**3. Data sources and analytical techniques**

The present study mainly used primary data to achieve the objectives. The primary data was collected from Jashore region (**Figure 1**) of Bangladesh due to the higher concentration of off-season summer tomato cultivation [6, 8]. At first, summer tomato cultivating villages was selected and for those villages a complete list of the off-season summer tomato growers was prepared taking help from local agricultural extension office. From that list, a total of 100 farmers were selected randomly as growers of summer tomato to collect the information regarding offseason tomato cultivation. These farmers were trained by different organizations on management aspect of summer tomato cultivation. Besides 150 farmers who

**Figure 1.** *Location map.*

did not cultivate off-season summer tomato but had suitable land and interest in growing summer tomato were selected randomly for interview as non-growers of the technology. The non-growers grew winter tomato and also did not receive any training on summer tomato cultivation. Thus, a total of 250 farmers were selected randomly for the face-to-face interview.

The present study employed propensity score matching (PSM), inverse probability weighting (IPW), and inverse probability weighted regression adjustment (IPWRA) techniques to achieve the objectives. PSM constructs a statistical comparison group that is based on a model of the probability of participating in the treatment, using observed characteristics [12]. According to Heckman et al. [13], the basic assumption of using a counterfactual is that the untreated samples approximate the treated sample if they had not been treated, that is, E (Y0i T = 1). The validity of PSM depends on two condition; conditional independence assumption (CIA) and sizable common support in propensity score across the growers and non-growers. The CIA argues that program outcomes are independent of program participation conditional on a set of observables (X). When CIA condition is not met, it is assumed that may be unobserved factors affect the outcome and treatment assignment, leading to a hidden bias [14]. Under the CIA, the average treatment effect on treated (ATT) was computed as:

$$\text{ATT} = \text{E}(\text{Yz} - \text{Y} \bullet | \text{X, T=1}) = \text{E}(\text{Yz} | \text{X, T=1}) - \text{E}(\text{Yo} | \text{X, T=0}) \tag{1}$$

Balancing properties need to be satisfied for PSM to be valid which implies that for observation with the same propensity score, the distribution of pretreatment characteristics must be same across growers and non-growers' group. Another requirement for PSM is common support or overlap condition. It implies that farmers with same X values have positive probability of being both grower and nongrower. Three matching algorithms: nearest neighbor, radius matching and kernel matching were used to present the findings of the study.

IPW uses the inverse of the propensity score as weights in calculating the average value of the outcome variable [15, 16]. IPW does not match off-season tomato growers with non-growers. In IPW, farmers with low predicted probability receive a lower weight while farmers with high predicted probability of adoption receive a higher weight.

True measurement of impacts requires controlling of sample selection bias through random assignment of individuals into treatments. However, ATT from PSM and IPW can still produce biased results in the presence of mis-specification in the propensity score model [17, 18]. To overcome the problem, the present study used IPWRA which has the double-robust property that ensures consistent results as it allows the outcome and the treatment model to account for mis-specification. ATT in the IPWRA model was estimated in two steps. In the first step, we estimated the propensity scores using binary probit model and in second step, linear regression was used to estimate the ATT.

To assess the impact three outcome indicators were selected. Income from offseason tomato (Tk./ha): The sum of crop output minus the value of variable inputs (fertilizers, pesticides, seeds, hired labor, etc.) and fixed inputs. This is the net income households receive from off-season tomato cultivation (Tk. is Bangladeshi currency, 1 USD = Tk. 85). Consumption expenditure (Tk./adult): Total expenditure on consumption per adult per year was calculated. Food security status: Food security status of the farmers was assessed by using Food Consumption Score (FCS). The FCS of a household is calculated by multiplying the frequency of foods consumed in the last seven days with the weighting of each food group [19].

*Impact of Off-Season Summer Tomato Cultivation on Income and Food Security of the Growers DOI: http://dx.doi.org/10.5772/intechopen.93674*
