Home Blockchain Microfinance and alternative data meets the world of Blockchain

Microfinance and alternative data meets the world of Blockchain

10
0

I interviewed modern-day Microfinance institutions that have issued a combined $750m in micro loans to over 4.5m customers globally, and here’s what I found.

Source: Shutterstock

What could distributed ledger technology mean for unbanked and underbanked populations? In particular, how might distributed ledger technology revolutionize the Microfinance industry? Is the Microfinance industry even in need of change?

These are questions that I’ve been asking myself repeatedly over the last 18 months. Rather than sit in my bedroom and ponder, I decided to spend a few months doing some research. I interviewed a selection of modern-day Microfinance institutions (MDMFIs) that have, in recent years, issued a combined $750m in micro loans to over 4.5m customers in emerging economies across the globe. First, I assessed MDMFI performance and impact. Second, I examined how Blockchain technology could increase the scope of impact of MDMFIs. Third, I explored the key challenges to Blockchain adoption.

Summary of Key Findings

Performance and impact

MDMFIs have shown that they are adept at using mobile phone data to draw signals and inferences about the behaviors that drive loan repayment. They have achieved, on average, repayment rates of over 90%, and their loan acceptance rates of 50% are significantly higher than those of traditional financial institutions.

With loan decisions made in mere minutes, MDMFIs algorithmic data driven approaches have acted as a catalyst for greater financial inclusion, allowing millions of un(der)banked individuals to grow their businesses or smooth their income streams.

Opportunities for Blockchain technology

Blockchain technology has the potential to foster growth in Microfinance and provide more effective solutions in four key areas:

  1. offering a new and innovative way of verifying a borrower’s identity
  2. creating shared and trusted credit histories
  3. enabling the sharing and maintenance of sensitive data in more secure ways
  4. allowing for cheaper and quicker flows of capital to and from borrowers.

Challenges to Blockchain adoption

Despite offering improvements to the Microfinance business model, Blockchain technologies face a number of challenges to become a viable solution. These include:

  1. a lack of interest from relevant stakeholders and an inability to change processes due to legacy contracts and infrastructure
  2. the infancy of the technology and the absence of integrated and interoperable solutions that can be easily incorporated into existing operational models
  3. the significant regulatory uncertainty, particularly relating to public accessibility of sensitive data (e.g. GDPR).

Background

A history of the Microfinance industry and the rise of MDMFIs

Microfinance institutions (MFIs) have been around for a long time. Initially coined in 1970s following the success of the Grameen Bank of Bangladesh, the basic objective of an MFI is to enable traditionally unbanked individuals (and businesses) to obtain access to capital through the provision mini or micro loans. MFIs seek to achieve financial inclusion through developing solutions that a) seek to find innovative ways to address the challenges of nonexistent credit history, and b) build partnerships with a broad range of organizations, from governments through to non profit foundations to source and secure capital flows.

Whilst the mechanics of how micro loans are provided differ based on the relative sophistication of financial systems in different locations (e.g. traditional cash vs. mobile money vs. peer to peer lending), and the specific client base (e.g. unbanked, SMEs, social enterprises, underbanked), the essence of what a MFI does has remained unchanged for decades:

assess creditworthiness and provide capital to the underserved.

Historically, MFIs were operated as quasi-nonprofits. Backed by organizations such as the World Bank, the IMF, international foundations, in addition to national governments and global nonprofits, their main goal was to provide access to credit. There was no inherent need to generate the high Return on Investment (ROI) required by commercial credit providers because MFIs’ key funders did not hold investment returns as their primary objective. Rather, they were concerned with social good. Borrowers were sourced through intimate knowledge of local communities, using information obtained through interviews and experimentation, with limited focus on the benefits of using recordable digital data.

Over time, MFIs have innovated and become more akin to traditional profit driven providers of credit, lending at market rates of interest. Traditional MFIs are now widely accepted as having contributed significantly to global financial inclusion by serving an estimated 200m borrowers. However, MFIs face significant criticism for often charging high interest rates to borrowers, something that is widely accepted to be a consequence of an operationally heavy and labour intensive business model.

Modern-day Microfinance institutions

Modern-day MFIs (MDMFIs), in the context of this post, are MFIs that use alternative data, specifically mobile phone data from smartphones, as the underlying source and foundation of their credit solutions. MDMFIs have lower operating costs and are digitally native, allowing them to offer cheaper and more efficient credit solutions to borrowers.

It is estimated that two thirds of the world’s unbanked have mobile phones. Through the use of data generated from a device that has now become ubiquitous in both the developed and the emerging worlds, MDMFIs are able to assess creditworthiness of potential borrowers in a manner that is automatic, automated, and driven by big data.

My research

I interviewed a number of MDMFIs, which combined, serve over 4.5m customers in emerging economies across Africa, LATAM and Asia, and have issued over $750m in loans over the last few years.

How the MDMFI model works

Whilst there are variations in the loan process depending on geography and the particular MDMFI, a simplified model of the typical customer process is as follows:

  1. Borrower installs the MDMFI’s smartphone app. This app gives the MDMFI complete access to the borrower’s mobile phone (from call duration records through to words included in a SMS).
  2. Borrower uploads relevant ID documents through the app. This allows the MDMFI to create a unique record under which, the mobile phone is associated to a specific borrower, and local Know-Your-Client obligations are satisfied
  3. The MDMFI analyzes the borrowers mobile phone data and, if possible, any available credit data from local credit bureaus based on the borrowers ID, to assess credit worthiness.
  4. Upon the borrower’s request for credit, the MDMFI uses algorithms to analyze mobile phone data in order to generate a probability of repayment score. The MDMFI then offers, within minutes, a line of credit to the borrower. This line of credit is often for 60 days or fewer and typically for less than $350.
  5. The borrower’s repayment of the loan (or lack thereof) in addition to mobile phone usage and habits are analyzed repeatedly by the MDMFI and used to inform the MDMFI’s algorithm.

Key data points assessed

MDMFIs typically gather between 1,200 and 2,000 mobile phone data points to generate a probability of repayment score. This includes mobile phone usage data (e.g. regularity of phone usage, regularity of phone charging, times of day that the phone received SMS messages), mobile phone geosensing data (e.g. where the phone was physically used, variation on location of usage, consistently of location over time), and borrower psychometric data (e.g. how long it took borrower to complete loan application, answers on loan application etc), which are grouped and analyzed within the context of 3 main buckets:

  • Own data — data “controlled” by the individual borrower. This focuses on the borrower in isolation.
  • Inferred data — data on the social interactions and social networks that the borrower is associated with.
  • Relative data — data on the borrowers “metrics” relative to the rest of their cohort and relative to all other borrowers on the MDMFI’s platform.

Whilst not all of the 1,200 to 2,000 data points that comprise these three buckets have equal levels of importance, when considered in aggregate this data has predictive qualities. The core to an MDMFI’s success is an ability to use mobile phone data points to identify the particular behaviors or behavioral traits that offer the most accurate and consistent predictive signals about an individual borrowers likelihood to repay credit.

Anecdotal examples of mobile phone data that provide stronger and more robust predictive signals about a borrower’s likelihood of repaying loans, include the following:

  • Borrowers who include both first name and surname when saving a contact to their mobile phone address book offer higher repayment rates
  • Borrowers who maintain their phones fully charged for longer, offer higher repayment rates
  • Borrowers who have fewer gaming apps than productivity apps on their phones offer higher repayment rates
  • Borrowers who request credit between the morning hours of 5am — 8am, offer higher repayment rates

Traditional MFIs and banking institutions do not consider or analyze mobile phone data to this level of granular detail when assessing risk. Rather, an individual’s income, history of recorded credit repayments and residence address are determining factors.

Whilst income levels and history of recorded credit repayments are, of course, strong predictors of repayment, mobile phone data has been found to be more effective than credit bureau methods at predicting those who are more likely than not to repay funds. The assessment of the day-to-day interactions, in addition to the psychological and sociocultural context of borrowers is where MDMFIs have created a great niche. Given that mobile phone data is real time, MDMFIs can also assess quickly the impact that changes in borrower behavior are likely to indicate about the ability to repay, and adjust their credit amounts accordingly.

Progress towards meeting objectives

MDMFIs core lending businesses have been exceptionally successful using mobile data to assess a borrower’s ability and willingness to repay in a low cost manner. Given the lack of brick and mortar branches and low fixed overhead, the marginal cost of adding an additional borrower is minimal. This has allowed MDMFIs to offer loans to new and larger populations of borrowers. In particular, to individuals who operate in the informal economy, are geographically distant from traditional lending options, or cannot afford the interest rates offered by traditional MFIs.

From those MDMFIs that I spoke to, the following headline numbers were identified:

  • Typical MDMFI repayment rate is over 90%
  • More than 50% of MDMFI loan applicants receive a line of credit
  • Interest rates are typically materially cheaper that those from comparable payday loans or traditional MFIs
  • Loan approval occurs in mere minutes

These results show that commercially, the MDMFI model is working. MDMFIs are allowing a greater number of people to access credit, smooth their income and obtain cheaper working capital (through lower priced short term loans) in the process. However, they have faced many criticisms, particularly, for creating an increased dependence on debt, given the relatively low proportion of first time borrowers (typically around 7%) and high proportion of repeat borrowers (typically 90%+). This is particularly pertinent given widespread reports of individuals using loans from one MDMFI to repay loans obtained from another MDMFI.

If MDMFIs really want to deliver on their core goals of financial inclusion they will need to offer products, such as savings, investments or planning, that help borrowers think to the future.

Source: Imperial College Business School

Implications for Blockchain technology

What is Blockchain?

Blockchain or Distributed Ledger (DLT) technology is, at its core:

  • A shared database / ledger of chronologically recorded transactions, secured by cryptography;
  • That is immutable and tamper-resistant, allowing users to append and record data, but not delete or edit;
  • With no central owner and, instead, run by a network of distributed computers, each holding a copy of the database / ledger of transactions;
  • Where the network reaches consensus on the true state of database / ledger through a predefined consensus mechanism.

In more simple terms, Blockchain/DLT allows the creation of a shared database held and controlled by a network of distributed computers, where data can only be added to the database, and the database’s veracity is secured by complicated cryptography.

Use cases given research observations

Based on my interviews, MDMFIs see Blockchain technology as something for the future. That being said, practical use cases for Blockchain technology based MDMFI pain points include:

  • Credit histories — Whilst many MDMFIs create “probability of repayment” scores based on mobile phone data and then record borrowers’ payment histories on the MDMFI’s own individual platform, this information is not publicly shared across platforms in such a way that a borrower is able to develop a global verifiable history of credit repayments.
  • Blockchain’s immutable ledger offers the ability for a borrower’s repayment history to be permanently recorded to a public database that all potential lenders can access. This could cover all “credit like” repayments and would enable MDMFIs to implement risk based pricing to all borrowers through the aggregation of data across providers (many MDMFIs do not currently offer risk based pricing and instead offer flat interest rates to all applicants). The credit history could be linked to a digital ID (see below) with the solution implemented in conjunction with local credit bureaus. Such a partnership with local credit bureaus would ensure that borrowers are able to eventually access the traditional financial system (both locally and internationally) using MDMFI data (i.e. using proven and verifiable evidence of debt repayment, which is what traditional credit providers require). Existing Blockchain solutions that seek to blend credit history with digital ID include The Kiva Protocol, Bloom, BanQu, and Colendi.
  • ID verification — A common challenge for MDMFIs is obtaining valid ID documents when seeking to identify their borrowers, particularly in locations where more than one form of ID document is issued.
  • Blockchain solutions such as uPort, Civic, CULedger or those developed by the UN backed ID2020 Alliance, have attempted to address this issue, either through digitizing paper based records, or creating digital passports. These decentralized digital IDs could be linked to an individual’s biometric data or unique public-private key pairing, and would be more easily transferable, sharable and accessible. This would enable to MDMFIs to onboard a greater number of borrowers more quickly and effectively.
  • Disbursements and collection — MDMFIs work with local payment processors and telcos to distribute and collect funds. Given that most well funded MDMFIs operate in a number of jurisdictions, funds need to be regularly moved across international borders to disburse to borrowers, leading to associated FX risks. MDFIs use the existing financial infrastructure, which is slow, does not not offer real time information and often requires nostro accounts to be held at correspondent banks when transferring money across borders.
  • Blockchain based payments systems such as Stellar and Ripple seek to increase the speed and information associated with cross-border payments, in addition to eliminating the need for tying up funds in nostro accounts, and reducing FX risk through the use of stable coins (digital assets pegged to the value of non-digital assets) or atomic swaps. As MDMFIs disburse increasing amounts of capital each day to borrowers, having greater control of cash in real time and reducing reliance on telcos and payment processors will become imperative to continuing to offer cheap services.
  • Data sharing and protection — MDMFIs obtain and share data from/with partners (e.g. Android, telcos) in order to create and develop their algorithms and improve machine learning. This requires data sharing agreements with a broad range of parties across the globe. Privacy considerations are often the main concern of companies when it comes to sharing data. Currently, this data is held in centralized servers or cloud providers (e.g. AmazonAWS, Microsoft Azure, Google Cloud).
  • Blockchain providers such as Enigma, Oasis Labs seek to create solutions to allow privacy-secured sharing and storage of data. This will allow MDMFIs to identify and track who has access and has accessed which pieces of data at all times, ensuring that only the relevant people have access to sensitive information.

Challenges to use cases

Whilst Blockchain technology could facilitate the scaling of MDMFIs, there are a number of challenges, both practical and technological that will need to be addressed.

  • Adoption — Many of the solutions require the acceptance of new blockchain based infrastructure by multiple parties. Given the diversity of stakeholders, including regulators, governments and businesses, with conflicting objectives (e.g. maximization of profits vs. improving social cohesion vs. ensuring consumer protection), wide scale adoption may be difficult to achieve.
  • Additionally, the technical stack required for Blockchain technology to be feasible for all day to day transactions is still a way away from being fully built out. That being said, we are now seeing companies develop blockchain based mobile phones (for example, HTC and Sirin labs) which would enable seamless integration between mobile phone data and blockchain based technologies (for example, ID verification and credit histories), albeit that these phones are way outside the price point of the average MDMFI borrower. As the cost of blockchain based smartphones decrease, adoption of the underlying technology is likely to increase.
  • General Data Protection Regulation (GDPR) — Blockchain technology’s core benefit is that the data stored on the chain is immutable. Once it has been recorded, it cannot be deleted. This is contrary to the rules of the GDPR, which requires that businesses can delete data sorted on their customers, if requested by customers.
  • To the extent that local or national governments across the board introduce similar legislation, this could create regulatory challenges when using Blockchain technology to record data.
  • Technological — To create a system that covers credit data, ID, payments and privacy will require Blockchain interoperability and scalability. Interoperability is the area where Blockchain technology remains most in its infancy. Therefore, until this has been resolved (organizations such as Cosmos are intending to build a solution), the technology will struggle to achieve high levels of adoption, as integration into existing MDMFI operational models remains impractical.
Source: Kim Ruyle

Conclusions

MDMFIs have shown that they are adept at using mobile phone data to draw signals and inferences about the behaviors that increase an individual’s likelihood to repay loans. As a result, they are able to provide cheap credit funding to millions of un(der)banked individuals who otherwise would have no reasonably priced access to the loan funding required to grow their businesses or smooth their income streams. In this way, MDMFIs are clearly continuing the work of their MFI predecessors of assessing creditworthiness and providing capital to the underserved.

A major criticism of MDMFIs is that they encourage the use of credit where it otherwise wouldn’t be required, drawing people into a cycle of debt dependency. Given that over 93% of borrowers repay loans, however, it is possible that the cycle of debt claim may be overstated.

Blockchain technology has the potential to foster MDMFI growth and provide more effective solutions to larger client bases through a) offering a new and innovative way of verifying a borrower’s identity; 2) creating shared and trusted credit histories; 3) enabling the sharing and maintenance of sensitive data in more secure ways; and 4) allowing for cheaper and quicker flows of capital, both to and from borrowers. However, challenges for the implementation of Blockchain technology are abound, particularly those relating to technological infancy, regulatory uncertainty.

About Bosun

Bosun Adebaki is a Business Consultant at Blockchain at Berkeley and an MBA student at UC Berkeley’s Haas School of Business. He believes in using FinTech to create a more accessible financial system.


Microfinance and alternative data meets the world of Blockchain was originally published in Blockchain at Berkeley on Medium, where people are continuing the conversation by highlighting and responding to this story.