Tracing Migration Patterns Across Indonesia
The rapid pace of rural to urban migration in recent decades has added new dimensions and challenges for urban planners and social policymakers. Gaining an understanding of the general trends associated with these movements, as well as the more nuanced push and pull factors is crucial for national development planning. To better understand the experiences of citizens across Indonesia who migrate from rural to urban communities, last year Pulse Lab Jakarta collaborated with Empatika on research commissioned by the World Bank. On the heels of the recent mobile phone data for official statistics workshop in Jakarta organised by the UN Global Working Group on Big Data and Statistics Indonesia (BPS), our team at the Lab decided to revisit the study which analysed anonymised mobile network data to uncover insights on migration in Indonesia and which were subsequently used to inform the sampling of qualitative research.
Titled “Urban Villages — Perspectives and Experiences of Migrants and Their Families on Moving from Villages to Cities”, the study combines qualitative research methods with big data analysis. The qualitative approach places emphasis on immersion in order to provide researchers with an opportunity to gather in-depth, multidimensional insights from people who make the decision to move from a rural to an urban community, and from those who choose not to make the move. The study sought to better understand the following aspects:
(a) General trends associated with urbanisation in Indonesia
(b) High-level trends in rural-urban migration and its socioeconomic opportunities and risks
(c) Determinants of economic mobility for migrants
(d) Socioeconomic and geographical dimensions of urban poverty
(e) Policy implications and lessons that can be drawn from these experiences
The scope of the study was outlined with a view to assisting the Government of Indonesia develop a comprehensive understanding of the phenomenon, to then design relevant policies to drive economic growth, reduce poverty and address a range of related challenges.
To capture a cross-sectional view of rural to urban migration in Indonesia, the World bank team selected four destination cities from across the Indonesian archipelago as case studies: Jakarta, Medan, Tangerang and Makassar (two areas were included within Makassar — one central and one peripheral). Based on surveys and previous research, these cities are among the fastest-growing in the country in terms of inbound migration rates. Sampling the source communities where these migrants come from, however, was more challenging due to a lack of up-to-date data, for example surveys and the census become outdated on this topic shortly after data collection.
Big Data Meets Thick Data
To help close this knowledge gap in terms of the availability and timeliness of information about where migrants are moving from, Pulse Lab Jakarta analysed aggregate geo-signatures from mobile phone data to identify migrants source communities.
How did we conduct this analysis? We teamed up with two of Indonesia’s leading mobile network providers, who were generous in giving us access to insights from their subscribers’ anonymised mobile network data for the period August to November 2017. The datasets included around 20 million individual network subscribers, accounting for roughly three-quarters of the total population in the four destination cities. The datasets were analysed in the following way:
Inferring subscriber home location: identify the cellular tower used most frequently by each subscriber between midnight and 5 AM over the course of a month to serve as each subscriber’s “home cell tower”. This is based on the assumption that people would be at their regular places of residence at that time of day. Also, by using a month we hoped to control for the effect of business trips and holidays.
Building User “Home Cell Tower” Profiles: a database was created capturing user “home cell tower” locations each month, which included information such as anonymised user, year, month, home cell tower.
Determining Source Community: each cell tower was paired with the respective name of the administrative village (desa) in which it is located. In other words, the cell towers were made to represent the desa in which they stood. Therefore, the “home cell tower” of each subscriber was also understood to be the desa where the subscriber resides.
Distinguishing Residents from Non-Residents: the “Home Cell Tower” dataset was further filtered to remove users who had not spent at least a month in the particular city of interest. By applying this filter, the datasets would now only contain individuals who could be considered to be residents of the city at some point during the timeframe.
Identifying Migrants: to determine which individuals could be considered as “migrants”, we filtered the city-specific anonymised user home tower database for anonymised users that have had home cell towers outside of the city of interest during the timeframe.
Generating A Migrant Database: for those identified as migrants, a new dataset was made that included their villages of origin and the current city in which they reside (destination), which we then summed and scaled to get a sense of the relative importance of different source communities of migrants.
Making Sense of the Numbers: The visualisations below show the source communities that citizens migrate from (a separate visualisation is generated for each of the four destination cities).
Using the Insights: Empatika then used the insights to inform the sampling of the rural communities for the qualitative research.
Uncovering New Insights
Beyond the findings of this study, the approach used in this research demonstrates how big data and non-conventional datasets (in this case mobile network data) can be used to better understand internal migration across Indonesia in an efficient, timely and cost-effective manner. In addition to identifying migrant source communities and destination cities, the results of this research also revealed essential insights on the volume and directional movements of rural to urban migration across Indonesia’s vast archipelago.
The maps of the source communities show unprecedented granularity, for instance indicating village-level migrant distributions for each destination city. As one would expect, cities such as Medan and Makassar drew most of their migrants from regions that were within close proximity, while migrants to Greater Jakarta came from a broader range of source communities. It is interesting to note that the maps for Medan and Makassar also show Greater Jakarta as a migrant source community, which suggests that the study may also have identified returning migrants.
One of the broader implications of the research results is that it is likely that existing metrics and categories significantly underestimate the scale of short-term rural to urban migration in Indonesia. For example, there’s a noticeable difference between the rates of short-term migration observed in this research, compared to existing Susenas data. The mobile network data analysis also generated timely results that were used to sample the qualitative research, enabling ethnographic researchers to visit major source communities of migrants and learn more about the various push and pull factors that contribute to the decisions made by people to migrate, and construct different profiles of migrants.
With mobile phone usage across Asia and the developing world increasing, the research approach can be applied elsewhere to provide useful policy insights to understand the needs of migrants and destination communities and improve areas of public service delivery.