The rice crop was well established before the incidence of heavy rains in September 2013, which caused rivers to overflow, inundating paddy fields in many villages including the two villages selected for the demonstration of SS1. The crop remained submerged for 10–16 days depending upon the elevation of the land. Due to the Phailin cyclone in October 2013, the crop was again submerged for a few days. SS1 started to recover immediately after the water receded and was left relatively unimpaired. It was only after 13 years, many farmers from these villages have harvested a bounty crop of paddy in the wet season. Yield of SS1 was relatively high when the submergence period was less than 10 days and was still about 2.5 t ha−1 even after the two weeks’ submergence (Fig. 2). These observations are in accordance with the published data of agronomic trials conducted in experimental stations and farmers’ fields during 2008–2011 (Ismail et al. 2013). SS1 has withstood submergence for up to 17 days in research trials (Singh et al. 2009; Sarkar et al. 2009).
Choice of crops
Farmers have overwhelmingly preferred to cultivate rice paddy in the wet season and legumes like blackgram (urd bean, Vigna mungo) and greengram (Vigna radiata) in the dry season during the three years of the study. The percentage of plots under rice cultivation during the wet season of 2012 was 70%, rising to 90% and 86% in 2013 and 2014, respectively (Fig. 3). This increase was primarily due to fallow plots being brought into paddy cultivation. Out of 1351 available plots, 394 were left fallow in the wet season of 2012 and from these 275 (about 70%) were cultivated with paddy during the wet season of 2013.Overwhelmingly, 96% of the paddy in these plots was planted with SS1. During 2014, 229 out of these 275 plots were re-cultivated with SS1, indicating a re-adoption rate of about 83%. It can be inferred from the data that plots left fallow in the wet season of 2012 were probably situated on lower lands with poor drainage, making them highly susceptible to both flooding and prolonged submergence. The introduction of flood tolerant SS1 permitted cultivation on these plots, expanding the local agricultural area. Choice of crops by farmers during the dry season was very consistent, with about 70% being allocated to legumes over all 3 years. The remainder of the plots were mostly left fallow and there was no paddy cultivation during the dry season.
In the wet season of 2012, just fewer than half the farmers cultivated their plots with traditional varieties (Fig. 4). About a quarter opted for high yielding varieties (HYVs), such as Puja and Swarna. None of the beneficiaries had access to SS1 seed during this season. During the wet seasons of 2013 and 2014, the majority of the farmers switched to SS1. In the wet season of 2013, over 80% of the plots were cultivated with SS1. However, the number was reduced to about 65% in the wet season of 2014. The reduction in the number of plots cultivated with SS1 in 2014 as compared to the previous year was somewhat anticipated. In 2013, the beneficiaries received the SS1 seeds free of cost along with other inputs and were then more willing to grow it on the major portion of their holdings. In the next year, however, when the seeds were no longer freely available, farmers had to make a choice either to purchase the seeds or use seeds retained from the previous year, or fall back on the traditional varieties with which they were quite familiar. Moreover, although the price of SS1 seed is almost the same as that of traditional and other HYVs, farmers who had not saved seed from the previous year’s SS1 produce might have been reluctant to purchase new seed from market. Furthermore, two-thirds of the plots on which SS1 was discontinued during the wet season of 2014 recorded no output in 2013. These plots were probably unsuitable for SS1 cultivation, due to their low elevation and severe flooding. This is further supported by the fact that on nearly three quarters of these plots, no output was obtained during the wet season of 2012. The other contending explanation could be that the heavy and prolonged submergence that occurred in the aftermath of the Phailin cyclone in 2013 destroyed the paddy crop, resulting in loss of confidence in SS1, causing farmers to switch back to their traditional varieties. We argue this from the observation that average yields of about 2.75 t ha−1 were obtained during the wet season of 2013 on the plots where SS1 was continued in the wet season of 2014. This is about 15% more than the overall average yield (2.45 t ha−1) during the wet season of 2013. An average yield of 1.87 t ha−1 was obtained during the wet season of 2013 on plots where SS1 was grown, but was replaced by traditional varieties in 2014.
The majority of the farmers preferred to persist with SS1 in 2014, indicating that they found this variety more profitable than the landraces and HYVs: 345 out of the 355 farmers interviewed cultivated SS1 in the wet season of 2013 and about 75% of them persisted with SS1 in 2014. This trend is similar to the percentage of plots on which SS1 was continued in the wet season of 2014. Out of the remaining 25% of plots on which SS1 was discontinued in 2014, a quarter were left fallow and about a third were sown with traditional varieties. The remainder were sown with HYVs other than SS1. Many farmers preferred to cultivate low yielding traditional varieties over HYVs owing to their flood tolerance. The direct substitution of traditional varieties with SS1 has the potential to double the rice yields not only in flooding years, but also in years when there is no flooding. This is supported by the results of Emerick et al. (2016) who observed the shift from traditional varieties to SS1.
The Indian Remote Sensing Satellite (IRS) L3 images of Amathpur, Asalpur and adjoining villages situated on the bank of the Birupa river in the Jajpur district of Odisha are presented in Fig. 5. The images reveal the changes in cropping patterns during 2012–2014 due to introduction of SS1 in the area. The vegetation on these False Colour Composite images is red, rivers and flooded/waterlogged areas are in blue to green, whereas fallow areas are in grey-green. Village habitation surrounded by trees appears in bright red. Image (A) shows blackgram (BG) crop at the vegetative stage in these villages during the wet season of 2012. In 2013, cluster demonstrations of SS1 were conducted in a 24 ha area shown by a black outline on image (B). Due to the Phailin cyclone in October 2013, severe flooding occurred, the effect of which can be seen by the bluish colour in the eastern part of the demonstration site. Despite severe flooding, the success of SS1 in 2013 encouraged farmers of the two villages to grow rice on a larger area during the wet season of 2014. Image (C) shows that SS1 cultivation was extended by 35 ha as indicated by the yellow outline on the image. These images show that farmers used to keep their fields fallow during the wet season through 2012 and cultivate blackgram in the dry season. Rice cultivation on these fields was introduced in 2013 when SS1 become available.
Plot yields
There has been a steady and significant increase in paddy yields per unit area between 2012 and 2014. During the wet season of 2012, average plot yields for paddy was about 1.3 t ha−1, almost doubling to 2.5 t ha−1 in the wet season of 2013, which was further doubled to around 5 t ha−1 during the wet season of 2014. This nearly four-fold increase in average yields of plots is attributed to the introduction of SS1. The last figure is consistent with the yields obtained in controlled field experiments with SS1 (Singh et al. 2016). The lower yields in the wet season of 2013 is primarily due to the devastation caused by severe flooding caused by cyclone Phailin, which hit the eastern coast of India in October 2013.
Yields obtained from traditional landraces and non-sub1 HYVs were comparable, and were substantially lower than those of SS1. During the wet season of 2012, the average yield of traditional landraces was around 1.3 t ha−1 and non-sub1 HYVs was about 1.5 t ha−1. This trend was exactly reversed in 2013. The yield enhancement in traditional varieties during 2013 is attributed to the occurrence of severe floods in the aftermath of the Phailin cyclone. Under heavy flooding, local landraces generally produce higher yields than popular HYVs (such as Swarna and Puja) due to the high sensitivity of the modern cultivars towards submergence. In the wet season of 2014, the overall yields for both the types of varieties were much higher.
The yields obtained in the wet season from four major rice varieties (i.e., SS1, Swarna, Puja and Mugdi) were also compared (Fig. 6). Swarna and Puja are the two major HYVs which are commonly cultivated in this area, while Mugdi is the dominant landrace of the region. In the wet season of 2012, the average yield of Swarna was 50% higher than Mugdi and 30% higher than Puja. Puja yielded about 1.4 t ha−1 in 2012, which is approximately 15% higher than that of Mugdi. In the wet season of 2013 with heavy flooding in October as a result of cyclone Phailin, Mugdi produced higher average yields than Swarna and Puja. While SS1 produced 2.5 t ha−1, Mugdi yielded about 1.7 t ha−1, which is 30% and 42% higher than that of Swarna and Puja, respectively. The lower yields of these two HYVs is due to their extreme sensitivity to flooding. Thus, in the absence of a reliable flood tolerant high yielding variety, such as SS1, farmers find it more beneficial to cultivate traditional varieties (such as Mugdi) in flood prone areas. Replacing traditional varieties with SS1 has the potential to double the yields in flood prone areas, irrespective of the occurrence of floods. In 2014, with relatively low flooding, a trend similar to the wet season of 2012 was observed. Mugdi produced the lowest yield of about 1.9 t ha−1. This is less than half the yield of Swarna and about 3 times lower than the yield of SS1. Puja with a yield of more than 3 t ha−1 produced 60% higher yields than Mugdi. Swarna outperformed Puja by a similar margin. The reason for the low average yields during the wet season of 2012 is that virtually half the plots produced no yield during this period, probably due to being at extremely low elevation and therefore ill-suited to HYVs combined with the low use of inputs and poor management practices.
The present study compared the performance of different varieties under similar flooding intensity during a specific period of time. The data for one year before and after the intervention was evaluated for better comparison of varietal dynamics. Although, no floods were observed in either 2012 or 2014, floods of a similar magnitude to those of 2013 occurred in the sample area during 5 of the previous 10 years. As demonstrated in the present investigation and reported by previous studies (Mackill et al. 2012; Ismail et al. 2013), SS1 shows substantial yield advantages during both flooding and non-flooding years. Swarna is being cultivated over a large area and can be safely replaced by SS1 irrespective of the occurrence of flooding, considering the fact that floods are becoming less predictable in time and space in the present scenario of climate change. However, the research findings by Lybbert and Bell (2010) suggest that the spread of submergence tolerant varieties can be rapid only if the exposure to flooding is more frequent.
Household level output and sales
Household sales serve as a proxy for the income of farmers from farming activities. A survey was carried out at the end of January 2015, by which time many households had not completed the sales of crops harvested in the preceding wet and/or dry season. To avoid any ambiguity arising from the incompleteness of data, analysis of sales and total household output was restricted to those households which had completed their sales by that time. This amounted to 82 households in the wet seasons and 71 in the dry seasons of all 3 years.
Average household income from paddy sales during the wet season of 2012 was about 3000 rupees, with the corresponding average household output of around 1.2 tons (Fig. 7). Despite the occurrence of flooding, sales were increased to over 11 thousand rupees (nearly four-fold) due to the doubling of paddy output to over 3 tons in 2013. Income from paddy sales in the wet season of 2014 was boosted still further to nearly 14 thousand rupees, with the average household paddy output of about 4 tons. The sharp hike in average household paddy output and sales is consistent with the increase in paddy yields per unit area. As the price of paddy remained almost the same at 10,000 rupees per ton of grain, the increase in sales was due to the increase in quantity sold as reported previously by Dar et al. (2013).
During dry seasons, the farmers incurred relatively higher income from crop sales. In the dry season of 2012, the average of 6000 rupees per household was realized from the sale of blackgram. This increased significantly to over 9000 rupees in the dry seasons of both 2013 and 2014. The hike in sales was primarily attributed to the willingness of farmers to sell a larger proportion of their produce. Farmers sold the 50% of their output in the dry season of 2012 and about 60% in subsequent years. Another reason for the increase is the introduction of new and high yielding varieties of mung bean.
Empirical analysis
Plot fertility, farmers’ preferences for crops and paddy varieties, availability of irrigation and seasonal variation were not controlled in the data reported, all of which could have significantly influenced the yields of SS1. For instance, farmers might have planted SS1 on their most fertile plots. Therefore a rigorous empirical analysis was carried out to control these factors in order to identify the actual output and yield premium derived from SS1. A fixed effect regression model was used to estimate the output and yield gains from SS1 similar to that described by Dar et al. (2013).
$$ {Y}_{ij kt}={\alpha}_{ij}+{\delta}_{kt}+{\beta}_1{SS}_{ij kt}+{\beta}_2{HYV}_{ij kt}+{\beta}_3{Gram}_{ij kt}+{\beta}_4{Fallow}_{ij kt}+{\beta}_5{Other}_{ij kt}+{\beta}_6{Irrigated}_{ij kt}+{\beta}_7{PlotArea}_{ij kt}+{\beta}_8{TotalLand}_{ij kt}+{\varepsilon}_{ij kt} $$
In the above equation, ‘i’ represents the plot, ‘j’ the farmer, ‘k’ the season and ‘t’ the year. Separate intercepts were specified for each farmer-plot and season-year combination (α and δ). This allows control of unobservable plot-characteristics, such as short-term fertility and the incidents common to all farmers in a specific season of a year, such as flooding due to heavy monsoon rains. Results are based on a comparison of output and yields of the same plots (within-plot) over years. ‘Y’ represents the plot output or plot yields.
The unit of observation in this case is farmer-plot-season-year. ‘SS’ is a categorical variable whose value is 1 if the plot is cultivated with SS1, otherwise it is 0. ‘HYV’ is another categorical variable, with the value of 1 if the plot is cultivated with a high-yielding paddy variety (except SS1) and it 0 otherwise. Similarly, ‘Gram’, ‘Fallow’ and ‘Other’ are also categorical variables equaling 1 or 0, depending upon if the plot is cultivated with blackgram, left fallow, or cultivated with any other crop (apart from paddy and blackgram). The category omitted here is the traditional variety which acts as benchmark for the comparison of results. The control parameters included whether the plot was irrigated (Irrigated), the area of the plot (PlotArea) and the total land (TotalLand) available to the farmer. The standard errors were clustered in each specification at the farmer level to allow for the potential correlation in the choices and decisions of farmers.
The primary coefficients of interest are β
1
and β
2
, which estimate the benefit (or loss) in output (yields) after switching from traditional to high yielding varieties (SS1 or non-SS1). As the distribution of seeds to recipients was carried out on a non-random basis, we cannot attribute a causal interpretation to β
1
or β
2
. Thus, our estimates are correlational, identifying the additional output or yields obtained in cultivating SS1 (or any other HYV) in lieu of the traditional varieties. This is conditional on the farmer-plot and season-year fixed effects and other controls included in the specification. The causal interpretation of the coefficients need to impose the additional assumption on the econometric model that there is no other unobserved factor correlated with the introduction of SS1 and also with the observed plot-level yields and output. Thus, if training pertaining to management practices is provided to farmers in addition to the introduction of SS1, it would possibly bias the coefficient on SS1 upwards, and the true causal impact of SS1 introduction on plot-level output and yield per unit area would not be estimated.
The results revealed that the plots cultivated with SS1 offers nearly 0.25 tons of rice extra over those cultivated with traditional varieties (Table 1). Similarly, plots cultivated with non-SS1 HYVs furnish 0.20 tons of additional output compared to the traditional varieties. Both the values are highly significant. In comparison with traditional varieties, an additional 0.83 t ha−1 of output is obtained from the cultivation of SS1. Likewise, the non-SS1 HYVs produce about 0.65 t ha−1 over the traditional varieties. Both the values are again highly significant. The equality of coefficients for SS1 and other HYVs for either plot output or yield per unit area cannot be rejected. This implies that after conditioning on our set of controls and controlling the plot characteristics, seasonal weather fluctuations, farmer attributes and other time-invariant factors, SS1 and non-SS1 HYVs are indistinguishable in terms of plot-output and yield per unit area. Therefore, the yield premium of SS1 over other HYVs is observed only during flooding years. This affirms the previous findings regarding SS1, particularly when compared with the HYV Swarna (Mackill et al. 2012; Ismail et al. 2013; Singh et al. 2013).
Table 1 Impact of the introduction of SS1 on plot wise output and yield per unit area
The regression results are in broad agreement with the descriptive statistics, particularly with regard to the plot level output and yield per unit area of SS1. However, unlike the descriptive trends where plot output and yields per unit area for traditional varieties and non-SS1 HYVs were comparable, the regression results revealed that the latter significantly outperform the former. This could be due to the fact that a number of factors are controlled in this specification such as plot-level fertility and seasonal weather fluctuations, which were not considered in the descriptive statistics. Commuting the regression coefficients into monetary values with the assumption that a kilogram of rice is priced at 10 rupees, it can be estimated that cultivating one hectare (ha) area can accrue an additional benefit of over 8000 rupees to a farmer if he switched from a traditional variety to SS1. The additional monetary gain from cultivating a non-SS1 HYV relative to a traditional variety is approximately 6000 rupees per hectare.
The monetary benefits accruing to farmers from switching over to SS1 is substantiated by the follow-up survey of bank deposits in two villages, covering 230 of the 355 beneficiary farmers. The average number of bank deposits per year was significantly boosted from about 2 in 2012 to nearly 4 in 2014. As 90% of the households depend on agriculture as their primary occupation, the increased number of bank deposits can be attributed to higher agricultural incomes resulting from the introduction of SS1. This was complemented by the increased willingness of farmers to sell a larger proportion of their produce. Some more suggestive evidence regarding the positive impacts of switching over to SS1 is discerned through the trends in loan-taking of these 230 households. In 2012, the average loans taken by the households was about 7000 rupees, rising by about 5% in 2013, and subsequently by over 25% to about 9000 rupees in 2014. This can be explained by the improvement in repayment abilities of the farmers in view of their increased agricultural income. As reported by the households, almost all the loans were borrowed for agricultural purposes and the introduction of SS1 can be linked to the additional use of fertilizers, increase in area cultivated, additional hiring of labor and improvements in agricultural technology. All of these have positive impacts on crop yields, thereby strengthening the rural economy.