The Asian Financial Crisis of 1997 brought about complex challenges, which compelled governments across the region to shift from traditional to non-traditional approaches of shaping economic development and financial stability. In Indonesia, examining household level activities was important for understanding changes at the subnational and national levels. These insights proved useful for informing long-term development planning and efforts to increase resilience . In partnership with the macroprudential policy department at Bank Indonesia, Pulse Lab Jakarta recently brought together selected researchers across government, academia and the private sector for its 9th Research Dive for Development to undertake further research on this topic. This unique collaboration, which combined the Lab’s advanced data analytics capacity with Bank Indonesia’s domain expertise, facilitated a series of thematic research sprints to address important policy questions.
While exposure to poverty is generally understood as one of the threats to a household’s financial stability, multiple factors help to determine the measure of a household’s vulnerability. Household vulnerability is multidimensional and holistically takes into account the financial system of a household to identify ways to avoid risks that may result in extreme shocks to the welfare of the individuals.
As digital technologies become more accessible in both urban and rural communities across Indonesia, the way citizens go about their everyday lives has transformed significantly. From e-commerce to other fintech-enabled services, the adoption and use of new technologies has also led to an increase in the availability of alternative data sources. Such opportunities have opened up new areas for big data research relevant for macroprudential policy making in Indonesia.
Bank Indonesia, Indonesia’s central bank, is tasked with formulating macroprudential policies to mitigate systemic risks and help strengthen the overall financial system. Recognising the potential of big data and how it can be harnessed effectively to better inform its ongoing and future work, Bank Indonesia was interested in exploring how emerging data sources can be leveraged to uncover insights that might complement traditional approaches.
Given the format of PLJ’s research dives, which regularly convene researchers across Indonesia from different professional backgrounds to analyse multiple data sets in a hackathon-style, the venue was ideal for conducting data analysis research and engaging in policy discourse on household vulnerability as a multidisciplinary subject. Through a competitive process, a diverse cadre of researchers was selected, which was then divided into four groups based on the participants’ research interests.
Each group worked closely with an expert advisor, who provided feedback on how to approach assigned research tasks and what datasets may be useful based on the datasets available, the type of research methodology that might be most applicable, as well as sharing contextual knowledge about the task at hand. After spending several days analysing different data sets with the goal of gleaming useful insights, the research participants took the stage during final presentations to share their findings with representations from banking institutions, universities, embassies, private companies and the Indonesian Government in attendance.
The four areas of research on household vulnerability included:
(1) Understanding the housing mortgage default rate in Indonesia
Throughout developing countries, the practice of taking out a mortgage from a formal banking institution to assist with the purchase of a home is significantly lower compared to the rates in industrialized countries. Indonesia, which has more than 260 million people, requires over 800,000 new housing units per year; however the mortgage sector currently only finances around 200,000 units per year. To look into this issue, the team analysed Bank Indonesia’s cross sectional data from 2016 to 2017, 2017 National Socio-Economic Survey (Susenas) data and OLX’s property advertisement data from 2016 to 2017 to model the default rate of housing mortgages in Indonesia (the percentage of outstanding loans) at the city level. They found that among the many factors that contribute to the default rate, the loan to expenditure per capita ratio and aggregate interest rate were most dominant. Their findings indicated that the threshold estimation for loan to expenditure per capita ratio was 85 per cent, while the aggregate interest rate was marked at 11.5 per cent. These are two significant variables that affect the default rate at the household level, which means if either exceeds the estimated threshold, then a mortgage is likely to fall into default.
(2) Identifying indicators of household indebtedness at the provincial level
With the aim of identifying factors that affect household indebtedness and the reason for the loans, the second team examined Susenas data from 2018, 2018 to 2019 lending data from Bank Indonesia, and Julo’s fintech customer loan data from 2018 to 2019. The team employed descriptive statistics, logistic analysis, and text mining methods for this research, but due to data availability, was only able to look at the comparison of factors affecting loans in three provinces. Their analysis confirmed that borrowers’ income level and job expertise factored in the need to borrow certain loan amounts. Some of the primary reasons for taking out a loan included home renovation, venture capital and educational expenses, with further indications that borrowers who are married also applied for more loans related to school registration compared to unmarried borrowers.
(3) Using fintech data to assess customers’ financial vulnerability
The third group investigated the differences in loan vulnerability between fintech customers and banking loan debtors, by analysing Julo’s customers lending data from 2018 to 2019, Bank Indonesia’s lending data from 2018 to 2019 and 2018 Susenas data. Using logistic regression and cross sectional regression to assess the loan vulnerability factors, they observed that although fintech lending vulnerability (risks that arise in fintech’s lending system) is lower than that of formal bank institutions, it does not necessarily mean that fintech loans are less vulnerable than banks. Based on the research, they discovered that fintech loan vulnerability varies according to income level, marital status, profession, age, house ownership, number of dependents and geographical location.
(4) Evaluating how natural hazards impact loans-at-risk
Considering Indonesia’s vulnerability to natural hazards and the potential long term impact on households, the fourth team was interested in evaluating how natural hazards influence customers’ credit worthiness. To do so, they looked at 2018 to 2019 aggregate loan data from Bank Indonesia, displacement tracking matrix data related to the Palu earthquake in 2018, post disaster needs data from UNDP Palu, 2018 InaRisk data from Indonesian National Board for Disaster Management (BNPB) and Statistics Indonesia (BPS) 2018 population data on Palu, Sigi and Donggala. Their findings indicated that natural hazards played a significant role in the assessment of credit risk. In measuring customers’ credit-worthiness, the team suggested that a risk index based on location should also be factored in as one’s place of domicile may be more susceptible than others to natural disasters, especially considering availability and conditions of infrastructure in place to accelerate financial recovery after a disaster.
Research Opportunities Ahead
This household vulnerability themed research dive, similar to other research sprints PLJ has organised, was strategically designed to gather qualified researchers from across the country to conduct experimental research on topics that are priorities for the Indonesian Government. Whilst these research sprints produce useful results, the broader goal is for them to serve as springboards for further in depth research. Aside from technical papers produced for the event, participants were also encouraged to refine their research findings based on the expert peer reviews received, and submit them to national and international journals and conferences for further peer review.
This Research Dive was made possible through the engagement and support of expert advisors and data partners to whom we are extremely grateful. Special thanks to Mr. Bagus Santoso from the Department of Economics at Gadjah Mada University in sharing his expertise in econometrics and macroeconomics; Dr. Muhammad Nur Aidi from the Department of Statistics at Institut Pertanian Bogor (IPB) University specialising in statistical modelling; and Faizal Thamrin from Pulse Lab Jakarta with his wealth of experience in humanitarian assistance and disaster relief. We are also thankful for the data support from Bappenas’ Directorate of Macro Planning and Statistical Analysis, Bappenas’ Data and Information Centre, JULO, OLX, IOM, UNDP and Bank Indonesia.
To the team at Bank Indonesia’s Macroprudential Policy Department, Pulse Lab Jakarta expresses our sincere appreciation for joining forces and co-organising this successful event. We look forward to future research collaborations that may help to support the critical work of Bank Indonesia!
Pulse Lab Jakarta is grateful for the generous support from the Government of Australia.