Drought coping strategies: An exploration of determinants of adoption by qualitative response model

Introduction

The State of Karnataka has 72 percent of the cultivable area is rainfed and only 28 percent is under irrigation (GOK News, 2023). The State is the second largest in terms of arid region, next only to Rajasthan in India. The State faced consecutive droughts during the years 2001-02, 2002-03 and 2003-04 resulted in a sharp decline in agricultural output (Nagaratna and Sridhar, 2004) [7] . Drought stress is the major limiting factor for rice production and yield stability under the rainfed crop ecosystem (Meenakshi et al, 2015) [5] . Karnataka faces a high risk of moisture stress at maximum tillering and reproductive stages of the crop, which may lead to yield loss of 25 to 100 percent (Hanamaratti et al., 2008) [4] . Drought is a worst kind of natural disaster that happens slowly and causes huge loss to agriculture and live stock. It develops over a long period of time and covers a wide area. Drought has different dimensions and can be defined or viewed from different perspectives. The severity of drought is expressed in the form of biological degradation and socio-economic distress. Drought is usually understood as a period of dryness due to lack of rainfall. According to the water balance concept, drought is a physical condition in which the amount of water available from precipitation and soil moisture is insufficient to meet the demands of evapotranspiration. The drought-areas are characterized by excess of evaporation over actual rainfall. Kolar district is one of the biggest districts in Karnataka State and experiences recurrent drought conditions. When drought occurs, the agricultural sector is usually the first to be affected. Even though the meteorological drought is over, the adverse economic impact of drought may persist for several years depending upon the nature of drought. For an instance, rice farmers suffer frequently from drought, they have evolved various strategies to cope with it. Based on the case studies and focus group discussions with farmers, the most commonly used adjustments in rice production practices during drought years are allocating more area to short-duration varieties, switching from transplanting to direct seeding, switching to traditional varieties which are considered to be more drought-resistant, etc (Serraj, et al, 2008; Meenakshi et al, 2015) [10,5] . The objective of the study was to identify the determinants of the adoption of drought coping mechanisms in order to balance and stabilize the farm income of the stakeholders. The study also helps to know how to mitigate the effect of drought on farmers' livelihood.

Reason for selection of Kolar district as a study area

Kolar region was one of the drought prone areas in eastern dry zone of Karnataka where most of the farmers were involved in rainfed agriculture. Kolar district was purposively selected based on important parameters like rainfall distribution, Analysis of the rainfall distribution showed that there was a 50 percent probability of occurrence of drought in Kolar district than in any other districts of southern Karnataka. Hence adoption of certain coping strategies against drought in this region is the major solution to stabilize the farm income during the drought period. The drought occurs in Kolar district is majorly contingent drought, this occurs in climates which have rainfall of an irregular and variable type. Because Kolar district falls in rain shadow region of monsoon flow and located far away from the coastal zone, hence drought is more prevalent there and precipitation occurrence is less. However the district also stands frontline in dairy farming and live-stock rearing which leads to dairy products. Alternative solutions based on scientific assessment of land and water resources are essential to mitigate the losses from recurrent drought (Nagendra M and Ashok Hanjagi, 2019) [8] .

Scope and Objective of the study

Outcome of the study would be useful to the policy makers to implement drought relief programmes during the drought periods depending upon the intensity of drought and its impact on levels of income and productivity. Identification of factors determining the adoption of drought coping strategies would be helpful to focus more on the factors to speed up the adoption rate. In order to mitigate the effect of drought and stabilize the farm income in Kolar district of Karnataka, the objective of the study has been framed and the study aims to determine and analyze the factors influencing on adoption of drought coping mechanisms.

The role of fitting logistic regression

One of the problems of classification lies in the use of appropriate methods to fit the model depending on the nature of the data. It is well known that most of the data related to the adoption of any agriculture technology are ending up with qualitative response variables with two or more categories, which is a problem when using the traditional statistical methods, such as linear regression analysis because of not satisfying the assumptions of quantitative regressand (Halagundegowda et al, 2018) [2] . In such a case to measure the farmer's perception towards adoption of particular agriculture technology for evaluating the binary classification problems, logistic regression is a most powerful tool, because it accommodates all kinds of variables (both qualitative and quantitative variables) in the model. In addition to the above advantage, it is robust and void of all kinds of restrictions or assumptions to use.

Materials and Methods

The household data were recorded on socio-economic characters of farmers of Kolar district of Karnataka (India). The data is mainly related to coping strategies implemented against drought by the farmers of this region and was collected by employing the multi stage sampling design. The total sample size (farmers) is 150, to develop the model, it is necessary to split the set of data in two different subsets. Therefore, the training data, which consists of 80% of respondents (120 farmers) and test data consists remaining 20% respondents (30 farmers). Before training the model each attributes are normalized to zero mean and unit variances, which will improve performance of the model as well as cut down the learning time. The following table

gives the variable of interest and the unit of measurement used in this study.

Logistic Regression Model

The Logistic response function resembles an S-shape curve where the probability π initially increases slowly with increase in X, and then the increase accelerates, finally stabilizes, but does not increase beyond 1. The probability of willingness to adopt drought coping mechanisms approaches zero at slower and slower rate, as Xi becomes small and the probability approaches one at slower and slower rate as Xi becomes large. Since, Pi = E (Yi) = 1 (given Xi) non linearly increases with Xi, let us consider Pi to be a logistic function of Zi, given by Pi= 1 [1 + e (-Zi) ]

Where zi= A + ∑ Bi Xi Taking logarithm of this odds ratio to the base e, we will get,

Here, Log [Pi/ (1 -Pi)] is called the logit or the log of odds ratio. It follows logistical distribution. Here to find out the probability of willingness to adopt coping strategies, the mean of all the variables taken and multiplied with their respective coefficient to get the Zi. Estimation and testing are two important aspects of regression analysis. The usual method of estimation under logistic regression is maximum likelihood estimation method (MLE). In the present study Wald test and Likelihood Ratio tests have been used for testing the overall significance of the logistic regression model. The Hosmer-Lemeshow test used to assess goodness-offit of the final model. Further, once models are fitted and relevant goodness of fit measures are employed, judging the predictive ability of the model can be done. In Classification statistical modeling setup, predictive ability of models can be judged by employing various measures such as Somers'D, Gamma Statistics, Kappa statistics, Accuracy ratios, Kendall's Tau (Tau-a) and c.

Results and Discussion

The data recorded on adoption behavior of drought coping strategies across various socio-economic characters of farmers were analyzed using binary logistic regression. The logistic regression works on the principle of computation of probability for each respondent based on the agro economic attributes of farmers. The maximum likelihood model, in which the intercept is included without any predictor variables, such model is called Intercept Model or Null Model. This is basically interesting to calculate the Pseudo R² that describes the goodness of fit for the logistic model, whereas full model considers all the explanatory variables in the equation along with constant. Here, the analytical framework carried out only for full model. ): if one extra crop grown in the farm then on an average estimated odds of adoption of drought coping strategies will significantly increase by 5.54 times. Income status (Wald= 5.84, P=0.016, Exp (B) = 1.00): if one rupee increase in the farm income then on an average estimated odds of adoption of drought coping strategies will significantly increase by 1times. The variable Liquid assets (Wald= 2.80, P=0.034, Exp (B) = 1.00): if one rupee worth increase as a liquid asset then on an average estimated odds of adoption of drought coping strategies will significantly increase by 1times. crop insurance (Wald= 2.90, P=0.048, Exp (B) = 1.00): if one extra rupee got as a crop insurance then on an average estimated odds of adoption of drought coping strategies will significantly increase by1times.

The table also provides negatively significant coefficients such as age and intercept, age (Wald= 5.45, P=0.02, Exp (B) = 0.67) if one year increase in the age of farmers then on an average estimated odds of adoption of drought coping strategies will significantly decrease by 0.54 times. The other literature on adoption of drought coping strategies, Francis Ndamani et al (2016) [1] , Stephen Abugri et al (2020) [11] and Mequannt Marie et al (2020) [6] are used logistic regression and depicted that education, household size, annual household income, access to information, credit and membership of farmer-based organization are the most important factors that influence farmers' adaptation to drought coping strategies.

Table 4 provides the classification matrix for both training data set and testing data set, the classification matrix helps to assess the performance of the model by cross tabulating the observed response categories with the predicted response categories and shows how well our full model correctly classifies the cases. The rules works for each case such that the predicted response is the category treated as 1, if that category's predicted probability is greater than the user-specified cut-off (0.5), otherwise it's treated as 0.

Conclusion

Adoption of drought coping strategies is considered as categorical response variable and it's a psychological perception, in which farmers deciding whether to adopt the strategies or not. This process is affected by various socio economic, agro ecological, institutional and resource based factors. Logistic regression was fitted to predict and identify the factors influencing on adoption of drought coping strategies by the farmers in the study area. The result illustrates that the predictors like age, education, farm size, extension visits, crop diversification, income status, liquid assets and crop insurance were significantly contributing for adoption of coping mechanisms. The variables such as education, farm size, extension visits, crop diversification, income status, liquid assets and crop insurance are act as positive drivers and the variable like age is negatively influencing on adoption of drought coping mechanisms by the farmers.

References

1. Ndamani F, Watanabe T. Determinants of farmers adaptation to climate change: A micro level analysis in Ghana. Scientia Agricola. 2016;73(3):201-208.
2. Halagundegowda GR, Singh A. Multilayer perceptron method of artificial neural network for classification of farmers based on adoption of drought coping mechanisms. International Journal of Pure and Applied Bioscience. 2018;6(2):1408-1414.
3. Halagundegowda GR, Singh A, Meenakshi HK. Discriminant analysis for classification of farmers based on adoption of drought coping mechanisms. Agriculture Update. 2017;12:2643-2648.
4. Hanamaratti NG, Prashanthi SK, Salimath PM, Hanchinal RR, Mohankumar HD, Parameshwarappa KG, Raikar SD. Traditional land races of rice in Karnataka: Reservoir of valuable traits. Current Science. 2008;94(2):242-247.
5. Meenakshi HK, Selvaraj KN, Nagaraja MS. Adoption of drought coping mechanisms (Ex-Ante and Ex-Post) and estimate the economic costs of adoption. Trends in Biosciences. 2015;8(18):4894-4899.
6. Mequannt M, Yirga F, Haile M, Tquabo F. Farmers choices and factors affecting adoption of climate change adaptation strategies: evidence from northwestern Ethiopia. Heliyon. 2020;6:1-10.
7. Biradhar N, Sridhar. A study on consequences of 2003 drought in Karnataka with reference to livestock and fodder. Journal of Human Ecology. 2004;26(2):123-130.
8. Nagendra M, Hanjagi AD. An assessment of drought in Kolar district in Karnataka State. Review of Research. 2019;8(4):1-9.
9. Panahi F. Analysis of factors affecting the adoption of modern methods of irrigation in the agricultural system of Iran. World Applied Sciences Journal. 2013;21(11):15521559.
10. Serraj R, Bennett J, Hardy B. Drought frontiers in rice: crop improvement for increased rainfed production. International Rice Research Institute; 2008.
11. Abugri SA. A study on farmers perception and preference in subscribing to drought-index crop insurance in the northern region of Ghana. Regional Economic Development Research. 2020;1(2):63-76.