Introduction

Globally, the consumption of wheat has increased by 19 per cent from 54.9 kg yearly per capita in 1961 to 65.43 kg in 2013. Wheat consumption has also increased in Asia’s rice economies, including leading rice-producing and consuming countries such as India, Bangladesh, and Indonesia (Table 1). For example, in India, the per capita wheat consumption has increased to 117.37 per cent from 36.23 kg in 1961 to 60.56 kg in 2013. In contrast, the per capita rice consumption in India has increased by 4.09 per cent from 66.76 kg in 1961 to 69.49 kg. During the same period, the yearly per capita wheat consumption increased from 8.62 to 17.47 kg (+102 per cent) in Bangladesh, 11.94 to 50.46 kg (322.61 per cent) in Nepal and from 1.65 to 25.45 kg (+1442 per cent) in Indonesia.

Table 1 Wheat and rice consumption (kg/capita/year) and the percentage changes in the consumption (base 1961) in the selected Asian rice economies from 1961 to 2013
Full size table

The increasing wheat consumption in developing countries can be partly explained by changing consumption patterns associated with economic development (e.g., Huang and Bouis, 1996; Hossain, 1998; Pingali, 2007). With an increase in income, households tend to consume proportionately more higher-value food, including wheat-based products, and protein. Table 1 clearly presents this tendency in a number of Asia’s developed countries. For example, in 1961, the yearly per capita rice consumption in Japan, Malaysia, and South Korea was 122.78, 120.61, and 98.85 kg, respectively. Compared to 1961, by 2013, in Japan, the per capita rice consumption had declined by 153.33 per cent (to 59.85 kg); in Malaysia, it was reduced by 32.63 per cent (to 81.25 kg), and in South Korea, it was reduced by 13.83 per cent (to 85.18 kg) (FAOSTAT, 2017). In contrast, during the same period, wheat consumption in Japan has increased by 34.73 per cent from a per capita yearly consumption of 33.43 kg to 45.04 kg; in Malaysia, it has increased by 85.87 per cent from 27.46 to 51.04 kg; in South Korea, it has increased by 336 per cent from 11.66 to 50.84 kg per capita/year (FAOSTAT, 2017).

The increasing wheat consumption in Asia’s rice economies warrants careful investigation and policy attention. The fact that these are rice economies in the first place reflects a tradition of growing and consuming rice – and coincides with their prevailing (sub) tropical climate. Wheat is a temperate crop and can extend well into the sub-tropics, but it is often confined to where temperatures remain more amenable, including the cooler winter season and highlands. In South Asia, the wheat belt stretches across the northern sub-tropical latitudes. In the Indo-Gangetic plains, this includes the rice–wheat systems which extend from the Green Revolution heartland of Punjab and Haryana in north-western India and Pakistan Punjab eastward into the lower Gangetic plains of eastern India, Nepal terrai and north-western Bangladesh (Timsina and Connor, 2001; Erenstein and Thorpe, 2011). The rice–wheat systems of the Indo-Gangetic plains are of particular interest as they extend from the north-western plains where wheat was the traditional staple crop, into the eastern plains where rice was the prevailing staple. With wheat encroaching into the eastern rice bowl, this raises the prospect of producing wheat to meet increasing domestic wheat demand, and questions about comparative advantages and whether this merits investments in wheat research and development to ensure local adaptation and sustainable intensification.

Two decades ago, studies on the economics of wheat in Bangladesh found boro (winter) rice to be more profitable than wheat in irrigated zones, but wheat often generated the highest returns in non-irrigated zones and in areas that are unsuitable for boro rice production (Morris et al, 1996, 1997). These studies highlighted that should Bangladesh become self-sufficient in rice, wheat production would become even more attractive and wheat may then merit further promotion and research investment (Morris et al, 1996). Given that Bangladesh has subsequently successfully achieved self-sufficiency in rice production since these earlier studies, the issue of the prospects of wheat production in Bangladesh merits revisiting. Empirical studies are scarce on this important issue for Bangladesh’s wheat food security. Although an important open question, the present study does not envisage a detailed analysis of wheat’s comparative advantages. Instead, we focus on a subset of questions, particularly looking into Bangladesh’s current wheat production patterns and producers and the resultant implications with the hope of shedding further understanding on the prospects for producing wheat in Asia’s major rice economies.

While Bangladesh has been highly successful in achieving self-sufficiency in rice production (Mottaleb et al, 2015), the increasing wheat consumption is met mainly by import. More than 65 per cent of the total wheat consumption in the country is currently met by imports (Figure 1), and the inflow of wheat food aid to Bangladesh has been decreasing since the 1990s. In 2015–2016, it was estimated that the domestic production of wheat in the country would be 1.40 MT, and from July 2015 to February 2016, the country imported an additional 2.58 MT from abroad to meet consumption demands (GOB, 2016).

Figure 1
figure 1

Sources: Authors’ calculation, GOB (2009, 2016).

Inflow of wheat in the form of food aid and imports (‘000 MT) into Bangladesh, 1972–2015 and self-sufficiency index [{(domestic production/Total consumption)} × 100].

Full size image

With the overall wheat consumption in the country increasing steadily over the years, the self-sufficiency in wheat production has declined particularly after 2000. Currently, domestic production can meet less than 35 per cent of the total consumption (Figure 1). In this critical backdrop, questions arise as to the potential role of domestic wheat production to feed Bangladesh’s burgeoning population and changing consumption patterns. To help think through the policy implications, the farmers who produce wheat in a country such as Bangladesh where 75 per cent of the total land is allocated for rice production only need to be characterized. Using a large nationally representative data set, this study identifies influential socioeconomic and biophysical factors that affect the decision to produce wheat and the allocation of scarce land for wheat production at the household level in Bangladesh. As the empirical model controls for district-level influences by explicitly including district dummies in each division, this study provides district-level implications on how to encourage farmers to produce wheat at the district level.

Brief History of Wheat Production in Bangladesh

Bangladesh’s wheat production environment is characterized by autumn-sown wheat, with high tropical rainfall, high humidity, and minimum temperatures ranging between 11 °C at the lowest to 16 °C at the highest in the coolest quarter of a year (classified by the International Maize and Wheat Improvement Center (CIMMYT) as wheat mega-environment 5 – similar to Paraguay and parts of Nigeria and Sudan). The major biotic and abiotic stresses identified in this mega-environment are heat stress, spot blotch, and leaf and stem rust (Wheat Atlas, 2016). However, in 2016 for the first time in history, wheat blast disease has been reported in Bangladesh (Malakar et al, 2016).

Before independence, wheat cultivation in Bangladesh was sporadic. After independence in 1971, initially the government discouraged wheat production, as eating wheat was considered part of Pakistani culture (Ahmed and Meisner, 1996). Realizing the importance of wheat for ensuring food security, the attitude towards wheat production changed, and in 1975, the government imported high-yielding wheat seed from India and Mexico and trained agriculture extension officers to help farmers cultivate wheat. It was the first well-organized government initiative in Bangladesh to expand wheat production in the country (Ahmed and Meisner, 1996).

Thirty-five wheat varieties have now been released in Bangladesh since 1974 (Table 2). Modern wheat varieties in Bangladesh increasingly take less time to mature and a have higher yield potential than the early varieties (Table 2). Partly because of the availability and the adoption of high-yielding modern wheat varieties, even with progressively declining land allocation for wheat cultivation, Bangladesh has been able to increase wheat production from 1961 to the present (Figure 2).

Table 2 Wheat varietal releases in Bangladesh, 1974–2014
Full size table
Figure 2
figure 2

Source: FAO (2017).

Trends of wheat area (000 ha), production (000 tons), and yield (ton/ha) in Bangladesh from 1961 to 2014.

Full size image

While wheat production in Bangladesh has increased, production failed to keep pace with the noticeably increasing consumption of wheat by the growing population. In 2005, the national average per capita yearly wheat consumption in Bangladesh was only 4.42 kg, but by 2010, it had increased by more than 115 per cent to 9.49 kg (Table 3). Importantly, in both rural and urban areas, while wheat consumption has increased significantly, per capita yearly rice consumption has decreased from 2005 to 2010. For example, in 2005, on average, the per capita/yearly rice consumption was 160.45 kg nationally which has been reduced by 5.37 per cent to 151.84 kg in 2010.

Table 3 Average daily food intake (yearly/per capita/kg) by residence status in Bangladesh, 2005 and 2010
Full size table

Although Bangladesh meets the growing wheat consumption mainly by import, such increasing dependency on international markets for meeting this growing demand provides challenges considering the recent volatility of international commodity prices. The rice–wheat production system provides a potentially feasible option to further increase wheat production. For instance, winter fallow land after rain-fed Aman rice and other crops have been harvested, or the land under low-intensity crops during the dry season can be brought under wheat cultivation. As the profitability from dry-season boro rice is declining due to a recent bumper production accompanied by high input costs (Mahmumd, 2016), farmers can recover costs and earn extra profits by producing wheat instead of producing boro rice wherever possible.

The expansion of wheat production in the current boro rice areas can bring ecological benefits in addition to saving foreign exchange by reducing imports. At present, out of the 8 M ha of cropland1, 67 per cent of the cropland (5.37 M ha) of the country is under irrigation, of which more than 77 per cent of the total irrigated land is ground-water based. A recent estimation shows that to produce one kilogram of boro rice, farmers need to apply on average 3200 L of water (Mahmumd, 2016). Consequently, the total fresh water withdrawal in Bangladesh in 2015 was 35.9 billion m3, of which 88 per cent was used only for crop irrigation (World Bank, 2016). Due to the massive extraction of ground water, the ground-water level has been declining gradually, particularly in Rajshahi Division and to some extent in Khulna Division by between 0.01 and 0.05 m yearly (Shamsudduha et al, 2009; Dey et al, 2013). The expansion of wheat production instead of irrigated boro rice, where possible, can significantly reduce the extraction of ground water, as wheat requires significantly less water than boro rice, and thus can bring positive environmental externalities.

Materials and Methods

This study relies on data sets made available by the Bangladesh Bureau of Statistics (BBS) and other national government and international institutes, such as the Food and Agriculture Organization of the United Nations (FAO) and the World Bank. The household-level information on the number of family members, sex of the household head, decision to produce wheat and land allocated for wheat production are collected from the Agricultural Census 2008 conducted by BBS. The Agricultural Census 2008 was deployed May 11–25 in 2008, in which a total of 28.69 million farm households were surveyed, of which 25.35 million were rural. Of these, 1.73 million were from Barisal, 4.88 million from Chittagong, 9.46 million from Dhaka, 3.43 million from Khulna, 7.66 million from Rajshahi and Rangpur divisions, and 1.53 million were from Sylhet Division. Although the 2008 census covered all farm households in Bangladesh, the BBS provides access only to a randomly generated 5 per cent sample of the entire census data, resulting in 1,163,147 households in 480 sub-districts in 64 districts in all seven divisions of Bangladesh. In addition, data on sub-district level crop areas (ha) that are very suitable, suitable, and moderately suitable for wheat production are collected from the Bangladesh Agricultural Research Council (BARC). Figure 3 shows the wheat suitability map, as well as the total number of sampled households by division considered in this study, and the percentage of the households that produce wheat at the division level. Note that for 20 newly declared sub-districts located in Brahmanbaria (B. Baria), Bogra, Chandpur, Comilla, Cox’s Bazar, Feni, Laxmipur, Noakhali, Maulavibazar, Meherpur, Patuakhali, Pirojpur, Sunamganj, Sylhet, and Tangail districts, the information on suitable areas for wheat crops is unavailable. In that case, the wheat-crop suitable area and the total crop area of the older sub-districts were used and replaced in the corresponding newly declared sub-districts.2

Figure 3
figure 3

Sources: BARC (2016a) and BBS (2008).

Wheat suitability map for Bangladesh and 2008 sample distribution by administrative division.

Full size image

To capture the influence of the flood-water inundation on wheat production at the sub-district level land allocation for wheat production, we collected and used land elevation data at the sub-district level from BARC. Based on the depth of the flood water in centimeters (cm), BARC has classified total land area of the country at the sub-district level into highland (less than 30 cm), medium highland (30–90 cm), medium lowland (90–180 cm), low land (180–300 cm), very low land (more than 300 cm), and miscellaneous types of land in ha. This study exclusively considered the extent of highland (ha) at the sub-district levels in which flood water should not stay long, thus allowing farmers to plant wheat at the correct time.

To see the impact of the availability of seeds, the information on the number of government-registered seed dealers (numbers) at the sub-district level was collected from an online database made available by the Ministry of Agriculture (MOA). This study used a distance variable (km) that shows the distance between the sampled sub-districts and the corresponding district headquarters to see the effects of large cities on the production of wheat in the nearest suburbs. To capture the influence of the number of households, literacy rates, and physical infrastructure such as the availability of electricity on the allocation of land at the sub-district level (if any), the data on the number of households, % of households connected to electricity, and the literacy rate (%) at the sub-district level were collected from the population and housing census of 2011. Finally, to capture the influence of the extent of irrigated areas on wheat production, data on irrigated land are collected at the sub-district level (ha) from minor irrigation survey reports from 2012 to 2013 of the Bangladesh Agricultural Development Corporation (BADC, 2013). Note that the use of 2011 census data on the number of households, the literacy rate (%) and the irrigated area in 2012–2013 to determine the proportion of wheat land at the sub-district level in 2008 may generate a data-matching problem. However, it is expected that within two to three years, one cannot expect a big change in the literacy rate, electrification, and population growth in a developing country such as Bangladesh, where, because of shortages of electricity and natural gas, the government has almost stopped the rural electrification and gas connection programs. In fact, this is the best available data and it can be used to control for sub-district-level influences on the allocation of land for wheat production at the sub-district level.

Model Specifications and Estimation Techniques

The production of wheat in a major rice economy can be treated as a decision to apply an innovation with a productive purpose (Rogers and Shoemaker, 1971). Consequently, the decision to produce wheat (si=1 or 0) and the allocation of land for wheat production (yi) by the ith farm household, can be influenced by a number of factors. As the decision to produce wheat, and land allocation for it are two different but interrelated issues, following the Heckman selection model (Heckman, 1976), it is assumed that there exists an underlying regression relationship between land allocation for wheat production and the wheat production decision as follows:

$$y_{i} = x_{i} \beta + u_{1i},$$
(1)

where yis is the land area allocated to wheat. The caveat in Eq.(1) can only be observed if

$$s_{i} = z_{i} \gamma + \delta_{i} ({\text{PWA}})_{id} + u_{2i} = 1,$$
(2)

where si is the decision to produce wheat (yes=1, no=0) by the ith farm household, and z is the vector of independent variables that includes all of the variables included in vector x in Eq.(1); PWA is the proportion of wheat area (total wheat area/total crop area) in a sub-district d where ith farm household is located. In regression Eq.(1), the dependent variable yi is the size of farm land allocated for wheat production at the household level; x is a vector of variables including a medium-sized farm dummy variable that assumes value 1 if a farm household is endowed with 1–2.99ha of land, or 0 otherwise; a large-sized farm dummy variable that assumes value 1 if a farm household is endowed with three or more hectares of land, or 0 otherwise; number of family members, a sex dummy that assumes value 1 if a farm household is headed by a female, or 0 otherwise; a rural household dummy that assumes value 1 if a farm household is located in rural areas, or 0 otherwise; an agriculture labor household dummy that assumes value 1 if a farm household’s major source of income is agriculture labor work on farms owned by others, or 0 otherwise; and district dummies, in which the number of dummies depends on the number of districts in each division. In Eq.(2), the variable PWA, which is the proportion of cropland under wheat in a sub-district d (wheat cropland/total cropland), captures the local influence of wheat production. It is expected that PWA can be influenced by a number of sub-district-level factors. To disaggregate the factors that affect the proportion of wheat land (total wheat land/total cropland) at the sub-district level, Eq.(3) is developed as follows:

$$\begin{aligned} ({\text{PWA}})_{d} =\, & \beta_{0} + \beta_{1} ({\text{No}}.\,{\text{of}}\,{\text{households}}_{d} ) + \beta_{2} (\% \,{\text{Households}}\;{\text{connected}}\,{\text{to}}\,{\text{electricity}}_{d} ) \\ & + \beta_{3} \left( {{\text{Literacy }}\,{\text{rate}}\%_{d} } \right) + \beta_{4} \left( {{\text{Distance }}\,{\text{from}}\,{\text{the}}\,{\text{district}}\,{\text{headquarter}}\,({\text{km}})_{d} } \right) \\ & + \beta_{5} ({\text{Land }}\,{\text{classified}}\,{\text{as}}\,{\text{highland}}\,({\text{ha}})_{d} ) + \beta_{6} ({\text{No}}.\,{\text{of}}\,{\text{registered}}\,{\text{seed}}\,{\text{dealers}}_{d} ) \\ & + \beta_{7} ({\text{Proportion}}\,{\text{of}}\,{\text{wheat }}\,{\text{suitable}}\,{\text{area}}_{d} ) + \beta_{8} ({\text{Irrigated }}\,{\text{land }}\, ( {\text{ha)}}_{d} ) \\ & + \mathop \sum \limits_{dd = 1}^{63} \theta_{dd} ({\text{DisDum}})_{d} + \epsilon_{d}, \\ \end{aligned}$$
(3)

where PWAd is the proportion of wheat area, which is calculated as the total wheat area divided by the total crop area in a sub-district d in 2008. The variable DisDum (63 district dummies for 64 districts) is included in the model to capture the district level unobserved characteristics that affect wheat-land allocation at the sub-district level; βi and θi are the parameters to be estimated and ϵd is the random error term. The proportion of the crop area suitable for wheat cultivation is calculated as areas very suitable, suitable, and moderately suitable for wheat cultivation, divided by the total crop area in sub-district d. As the dependent variable is the proportion of land under wheat cultivation, the value of which ranges between 0 at the minimum to 1 at the maximum, we applied a two-limit tobit estimation method to estimate Eq.(3). The estimated value of the proportion of wheat area (PWA) at the sub-district level d is plugged in into Eq.(2) to capture the sub-district-level influence on the household-level decision to produce wheat (si). To correct for the standard error, we applied the bootstrapping method and replicated the regression estimation procedures 500 times.

General Findings

Table4 shows that, on average, in each sub-district out of the sampled total of 956ha of cropland, 2.44per cent was under wheat crop in 2008. Thus, on average, in each sub-district, more than 26ha of land were under wheat production3. On average, more than 61per cent of the total crop area in a sampled sub-district is classified as suitable for wheat production, and each sub-district is equipped with nearly 11,000ha of irrigated land and 7.79 thousand hectares of highland. Table4 shows that, on average, more than 61,000 households reside in each sub-district, of which 47per cent were connected to electricity and 48per cent were literate (can sign their name at least). On average, each sub-district is located more than 21km away from the respective district headquarters, and there were more than 22 government-registered seed dealers to sell seeds to farmers.

Table4 Sub-district level indicators in Bangladesh by division
Full size table

A closer scrutiny of Table4, however, shows that the sub-districts located in Barisal and Sylhet divisions have the lowest number of government-registered seed dealers and the sub-districts in Barisal Division have the least amount of highlands, on average. This is probably the reason, reflected in Table4, that the sub-districts located in Rajshahi and Rangpur divisions allocated a larger proportion of land for wheat production; in contrast, the sub-districts located in Barisal and Sylhet divisions used the lowest proportion of land for wheat production in 2008.

Table5 presents household-level information on land endowment and demographics by division. The first column shows that, on average, more than 96per cent of the sampled households are male-headed households, nearly 80per cent of them live in rural areas and 33per cent of them are agriculture labor households, who mainly depend on agriculture wage income as their major source of income, earned from working on farms owned by others. On average, more than 6per cent of the households produced wheat in 2008, and a sampled household has more than three family members and 0.39ha of land, of which 0.01 hectare was allocated for wheat production in 2008. Nearly 90per cent of the sampled households can be classified as small-sized farm operators who operate less than one hectare of land. In contrast, 9per cent of the total sampled households can be classified as medium-sized farm holders (operating 1–2.99ha of land) and only 1per cent of the sampled households can be classified as large-farm operators (operating on more than 3ha of land). As nearly 90per cent of the total farm households in Bangladesh farm less than one hectare of land, they are forced to allocate their scarce land mainly to produce rice primarily for self-consumption.

Table5 Farm household-level indicators in Bangladesh by division
Full size table

Table5 shows that more than 10per cent of the sampled farm households in Khulna and Rangpur divisions and nearly 10per cent of the sampled farm households in Rajshahi Division produced wheat in 2008. Because of the favorable cooler weather for wheat production in Rajshahi and Rangpur divisions, which are located in the northern part of Bangladesh, farmers are encouraged to produce wheat. Table5 reveals that the percentage of medium- and large-sized farms, and the average farm size in Rajshahi, Rangpur, Khulna, and Sylhet divisions are larger than that in other divisions. The percentage of agricultural labor households is relatively higher in Rajshahi, Rangpur, and Khulna divisions, where both the incidence and intensity of wheat production are higher than that in other divisions. Probably, as agriculture is the mainstay in these divisions and crops are more intensively produced compared to Dhaka, in Chittagong and Sylhet divisions, farm households with surplus labor can find agriculture wage income opportunities by working on farms owned by other households.

Econometric Findings

Table6 presents the estimated functions applying a two-limit tobit estimation process, explaining the proportion of land allocated in a sub-district for wheat production in 2008. The population density measured by the number of households in a sub-district, the extent of highland, and the proportion of the cropland identified as suitable for wheat production positively and significantly (p<0.00) influence the allocation of land for wheat production at the sub-district level. Table6, however, reveals that the presence of government-registered seed dealers and the extent of irrigated land negatively and significantly (p<0.00) influence the allocation of land for wheat production at the sub-district level. As the expansion of irrigation influences mainly the cultivation of boro rice, the major competing crop of wheat, the extent of irrigated land area negatively and significantly affected land allocation for wheat production at the sub-district level. Although previous studies found that the presence of government-registered seed dealers positively influences the adoption of hybrid rice at the household level (e.g., Mottaleb et al, 2015), the estimated function in Table6 shows that the number of government-registered seed dealers at the sub-district level negatively influences the land allocation for wheat production, probably because wheat cultivation is still not as pervasive as rice, jute, and other crops. Thus, the presence of government-registered seed dealers does not necessarily mean the availability of wheat seed in time. Table6 demonstrates that electricity connection at the household level, the literacy rate, and the distance from the nearest large city do not have any significant impact on wheat production intensity at the sub-district level.4

Table6 Estimated function explaining wheat area share at the sub-district level in Bangladesh, 2008 (two-limit tobit)
Full size table

Table7 presents the estimated function explaining the decision of a farm household to produce wheat (yes=1, no=0), and Table8 presents the estimated function explaining the allocation of land for wheat production at the household level by division in 2008. The econometric results depict that the farm size positively and significantly affects the decision to produce wheat and the land allocation for wheat production at the household level in all divisions except the Chittagong Division.

Table7 Estimated functions explaining wheat production incidence at the division level in Bangladesh, 2008 (Heckman two-step selection model, first step)
Full size table
Table8 Estimated functions explaining wheat production intensity at the division level in Bangladesh, 2008 (Heckman two-step selection model, second step)
Full size table

Households with medium- and large-sized farms are more likely to produce wheat and allocate more land for wheat production than households with small farms (p<0.00). The findings indicate that the medium- and large-sized farm households are more risk-taking households, compared to households with small farms. The household size measured by the number of family members is significant and positive across the divisions (p<0.00) in Table7. The availability of family members and, thus, family labor significantly and positively affects a household’s decision to produce wheat. In Dhaka, Khulna, Rajshahi, and Rangpur divisions, where the incidence of wheat production is relatively higher than other divisions, the number of family members affects the allocation of land for wheat production positively and significantly (Table7).

Table7 shows that a female-headed household is less likely to produce wheat compared to a male-headed household in 2008 in all seven divisions in Bangladesh. This finding indicates that female-headed households are less likely to take the risk of producing a non-traditional crop compared to their male counterparts. There might be a number of socioeconomic factors behind such a risk-averse attitude in female-headed households including the higher incidence of poverty and the ubiquity of small-farm holders among them; however, further research is necessary with gender disaggregated data to generate any suggestive conclusion on this important issue.

Table7 shows that the incidence of wheat production is high among rural households in all divisions except Chittagong and Rajshahi, and for Sylhet Division the rural household dummy has no significant influence on the incidence of wheat production at the household level. Table8 shows that except Barisal and Dhaka divisions, the households located in rural areas also allocate more lands for wheat cultivation than the households located in urban areas. Households that sell their labor for wages to farms owned by others are less likely to produce wheat, in all divisions except Chittagong, where the dummy for a household that sells labor for wages to farms owned by others actually is more likely to produce wheat.

Table8, however, clearly shows that the households that supply labor for wages to farms owned by others allocate less land for wheat production than others, probably because agriculture labor households that depend mainly on income from selling their labor to other farms are more likely to be small and marginal farm holders. They are more likely to allocate land for rice production to ensure rice self-sufficiency first. Thus, households that supply labor for wages allocate less land for wheat production. In Table8, the predicted wheat production intensity measured by the land allocation for wheat production at the sub-district level shows that, except for Chittagong Division, local level factors have positive and significant impacts overall on the decision to produce wheat at the household level (p<0.00).

Table8 presents the estimated coefficients and the level of significance of the Mills λ ^i that presents the correlation between the unobservable characteristics in the adoption of wheat and the allocation of land for wheat production. The significant and positive Mills lambda in the estimated functions explaining land allocation to wheat production in the case of Dhaka, Rangpur, and Sylhet divisions indicates that the unobserved characteristics of the households increase the probability of allocation of land to wheat production in these divisions; in contrast, the significant and negative Mills lambda in the estimated functions explaining land allocation to wheat production in the case of Chittagong and Rajshahi divisions indicates that the unobserved characteristics of the households decreases the probability of allocation of land to wheat production in these divisions.

Conclusion and Policy Implications

Globally, wheat consumption has been increasing steadily over the years. Even in Asia’s traditional major rice economies, households are increasingly consuming more wheat. Until now, the growing demand for wheat in Bangladesh has been met by imports, extending the rice–wheat cropping system may provide a promising option.

The study indicates wheat production can be expanded in a number of divisions in Bangladesh by facilitating and encouraging farmers to produce wheat – for instance in Barisal, where cropping intensity is lower than the national average (e.g., Mottaleb and Krupnik, 2015), and Chittagong and Sylhet divisions. This can have positive impacts on the poverty reduction efforts of the government by allowing resource-poor farmers to grow one additional high-value crop (e.g., McArthur, 2015), for instance in the poverty-stricken Barisal Division by bringing fallow land under crop cultivation. A recent study by Gumma et al. (2016) demonstrates that there are more than 1.90Mha of land in Bangladesh that farmers keep as winter fallow (from December to February, after Kharif season (wet season) rice and other crops). By bringing this land into short-duration wheat cultivation, Bangladesh can boost its domestic supply of wheat, and lessen its import burden. Moreover, the expansion of wheat cultivation in Rajshahi and Khulna divisions in place of irrigated boro rice can also mitigate the adverse effects of over exploitation of ground water in these regions.

This study shows that the incidence of wheat production among the agriculture wage-based households in Chittagong Division is high, and in Barisal Division, the land allocation for wheat cultivation between agriculture wage-based households and others is not significantly different. This indicates that even resource-poor small-sized farm households in poverty-stricken regions are more likely to produce a high-value crop such as wheat. Therefore, the extension of wheat production in Barisal and Chittagong can bring positive impacts to the livelihoods of resource-poor small-sized farm households and even agriculture labor households who depend on income from working on farms owned by others. The econometric findings of the study confirm that the incidence of wheat production among farm households headed by females across divisions is low compared to households headed by males and this warrants further study on the underlying factors including the potential role of risk-aversion.

Finally, this study demonstrates that the extent of highlands – associated with the length of time that flood water lingers – can increase the production of wheat at the household level. It indicates that proper flood-water control can have a significant impact on wheat production in Bangladesh. Effective management of flood water and drainage systems can bring other positive benefits to the country as a whole. International donor agencies together with the national government should invest in effective management and control of flood water and drainage systems, not only to expand wheat production in Bangladesh, but for the overall development of the country and in the face of climate change.

Notes

  1. 1.

    While total gross cultivated land (aggregate crop area across the year’s seasons) was 14.41 millionha, net cropland (effective land base) was 7.94millionha in 2010 (BBS, 2011b).

  2. 2.

    For example, in B. Baria district, the information on the crop area suitable for Akhaura sub-district is available but not for the Ashuganj sub-district which was a part of Akhaura. In this case, we replace the information of Akhaura sub-district for Ashuganj sub-district.

  3. 3.

    Total crop area and wheat area at the sub-district level are computed by adding up the total cropland and total wheat land operated by the sample 5per cent census households.

  4. 4.

    In the Supplementary Material, the district dummies are analyzed in detail.