Does A Spatial Influence on Fertility Rates of Ethnic Groups in Indonesia? A Spatial Analysis Using National Data

1. Introduction

The total fertility rate (TFR) for the past 50 years from 1971 to 2017 has continued to decline. This decline proves the success of the family planning program in suppressing the rate of population growth in Indonesia. This can be proven from the results of the 1971 Population Census (SP) which showed a TFR of 5.61 and the TFR figure continued to show a decline in the 1980 SP of 4.68 and the 1990 SP of 3.33. Furthermore, the Indonesian government through the Central Statistics Agency (BPS) surveyed to measure the TFR every 5 (five) years. The 1991 Indonesian Demographic and Health Survey (DHS) found that the TFR figure had decreased again to 3.02. In the 1994 DHS, the TFR figure also decreased again to 2.85 and continued to decrease in the 1997 SDKI to 2.78. Based on the TFR figures from the 1994 DHS to the 1997 DHS results, the decline in the TFR figure began to show a downward trend. The TFR rate that declined in the 1994-1997 DHS fell again to 2.6 in the 2002 DHS. However, the decline in the TFR rate from the 2002 DHS to the 2012 DHS stagnated for 10 years and based on the 2017 DHS, the TFR rate fell again to 2.4 per woman of childbearing age (BPS, 1987.; BKKBN et al., 1992; BKKBN et al., 1995; BPS et al., 1998; BPS et al., 2003; BPS et al., 2008; BPS et al., 2012; BPS et al., 2018).  

The government's success in controlling the population in Indonesia is faced with challenges in implementing the family planning program. Since regional autonomy was implemented, each district and city has been given direct authority to regulate policies at the regional level which have previously been vertically centralized. This has caused the policy authority which is an indicator of National Development performance to not be fully implemented optimally in each region, one of which is the family planning program as an effort to control the population and maintain reproductive health  (BKKBN, Bappenas, Kemenkes, UNFPA, Kedutaan Canada, 2021). 

The challenges faced after the decentralization include causing significant variations in the implementation of family planning programs throughout Indonesia. The Central BKKBN as a government institution tasked with controlling the population through various programs, one of which is the family planning program, continues to set policies to control the birth rate, but the implementation of these policies can be adjusted by district and regional authorities. The family planning institution at the district/city level which is combined with other institutions also influences the implementation of the family planning program in the community. The quantity and quality of contraceptive services vary depending on the commitment of the regional head to family planning, (United Nation Population Fund Indonesia, 2012). The impact of the implementation of regional autonomy has made policies vary in each region. This causes geographical or spatial factors to possibly influence birth rates based on ethnicity in a region (Ekoriano et al., 2025). The results of a study conducted in Taiwan tested the possibility of a relationship between regions on the variables of education level and decreasing mortality rates (Li, 1973). Other studies have also concluded that geographical aspects are something that can cause variations in birth rates (Goldstein dan Klüsener, 2014; Mansour et al., 2021). The spatial or geographical influence that occurs between regions is also due to the association between regional transformation and degree of access, which means that high levels of transformation will cluster in areas that also have high degrees of accessibility (Giyarsih, 2010). Therefore, this study aims to analyze the influence of regions on fertility rates in ethnic groups in Indonesia.

2. Method

2.1 Source of Data

This study uses secondary data sources from the 2020 Population Census (PC) Long Form, the data collection of which was carried out in 2022 by BPS with a total sample of 4,294,896 households with an estimated level in 514 districts/cities (BPS, 2023). The estimated sampling error is around 2.2 percent for the TFR estimate at the district/city level with a sampling stage using multistage techniques and a sample fraction of 5 percent (n/N), https://sensus.bps.go.id/metadata_kegiatan/index/sp2022/desain%20penarikan%20sampel

(accessed May 25, 2023).

2.2 Data Analysis

In this study, birth history data was not available. An indirect method was used to calculate the total fertility rate (TFR). Calculation of TFR using an indirect method consists of several calculation methods, including the Rele method, the Palmore method, the Gunasekaran-Palmore method, and the own children method (BPS, 2012). The TFR calculation in this study used one of the indirect methods, namely the Rele Method. The Rele method uses the ratio of children and women/child women ratio (CWR) and estimated mortality so that the birth rate can be known in one, two to 5 years before the census. The Rele method is used to estimate the GRR obtained from the Child Women Ratio (CWR) value and life expectancy. The data needed is quite simple, such as the number of children aged 0-4 and 5-9 years; the number of women aged 15-44, 15-49, 20-49, 20-54; total life expectancy; sex ratio at birth. The results obtained Gross Reproduction Rate (GRR) and TFR (United state cencus bereau, UNFPA, 2019;Gunawan et al., 2017).

BPS has published TFR figures at the national level to the district/city level using the own children method, so in this study the TFR figures are calculated based on the 15 largest ethnic groups at the district and city levels. As for calculating TFR. The calculation of TFR using the indirect method (rele method) is based on the following justifications, (1) the sample of women based on 15 ethnic groups in each ethnic group at each birth age group is not sufficient for the minimum sample at the district/city aggregation level, (2). The number of samples in each age group in each ethnic group is insufficient. Researchers only focus on the total birth rate (TFR), so the birth rate per age group (ASFR) is not needed, the relay method is a method with the closest birth rate calculation results to the biological child method compared to other indirect methods BPS DIY, https://dinp3ap2kb.slemankab.go.id/wp-content/uploads/2020/07/Paparan-TFR-Kab-Kota2020-2.pptx (accessed February 29, 2024). To facilitate and reduce human error, the TFR calculation uses the FERT mini software developed by the East-West Population Center in 1992. This software has been implemented so far by BPS in calculating indirect methods, one of which is the relay method (Salim et al., 2018). Geographical or spatial data plays an important role in life, from the simplest to the most complex. The simplest thing is a residential address that contains spatial information such as a postal code. Information on a location that includes the coordinates of the depth location is an example of the use of spatial data. Spatial data can be interpreted as everything related to the earth's space marked by geographic attributes such as location coordinates and other points so the Geographical Information System (GIS) has an important role in spatial studies (Lioyd, 2010).

This study aims to present and explore spatial data using the Geoda application. Geoda is software that was first introduced by Luc Anselin in 2002. Geoda software was developed to facilitate simple spatial data analysis to data analysis exploration, presenting both global and local autocorrelation visually spatially and ending in spatial regression (Anselin et al., 2006).

Spatial analysis in this study using Geoda software emphasizes the ability to visualize maps, namely maps with outlier categories, map smoothing, cortogram maps, and animated maps. Data exploration is presented in the form of statistical graphs, parallel coordinate plots, and conditional plots. Geoda's capabilities are also able to calculate spatial autocorrelation univariately and bivariately. Geoda facility can create a spatial weight matrix, according to contiguity, distance, and K-nearest neighbor criteria (Anselin, 2003).

The most important part of determining spatial autocorrelation is determining the most relevant neighbors of an area or region to be studied and where these neighbors are expected to influence the observation value of their neighboring areas. Measurement techniques to determine these neighbors are contiguity, distance, K-nearest neighbors, and Customized contiguity (Insee- Eurostat, 2018). Some characteristics of spatial autocorrelation are if there is a systematic pattern in the spatial distribution of a variable, it can be said to be spatially autocorrelated, if adjacent or neighboring areas are the same, this is called positive spatial autocorrelation, but if adjacent or neighboring areas are not the same, it is said to be negative spatial autocorrelation and random patterns indicate no spatial autocorrelation.

Moran's index value ranges between -1 and 1. Identify patterns using the Moran's I index value criteria,

a.      If the value of I > E(I) then it has a clustered pattern,

b.     If I < E(I) then it has a spread pattern,

c.      If I = E(I) then it has an uneven spread pattern

The higher the local Moran value, the adjacent areas have almost the same value or form a clustered distribution.

3. Results

Fertility rate (TFR) varies across the 15 largest ethnic groups in Indonesia. The TFR of the Javanese, Madurese, Balinese, Dayak, and Chinese ethnic groups has been below the national TFR value, which is 2.18 per WUS aged 15-49 years, but the TFR values ​​of other ethnic groups are still above the national TFR value. The Sasak ethnic group has the highest TFR value of 2.53, then the Acehnese ethnic group, at 2.43, and the Banten ethnic group at 2.39. The TFR value of the Chinese ethnic group is the lowest, at 1.49. Then the Madurese ethnic group, at 1.92, and the Balinese ethnic group, at 1.93 It is very important that culture can be used to provide information about differences in fertility and to consider program policies in population control. It is very important, cultural aspects are also used to inform people about differences in fertility.

In addition, cultural aspects can be used as part of government policies to control population growth and provide information to the public about the importance of reproductive health. The fertility of the Batak ethnic group (2.37), Minangkabau (2.29), Bugis (2.28), Banjar (2.24), Betawi (2.23) and Sunda (2.20) also requires control. The size of fertility in a particular culture shows how social construction is built into a civilization in viewing fertility in a particular culture in a region. To also find out the influence of culture in each region on fertility in 15 ethnic groups, a spatial analysis of fertility based on ethnicity in Indonesia is carried out, namely the Sasak ethnic group, Acehnese ethnic group, Banten ethnic group, Malay ethnic group, Batak ethnic group, and Minangkabau ethnic group. This section will then present and test how the fertility value based on ethnicity is globally through the global Moran's Index value and P-value so that it can be seen that each region influences the adjacent or bordering areas through the Moran's Index spatial test (Moran's I) and Local indicator of spatial autocorrelation (LISA) using Geoda software.

Table 1: Global Moran Index Value, E(I), Standard Deviation, and P-Value of Ethnic Fertility in Districts/Cities in Indonesia 2022

Source: Geoda, data processed, 2024

Based on Table 1, the results of the univariate test with a global significance value (<0.05) are obtained, namely the Sasak ethnic group (0.042), Banten ethnic group (0.037), Malay ethnic group (0.001), Javanese ethnic group (0.001), Sundanese ethnic group (0.001), Madurese ethnic group (0.001), Banjar ethnic group (0.014), Balinese ethnic group (0.002), Dayak ethnic group (0.001) and Chinese ethnic group (0.018) which means that there is spatial autocorrelation between the fertility of the Sasak ethnic group, Banten ethnic group, Malay ethnic group, Javanese ethnic group, Sundanese ethnic group, Madurese ethnic group, Banjar ethnic group, Balinese ethnic group, Dayak ethnic group and Chinese ethnic group with the fertility of the Sasak ethnic group, Banten ethnic group and Malay ethnic group, Javanese ethnic group, Sundanese ethnic group, Madurese ethnic group, Banjar ethnic group, Balinese ethnic group, Dayak ethnic group and Chinese ethnic group in Indonesia. Table 1 also informs that the fertility of the Sasak, Acehnese, Malay, Javanese, Sundanese, Madurese, Banjarese, Balinese, Dayak, and Chinese ethnic groups shows positive spatial autocorrelation because they have a Moran Index I> E(I) meaning that the pattern of the relationship between the fertility of each ethnic group between districts/cities is clustered. Meanwhile, the Acehnese, Bataknese, Minangkabau, Betawi, and Bugis ethnic groups have a significant P-value (>0.05) which means that there is no spatial autocorrelation between fertility in each of these ethnic groups. The results of the autocorrelation test presented in Figure 1 globally on 15 ethnic groups using GeoDa software obtained Moran's I values ​​in the 15 ethnic groups showing values> E(I), which means that there is positive spatial autocorrelation in the six variables between sub-districts. The Moran's I value for the six ethnic groups is greater than E(1), so it can be concluded that it shows the pattern of fertility distribution in each ethnic group is clustered and has similar characteristics in adjacent districts/cities. Moran's scatterplot and I value can be seen in Figure 1.

Figure 1: Moran’s I Scatterplot of Variables

Source: data processed, 2024

The LISA statistical test in this study is only presented for the three highest ethnic groups as follows below such as Sasak ethnic groups, Aceh ethnic groups and Banten ethnic groups. Through Map 1, a LISA statistical test was obtained which informed that fertility in the Sasak ethnic group was the highest TFR nationally and spatial analysis showed through (a moran scatter plot) which was categorized into four categories as follows:

a.      Quadrant I (High-High), has positive autocorrelation because the observed value of fertility in the area is high and is surrounded by surrounding areas that are also high. This area is also called a hotspot area. The areas or regencies/cities included in this category are 5 areas, namely South Solok, Bandung City, Bandung Regency, and West Sumba Regency.

b.     Quadrant II (Low-High), This quadrant has negative autocorrelation, because the observed value of the area is low and is surrounded by surrounding areas that have high values. The areas included in this category are two areas, namely Purwakarta and Palopo.

c.      Quadrant III (Low-Low), has positive autocorrelation, this is because the observation value of the region is low and is surrounded by 4 low surrounding areas, namely Kampar, Bintan, Tanjung Pinang, and Tapin.

d.     Quadrant IV (High-Low), has negative autocorrelation. This is because the observation value of the region is high and is surrounded by low areas. There are three locations in this category, namely Langkat, North Jakarta, and Banjarmasin.

The results of the LISA statistical test in grouping regions into four quadrants with fertility based on ethnicity still require caution in concluding the spatial analysis in this study, because the adequacy of the minimum sample in each region in the calculation of fertility does not all meet the minimum limit (30 samples).

The following table 2 in the Sasak ethnic group below presents regions with categories of regions included in the high-high, low-low, low-high, and high-low categories. The LISA test presents Moran's I value and p-value in each region included in the category.

Table 2: List of Regions based on Quadrants I (1), II (3), III (2) and IV (4)

Sasak Ethnic Group

Source: Geoda, data processed

Based on Map 2, informs the fertility of the Acehnese ethnic group using spatial analysis showing through (a moran scatter plot) which is categorized into four categories as follows:

a.      Quadrant I (High-High), has a positive autocorrelation because the observed value of fertility in the region is high and is surrounded by surrounding areas that are also high. This area is also called a hotspot area. The regions or districts/cities included in this category are two regions, namely West Lampung and Purworejo.

b.     Quadrant II (Low-High), This quadrant has a negative autocorrelation, because the observed value of the region is low and is surrounded by surrounding areas that have high values. The regions included in this category are six regions, namely Bangka, Cilegon, Serang, Bengkayang, Balikpapan and Samarinda.

c.      Quadrant III (Low-Low), has a positive autocorrelation, this is because the observed value of the region is low and is surrounded by surrounding areas that are also low, one region, namely Ogan Ilir.

b. Quadrant IV (High-Low), has negative autocorrelation. This is because the observation value of the region is high and surrounded by low areas. There is not a single region in this category. The following is a list of regions based on the fertility of the Acehnese ethnic group in each cluster of the LISA statistical test results.

Table 3: List of Regions based on Quadrants I (1), II (3), III (2) and IV(4)

Aceh Ethnic Groups

Source: Geoda, data processed

Furthermore, Map 3 informs the fertility of the Banten ethnic group using spatial analysis showing through (moran scatter plot) which is categorized into four categories as follows:

a.      Quadrant I (High-High), has a positive autocorrelation because the observed value of fertility in the region is high and is surrounded by surrounding areas that are also high. This area is also called a hotspot area. The regions or regencies/cities included in this category are four regions, namely Aceh Tamiang, Asahan, Serdang Bedagai, and Batu Bara.

b.     Quadrant II (Low-High), This quadrant has a negative autocorrelation, because the observed value of the region is low and is surrounded by surrounding areas that have high values. The regions included in this category are two regions, namely Palembang City and Bandung Regency.

c.      Quadrant III (Low-Low), has a positive autocorrelation, this is because the observed value of the region is low and is surrounded by surrounding areas that are also low, one region, namely Depok.

d.     Quadrant IV (High-Low), has a negative autocorrelation. This is because the observation value of the area is high and surrounded by low areas. There is one area, namely Bandung City. The following table presents a list of 4 areas based on the fertility of the Banten ethnic group in each cluster of the LISA statistical test results.

Table 4: List of Regions based on Quadrants I (1), II (3), III (2) and IV (4)

Banten Ethnic Groups

Source:  Geoda, data processed

Figure 2: Fertility cluster Map of Sasak Ethnic Group

Figure 3: Fertility Cluster Map of Aceh Ethnic Group

Figure 4: Fertility Cluster Map of Banten Ethnic Group

4. Discussion

The results of the spatial analysis described above show the influence of regional or spatial factors on fertility based on ethnic groups in Indonesia. Several studies in the world also confirm this, such as in South Korea and 21 European countries, Jung et al. (2019); Campisi et al. (2020) explain the relationship between socio-demographic, economic, cultural, and spatial factors on births (TFR). Regional influences occur due to similarities in policies, reproductive health programs and cultural similarities in a region which can influence fertility measures in other regions. With the proof of fertility based on the 15 largest ethnic groups in Indonesia through spatial testing, the regional and local cultural factors in a region can be used as considerations for determining policies and programs, in providing interventions and planning in the future in controlling fertility in Indonesia. The influence of fertility measures on each ethnic group spatially or regionally is certainly directly related to the community in the region. The influence of fertility based on a particular ethnicity spatially occurs, and attitudes and behaviors increase along with the increasing spread of these attitudes and behaviors from one part of a community to another (diffusion process). Jhon B. Casterline (2001) explains that birth is a consequence of attitudes and behaviors that were initially very rare or not involved in a particular society, in other words, social interaction between individuals in a community occurs because they see and observe each other's fertility behavior. This will change their social environment, which in turn affects their decision-making about fertility.

Acknowledgments: Many appreciate our gratitude to the Directorate of Statistical Dissemination of Central Statistics Indonesia and the Data and Information Center of the National Research and Innovation Agency (BRIN) and Central Statistics (BPS) which have facilitated our access to the 2020 Population Census Long Form and national socio-economic microdata.

Funding: This work was supported by the National Research and Innovation Agency under Grant [number 22/STD/VII/2023 dated July 25, 2023 and number 42/STD/XI/2023 dated November 27, 2023.

Conflicts of Interest: The authors declare no conflict of interest.

Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Institutional Review Board Statement: Our research uses secondary data which has been approved for use by the Directorate of Statistical Dissemination of the Statistics Indonesia number 24/LADU/0000/07/23 and 49/LADU/0000/11/23; Data and Information Center of the National Research and Innovation Agency with number 22/STD/VII/2023 dated July 25, 2023 and number42/STD/XI/2023 dated November 27, 2023.