*3.2. Determination of the Most Influential Indicators on Food Security in Pamijahan District, Bogor Regency*

#### 3.2.1. Common Method Bias

A problem known as common method bias (CMB) occurs when the measurement technique used in SEM studies causes problems rather than a network of causes and effects among the latent variables in the model being investigated (Kock 2015). In this study, Smart PLS is used to identify CMB threats. The test indicates that the VIF element is lower than the 3.3 threshold. This shows that the model is free from CMB (J. H. Hair et al. 2017; Kock 2015).

#### 3.2.2. Model Measurement

The suitability of the measurements is checked using validity and reliability standards. The ability of a measuring device (or object) to consistently produce the same result is known as reliability. Validity is a measure of how accurately an understanding is measured by measuring instruments (items). Because there is a multicollinearity condition, actions that can be taken include reducing or eliminating indications with a high degree of association. The VIF/correlation matrix measurement results at the manifest variable level for all latent variables in Table 1 are listed below, while the summary of model measurements after the multicollinearity test is shown in Table 2.

The assumption that must be met when analyzing the outer model is that there are no multicollinearity problems. Multicollinearity is a problem of interconnection or strong correlation between indicators. The multicollinearity correlation value is indicated by a correlation value of more than 9, which is indicated by a variance inflating factor (VIF) value of more than 5.


**Table 2.** Correlation matrix information.

Source: Compiled by the author. Remarks: AP: access to food; PC: COVID-19 outbreak; ETC: food security; PP: utilization of food; KP: food availability; PI: climate change.

If there is a latent variable VIF value of more than 5, then there is multicollinearity. The consequent actions that can be taken include dropping or removing indicators that have a strong correlation. Following are the results of VIF measurements at the manifest variable level for all latent variables shown in Table 3.



Source: Compiled by the author.

All of the statements in the questionnaire were declared valid at the 5% significance level, where the r count exceeded the r table based on the results of the validity and reliability of 64 samples (0.361). In this study, the value of Cronbach's alpha for each variable was greater than 0.06, which indicates the dependability of the variable.

The Fornell-Larcker criterion, measuring the degree of anticipated "difference" between items for various factors, was used to test for discriminant validity. The square of the correlation was compared with the AVE of each factor to assess the discriminant validity of the model. In the other case, the correlation coefficient between factors is considered to have very good discriminant validity when the AVE is greater than the correlation coefficient between factors and other factors. The values on the diagonal represent the square root of AVE (J. H. Hair et al. 2017), while the values outside the diagonal are correlations. The results of discriminant validity are shown in Table 4.


**Table 4.** Discriminant validity matrix.

Source: Compiled by the author.

#### 3.2.3. Result Analysis

With a loading factor value of 0.50 and no multicollinearity problems, 26 of the 112 tested indicators passed the convergent validity test, according to the outer model assessment (Table 1). In the discriminant validity test, each latent variable has an AVE value greater than 0.50. Figure 6 depicts the final model.

**Figure 6.** Final research model.

All latent variables in the composite reliability test are known to have a Cronbach's alpha value of 0.60, and these variables meet the requirements for the composite reliability test. This follows the concept that the research model can be accepted as valid and credible by eliminating eleven variables. The inner model is assessed by analyzing the R-square value and the path coefficient. In addition, this model uses the R-square value to determine how much influence the exogenous variables have on the endogenous variables.

The results of the convergent validity test are determined based on the principle that the measurement of a construct must have a high correlation (Joseph F. Hair et al. 2019). The convergent validity of a construct with a reflective indicator was evaluated by average variance extracted (AVE). The minimum AVE value is equal to or greater than 0.5, while an AVE value of 0.5 or more means that the construct can explain more than 50% of the item variance.

Analysis of *discriminant validity*, which is carried out to ensure that each concept of each latent model is different from other variables, refers to value cross-loading, or the *Fornell-Larcker criterion*, from the manifest variable to its latent variable. *Discriminant validity* aims to test to what extent the latent construct differs from other constructs. A high value indicates that a construct is unique and able to explain the phenomenon being measured. All latent construct values must be greater than the correlation with other constructs so that

the discriminant validity requirements in this model have been fulfilled. The calculation results of the *Fornell-Larcker criterion* for *discriminant validity* are shown in Table 5.


**Table 5.** Fornell-Larcker criterion for discriminant validity.

Source: Compiled by the author. Remarks: AP: access to food; PC: COVID-19 outbreak; KTP: food security; PI: climate change; KP: food availability; PP: food utilization.

All latent variable values in the model have findings that are greater than the correlation values of the other latent variable constructs, according to measurements using the Fornell-Larcker criterion, as shown in Table 5. As a result, this model meets the criteria for discriminant validity (J. F. Hair et al. 2019). Furthermore, it may be argued that this model is both usable and legitimate. To test and evaluate the inner model, R-square and path coefficient significance are used. The impact of exogenous variables on endogenous variables is measured using the R-square. Table 6 displays the R-square calculation's outcomes.

**Table 6.** Results of the R-Square.


Source: Compiled by the author.

According to Table 6, which is in the moderate category, the food access variable can explain the food security variable by 41.3%, and the food utilization variable can explain the food security variable by 31.9%. Food availability and food security have a 2.5% association, which is considered to be low. A 35.5% link exists between food security and the accessibility, utilization, and availability of food. There are additional elements that influence food security in this situation, including the stability of the food system.

The route coefficient significance test, which uses the bootstrap method, produced the original sample values, *p*-values, and t-statistic values to assess the research model and research hypotheses. The variables' relationships to one another are made evident by the initial sample values. A variable transforms from a negative to a positive state when it has a positive influence, and vice versa. A hypothesis test's significance can be determined by calculating the t-statistic value. The *p*-value is less than 0.05, which indicates that the hypothesis is supported. Table 7 displays the value of the route coefficient, which depicts the link between all variables.

Food security is positively impacted by access to food, with a *p*-value greater than 0.05. This shows that despite a variety of circumstances and conditions that might make it difficult to access food, residents of Cibunian Village and Purwabakti Village, Pamijahan District, will continue to make an effort to meet their food needs, so H1 is not accepted.

With a *p*-value greater than 0.05, food availability has a beneficial but small impact on food security. As a result, H2 is not accepted. This shows that food is still available for the citizens of Cibunian Village and Purwabakti Village in the Pamijahan District, though the quantity and variety have declined at the home level and in food supply facilities.


**Table 7.** Path coefficients.

Source: Compiled by the author.

With a *p*-value less than 0.05, the COVID-19 outbreak has a positive and significant influence on the food access variable, indicating that the COVID-19 outbreak has caused a fall in the income of residents in Cibunian Village and Purwabakti Village. Furthermore, the COVID-19 outbreak has raised food prices, impeded food distribution, and forced the closure of food supply facilities. In the event of a COVID-19 outbreak, there is a policy of restricting community activities that influence the food delivery system, so H3 can be accepted.

With a *p*-value greater than 0.05, the COVID-19 outbreak has a positive but insignificant influence on food security. This demonstrates that, despite the COVID-19 outbreak, Pamijahan residents continue to consume food, but the amount consumed has dropped, particularly consumption of animal food, causing family members, including toddlers, to feel unwell more frequently. As we know, the output of food security is good nutritional and health status, which is shown in the monthly increase in the weight of children under five and the consumption of food by the needs for an ideal healthy life, resulting in H4 being accepted.

The COVID-19 outbreak has a positive but insignificant effect on food availability, with a *p*-value of more than 0.05. This shows that food is still available at the household level and in food supply facilities, but the amount and type have decreased. Apart from that, the COVID-19 outbreak has also caused the reserve fund to buy food to decrease, so H5 was not accepted.

The COVID-19 outbreak had a beneficial and substantial effect on food utilization, with a *p*-value less than 0.05. Because of the COVID-19 outbreak, the variety and amount of food consumed have decreased, while the use of rice fields for food sources has increased, allowing H6 to be accepted.

Food utilization has a positive but insignificant effect on food security, with a *p*-value greater than 0.05. This suggests that residents of Cibunian Village and Purwabakti Village continue to eat every day, but the frequency and kind of food consumed have decreased. Furthermore, the quality of drinking water has deteriorated, causing H7 to be accepted.

Climate change has a favorable and significant impact on food access, with a *p*-value less than 0.05. This demonstrates that floods caused by climate change reduce the amount of money available for food purchases. Flooding also forced food providers to change modes of conveyance due to a lack of road access. Furthermore, during flood conditions, food supply facilities such as booths and shops are closed, so H8 can be accepted.

With a *p*-value less than 0.05, climate change has a positive and substantial effect on food security. The nutritional and health status of a community reflects the region's food security situation. This suggests that floods caused by climate change have made residents of Cibunian Village and Purwabakti Village sicker more frequently; hence, H9 is accepted.

With a *p*-value greater than 0.05, climate change has a positive and substantial effect on food availability. This demonstrates that food is still available during floods caused by climate change, but the amount and type at the household level, as well as the food provider's advice, are reduced, hence H10 is refused.
