-
Maragheh township, with 21 000 ha of orchards, is the center of fruit production in eastern Azerbaijan. The local government of Maragheh township with the collaboration of the department of agriculture has implemented the HCT project in the villages of Nova, Divrazm, Saeed Abad, and suburbs of the township, which cover 300 ha of orchards in the township, 8 HCTs also will be installed and put into operation in future. In the province of eastern Azerbaijan, 47 active HCTs have covered 4700 ha of orchards, as 700 million Rials have been allocated to cover the cost.
-
Table 1 displays the distribution of rural areas where the survey was done (n = 249), as farmers were randomly selected. The criteria for grouping respondents into low-risk and high-risk areas was based on the experience of these areas being encountered with risk of hail, as officially reported, and also, being under the cover of anti-hail devices installed in the high-risk regions.
Rural areas Frequency Percentage / % Nova 43 17.3 Saeid-Abad 26 10.4 Divrazm 31 12.4 Aghajeri 35 14.1 Taleb-Khan 20 8.0 Haji-Kord 26 10.4 Chekan 13 5.2 Alavian 22 8.8 Tazeh-Kand Sofla 18 7.2 Tazeh-Kand Olya 15 6.0 Total 249 100.0 Table 1. Frequency distribution of selected rural areas
-
To make sure the sufficient value of internal consistency of constructs has been achieved, we benefited from Cronbach’s Alpha, Composite Reliability (CR), and Average Variance Extracted (AVE), the cutoff measures of these indices were met (Cronbach’s Alpha > 0.70; CR > 0.70, AVE > 0.50; and CR > AVE). To measure CR and AVE, the following formulas (1) and (2) (Abadi et al., 2020) were used:
$$ \mathrm{C}\mathrm{R}=\frac{\displaystyle\sum {\left({\mathrm{\lambda }}_{\mathrm{i}}\right)}^{2}}{\displaystyle\sum {\left({\mathrm{\lambda }}_{\mathrm{i}}\right)}^{2} + \displaystyle\sum {\left({\mathrm{\sigma }}_{\mathrm{i}}\right)}^{2}} $$ (1) $$ \mathrm{A}\mathrm{V}\mathrm{E}=\frac{\displaystyle\sum {\left({\mathrm{\lambda }}_{\mathrm{i}}\right)}^{2}}{\displaystyle\sum {\left({\mathrm{\lambda }}_{\mathrm{i}}\right)}^{2} + \displaystyle\sum {\left({\mathrm{\sigma }}_{\mathrm{i}}\right)}^{}} $$ (2) where λ represents the standardized factor loading, i is the frequency of the items of latent variables, and (σ) is the error variance of items.The constructs were assessed by a 5-point Likert scale from 1 (strongly disagree) to 5 (strongly agree).
-
We met the assumptions of linear regression as follows, as stated by Hahs-Vaughn (2016).
• Independence: The assumption pertains to the independence of observations, resulted from the randomly selected respondents and randomly collected data.
• Collinearity: The assumption indicates that high correlations between independent variables collapse the power of the model in predicting the variance of the dependent variable.
• Linearity: In logistic analysis, the assumption of linearity is realized when there are linear relationships between all pairs of independent variables in each group of dependent variables.
• Outlier: Out-of-bounds or out-of-range data are data that, due to the strong impact on the mean, lead to inaccuracies in inference and interfere with the inference of results.
• The scale of independent variables: In logistics analysis, the type of scale used to measure independent variables should be cumulative response scale, interval, or ratio scale, although two-category variables can also be entered into the logistic regression function.
• Sample size: There is no consensus among statisticians on the exact value of sample size relevant us logistics analysis; larger samples might be sufficient for logistic regression.
As believed by Pampel (2000), the benefit of logistic regression is no requirement of the method to the assumptions obliged to be fulfilled in the linear regression like linearity, additivity, normality, and homoscedasticity.
-
Table 2 displays the items of the questionnaire, mean, standard deviation, and internal consistency of constructs, as assessed by Cronbach’s alpha, CR, and AVE.
Indices and items Code in SPSS Mean $ \pm $SD Composite reliability (CR) Average variance extracted (AVE) Skewness/
(Kurtosis)1. Attitude towards the use of HCT $(\boldsymbol{\alpha }$ = 0.97) The use of HCT is a useful career ATT1 3.33$ \pm $1.44 0.74 0.52 –0.38(–1.21) The use of HCT makes sense ATT2 3.51$ \pm $1.45 –0.61 (–.99) The use of HCT is a good career ATT3 3.42$ \pm $1.42 –0.50(–1.06) The use of HCT is a wise thing to do ATT4 3.42$ \pm $1.45 –0.50(–1.11) 2. Social Norms $(\boldsymbol{\alpha }$ = 0.96) People who are important to me think that I should use an HCT SN1 3.10$ \pm $1.26 0.1(–1.07) If I use an HCT, people who are important to me will approve of it SN2 3.15$ \pm $1.29 0.85 0.56 –0.02(–1.13) People who are important to me think that using an HCT is appropriate and useful SN3 3.08$ \pm $1.32 0.009 (–1.16) 3. Perceived Behavioral Control $(\boldsymbol{\alpha }$ = 0.64) If I want, I can easily use the HCT PBC1 2.33$ \pm $1.21 0.48(–0.88) I have the time and skills to use an HCT PBC2 2.63$ \pm $1.24 0.17(–1.12) I have the necessary financial resources to use the HCT PBC3 1.96$ \pm $1.09 0.78 0.66 0.92 (–0.05) I can decide on everything that happens on the farm PBC4 3.51$ \pm $1.29 –0.54(–0.86) 4. Intention $(\boldsymbol{\alpha }$ = 0.77) I plan to buy and use an HCT in the next year INT1 2.61$ \pm $1.46 0.30(–1.35) I intend to encourage other farmers to buy and use HCT INT2 2.97$ \pm $1.51 0.86 0.66 0.05(–1.42) I plan to attend HCT training classes INT3 2.51$ \pm $1.38 0.37(–1.22) 5. Compatibility $(\boldsymbol{\alpha }$ = 0.84) Buying an HCT fits my life needs Comp1 3.08$ \pm $1.22 –0.16(–1.06) Buying an HCT is harmless to the nature and climate of the region Comp2 2.91$ \pm $1.45 0.07(–1.35) Purchasing an HCT is commensurate with improving the welfare of farmers Comp3 3.17$ \pm $1.25 0.81 0.63 –0.24(–0.96) Purchasing an HCT is commensurate with improving the financial situation of farmers Comp4 3.08$ \pm $1.37 –0.11(–1.23) 6. Visibility $(\boldsymbol{\alpha }$ = 0.93) The results of using an HCT are easily visible to me Obs1 3.04$ \pm $1.31 0.88 0.67 0.02(–1.15) I can easily explain the results of using an HCT to others Obs2 3.20$ \pm $1.30 –0.17(–1.10) The results of using an HCT are clear to me Obs3 3.02$ \pm $1.32 –0.05(–1.14) 7. Relative Advantage (RA) $(\boldsymbol{\alpha }$ = 0.95) Use of HCT Reduces damage to facilities, tools, farm buildings, and gardens RA1 3.38$ \pm $1.17 –0.88(–0.13) Reduces damage to crops and fruit trees RA2 3.91$ \pm $1.15 –1.01 (0.18) It turns hail into the rain and reduces the impact on crops and orchards RA3 3.61$ \pm $1.36 0.79 0.63 –0.57(–0.93) Table 2. Items of questionnaire, mean, standard deviation, and reliability of research constructs measured by composite reliability (CR), average variance extracted (AVE)
Continued Table 2 Indices and items Code in SPSS Mean ± SD Composite reliability (CR) Average variance extracted (AVE) Skewness/
(Kurtosis)Prevents stress on farmers and gardeners during the seasons with the possibility of hail RA4 3.62$ \pm $1.38 –0.64(–0.82) It reduces financial losses to farmers RA5 3.72$ \pm $1.38 –0.72(–0.78) It reduces food crisis food shortages RA6 3.38$ \pm $1.48 –0.53(–1.14) 8. Ease of use $(\boldsymbol{\alpha }$ = 0.82) Working with HCT is easy for me Ease1 2.63$ \pm $1.09 0.19(–0.90) I have enough knowledge to work with HCT Ease2 2.30$ \pm $1.03 0.89 0.56 0.50(–0.55) I have enough skills to work with HCT Ease3 2.56$ \pm $1.13 0.31(–.76) I think it is easy for me to learn how to work with an HCT Ease4 3.22$ \pm $1.32 –0.17(–1.07) 9. Financial support (FS) $(\boldsymbol{\alpha }$ = 0.93) Government subsidies encourage farmers to adopt HCT FS1 4.12$ \pm $1.13 0.71 0.54 –1.20(0.47) The provision of credit facilities (loans) by the government to farmers increases the acceptance of HCT FS2 4.17$ \pm $1.09 –1.27(0.77) 10. Educational support (ES) $(\boldsymbol{\alpha }$ = 0.93) Training programs on how to use an HCT encourage farmers to adopt it ES1 3.48$ \pm $1.16 –0.59(–0.64) Government education support encourages farmers to adopt HCT ES2 3.60$ \pm $1.17 –0.66(–0.51) The training support of consulting services organizations encourages farmers to adopt HCT ES3 3.33$ \pm $1.19 0.85 0.54 –0.29(–0.89) NGO training support encourages farmers to adopt HCT ES4 3.25$ \pm $1.23 –0.33(–0.93) 11. Institutional support (IS) $(\boldsymbol{\alpha }$ = 0.95) Organizing farmers’ unions and associations, coalitions make it easier for farmers to buy and use HCT IS1 3.46$ \pm $1.28 –0.55(–0.76) Organizing farmers’ cooperatives make it easier for farmers to buy and use HCT IS2 3.36$ \pm $1.25 0.78 0.57 –0.37(–0.85) Organizing NGOs makes it easier for farmers to buy and use HCT IS3 3.24$ \pm $1.30 –0.26(–0.99) Notes: Skewness and kurtosis are at the range of –2 and + 2, SNs: Subjective Norms, PBC: Perceived Behavioral Control -
To find out the drivers of the behavioral intention to accept the HCT, we took the measure of the hypotheses using the linear regression model.
-
The nonlinear linkage of the dependent variable with the independent variables is converted into a linear relationship by turning probability changes into the log odds changes (Hahs-Vaughn, 2016). A unique coefficient of Exp (B) is accordingly obtained.
-
As maintained by Hahs-Vaughn (2016), there is a group of tests in logistic regression, called the omnibus tests of model coefficients. The obtained results of this type of test show that the value –2LL for the constant only model calculate by the summation of the chi-square value and –2LL of the model for Model 2 (high-risk condition), and the measure of 343.415 for Model 3. A point necessary to remember is that the properly fitted model is attained when the value –2LL for the whole model is less than the value –2LL of the constant model, as achieved in the study. These tests inquire about the appropriateness and qualification of the independent variables entered into the logistic regression function. In this case, a significance level of less than 5% indicates that the null hypothesis is not accepted and the independent variables are tailored that improve the predictability of the model. The chi-square value and the critical value demonstrate that the null hypothesis is rejected, as it assumes that the best prediction model is the only constant model. To better understand, the full model with predictors has the best performance in predicting the adoption behavior of the HCT.
The Cox & Snell and Nagelkerk multiple correlation coefficients are pseudo-multiple correlation coefficients and function as the same R2, also interpreted as similarly as multiple R2. These coefficients serve as effect size for logistic regression, and the Cohen correlation interpretation is used for interpretation. The measures of the Cox & Snell R square are 0.29, 0.35, and 0.25 for the three models. These values indicate that independent variables predict 25%–35% of changes in the logistics function.
The Hassmer-Lemshoff test is also used to check the accuracy of the classification of the variables. The statistical non-significance of the test is an indication of the appropriateness of the model. In other words, for a significance level greater than 5%, the Hassemmer-Lemshoff fitness test is not rejected, therefore, expressing that the classification is consistent with the observations. It is worth noting that this test is affected by the small size of the sample and if the sample size is less than 50, the interpretation of the results should be used with caution.
-
For the respondents that come under the category of low-risk of hail, as being pertinent to Model 1, the significance level of less than 5% for the land size (B = 0.59, WaldBHV = 4.314, Exp(B) = 1.181, P < 0.05) and less than 1% for the educational support (B = –0.87, WaldBHV = 6.785, Exp(B) = 0.417, P < 0.01) signify that these two variables are suited for predicting the variation of the acceptance of HCT. The value of B for the land size variable is equal to 0.59, which indicates that per a one-unit increase in land size, a 0.59-unit increase would occur in the acceptance of HCT. The Exp(B) values represent odds ratios, the odds ratio of 1.181 for the land size variable makes manifest that per a one-unit increase in the land size variable, the chance of accepting the HCT would become 1.181 times. Notwithstanding, the variables of attitude (B = 0.29, P < 0.05), compatibility (B = 0.38, P < 0.001), financial support (B = 0.18, P < 0.01) are the predictors of behavioral intention to accept the HCT.
-
Model 2 bears upon the respondents who are at high risk of hail, for these cases, the variables, such as compatibility (B = –0.71, WaldBHV = 4.185, Exp(B) = 0.489, P < 0.05), visibility (B = 0.50, WaldBHV = 4.210, Exp(B) = 1.659, P < 0.05), and educational support (B = 0.63, WaldBHV = 3.952, Exp(B) = 1.895, P < 0.5) contribute to explaining the variation of the acceptance behavior of the HCT. The value of B for the visibility is 0.50, indicating that for a one-unit increase in visibility, a 0.50-unit increase would occur in the acceptance of the HCT and the odds ratio of 1.659 for this variable shows that for one-unit increase in the visibility variable, the chance of accepting the HCT would become 1.659 times. For the educational support variable, the odds ratio is 1.859, showing that for one-unit increase in the educational support variable, the chance of accepting the HCT would be 1.859 times. Additionally, age (B = –0.25, P < 0.05), experience (B = 0.34, P < 0. 01), land size (B = 0.29, P < 0.01), compatibility (B = 0.22, P < 0.05), visibility (B = 0.16, P < 0.05) are determinants of behavioral intention to accept the HCT (Table 3).
Variables Low risk (n1 = 43, n2 = 61) High risk (n1 = 71, n2 = 74) Wald coefficient
(WaldBHV)Regression coefficient
($ {\; \beta }_{INT} $)–2LL Cox & Snell R2 Wald coefficient
($ {\mathrm{W}\mathrm{a}\mathrm{l}\mathrm{d}}_{BHV}) $Regression coefficient
($ {\; \beta }_{INT} $)–2LL Cox & Snell R2 Owner of system 0.001 105.90 0.29 0.001 137.53 0.35 Age (yr) 0.704 –0.16 1.551 –0.25* Educational attainment 0.005 1.008 Family size 2.653 0.12 0.261 –0.07 Experience (yr) 0.045 0.28 4.992* 0.34** Owner of land 0.043 0.023 Total land (ha) 4.314* 0.14 0.839 0.29** Membership in Cooperative 0.722 0.943 Extension contact 0.001 0.401 Access to credit 0.024 0.691 Income (US $) 0.856 –0.21 0.000 Attitude 0.016 0.286* 0.000 0.08 Subjective Norms 0.252 –0.108 0.576 0.05 Perceived Behavioral Control 2.626 0.070 0.392 0.02 Behavioral Intention 0.110 – 1.325 Compatibility 1.718 0.380*** 4.185* 0.22* Visibility 1.061 0.099 4.210* 0.16* Relative advantage 0.229 0.163 0.270 0.12 Ease of use 1.209 0.039 0.037 0.02 Financial support 1.041 0.183** 0.094 –0.01 Educational support 6.785** 0.066 3.952* 0.11 Institutional support 1.557 –0.024 2.732 0.13 Notes: Regression coefficient $ \; \beta $ is related to the estimation of intention, and –2LL pertains to the estimation of behavior. * significant at P < 0.05, ** significant at P < 0.01, *** significant at P < 0.001, R2 of Model 1 is 0.579, R2 of Model 2 is 0.472 Table 3. Results of the estimation of the model for BHV and INT, under low risk condition, non-adopter (n1 = 43) and adopter (n2= 61), under high risk, non-adopter (n1 = 71) and adopter (n2 = 74)
-
As displayed in Table 4, the results of running of Model 3 manifests that land size (B = 0.34, WaldBHV = 5.319, Exp(B) = 1.410, P < 0.05), compatibility (B = –0.49, WaldBHV = 4.304, Exp(B) = 0.612, P < 0.05), and visibility (B = 0.38, WaldBHV = 4.508, Exp(B) = 1.465, P < 0.05) are the drivers of predicting the acceptance behavior of the HCT. In this group, the highest value of B has a bearing on visibility, showing that per a one-unit increase in visibility, a 0.38-unit increase would occur in the acceptance of HCT. The odds ratio of 1.465 relating to the visibility variable refers to a one-unit increase in the visibility, the chance of accepting the HCT would become 1.465 times. Furthermore, the variables of age (B = –0.18, P < 0.05), experience (B = 0.26, P < 0. 01), land size (B = 0.18, P < 0.01), income (B = –0.16, P < 0.05), attitude (B = 0.16, P < 0.05), compatibility (B = 0.32, P < 0.001), visibility (B = 0.13, P < 0.05), relative advantage (B = 0.16, P < 0.05), and financial support (B = 0.10, P < 0.05) account for the variation of the behavioral intention to take an interest in the HCT.
Variables Adopters (n1 = 114) & Non-Adopters (n2 = 135) Wald coefficient (WaldBHV) Regression coefficient ($ \;{\beta }_{INT} $) –2LL Cox and Snell R2 Owner of system 0.001 268.60 0.25 Age (yr) 0.205 –0.18* Educational attainment 0.947 Family size 1.790 –0.01 Experience (yr) 2.844 0.26** Owner of land 0.059 Total land (ha) 5.319* 0.18** Membership in Cooperative 1.534 Extension contact 0.090 Access to credit 0.428 Income (US $) 0.135 –0.16* ATT 0.056 0.16* SNs 0.009 –0.004 PBC 0.112 0.06 INT 1.797 – Compatibility 4.304* 0.32*** Visibility 4.508* 0.13* Relative advantage 0.015 0.16* Ease of use 0.700 0.02 Financial support 1.530 0.10* Educational support 0.163 0.07 Institutional support 0.073 0.03 Notes: Regression coefficient $\; \beta $ is related to the estimation of intention, and –2LL pertains to the estimation of behavior. * significant at P < 0.05, ** significant at P < 0.01, *** significant at P < 0.001. R2 of model 3 is 0.61. ATT: Attitude, SNs: Subjective Norms, PBC: Perceived Behavioral Control, INT: Intention Table 4. Results of the estimation of the model for behavior and intention (I), for two groups of the adopter (n1 = 114) and non-adopter (n2 = 135)
The Cox and Snell pseudo R2 serves as an index to estimate the value of the adoption behavior, the measures of Cox & Snell pseudo R2 referring to two conditions of low and high risk and without low-high risk conditions are illustrated in Table 5, also the results of adjusted R2 of two conditions of low and high risk as well as without low-high risk conditions are seen in Table 6.
Models Low riskAdopter and Non-Adopter High riskAdopter and Non-Adopter Adopter and Non-Adopter –2LL Cox & Snell R2 –2LL Cox & Snell R2 –2LL Cox & Snell R2 Model 1 105.90 0.29 Model 2 137.53 0.35 Model 3 268.60 0.25 Note: Adopter and Non-Adopter indicated adopters and non-adopters of HCT Table 5. Cox and Snell pseudo R2 for the two conditions of low and high risk and without low-high risk conditions, three logistic regression models
Models Low riskAdopter and Non-Adopter High riskAdopter and Non-Adopter Adopter and Non-Adopter R R2 R R2 R R2 Model 1 0.800 0.579 Model 2 0.726 0.472 Model 3 0.784 0.590 Note: Adopter and Non-Adopter indicated adopters and non-adopters of HCT Table 6. Adjusted R2 of two conditions of low and high risk and without low-high risk conditions, three linear regression models
-
In order to find out about the interactions of the components of the conceptual model, we run a system dynamics model in the interface of Netlogo software, simulated the hypothesized interactions to visualize the interactions of farm-level and natural environments. A system dynamics model is a mathematical representation and realization of the developed interactions among system variables over time (Rehan et al., 2013).
In this section, we show the results of a system dynamics model of adoption intention, as three predictors of attitude, compatibility, and visibility were constant at the real range, and other variables were minimized by at least value. The best value for intention obtained as the interactions of the components of the model was tested by three variables of attitude, visibility, and compatibility.
Drivers of Forecasting the Behavioral Intention and Acceptance Behavior of the Hail Canon Technology (HCT): Using Logistic and System Dynamics Modeling
doi: 10.1007/s11769-023-1346-0
- Received Date: 2022-03-10
- Accepted Date: 2022-07-17
- Available Online: 2023-05-05
- Publish Date: 2023-05-05
-
Key words:
- climate change /
- resilience /
- Hail Canon Technology (HCT) /
- system dynamics /
- Iran
Abstract: Globally known about the drivers through which farmers are instigated to uphold and use Hail Canon Technology (HCT) is lacking. Therefore, this article intended to examine the drivers of forecasting the behavioral intention and acceptance behavior of the HCT, 249 apple farmers from northwestern Iran were recruited, including adopters (n1 = 114) and non-adopters (n2 = 135). The conceptual foundation included demographic theory, resource-based theory, theory of planned behavior, innovation diffusion model, and institutional support model. We also used the system dynamics model (SDM) in the Netlogo to assess the results of the conventional statistical approach (i.e., the logistic model). Authenticated the fitness of conceptual model with the data, logistic model manifests that the most outstanding determinants of the acceptance of HCT entail age, experience, total land size, income, attitude, compatibility, visibility, relative advantage, and financial support. Using the SDM, it was also shown that the results of the logistic model are confirmed by the SDM. In conclusion, management implications are available for the university extension to eliminate the adoption obstacles and stir up farmers to join in applying HCT, furthermore, researchers would avail themselves of remarks for future research.
Citation: | ABADI Bijan, HAGHANINIA Mohammad, 2023. Drivers of Forecasting the Behavioral Intention and Acceptance Behavior of the Hail Canon Technology (HCT): Using Logistic and System Dynamics Modeling. Chinese Geographical Science, 33(3): 549−564 doi: 10.1007/s11769-023-1346-0 |