly composed of low-lying areas sketched below the northeastern Himalayan foothills (Garo Hills and Meghalaya state of India), in the large delta formed by the Brahmaputra River (Banglapedia 2003).

Fig. 1
figure 1

The target districts’ map in Mymensingh division where the primary sampling units have been plotted

Full size image

Mymensingh division primarily encompasses a flat area, subject to riverine flooding and vulnerable to flash floods, almost every year. Typically, a region that is close to a major river possesses a greater risk of floods (Akter et al. 2016). Floods occur on a transnational level in Bangladesh, in the sense that river systems run across countries like China and India and then flow into Bangladesh. A large portion of Bangladesh’s economy is based on agriculture, and flooding presents a short-term problem—a natural disaster, such as a flood, damages crops and plants. The destruction of crops and loss of land value in the floodplain make people economically vulnerable. Additionally, a shortage of food supplies can increase food prices in the country during floods. Agriculture is the broad economic sector in Mymensingh division and contributes to the national GDP, despite the substantial loss of crop due to flash flood (CPD 2018). Extensive flooding merely damages the infrastructure, and a larger section of the rural population is displaced by the increasing water levels and inundation of land, which further enhances poverty (about 26.4% of rural areas) (BBS 2017). There is no defensive infrastructure to protect agricultural crops against flooding, unlike those present in coastal areas against cyclones; however, floods affect people who use public schools and high roads, and city areas (mainly of Jamalpur town and Mymensingh town) have been shielded by heavy constructions besides the Brahmaputra river. Although the four districts of the Mymensingh division lie on the same delta formed by the Brahmaputra River, there are some socio-economic and environmental differences among the districts.

Jamalpur: Jamalpur district is located between 24°34′ and 25°26′ north latitudes and 89°40′ and 90°12′ east longitudes (Fig. 2). The district is well connected by road, rail, and river with the capital city of Dhaka and with the rest of the country. The Jamuna and old Brahmaputra rivers flow through this district; therefore, flood frequency is greater, and the area comprises all the hotspots in the river system and estuaries (BDP 2100 2018). Due to the low literacy rate and education (31.8%), agriculture is the primary source of income. In addition, the economy also comprises non-agricultural labor, industry, commerce, transport and communication service, construction, religious services, and other sources (Banglapedia 2003). Local rice, sugarcane, jute, tobacco, vegetables, mustard crops etc. grow very well in this district. Traditionally, nakshi kantha (embroidered quilts) making is a vital occupation of the rural residents in Jamalpur.

Fig. 2
figure 2

The four districts of Mymensingh division used in this study (Banglapedia 2003)

Full size image

Sherpur: It is located between 24°18′ and 25°18′ north latitudes and in between 89°53′ and 90°91′ east longitudes (Fig. 2). According to the Bangladesh delta plan 2100, this district is relatively less prone to hazard (BDP 2100 2018). The average literacy rate is 31.98%: the literacy rates for men and women are 29.51% and 51%, respectively. However, the economy of this district is agro based and agriculture takes the leading position among the income sources; 65.71% of total income comes from this sector, although non-farm income activities performing a substantial share (Banglapedia 2003). Since the annual average temperature of this district varies from a maximum of 33.3 °C to a minimum of 12 °C, farmers here grow local and high-yielding varieties of rice, jute, potato, pulses, vegetables, and fruits. Besides crops, livestock and fishery are important sources of household income in Sherpur as well.

Mymensingh: This district is located between 24°15′ and 25°12′ north latitudes and 90°04′ and 90°49′ east longitudes (Fig. 2). In addition to that, the district headquarters are very close to the capital city of Dhaka and only 121 km north, which is connected by road, rail, and river and the rest of the country as well. The primary sources of income include: agriculture 64.14%, non-agricultural laborers 3.65%, industry 0.82%, commerce 11.40%, transport and communication 3.53%, service 6.21%, construction 1.33%, religious service 0.21%, rent and remittance 0.46%, and others 8.25%. Average literacy rate is 39.1%, consisting of 41.7% and 36.3% for men and women, respectively (Banglapedia 2003). Besides, in many parts of this district, farmers reformed their common farming practices, changed paddy fields to ponds, and are cultivating fish. The climate of Mymensingh is moderate and the monsoon starts in May or June and continues until August. During the monsoon, temperatures vary between 15 and 20 degrees. The maximum temperature (about 40 °C) is felt in the April–May period, while lowest temperature falls below 15 °C in December–January.

Netrokona: Netrokona district is located between 24°34′ and 25°12′ north latitudes and 90°00′ and 91°07′ east longitudes (Fig. 2). This district is also connected with the capital city and other parts of the country by rail and road. However, some parts are isolated and traveling by boat is the only way of transportation. The Bangladesh delta plan 2100 mentioned that the district is the hotspot of haor and flash floods (BDP 2100 2018). Average literacy rate of Netrokona is 34.9%, consisting of 37.9% and 31.9% for men and women, respectively. Around 72.43% of total income in the district comes from the agriculture sector. The primary sources of income include non-agricultural laborer 3.40%, industry 0.53%, commerce 10%, transport and communication 2.11%, service 3.92%, construction 0.91%, religious service 0.26%, rent and remittance 0.23%, and others 6.21% (Banglapedia 2003).

Sources of data

To achieve the objectives of this study, we used cross-sectional data from this study site and performed an empirical analysis. Farm household’s plot wise data were obtained from International Food Policy Research Institute (IFPRI 2013) that conducted Bangladesh Integrated Household Survey for the country during the 2011–2012 period, with multi-stage stratified random sampling method. To fill knowledge gaps on critical food security and agricultural development issues, IFPRI conducted a comprehensive and nationally representative survey in Bangladesh. The main aim of this survey was to provide policy options and advisory services to decision makers and stakeholders, collaborate with national institutions to strengthen analytical capacity within the country, and stimulate policy dialogue. For this survey, IFPRI followed stratified sampling in two stages: first, villages (primary sampling unit) and then households within each village were selected, from the 2001 population census data of Bangladesh. The survey covered the rural areas of the entire country. In the first stage of sampling, 325 villages were allocated among the eight divisions of Bangladesh with probability proportional to size (size being the number of households in each division). In the second stage, 20 farm households were randomly selected from each village in the 2001 population census list of Bangladesh. A total of 6503 sample farmers were selected across Bangladesh and plot-wise crop production data were collected via semi-structured questionnaire by the IFPRI, from the survey period of December 1, 2010, to November 30, 2011. From the 2011–12 dataset of IFPRI, we chose the Mymensingh division, where the distribution of primary sampling unit was 30 villages, and selected 20 farms from each village that contained a total of 600 sample farmers; this was used as the data source for the current study. The summary statistics for the data are presented in Table 1. These plot-wise data were then converted as per household income from all types of sources, including each crop income. Aggregate farm income data by each household for the same period (2011–12) was obtained from each income source. According to rice production statistics in Bangladesh, 2010 and 2011 registered no severe crop loss, which reflects the typical agricultural year for the data collection of this study in the Mymensingh division (BBS 2015).

Table 1 Summary statistics for the sampled farms
Full size table

Data compilation

According to the objectives of our study, we compiled the data set. Then, we selected the quantitative variables. The data were entered into MS Excel and analyzed using R (version 10; 1999), and the study area map prepared by ArcGIS 10.4R (Vector layers were projected in WGS 1984 UTM zone 45 N) and the data source was 2020, LGED, Bangladesh. The quantitative parameters included family size, age of household head, education year, total farm size, production, cost and income from each crop, and income from different major sources. For the study’s economic analysis of data, we treated all the farms equally and placed a rental equivalent value for owned (included mortgaged) land for each farm (Paul 2011). To ensure that no double counting of costs is incurred in the account, interest charges on borrowed capital are excluded.

Empirical methods

We used descriptive statistics, correlations, and multivariate statistical methods for our study. The correlations between income sources and costs of inputs from crop production were analyzed. For the purpose of the study, statistical tests are undertaken across the districts to test the mean difference so that there is no difference in the per capita income of farms households (ANOVA). In addition, Gini coefficient for income inequality, Cluster analysis for similar groups, decomposition of variance for inter-constituency differences, and lognormal distribution for poverty measurement were used.

Analysis of variance (ANOVA)

To examine the differences among the farm households in terms of mean per-capita income, this study conducted a single-factor analysis of variance (ANOVA) among the districts of Mymensingh division.

Gini coefficient

To examine the income inequality among the districts we used Gini coefficient. The Gini coefficient is the most used measure of inequality of a distribution in economics; it is also used to measure income inequality. Corrado Gini, an Italian statistician, developed Gini coefficient and it was published in 1912, in the paper Variability and Mutability (Bellu and Liberati 2006). It is also defined as a ratio of the areas on the Lorenz curve. If the area between the perfect equality line and Lorenz curve is A, and the area under the Lorenz curve is B, then the Gini coefficient is,

$$G = \frac{A}{A + B}$$
(1)

For estimating the Gini ratio, we used the following equation,

$$G = \frac{{\sum\nolimits_{i = 1}^{n} {\left( {2i - n - 1} \right)x_{i} } }}{{n^{2} \mu }}$$
(2)

where i = 1, …, n individuals (ascending order), xi = income for individuals (x1 ≤ x2 ≤ ······ ≤ xn), n = Total number of individuals, and μ = mean income.

Cluster analysis

Cluster analysis has been widely used in many applications including social sciences. The aim of the cluster analysis is to find similar groups unveiling parallel income source patterns and similarity between each pair of income source patterns. In this study, it was used to determine the similar districts based on the primary and agricultural income sources in four districts of Mymensingh division. With this analysis, it is easy to identify the homogenous groups of districts in terms of previously unknown income sources. To determine the most appropriate clusters regarding the income sources in each district, Ward’s hierarchical method, and the partitioning method is used (Murtagh and Legendre 2014). The results are usually presented in a tree diagram or dendrogram of the hierarchy of nested cluster explanations, which is a manifestation of Ward’s method, and it provides clues for finding the preferable number of clusters, regarding income sources.

Decomposition of variances

To understand the main reasons for the inequalities among the districts, we decompose the variance of net cost and net income to the observed total variance of input cost and income. The per hectare crop yields, prices, and the costs at the farm level are included in this decomposition. We used the following relations.

$${\text{V}}\left( {P \pm Q} \right) = {\text{V}}(P) + {\text{V}}(Q) \pm 2{\text{Cov}}(P,Q)$$
(3)

where P and Q are stochastic variables, such as the costs of inputs or incomes from different sectors; V () is variance and Cov () is covariance.

Projection of income distribution

Projections can be useful when they are used for benchmarking or policy simulations. This study project the future poverty by using the lognormal distribution. Lognormal distribution is widely used for expression of income distribution, and it has certain similarities to the normal distribution. A random variable is lognormally distributed if the logarithm of the random variable is normally distributed. To find the future income distribution we calculated projected income of farm households by using this distribution. The lognormal distribution is a 2-parameter distribution with parameters mean (μ) and standard deviation (σ). If x is distributed log-normally with these parameters, then log(x) is distributed normally with μ and σ. The lognormal distribution is applicable when the quantity of interest is positive, since log(x) exists only when x is positive.

The probability density function of the lognormal distribution is

$$f\left( {x|\mu ,\sigma } \right) = \frac{1}{{x\sigma \sqrt {2\pi } }}exp\left\{ {\frac{{ - \left( {\ln x - \mu } \right)^{2} }}{{2\sigma^{2} }}} \right\};\quad x > 0$$
(4)

If we substitute a poverty line into x and integrate the probability density function up to x, we can obtain a poverty rate. We used the poverty line, which was estimated by the World Bank (Li et al. 2016; Poverty and Inequality in Bangladesh 2015). From all the sample families, we estimated the per capita income (BDT) and, considering the assumption of rice crop loss by flood, we drew the distribution of the estimated income as if it followed lognormal distribution. The mean and standard deviation of ln(x) was essential to draw lognormal distribution. Meanwhile, we obtained the actual distribution of per capita income by using the lognormal distribution from the actual per capita income of household members in the study areas. Following, we projected the rice yield loss due to flooding from the literature reviews’ assumption and estimated the projected per capita income. By using lognormal distribution on it, we obtained the estimated distribution of per capita income and finally obtained the poverty rate graph by simulating these two distributions.

Results and discussion

Socioeconomic status of farm families among the districts

Table 1 summarized the farm households’ characteristics in different districts of Mymensingh division. Of the 600 farms households, 120 farms were under the Jamalpur, 260 farms in Mymensingh, 120 farms in Netrokona, and 100 under the Sherpur district. Education is a key indicator of a society’s development; however, we found that in the study areas, most household heads have not attended schools. In addition, Jamalpur district faces the worst situation in terms of education. The percentage of household heads with an average of nine years of schooling in Sherpur district was 17%, which implies that the sampled farmers had secondary level of education (Alamgir et al. 2017). However, the educational status of this area is a little better than that of the other districts. Farm size is an important dimension for comparing farm income variability (Key et al. 2017). In Bangladesh, most agricultural farms are small scale and farm household obtain most of their income from off-farm sources. The average farm size in Jamalpur was 0.17 ha, which is marginally lower than the regional average (0.35 ha) (Table 1).

The total household income combines income from on-farm and off-farm sources of each household (Key et al. 2017). Farm income is the sum of all household members’ net income in the data collection year; off-farm income comes from earned and unearned sources.

Table 2 shows the average household income of different sources in four districts of Bangladesh. The data analysis reveals that the farmers of Mymensingh district earned higher average per-capita income whereas that in Sherpur earned lower average per-capita income. This finding shows that the larger farm size (Table 1), which is belongs to the Mymensingh district has the higher average income compared with other districts.

Table 2 Household income (BDT/year and %) from various sources, by district
Full size table

Employment is the leading income source in most districts, followed by nonagricultural enterprise income and agricultural crops. Income from owned businesses that are not agricultural or self-employed, formal and informal employment, and income received from relatives and friends not presently living within the household are included in the employment income. Similarly, in nonagricultural enterprise, income consists of the income from fishing, nonagricultural day labor, construction, manufacturing, nurseries, food processing, retail, wholesale, wooden furniture, and other businesses. Land rent or property rent, life and non-life insurance income, shares, gratuities, lotteries or prizes, bank interest, cash or in-kind receipts for charity are included in the other income category. The share of agriculture in the total income varies by district. Among farming households, the agricultural crop share is highest in the Mymensingh district (25.38%).

Table 3 shows the percentage of agricultural crop share for the different crops in four districts. The results show that the total rice share is than 70% point in Jamalpur and Netrokona districts, while boro (dry season irrigated) rice share is more than 50% point compared with the aman (wet season rain fed) rice. This result is consistent with previous reports where it is mentioned that aman and boro rice are the primary crops in Bangladesh (Banglapedia 2003; BBS 2015; BER 2017). Similarly, rice share in Sherpur district is slightly lower at 66% point. However, other crops share in Mymensingh district is 51% point in total net agricultural income. This result shows that rice is the key crop for the income sources of farmers, in this district.

Table 3 Each agricultural crop’s share in total net agricultural income (%), by district
Full size table

Table 4 presents the mean, median, standard deviation of per capita income, and poverty rate as well. The results show that, among the districts, the Mymensingh district has the highest average per capita income and standard deviation. Ferreira et al. (2015) studied global poverty based on estimations from 1990 to 2012, by using the 2011 purchasing power parity (PPP) and the international poverty line. This study mentioned that the international poverty line in US dollars was $1.90 a day in the 2011 PPP (PovcalNet 2020). According to the World Bank and International Monetary Fund, the purchasing power parity in 2011 for BDT was 23.145. The poverty line (BDT 16,051.05) was applied to the yearly per-capita income for estimating poverty rates. For validation of this income difference in four districts, we perform ANOVA and report the results in Table 5.

Table 4 Mean, median, and standard deviation of per-capita income (BDT/year), by district
Full size table
Table 5 ANOVA mean differences across districts
Full size table

Table 5 displays the mean difference across districts by ANOVA. The results of this table show that there is a significant per capita income difference in farm households of Mymensingh division’s districts.

Inequality of income by districts

To understand the income inequality in the districts of Mymensingh division, we measure the Gini coefficient from household per capita income; the results are shown in Fig. 3.

Fig. 3
figure 3

Graphical representation of Gini coefficient by districts

Full size image

The income inequality is derived from the household per capita income in the districts of Mymensingh division. It shows that the lowest inequality exists in Sherpur district and the highest in Netrokona. It implies that the income inequality is lower for those areas where the agricultural income contribution is higher.

The Gini ratios are shown in Fig. 3, where the blue color (Gini ration is 0.42) indicates that Sherpur district has the lowest inequality and the orange color (Gini ration 0.47) indicates that Netrokona district has the highest inequality among the districts of Mymensingh division. Flash floods experienced by Netrokona district may be a possible reason for its high inequality compare with others. During the boro rice (dry season irrigated rice) harvesting season in Bangladesh, some northeastern districts are affected by flash floods, almost every year, and Netrokona is one of the affected these districts. This unexpected sudden flood, damages the boro rice, reduces the farmer’s income, and increases the income inequality.

Comparison of income sources among the districts

This section classifies the districts of Mymensingh division by their income sources to compare the characteristics among these districts. Household income sources from Table 2 are analyzed by cluster analysis, which are shown in Fig. 4.

Fig. 4
figure 4

Dendrogram of household income sources, by districts

Full size image

To illustrate the group of districts based on the primary income sources for households, we use dendrogram (Fig. 4); Table 6 also shows the results of a similar analysis. Agricultural crops, non-agricultural profit, and employment are the chief income sources, where employment is dominant source in most of the districts.

Table 6 Household income sources, by districts
Full size table

The household income sources share of agriculture from Table 3 are analyzed by cluster analysis and presented in Table 7, which is also represented in Fig. 5. Figure 5 and Table 7 show that rice is dominant crop for agricultural income in Jamalpur and Sherpur; however, other crops are dominant in Mymensingh district. This result implies that crop diversification exists in Mymensingh district, which helps in achieving higher agricultural income.

Table 7 Agricultural income sources, by districts
Full size table
Fig. 5
figure 5

Dendrogram of agricultural income sources, by districts

Full size image

Inequality of income by sources in different districts

This section tries to explain the income inequality by sources and illustrates the alternative income sources on total household income in different districts of Mymensingh division. To check the inequality of income, we calculate the Gini ratio for total farm family income from selected sources, which is graphically represented through the Lorenz curve in Fig. 6.

Fig. 6
figure 6

Graphical representation of Gini coefficient by sources of income in total division and other districts

Full size image

From Fig. 6, we can see that the total farm income was decomposed into seven sources of income and the inequality of these sources. We found the lowest inequality in employment sources, where the Gini coefficient is 0.59 in total Division and 0.94 (the highest) in Remittance. It shows the contribution of higher remittance income in total farm income in those areas, which induced higher income inequality. Overseas employment and remittances play an important role for the economic development of Bangladesh, reducing unemployment and poverty, and increasing the foreign currency reserve as well. The amount of remittances in Bangladesh varies regionally and nationally by considering the main income source from remittance in Sylhet (BER 2018). However, some other parts of the country such as Comilla and Chittagong are also familiar with foreign remittance. Since the context of national remittance varies, the Mymensingh division did not acquire a significant position. However, many resident members from this division send remittance to the country, although many families cannot afford to send a family member to another country.

From Table 2, remittance share in the Jamalpur district is highest among all the districts in the Mymensingh division. It confirms that more family members within this district live away from the country and send remittance to their family. Nevertheless, when we analyze inequality among the sources of income in each district, remittance inequality is significant due to remittance flow by family members. Skilled workers living in Western countries and their remittance flow is higher than those living in the Middle East and other countries (BER 2018). Similar results were also found in other districts, where the lowest income inequality was seen in employment and preceded by agriculture (Fig. 6). These results indicate that inequality exists in the total farm income among the farmers in the Mymensingh division, suggesting that agriculture and employment income can reduce the inequality of farm income.

Total household income distribution in the districts

To understand the diversity of income from different sources in each district, we applied decomposition of variances. The decomposed variances share was derived from annual per capita income from different income source sectors. Table 8 represent the decomposition of variance in farm household income and shows that majority of the variability in farm household income originates from employment income in different districts.

Table 8 Variance decomposition of household income share (%) by sources
Full size table

However, agriculture plays an important role in the income variability of farm income in Mymensingh division. This is normal because off-farm income contributes significantly more to the annual household income. Across Mymensingh division, differences in remittances, other income, and employment are important factors that contribute to income differences. If a family can find good employment inside and outside its districts, it can become relatively rich. We found from the Table 8 that in Mymensingh district agriculture is one of the main contributors to income differences. This result also denotes that remittance is the most important sector to induce income disparity in Jamalpur and employment in Netrokona and Sherpur.

Agricultural income distribution in the districts

The main elements of agricultural income differences, shown in Table 9, were obtained by the decomposed variance method. From Table 6 and 7, we identified that agriculture is one of the main reasons for income differences in Mymensingh division. Now the empirical estimates of Table 9 indicate that the key variation in agricultural income comes from boro HYV rice and all other crops.

Table 9 Decomposed variances share (%) of crops in total agricultural income, by districts
Full size table

Rice is the leading crop and it dominates the cropping pattern throughout Bangladesh, accounting for more than 90% of total cereal production and covering 75% of Bangladesh’s total cropped area (Banglapedia 2003; BBS 2015; BER 2017). For Jamalpur, variances in boro HYV rice are the highest. For other districts, variances in boro HYV are high, including aman HYV rice.

The results of Table 9 shows the factors that are responsible for the large variation of income from aman HYV and boro HYV rice. From the Table 10, we can grasp the cost share of the rice productions and per ha income in each district that had an influence on the agricultural income. This study found that the rental cost for land, along with costs of seed, chemical fertilizers, and hired labor were the key costs for aman and boro HYV rice cultivation; it should also be mentioned that irrigation cost is also a major cost for boro HYV production (Table 10).

Table 10 Costs and income (BDT/ha) associated with aman HYV and boro HYV rice production, by districts
Full size table

The highest net income comes from aman HYV production in Netrokona. Table 10 also presents the highest net income in the Jamalpur and Mymensingh districts from boro HYV rice production. However, farmers of the Sherpur district used higher input than other districts. Generally, seed, irrigation, chemical fertilizer, rental cost for tools and machinery, and hired labor are the important costs of boro rice production. In Jamalpur, both the costs of irrigation and chemical fertilizer are higher than the Mymensingh district, since the yield of boro rice is also higher in Jamalpur—similar to Netrokona. However, in the Sherpur district, irrigation, and hired labor costs are high but production is low. Sometimes the irrigation cost depends on the owners of the power pump for the monopoly of irrigation and it varies among the districts. In addition, particularly for the planting and harvesting times of both aman and boro, hired labor in Bangladesh is scarce; this shortage increases the price of hired labor, which is more in the Sherpur district according to our study results (Table 10). On the other hand, seed cost is an important input cost along with chemical fertilizer and hired labor costs for aman rice cultivation, even though irrigation cost is not significant. Nevertheless, agricultural labor demand is currently high across Bangladesh, since most of the day laborers have now changed their occupation and moved to the city—mainly the capital city of Dhaka—for income generating activities. Therefore, it is essential to know the factors that affected the net income variation for aman and boro HYV rice cultivations.

Now, we can determine the factor that causes the net income differences of aman and boro HYV production. From decomposed variance of gross income and costs we found in Table 11, we found that the gross income is the chief factor causing net income differences.

Table 11 Decomposed variances share (%) of gross income and gross cost of aman and boro HYV rice, by districts
Full size table

It implies that even though farmers in same districts cultivated these two rice varieties, their gross income was different. These gross income differences mainly induce the net income disparity in Netrokona and Sherpur; however, for aman HYV rice the gross cost induce the income disparity in Jamalpur and Mymensingh. In contrast, for boro HYV, the gross income is the primary factor causing net income differences, except Jamalpur and Netrokona, where net income disparity was induced by gross cost. This implies that adaptation strategies have priorities on large gross income variances of aman and boro rice cultivation.

Now we want to know what costs are the main factors for income differences in aman and boro HYV rice production. Table 12 shows the decomposed variances shares in cost expenditures of these rice productions.

Table 12 Decomposed variances share (%) of costs for aman and boro HYV rice, by districts
Full size table

The results of Table 12 show that for aman HYV rice production, variances in chemical fertilizer and hired labor costs are high. These costs were the main factors to induce the income differences in aman HYV rice production.

We also found that, for boro HYV rice, variance in irrigation, chemical fertilizer, and hired labor costs are high in all districts and estimated the net income differences. It is also important to mentioned that variance in hired labor cost is highest in Sherpur, while it is lowest in Netrokona district, for boro HYV rice. This result implies that reduction of input cost variances will ensure the low net income differences, where farming knowledge and easy input access can also contribute to better rice cultivation.

Poverty projection for future

Agricultural production is fluctuating worldwide, including Bangladesh, because of climate change impact. Rice is the mainstay of Bangladesh’s agriculture. If rice production reduces by climate change impact, it will increase the rice price. Higher food prices can contribute to political instability and increase the poverty rates (Council 2013). Furthermore, poverty will also increase, if the climate changes increase the weather volatility (Ahmed et al. 2011). Floodplains constitute about four-fifths of the landmass in Bangladesh and floods occur every year (Mallick and Rahman 2009). In 2009, rice production faced a crisis after the floods in Bangladesh and saw its yield decrease radically, which caused food insecurity among farmers. Mymensingh is the agriculture-based division in Bangladesh, and farmers of this Division may face severe problem if rice production declines (BBS 2011).

We estimated the poverty rate of farm households in four districts of Mymensingh division, based on the yearly per capita income. According to purchasing power parity and poverty line of the World Bank (Ferreira et al. 2015), poverty rates were estimated by applying lognormal income distributions. We found the highest poverty rate in Sherpur, which was followed by Netrokona (Table 13), as shown in Fig. 7. We also projected the poverty, if farm production loss occurs by climate change (flood) impact. Flooding has the most deleterious effects on the crop production of Bangladesh. Severe flooding occurs every four to five years in Bangladesh, affecting more than 60% of the country. Up to 30% of Bangladesh’s landmass is in danger of flooding during monsoon season. While prolonged floods would tend to delay aman plantation and boro harvesting, flash floods have become regular phenomena in the haor areas, resulting in a significant loss of potential in the aman and boro rice productions.

Table 13 Change in poverty rate following a loss of rice yield due to flood (80% loss of rice)
Full size table
Fig. 7
figure 7

Annual per-capita income (BDT) distribution of total division and different districts

Full size image

Mymensingh is an area of Bangladesh prone to flooding (Ruane et al. 2013), where the yearly damage affects almost 80% of crop production, including that of rice (IFPRI 2013; Islam et al. 2011). In Bangladesh, floods occur almost every year; in 1998, floods covered almost 70% of the total land area, causing maximum damage (Rahman et al. 2009). We assume that, if extreme floods occur, as that in 1998 (the magnitude of the 1998 flood was the maximum in Bangladesh,) farm production will decrease by 80% in the flood-prone regions, including the Mymensingh division. By lognormal distribution, we project the poverty rate that was affected by the income reduction by yield loss because of extreme floods. From the study result, it is found that poverty will increase by a maximum of 6% in Jamalpur and by 5.5% in Netrokona (Table 13). This result implies that the potential loss of agricultural production by the impact of climate change in the Mymensingh division will cause farm families to suffer adversely, which calls for appropriate adaptation policies.

Conclusions

By using different empirical methods, this study reveals the per capita income differences, income inequality, corresponding income sources, cost and income disparity of crop production, and poverty of rural farm households under climate change impact on agricultural production. Based on the above study, findings show that per capita income difference is significant in farm household of the districts and agricultural income variation play an important role in the per capita income, which is consistent with the previous studies (Alamgir et al. 2018). Undoubtedly, variations in agricultural income has positive or negative impacts, which increase or decrease the household income. Higher income from agriculture contributed to lower income inequality in the districts, even though employment income is dominant in most of the districts and non-agricultural profits contribute the important share of household income in Sherpur. In Mymensingh district, income share of other crops is high in total agricultural income while in most of the districts; rice income share had the highest contributions. Higher income inequality was found in high remittance income share of household income, among the districts that are lowest in employment and then agriculture. By decomposing the variation in income, it was found that agriculture a primary contributors of inducing income disparity of farm households and the chief variation in agricultural income comes from aman HYV and boro HYV rice, however, the main difference is within the production of boro rice. In the context of Bangladesh agriculture, boro is the main cultivating crop by the farmers in all over the country due to its high yield, and the revenue is more than the aman rice as well. Although the production cost is higher than the aman rice production. Regarding the costs of aman HYV and boro HYV rice production, the income of these crops is largely accounted for by enhanced yields, higher irrigation, chemical fertilizers, and hired labor use of per hectare land. The districts of Mymensingh division are highly vulnerable in terms of poverty; thus, the unexpected damage to agricultural production from floods would increase the poverty rate and income vulnerability. We can predict that adverse climate change impact negatively affects the cash crop production of farm households in the study areas. Considering this, we recommended the following.

This analysis has drawn on a unique data set for a single year, consider the existing rice varieties cultivated by the farmers in the study areas, and we quantified this in conclusion section of this study, which provides the range of factors examined in the current paper. The 2011–2012 year was a typical year for agriculture production in Bangladesh (BBS 2015); however, it would have been beneficial to have a full set of contextual information across more years, to test the robustness of these findings by using a panel data approach.

In 2018, the Bangladesh Delta Plan 2100 (BDP 2018) approved by the government of Bangladesh with the aim to achieve safe, climate resilient, and prosperous Delta by 2100. Ensuring long-term water and food security, economic growth, and environmental sustainability in Bangladesh while effectively reducing vulnerability to natural disasters and building resilience to climate change, are the main intervention of BDP 2100 (BDP 2100 2018). According to the target of BDP 2100, proper management of agricultural farms with appropriate technology interventions to address the issue of climate change, such as research and development for long–run food security, should prioritize rice production; introducing flood-tolerant and short-duration crop varieties could, in fact, be a good option (Kobayashi and Furuya 2011; Israt et al. 2016). Intensive training programs are conducted on raising awareness, sharing the latest knowledge and approaches to adaptation, deciding adaptation plans (short, medium, and long term), programs and measures (individual and integrated), and monitoring for climate-vulnerable communities to adapt to new situations (Sikder and Xiaoying 2014). Traditional coping strategies, such as changing crops and diversifying livelihoods by motivating farmers to pursue alternative income opportunities, could substantially reduce income inequality and losses of farm income from climate change impact in the Mymensingh division and in other locations, where similar conditions prevail.

Future research could also seek to expand on the aspects identified in this paper by considering the specific sub-groups of the data, for example, considering high chemical fertilizer use and increasing the family labor in rice production. Checking the price of crops and labor wage rate that reduce the gross income, or examining the impact of flash flood in this division. It is imperative to understand that agricultural research and support services are critically significant for effective adaptation to climate change and reducing the loss of farm production and poverty.