How Data Science Can Give Further Understanding on Urban Poverty
Making the right poverty alleviation strategy that fulfills people’s needs based on insights and predictions generated using data science
As mentioned in my previous article, data is the new gold. With real-time analytics, tech companies can quickly understand their customers; such as their product preference, reaction to marketing campaigns, and the acceptable price point; without spending time and money on extensive market research. These insights are essential for management to decide on focus products, marketing content and pricing model. As cool as this seems, I start to think how we can use this technology to answer bigger and more complex problems like urban poverty. If data science can help us understand consumer behavior, then it should be able to help us understand the actual needs of those struggling with poverty.
Before going into detail, we must align our definition of poverty. Theoretically, there are two main types of poverty: absolute and relative. Absolute poverty means lack of access to basic needs, such as food, clothing, shelter and sanitation. Relative poverty is the inability to enjoy a decent standard of living. As standards vary among countries, this term is reflective to inequality. Poverty emerges in urban areas when the population increase is unmet by the number of economic activities and infrastructure capacity. This would lead to city slums, informal work and higher crime rates with much less financial security compared to the middle-to-high income groups. Displacement of slum dwellers would further trap them into poverty, as they would be moved away from their sources of income.
Governments, non-profit organizations and aid agencies have made several attempts to alleviate poverty. Examples vary from providing cash transfers and public housing to microfinance and skill-building training. However, many have been unsuccessful in achieving the intended goal. For instance, in the late 2000s, aid agencies distributed free mosquito nets to prevent the spread of malaria in African villages. Instead of using them, most villagers ended up using the nets for fishing or even sold them to the markets as they were more in need of additional income. Another case are public housing projects for existing slum dwellers that have been done in various Latin American and Southern Asian countries. Despite the perceived quality of the new housing buildings, the targeted users were still reluctant to move from the slums as the new location did not provide the same level of economic activity. This failure is a result of planning and policymaking that was done based on subjective views or comparison with rich standards instead of actual understanding the needs of the poor. Most projects actually do conduct public hearings and focus group discussions, however this dialogue has limitations due to communication gaps and/or unrepresentative participants. This is where data science comes in handy, the science of using vast amounts of data to generate insights and predictions for decision-making.
I found a good use case of data from Bill Gates and Rashida Jones’ podcast called ‘Ask Big Questions’. In a discussion on inequality, they featured Raj Chetty, a professor in public economics from Harvard. Using big data, Chetty was able to form an algorithm that predicts the likelihood of young people to avoid crime in specific neighborhoods in Los Angeles. After identifying those with high rates, he was able to see the pattern and characteristics of these areas. He also found that areas like Compton, that previously fell in the high-crime segment, could significantly improve within a few years. With these insights, he could trace the factors that led to Compton’s success and the potential application in other neighborhoods.
This case gave me an idea for developing countries that are facing issues of urban slums. For starters, we can formulate datasets on utility (electricity and water) flows by digitizing current data on cable and pipe installments. To be more advanced, we can integrate the utility system with IoT technology to monitor the reliability of electricity and clean water provision. Having the dataset on the utility provision and quality, we can match with another dataset extracted from geographic information system (GIS) data to map out neighborhoods. This new datasource would give insight on the accessibility of reliable electricity and water for each neighborhood or, in other words, identify slum areas. After obtaining these insights, we can dive deeper into the GIS data to understand the typology of slums. Key elements for this would be the physical condition of houses, main economic activities, and connectivity to public facilities and places of employment. An immediate response would be: so let’s provide power and water for these neighborhoods. But for me, this information should be used to gain a deeper understanding of the target groups. This would be the next phase of analysis: know the dwellers.
For the second phase, we can visit or seek ways to contact the community leaders in selected slums to help identify the residents with the use of a mobile application or a simple google form. The data to be collected would include place & date of birth, gender, occupation, level of income, level of education, and information on current and former dependents (resident’s children). The purpose of the last data point is to indicate the development of previous young residents, such as going to higher education and secure employment. The other data points would explain their current condition and why they chose these settlements in the first place. Having this data would help predict the likelihood of stable employment for young residents of these slums. If possible, it would be interesting to digitize school data to see the correlation between living conditions and academic performance. The incentive for community leaders and residents for filling in these is a clear support program. This would be much more difficult than it sounds, the right way of communication would be the key. Completion of this data will enable us to continue to the third phase: experimentation.
Many poverty reduction initiatives are determined based on successful practices overseas and simplicity in implementation. The problem with this is, how to be sure that it would fulfill the actual need of the poor? Taking public housing projects again for example, would the slum dwellers rather move away to a newly constructed neighborhood with reliable utilities or upgrade their current settlement? If they were given a more decent place to live, would their productivity increase? Moreover, would they stay or sell the unit for a higher price for extra cash? Experimentation would help planning authorities understand the need of the poor, like it does for companies in understanding consumer preferences. To answer the questions on housing, we can pick two neighborhoods with similar typology to pilot an experiment. One would be offered capital, design guidelines and adequate infrastructure to upgrade their settlement, while the other would relocate to a public housing complex with space for business and employment. The cost of the prior would be lower but the latter would be easier to implement. Community leaders can help manage and update the progress using digital platforms. After a few months, we can assess both programs by measuring the productivity (increase in earnings per resident) and satisfaction of residents. The upgraded group could have higher productivity from having more economic activity in the current location or vice versa since the relocated group might adjust well to the new environment. This would help understand which program would work, factors that led to unexpected results and potential to escalate to other areas. The experiment output can also be incorporated into data scientists’ the machine learning algorithm to predict the economic output of planned or future initiatives.
To further understand the urban poor, planning authorities could also collaborate with tech companies to utilize their data capabilities. These companies would be fintech and sharing economy platforms that have the same mission of poverty alleviation. The pandemic has shown the importance of digital transition for those relying on daily earnings from offline activities. The question to be answered is whether the urban poor would benefit more from cash transfers or skill building? One group would be provided cash in form of an e-wallet balance and the other special training to excel in the sharing platform, whether as a driver or food merchant, which would receive earnings in the same e-wallet. Both groups would be required to spend only using the e-wallet so data can be generated for analysis. Using the tech partner’s data platform, authorities can gain visibility on the additional earnings and spending patterns from both groups to understand which is more sustainable. If results are both as effective or even ineffective, we can experiment deeper into what is the right amount and frequency for cash transfers and what skills or incentive schemes that need to be improved.
An important element in poverty alleviation is empowering the poor to influence policy making and government fund allocation. A practice that was first proven successful in Brazil is participatory budgeting. This is where low income groups have their voice heard by being involved in the local government budgeting process. Current technology can also be used here by developing a digital budgeting platform. Community representatives can fill in their aspired budget using this platform and submit to planning authorities. Using data analytics and science, we can see which areas this group deems the most urgent to improve.
In conclusion, data science would give insight on the actual needs of the poor to formulate the right poverty alleviation program. With technology and data, we can understand the target group’s current and potential source of income, provide a platform to capture their aspirations and identify the most valued elements for social security. This would help us decide which economic opportunities to promote, policies that empower these groups and the form and level of aid to be distributed. There are still limitations as machine learning algorithms still have elements of subjectivity. Nevertheless, data science enables us to dive deeper into the issues faced by the people and continuously monitor for improvement. My closing remark: there is so we can explore with data, including the best ways to serve those struggling to access basic needs.
