Neighbourhood Socio Economic Disadvantage Index’s Analysis of the Flood Disasters Area at East Jakarta in 1996 and 2016

Flood is one of natural disasters that have often happened in East Jakarta. Flood can give several negative impacts and it can affect all aspects of society lives such as economics, political, cultural, socials and others. East Jakarta is an urban area which continuously grows and establishes to become a rapid area. It can be seen from the highest population density in East Jakarta (BPS, 2016) and categorized into a region prone to flooding based on data Prone Flood Map in 1996 and 2016. The higher population exists in East Jakarta, the bigger possibility of the negative effects of disaster it gets. The negative impacts of flood disaster can affect societies especially with socio-economic disadvantage. One of the index to measure socio-economic disadvantage is NSDI (Neighbourhood socio-economic disadvantage index). However, to adjust indicators used in NSDI with Indonesia statistical data compatibility, it needs further assessment and evaluation. Therefore, this paper evaluates previous main indicators used in previous NSDI studies and improves with indicators which more suitable with statistical records in Indonesia. As a result, there will be improved 19 indicators to be used in NSDI, but the groups of indicators remain the same as previous namely; income, education, occupation, housing, and population.


Introduction
Recently, NSDI has been used to analyse the relationship between the public urban green open space availability and neighbourhood socio-economic disadvantage [1].
When being applied in spatial context, the relationship between NSDI and other phenomenon can be assessed by using Geographically Weighted Regression (GWR). Furthermore, by using Principal Component Analysis (PCA) socio-economic indicators within initial loading can be removed if not correlated with dependent variables [1,2,3]. Therefore, this study will discuss indicators used in previous NSDI application, and evaluate similar indicators which are more suitable with statistical records in Indonesia. The analysis within this paper is purposed to adapt NSDI indicators as possible approach to measure social deprivation in communities located in areas where flood occurred in East Jakarta from 1996-2016.

Methods
Firstly, all possible socio-economic indicators exist in East Jakarta statistical records will be explored especially those reflect disadvantage such as number of employment and education level. The data exploration mainly focus on East Jakarta statistical records published in East Jakarta Statistical Bureau website [4]. The next step is to list all possible indicators within the website and compare it with previously used indicators in NSDI (Table 1). If there are many indicators adaptable with previous studies, it will select only the same number of indicators within one group.

Systematic review and discussion
At first, Australia Bureau of Statistics (ABS) in 2011 developed an index that summarizes variables to illustrate relative disadvantage, this index is called Index of Relative Socio-Economic Disadvantage (IRSD) [3]. In form of vector maps, IRSD can be displayed to rank the regions from the most disadvantaged socio-economic condition to least disadvantaged regions.
If an area has a low score of IRSD, it means that this area has high proportion of relatively disadvantaged people, while high score of IRSD reflects that the area has low incident of disadvantage [3].
The dimension used by ABS to compute IRSD: 1. Income variables, 2. Education variables, 3. Employment variables, 4. Occupation variables, 5. Housing variables, and 6. Other miscellaneous indicators of relative advantage or disadvantage. However, for some Asian countries the indicators used for every variable by ABS seems out of context such as % people who do not speak English well. These differences can be found when Li and Liu (2016) conducted research to observe the relationship between Urban Public Green Spaces (UPGSs) and Neighborhood Socioeconomic Disadvantage Index (NSDI) [1]. Li and Liu (2016) instead used variables from ABS, they used 14 socioeconomic indicators which was grouped into 5 groups to compute NSDI.
Therefore, this paper develops new indicators for socioeconomic disadvantages based on data availability and compatibility with East Jakarta Municipality condition. The comparison of socioeconomic disadvantage indicators can be observed from Table 1.
where expressed the component I's eigen value; Lj is indicator of j's loading score; is indicator j's standardized value.
In this case to measure the area with disadvantage it will use the percentage of people living with poverty. The consideration behind this, because Indonesian Bureau of Statistics has measured several indicators to be used in poverty computation, hence no need to compute low dimension per capita income. People living in poverty are the people whom its daily calorie consumption around 2100 kilo calorie or below, plus can only buy certain commodities in housing, cloth, education and health [4].
Since the growth of non-agricultural population proportion reflects rapid economic development [2], therefore in this study, we include the number of population which works in agriculture sector as indicator of disadvantage.
Since the Indonesian Bureau of Statistics has also counted the employed person whom in sick-leave as temporary unemployment, so in this study we do not include temporary unemployment as indicator of socio-economic deprivation, as suggested by previous research [1,2].

Principal Component Analysis
Principal component analysis (PCA) is a multivariate technique that analyses a data table in which observations are described by several inter-correlated quantitative dependent variables.
The goals of PCA are to (a) extract the most important information from the data table, (b) compress the size of the data set by keeping only this important information, (c) simplify the description of the data set, and (d) analyze the structure of the observations and the variables [5].
PCA works by ranking the component in the variance-covariance which has the highest percentage of variance (%p) among other components. In the multivariate statistics, variance of each variable/factor can be computed by using: In PCA, variance and covariance can be formed to be a single matrix consists of eigenvalue (λpn) and eigenfactors. Basically, eigenvalue in j column is the variance, therefore the percentage variance of every column can be computed by using; where %p is the percentage variance in component j, λ pn is variance in component j, and ∑λ pn is total variance [6].
To maintain the factor/variable, it can be observed from the factor loading value, which expresses the correlation (Rkp) between variable k with component p. The formula to calculate correlation between variable k with component p is: where akp =eigenvector (factor loading) for variable k, principal component p; = p-th eigenvalue (component); var k = variance of variable k (diagonal values in the covariance matrix) (Jensen,1996).
In previous study, the socioeconomic indicators with factor loading more than 0.75was selected, and removed the factor with value less than 0.75 [1].

Proposes framework
Therefore, based on aforementioned systematic review and discussion, it would be best for implementation of NSDI to replace earlier indicators used by Li