The Development of Food and Nutrition Security Index at Provincial Level in Indonesia

This study aimed to develop a novel Food and Nutrition Security (FNS) index to be implemented at provincial level in Indonesia. Principal component analysis was used to calculate food and nutrition security index based on 23 indicators, data were obtained from provincial and national public report published in 2010 and 2013. These indicators were further grouped into three different dimension of food availability, food accessibility and food utilization. Principal Component Analysis (PCA) was then used to calculate each aspect and the FNS indices. The index was later used to categorize food and nutrition security performance of a region into one of three levels, namely secure, vulnerable or insecure. The application of this new FNS index found that most of Indonesia’s regions (87.5%) were categorized as vulnerable (50.0%) and insecure (37.5%) and only 12.5% had high food and nutrition security performances in 2013. Among all provinces, Bali had the best performance in 2013, while East Nusa Tenggara had the lowest performance in 2010 and 2013. In 2013, the food availability status of all the provinces were categorized as secure. However, their status on food accessibility and food utilization showed many of them fell into vulnerable and insecure categories. Hence, the effort to increase the national food and nutrition security performance should focus on improving food accessibility and food utilization.


INTRODUCTION
Global efforts to reduce hunger and nutrition problems, as well as to increase food security, have started since 1948 (Pangaribowo et al. 2013;Capone et al. 2014;Hjelm et al. 2016). Food and nutrition security concept is used by Food and Agriculture Organization (FAO) and other international to integrate nutritional aspect in food security development (Committee on World Food Security (CFS) 2012; Aliaga & Chaves-Dos-Santos 2014; Fanzo 2014).
A number of food security indicators have been established at global level, FAO classified the food security indicators into four dimension, of food availability, food accessibility, food utilization and stability (FAO 2016;Headey & Ecker 2013). World Health Organization (WHO) also determined 17 indicators to assess the nutritional acheivement (WHO 2015). Another study by Pangaribowo et al. (2013) classified food and nutrition security indicators into different level, of individual, household, and regional levels. In Indonesia, at the national level the Food Security Council issued a strategic policy document of Food and Nutrition Action Plan (SP-FNAP) in 2016, which determined the matrix of food and nutrition achievement covering 28 indicators (Suryana et al. 2016).
Food and nutrition security is multidimensional (Suryana 2014), thus the indicators to assess food and nutrition security achievement should include a set of indicators (Pemberton et al. 2016). The study results from Leroy et al. (2015) indicated that composite indicators were better for measuring food security achievement. Several indicies have been developed globally, there are The Global Hunger Index by International Food Policy Research Institute (IFPRI 2014), Rice Bowl Index (RBI), and Human Development Index (HDI). Related to assessment of FNS level, various measures have also been developed at global level (Herforth & Ballard 2016), one of them was the Global Food Security Index (GFSI) which was developed by The Economist Intelligence Unit Gantina et al. (EIU) (Pangaribowo et al. 2013). This measure is approriate to evaluate food security situation at country level, whereas for assessment at provincial level the GFSI needs to be adjusted due to the lack of data for some indicators.
The GFSI developed by EIU can only be used for country level. At national level, the food security indicators and measures have also been established by The World Food Programme (The Food Security and Vulnerability Atlas of Indonesia (FSC & WFP 2009& 2015a; Faharuddin (2012); Nurhemi et al. (2014). However, these tools and measures need to be redeveloped to suit relevant data available at the provinsial level. Based on the pausity of multi dimensional assessment tools for provincial level data, this study aimed to develop a novel FNS index, integrating the many food and nutrition security dimensions for provincial level in Indonesia. The index can be used by local government to evaluate the performance of their food and nutrition security progress and development.

Design, location and time
This was an explanatory study, using a cross sectional study design. The secondary data was collected between March to August in 2017. The research was conducted following several stages; identification of potential indicators, selection of indicators, assessment of index and clusteritation of provinces.

Sampling
This study used secondary data published in year 2010 and 2013. The sampling method was convenience sampling, involving data from 32 provinces which reflect the Food and Nutrition Security (FNS) dimension (food availability, food accessibility, and food utilization). The food utilization dimension data is available in the national basic health survey reported every three years (the latest are year 2010 and 2013), hence this study used data published in 2010 and 2013 for all the other FNS dimension.

Data collection
Food availability dimension data was collected from publication of National Food Balance Sheet from 32 Provinces published in 2010 and 2013, gathered from the Food Security Agency-Ministry of Agriculture of Indonesia. The data of food accessibility dimension was collected from The National Socioeconomic Survey by Statistics Indonesia year 2010 and 2013. Data on the rice prices were collected from publication of food prices in 2010 and 2013 gathered from the Ministry of Trade. Dimension of food utilization data was collected from publication and website of related institution. The food consumption level was taken from the Indonesian Food Directory published in 2015 by the Food Security Agency-Ministry of Agriculture. Access to clean water and sanitation were collected from Statistics Indonesia, and data on nutritional status was collected from the Basic Health Research published in 2010 and 2013. Data on food safety and foodborne diseases was collected from the National Agency of Drug and Food Control/NA-DFC yearly report published in 2010 and 2013.

Data analysis
Identification of potential indicators. Potential indicators were identified based on published literature (FAO, WHO, National Action Plan for Sustainable Development Goals, National Medium-Term Development Plan and SP-FNAP).

Selection of indicators.
There are 78 indicators identified in the first step, the research team then conducted a group dicussion and consultation to screen the and narrow down the indicators to 51 potential indicators as longlist indicators of food and nutrition security dimension. Further, the research team conducted a qualitative selection process to narrow down the potential indicators using the following criteria: the relevance to food and nutrition security concept, redundancy elimination, data availability and representation of meanings this process resulted in 23 selected indicators. These indicators were grouped into three dimension of food security, according to FAO-STAT food security indicators classification (FAO 2016) ( Figure 1).
According to FAO-STAT food security indicators classification (FAO 2016), 23 selected indicators then were grouped into three food security dimension: (a) five indicators for food availability dimension (X 1 : percentage of irrigated land; X 2 : energy availability level/EAL; X 3 :protein availability level/PrAL; X 4 :proportion Food and nutrition security index of energy availability from cereals (non-wheat) and tubers, X 5 : availability of animal source foods/ ASF); (b) six indicators of food accessibility dimension (X 6 : percentage of paved road; X 7 : GDRP/Gross Regional Domestic Product per capita; X 8 : CPI/Consumer Price Index of the foodstuffs; X 9 : proportion of people with energy intake >1800 kcal/day; X 10 : percentage of people living above the poverty line; and X 11 : stability of rice prices (1/% CV)); (c) 12 indicators of food utilization dimension (X 12 : energy intake level; X 13 : protein intake level; X 14 : Desirable Dietary Pattern Score; X 15 : percentage of households with access to clean water; X 16 : percentage of households with adequate sanitation access; X 17 : prevalence of non-wasted under-five children; X 18 : prevalence of non-stunted under-five children; X 19 : prevalence of non-underweight under-five children; X 20 : prevalence of exclusively breastfed infants at the age of <6 months; X 21 : percentage of non-underweight adults; X 22 : percentage of healthy and safe food; and X 23 : frequency of foodborne diseases).

Index calculation.
The Principal Component Analysis (PCA) was used to calculate the food availability index (5 indicators), food accessibility index (6 indicators), food utilization index (12 indicators), and the provincial food and nutrition security index (23 combined indicators of all dimension). Index development using PCA had also been used by Ariawan (2006) to calculate socioeconomic index using Indonesia Demographic and Health Survey data in 2002-2003. Napoli et al. (2011 used PCA to calculate the food insecurity multidimensional index of 61 countries in the world using 1995-2009 time series data. PCA method is an algorithm with an orthogonal principle, which is used to transform the allegedly correlated variables resulting in a set of uncorrelated linear values (Karamizadeh et al. 2013;Jollife & Cadima 2016).
The calculation of food and nutrition security index was performed through the following steps: (1) a set of food and nutrition security variables was analyzed using PCA, resulting in αij eigenvector value (indicator coefficient/weight); (2) the values of origin variables were standardized to z-score with the formula in which z value was the standard variable, x ij was the initial variable, was the mean of i th variable j th matrix, and s j was the standard deviation of the j th matrix; (3) calculation of food and nutrition security index total score of each province; i.e. multiplication result of standard variable (z-score) with its variable coefficient (α ij ). Normalizing the variable into z-score resulted in zero mean and standard deviation of 1, with negative to positive total score (index). The mathematical models used to calculate the Availability Index (AV), Accessibility Index (AC), Utilization Index (UT), and Food and Nutrition Security Index (FNSI) were written in equations 1-4.
The first principal component always performed the determination of PCA, although it was subjective (Jollife & Cadima 2016). According to Jollife and Cadima, this study used the first principal component (PC1) with a maximum variance value, it could explain the original information. The food and nutrition achievement were classified based on each indicator cut off. The cut off was decided by desk study analysis referring to the standards or targets of FNS indicators.
Dimension of food availability cut off referring to Food and Nutrition Action Plan 2010-2014, FSVA gathered from Food Security Agency (FSA) 2009 & 2015a, The cut off of food  (WHO 2012), and Strategic Plan of NA-DFC (2010). Based on these standards or targets, each indicators of this study were categorized into: high (secure), middle (vulnerable), and low (insecure). The value of each categorized were classified based on researcher analysis. Then all of data of each categorized proccesed with PCA and resulted FNS criteria, as seen in Table 1. These criteria were then used to classified the provincial food and nutrition security achievement. Data processing and analysis were performed using Microsoft Excel version 2010 and SAS program version 9.4.

Food and nutrition security index
The index was calculated for each pillar and for the food and nutrition security composite.
Food availability index (AV). Food availability index was calculated using the following mathematical models: AV 2010 =0.232 Irrigated Land + 0.564 EAL + 0.549 PrAL + 0.519 Cereals + 0.239 animal protein AV 2013 = 0.086 Irrigated Land + 0.573 EAL + 0.57 PrAL + 0.539 Cereals + 0.214 animal protein The national food availability index was -0.002 in 2010 and decreased to -0.29 in 2013 (Table 2). National level assessment showed that food availability domain is secure. However, provincial level assessment showed diverse result with Gorontalo Province occupied the top position in food availability aspect in 2010 and 2013. The main drivers were the high energy and protein availability levels, high proportion of energy availability from cereals (non-wheat) and tubers, as well as the animal protein availability.
Food accessibility index (AC). The calculation of food accessibility index used the following mathematical models: AC 2010 =0.489 paved road + 0.221 capita GRDP -0.486 CPI + 0.256 food secure + 0.532 non poor -0.352 rice price AC 2013 =0.547 paved road -0.132 capita GRDP + 0.293 CPI + 0.627 food secure + 0.415 non poor -0.178 rice price X 12 -X 23 =variables' values as stated in method The national food accessibility index was slightly increased in 2013 (Table 2) Table 2). The low proportion of the poor population in the two provinces indicated that the economic access to food was good. The increased of Bali GDRP/ capita (IDR 28,129.7 in 2013) and the decreasing proportion of people living below the poverty line influenced the Bali province food accessibility achievement (Finkayana & Dewi 2016).  (2013) and Riau Islands Province (2010) are the provinces with the highest food utilization performance. This due to their energy and protein intake levels (FSA 2015b), people's access to adequate sanitation and high prevalence of non-stunted under-five children in the two provinces (MoH 2013). The lowest performance of food utilization was found in East Nusa Tenggara Province (Tabel 2). Two-thirds of all households in Papua and East Nusa Tenggara had no adequate access to sanitation (proportion of >70%) (MoH 2013). A study by Tono et al. (2016) also found high ratio of households without adequate toilet facilities in most villages in the province which associated with the low nutrition status in the region.
The application of this novel index showed that four provinces (Bali, Jakarta, Yogyakarta and Riau Island) have the highest performance on food and nutrition index in 2013 ( Table 2). The high FNS achievement in the provinces were influenced by the relatively good achievement on the dimension of food accessibility. This condition indicated that food accessibility aspect is a significant determinants for food and nutrition security achievements, after the food availability aspect was fulfilled. While at the same time in these provinces, the slightly good achievement of food accessibility dimension also contributes to their achievement in the food utilization dimension.

Profile of food and nutrition security at provincial level in Indonesia
At the national level, most of the regions in Indonesia (84.4%) were categorized as vulnerable and insecure in 2010 ( Figure 2). The low score was associated with the low achievement in the food accessibility dimensions as reflected by the proportion of food-secure population, people living below the poverty line, and the increase in food prices that triggered the inflation rate, especially in Papua, Maluku and Nusa Tenggara.
Food and nutrition security achievement in 2013 was slightly better than in 2010, as can be seen in Figure 2 and 3 that the food and nutrition security index scores increased at the national level and in some provinces in 2013. The improvement from insecure to vulnerable categories was found in 16 provinces, and the secure category was found in four provinces. However, there were still 12 provinces that fell into insecure category (Figure 3).
The low achievement in food accessibility dimension hindered improvement in nutrition, thus affected the human resources development in the country. The Global Food Security Index (affordability, availability, quality and safety) report showed that Indonesia was ranked 71st (EIU 2016) and slightly increased to 69th (EIU 2017) the next year, this achievement still far below Thailand, Vietnam, and Srilanka (EIU 2017).
The performance of provincial food and nutrition security in 2013 slightly different with 2010, which were characterized by the high proportion of households with adequate sanitation access (PC1=0.378), high prevalence of non-underweight under-five children  i.e. 44.9 percent of villages belonged to severe food-vulnerable and food-vulnerable categories (FSA 2015a). This situation was also influenced by the low household access to electricity, clean water, sanitation facilities, and high levels of poverty (Tono et al. 2016). Moreover, based on basic health research (MoH 2010(MoH & 2013, East Nusa Tenggara Provinces had the lowest proportion of non stunted under-five children (less than 50%). Pangaribowo et al. (2013) stated that a healthy environment indicated by better access to sanitation and hygiene and health service infrastructure is also important. All of these aspects affect on the nutritional status of under five children.
There are several government policy and program to improve nutritional status such as the Food and nutrition security index (PC1=0.354), high prevalence of non-stunted under-five children (PC1=0.353), and the high proportion of households with access to clean water (PC1=0.346). This study showed that the provincial FNS achievement assosiated with nutritional status, food utilization dimension, and food accessibility dimension. Nutritional status of children under five was the outcome indicator of food security development and the key indicator of poverty and hunger alleviation as the global development targets in the achievements of MDGs and SDGs (WHO 2015;Suryana et al. 2016;FAO 2016).
The general characteristic of regions that were catagorized as food and nutrition secure indicated by the high proportion of households with adequate sanitation access. Jakarta and Special Region of Yogyakarta had the highest proportion (>80%) among other 32 provinces in 2010, while Bali Province had the highest proportion of water and sanitation access in 2013. had the best achievements, the nutritional problems in children under five were still found. The food availability dimension is generaly good nationally, however the food accessibility and utilization are still become the barriers for the food and nutrition security development. Hence, the action plan and strategy in midterm and longterm development to improve FNS performance should focus on improving access and utilization. Nutrition-sensitive program is the largest key component to improve the food accessibility performance and to sustain the food availability dimension, while nutrition-spesific program is a basic curative program to attain better food utilization performance.

ACKNOWLEDGEMENT
The authors would like to thank The Agricultural Extension and Human Resources Development Agency (BPPSDMP), Ministry of Agriculture as the scholarship sponsor. We would like to express our gratitude to National Institute of Health Research and Development, Statistics Indonesia, Food Security Agency, National Agency of Drug and Food Control of Republic of Indonesia, and Ministry of Trade for providing the research data. We would also like to give special thanks to Andi Alamsyah who has helped in the preparation of food and nutrition security maps.

AUTHOR DISCLOSURES
The author have no conflict of interest. In the past, Indonesia became a role model country for reducing high levels of malnutrition. However, nowadays there are times of stagnant, setback, loss of attention (World Bank Group 2018). Increase in income equality and purchasing power is one of a key strategy to improve food access. Applanaidu et al. (2014) explained that macroeconomic stability, economic growth and distribution are key factors in achieving food and nutrition security. Analysis from Purwantini (2014) also emphasized the importance of synergistic unification in food security development in Indonesia in line with the handling of nutritional problems. Victora et al. (2008) suggested that children aged two years that had suffered from stunting is the predictor for low-quality of human resources, which in turn would have negatively affected the nation's potential. Thus, the efforts to accelerate nutrition improvement require not only a nutrition spesific program but also a nutritionsensitive programs involving agricultural, social and education sectors, as well as food security (Ruel & Alderman 2013). Therefore, multisector collaboration and nutrition-sensitive programs are essential in the synergy of health, nutrition, food security and food safety developments to achieve sustainable development goals (Black et al. 2016).

CONCLUSION
PCA on 23 selected indicators of food and nutrition security showed that the food and nutrition security achievement was still far from the expectation, indicated from the low index score especially in the eastern part of Indonesia. The lowest food and nutrition security achievement was found in East Nusa Tenggara Province in 2010 and 2013, while the best achievement was found in Riau Islands (2010) and Bali (2013). Although the two provinces