*2.2. Data Analysis*

The data processing and analysis for this study employed partial least squares structural equation modeling, validity testing, reliability tests, descriptive analyses, and simulation of partial least squares structural equilibrium (PLS-SEM). In this study, Smart PLS 3.0 was used. Descriptive analysis was performed to obtain a wide description of the characteristics of respondents, including gender, age, education, and respondent profile, as well as to describe food security implementation using the mean. For the measurement, a 5-point Likert scale was used to determine the scale range. A Likert scale is used to assess respondents' attitudes, views, and perceptions of social issues.

Secondary data in the form of FSVA maps and SKPG maps of Pamijahan District from the Bogor District Government were analyzed in relation to the food security situation in the Bogor District. Primary data for quantitative analysis was collected using a survey method with questionnaire aids, as well as the results of in-depth interviews with key persons in Pamijahan District who understand the characteristics and local ecological knowledge that can maintain food security in the area, as well as residents and housewives who have babies or toddlers who are affected by floods and the COVID-19 outbreak. The population of this study was composed of 194 families affected by natural disasters in Pamijahan District, Bogor Regency. The locations affected by the disaster in Pamijahan District are Cibunian Village and Purwabakti Village. The determination of examples for research respondents is based on the formula from Frank Lynch as follows (Iskandar 2018):

$$n = \frac{NZ^2.p(1-p)}{ND^2 + Z^2.p(1-p)}$$

where:

*n* = Sample size (64)

*N* = Total population (194)

*Z* = Value in the area under the normal distribution curve (1.96)

*P* = Largest possible proportion (0.50)

*D* = Degree of deviation (10%)

By using a population-sample table based on the formula from Frank Lynch, for a total population (N) of 194 at a 95% confidence level and a degree of deviation (D) of 10%, the sample size (n) in this study amounted to 64.

### *2.3. Model Analysis*

The study began with an analysis of the food security situation in Pamijahan District, Bogor Regency, using secondary data from the results of the FSVA analysis and SKPG analysis from the Bogor Regency Food Security Service (Rimadianti et al. 2016). Termination of the indicators that most influence the status of food security after the disaster due to climate change and COVID-19 in Pamijahan District, Bogor Regency, uses path analysis with Smart PLS 3.0 software (Yudhanto et al. 2023). Then the results of the analysis above are compiled with the results of in-depth interviews to formulate managerial implications for maintaining food security in disaster-prone areas. This research uses a validity test, a reliability test, descriptive analysis, and partial least squares structural equation simulation (PLS-SEM).

In FSVA analysis, there was a change in the assessment indicators, which were originally nine indicators in 2017, to six indicators in 2019. The nine FSVA indicators in 2017 include the ratio of stalls to households, the ratio of shops to households, the ratio of people with the lowest welfare status, the ratio of households without access to electricity, the number of villages without adequate connecting access, the ratio of children not attending school, the ratio of households without access to clean water, the ratio of health workers to residents, and the ratio of households without residential facilities. Then the FSVA analysis was refined into six indicators covering the ratio of agricultural land area, the ratio of food supply infrastructure, the ratio of people with the lowest level of welfare, the number of villages with inadequate connecting access, the ratio of households without access to clean water, and the ratio of the number of villagers per health worker. The six indicators are weighted according to their level of importance, then a composite score is calculated and grouped according to the cut-off for each category.

The measurement model (outer model) and the structural model (inner model) are the two sub-models that make up the PLS-SEM analysis (Joe F. Hair et al. 2014). Construct convergent validity, discriminant validity, and reliability were assessed using the outer model (J. H. Hair et al. 2017). In addition, the inner model is used to assess the relevance of the path coefficient and R-square value. Two categories of variables are used in PLS-SEM. The first is the observed variable, sometimes known as the manifest variable because it can be viewed immediately. The second category is unobserved variables, sometimes known as latent variables because they cannot be observed directly (Joe F. Hair et al. 2014). Together with the six latent variables, there are 112 manifest variables. The model of this study can be seen in Figure 3. Previously, a content validation test was carried out on the research questionnaire to obtain validity (Ayu Dessy Sugiharni 2018), conducted by two experts, including academics and practitioners in the field of food security. The value of the content validity ratio (CVR) from the results of the questionnaire content validation test in this study was 0.89, so the questionnaire can be said to be valid.

**Figure 3.** Research models.

The hypotheses in this study can be described as follows:

**H1.** *Access to food has a positive and significant effect on food security in Pamijahan District*.

**H2.** *Food availability has a positive and significant effect on food security in Pamijahan District*.

**H3.** *The COVID-19 outbreak has had a positive and significant effect on access to food in Pamijahan District*.

**H4.** *The COVID-19 outbreak has had a positive and significant effect on food security in Pamijahan District*.

**H5.** *The COVID-19 outbreak has had a positive and significant effect on food availability in Pamijahan District*.

**H6.** *The COVID-19 outbreak has had a positive and significant effect on food utilization in Pamijahan District*.

**H7.** *Food utilization has a positive and significant effect on food security in Pamijahan District*.

**H8.** *Climate change has a positive and significant effect on access to food in Pamijahan District*.

**H9.** *Climate change has a positive and significant effect on food security in Pamijahan District*.

**H10.** *Climate change has a positive and significant effect on food availability in Pamijahan District*.

**H11.** *Climate change has a positive and significant effect on food utilization in Pamijahan District*.

Outer model evaluation and inner model evaluation are two evaluation models used in PLS-SEM data analysis (Cheung 2013). The external model is used to test the effect of latent variable indicators. Multicollinearity was used in this work to clarify the data without any visible bias before analysis. The absence of multicollinearity problems is a prerequisite for properly checking the outer model. Situations with substantial correlations or connectedness between indicators are called multicollinearity. A variance inflating factor (VIF) value of more than five indicates a multicollinearity correlation value, which is defined as a correlation value of more than nine. Multicollinearity occurs if the VIF value of the latent variable is greater than five. Actions that can be taken include reducing or eliminating indications with a high degree of association (J. H. Hair et al. 2017).

The evaluation of the outer model consists of three tests. The convergent validity test can be used to assess how well the manifest variable can explain hidden variables by looking at the loading factor above 0.50. When the average variance extract (AVE) results are greater than 0.50, the discriminant validity test is used to assess how much the latent and manifest variables differ from each other. Previous research explained the relationship between Cronbach's alpha above 0.60 and the composite reliability used to test composite reliability (J. H. Hair et al. 2017). The inner model is used to determine the effect of the independent variables on the dependent variable by comparing the coefficient of determination (R square) and the path coefficient (Ghozali and Latan 2015).

#### **3. Results and Discussion**

#### *3.1. Food Security Conditions in Pamijahan District*

The conceptual framework for regional food security considers food availability, access to food, and utilization of food, which guarantees that all individuals have the right to obtain food according to their needs. The opposite condition of food security is called food and nutrition insecurity.

Information related to food insecurity was analyzed using two analytical tools according to its causes. Based on the causes, food insecurity can be divided into chronic food insecurity and transient food insecurity. Chronic food insecurity was analyzed using the analysis map of Food Security and Vulnerability, or Food Security and Vulnerability Atlas (FSVA), and transient food insecurity using the Food and Nutrition Alertness System/SKPG analysis.

#### 3.1.1. FSVA Analysis

The FSVA analysis for Pamijahan District was carried out in 2017, 2019, 2021, and 2022. Table 1 shows the composite results of the FSVA analysis for 2017, 2019, 2021, and 2022 in the Pamijahan District.


**Table 1.** FSVA analysis of Cibunian Village and Purwabakti Village.

Source: Compiled by the author

The results of the FSVA analysis were divided into 2 groups, namely the food-insecure vulnerable group consisting of Priority 1 (very food-insecure), Priority 2 (food-insecure vulnerable), and Priority 3 (somewhat vulnerable to food insecurity). The food-secure group consists of Priority 4 (somewhat food-secure), Priority 5 (food-secure), and Priority 6 (very food-secure).

In 2017, Cibunian Village and Purwabakti Village were in Priority 1, namely in the vulnerable category of food insecurity caused by the low level of welfare of the population, the high number of children who are not in school, and the high number of households that do not have clean water facilities. There are differences in the approach and method of calculating the 2017 FSVA analysis, so it cannot be compared with the following year's FSVA analysis.

#### 3.1.2. SKPG Analysis

SKPG is an early warning system adopted from GIEWS, the Global Information and Early Warning System on Food and Agriculture (Shaw 2007). In the SKPG analysis, three aspects are used: the availability aspect (planting area and puso area), the aspect of food access (food prices), and the aspect of food utilization (nutritional status of toddlers). This study found an increase in food insecurity in terms of food availability and food access. Figure 4 presents the development of the results of the availability aspect analysis in Pamijahan District in the 2017–2022 period.

Based on the SKPG analysis, the food security status is classified into three categories: safe, alert, and food insecure. Food security generally looks stable in terms of availability in 2017, 2019, 2021, and 2022, but there has been an 8.3% increase in food insecurity. This

shows an increase in crop failures caused by natural disasters such as floods, droughts, and plant-destroying organism attacks as a result of climate change (UNFCC 2015). Figure 5 shows the situation of food security in terms of food access.

Changes in rice prices compared to prices in the previous three months are used to assess food access. The vulnerable category increased from 0% to 25% from 2017 to 2022, while food alert conditions increased from 2017 to 33.3% in 2022, indicating an increase in food costs beginning in 2019 owing to the COVID-19 pandemic. The COVID-19 epidemic poses a significant danger to food availability (Rume and Islam 2020). The COVID-19 pandemic has caused a 20% increase in global food prices (Laborde et al. 2020).
