An Initial Evaluation of the Unified Lending Interface

By
Peter Pakdeejit

Executive Summary:

The Reserve Bank of India (RBI) has developed the Unified Lending Interface (ULI), a revolutionary digital lending platform that promises to solve the dual problems of informal lending and credit frictions. The implications of a successful ULI are groundbreaking: chronic debt traps would cease to exist, rural and agricultural populations will have access to formal credit, accelerating investment, living standards, and productivity growth, whilst core-periphery inequality would greatly diminish. This paper aims to (1) introduce the problem of informal lending, (2) provide a preliminary overview of ULI’s proposed solutions, (3) evaluate the ULI’s risks, and (4) compare the results of other similar aggregators or scoring globally. We will finally recommend that further research into a ULI-like platform, additional psychometric indicators, and government stimulus be performed as a potential solution to Thailand’s rural, agricultural and MSME debt trap. Although more thorough quantitative econometric analyses will have to be performed, there is enough qualitative evidence suggesting the adoption of a ULI-like framework will boost productivity and propel economic growth.

Keywords: Credit Frictions, FinTech, Data Aggregator, Informal Loans, Agriculture, MSMEs

Credit Friction Background:

Credit frictions and informal lending pose one of the largest barriers for economic development globally. This is manifested through three main channels: reduction in private-sector productivity, rural and/or agricultural productivity, and increased inequality (Sathyanarayana, 2024.)

Regarding the former, Bailey argues that the evolution of MSMEs into stalwart enterprises is a core driver of global development (Bailey, et.al, 2013.) This is in accordance with neoliberal economic theory, stating that the organic advancement of local firms into large-scale enterprises will likely spur a virtuous cycle of better employment, higher income, and access to even better education and healthcare etc. They highlight local heroes like our ‘Charoen Phokpand Group’ and India’s Tata especially, for their impacts on local employment, education and infrastructure, and believe this cycle will likely repeat ad infinitum. Whilst the extent of ‘trickle down’ is likely an oversimplification, requiring considerable legislatory and enforcement efforts from government agencies, their core thesis that employment and productivity growth is a byproduct of enterprise development, rather than converse causality, largely remains true. Therefore developments, or lack-thereof, that constrain the evolution of local firms stymie economic growth. In Thailand in particular, credit frictions greatly prevent MSMEs from breaking out of the ‘survivorship stage.’ This is because an overwhelming majority of firms cannot secure attractive loans to fund working capital, invest in talent acquisition and technical machinery, or perform adequate research and development to compete with domestic and international rivals.

Regarding rural and agricultural communities, credit is a necessary cornerstone of everyday life. For them, loans are essential to maintain living expenses and working capital investments mainly due to seasonality, as agricultural workers earn income in lump sums after harvests. Thus, the number of loans spent on ‘unnecessary’ purchases (frivolity) is negligible. This is evidenced by the Puey Ungphakorn Institute’s stylised facts survey which states that rural loans are almost always used for economically productive activities (Pinjutsamut and Suwanprasert, 2022.) For farmers, over 60% of all informal loans are utilised for investment including purchasing fertilisers, ‘iron buffalo’ tractors and building irrigation infrastructure; the rest being deployed to fund necessities including education and healthcare or reservice existing debt. On the other hand, 74.9%, 69.9% and 69.7% of loans for private corporation employees, freelancers, and contract-based workers are used for necessities.

However, Thailand’s formal credit drought forces locals to turn to predatory informal loans instead of legal debt from financial institutions. The extent of this problem is severe: over 42.3% of the Thai population owe informal debt as a last resort, even though reputable surveys have stated they would much rather borrow from reputable financial institutions if they could (Moheldin and Wright, 2000.) Such informal agreements often contain illegally high interest rates of up to 4-5 times the legal limit of 15% YoY enforced through aggressive collection practices (Pinjutsamut and Suwanprasert, 2022), creating ‘debt traps.’ These occur when interest payments exceed total income, meaning households have to borrow more just to pay existing dues. Moreover, the Ungphakorn institute found that informal creditors are always regional ‘influential people’ or ‘powerful community figures’ (ผู้มีอิทธิพล), encompassing mafia, powerbrokers and local politicians, who use violence to enforce their contracts. Asset, property seizures and physical violence are commonplace for defaulters, thus households chained in debt traps often forgo essentials, education and investment, reducing future income growth. Thus, formal credit is vital to ensure rural communities can uplevel their businesses and themselves, improving productivity, living standards, and to decrease rampant core-periphery inequality.

Diagnoses of similar credit droughts globally reveal two common constraints: the traditional loan appraisal process and employee risk aversion. Traditional loan appraisal processes are designed to maximise expected returns, prioritising profits rather than serving those who need the money the most. Thus, the process, combining credit history records, physical documentation, income statements, tax records etc greatly favours risk minimisation rather than loan democratisation (Bailey, et.al, 2013.) Collateral is also generally required as a hedge, excluding most rural borrowers who either do not possess the assets period, or possess ‘informal’ assets. Although the process was not designed to isolate rural and agricultural directly, an unintended constraint is still a constraint. Rural debtors may not have formal credit histories and scores, earn income from the informal sector (which is non-taxable and does not appear on cashflow assessments), or lack the
title deeds to their land. In other words, they will fail these traditional assessments predicting NPL risk regardless of their true financial health purely because existing metrics are incomplete.

Furthermore, employees are risk averse. Under current compensation frameworks, employees are directly responsible for loan underwriting; when subject to disciplinary processes and performance evaluations that penalise NPLs more than returns, human behaviour understandably becomes biassed towards self-preservation. In addition, entire financial organisations can also be risk averse: according to Sathyanarayana, legacy banks adopt ‘procyclical’ credit supply patterns (Sathyanarayana et.al 2024.) Thus, during times of recession, when credit is required the most for underserved communities (because of income shortages and high inventory stock, due to lower consumption etc), traditional financial institutions are least likely to provide formal credit. Such times are precisely when rural households would take out their informal loans to fund necessities or working capital and end up in ‘debt traps.’

The ULI will solve this information asymmetry and bias through (1) utilising alternative metrics and machine learning to provide a more complete credit score and (2) producing an automated ‘good loan bad loan’ assessment from the model, reducing the downwards-skewed bias of bank employees due to risk aversion. However, an additional solution to curb the power of “   ผู้มีอิทธิพล ”

or regional influential people will also need to be implemented and enforced in collaboration with a responsible government agency to truly and structurally address the problem in the long term.

Unified Lending Interface:

Introduction:

Adopting a solution like the Reserve Bank of India’s Unified Lending Interface (ULI) will significantly reduce Thailand’s rural credit frictions through simultaneously removing the information asymmetry inherent in traditional loan appraisal processes and risk-aversion bias. Through digital integration across multiple agencies, the platform will revolutionise appraisal for the most vulnerable, aggregating large datasets, utilising machine learning models and alternate sources of data to predict credit viability through ‘scoring’ (Sathyanarayana et.al 2024.) This ‘one-stop’ platform will then allow financial institutions to compete on an online (Amazon-like) marketplace: given the same models (albeit they can set their own parameters), rival banks will be able to offer loan contracts to approved debtors, who will compare products and choose the ones that match their needs the most. Such competition is projected to lower interest rates for debtors, increase product differentiation and reward banks who manage risk more effectively.

Those who realise and capitalise on the largest information asymmetry will simultaneously generate returns and reduce credit frictions. Thus, the ULI promises to decrease the percentage of risk-averse credit rejections, mitigate the pro-cyclicality in loan supply, increase the availability of non-collateralised loans and data-collateralised loans for underserved communities and reduce

the prevalence of ‘debt-traps’ in rural areas, boosting long-run productivity, investment and living standards. According to RBI estimates, the ULI is projected to cost $2 billion.

Functionality:

The ULI workflow will work as follows.

1. Big Data Aggregator: The ULI will firstly perform the role of a ‘Big-data Aggregator’, using open APIs to pull datasets from multiple government and corporate agencies that consent to the policy. In practice, when debtors consent for a ULI loan, the platform will initially extract their personal data from sources including Aadhar and E-KYC, PAN DigiLocker, and State Land Records et cetera (Sathyanarayana et.al 2024.) Aadhar and E-KYC, India’s combined electronic identification system, combine to verify the debtor’s identity and reduce fraud. Digitised data, as opposed to traditional appraisal documents, also streamlines workflow and increases loan accessibility because rural debtors no longer have to fill in legal documents that they may not have the financial literacy or legal fees to compile. Potential debtors in very remote situations also no longer have to travel to large cities, reducing fees. On the other hand, DigiLocker (aggregated bank statements), PAN (Permanent Account Number), which contains income tax records, and State Land Records can be used in accordance with alternative data in appraisal. Thus, the aggregator secures data onto one platform used for security verification and in appraisal modelling.
   
2. Automated Appraisal: The ULI will also contain an algorithmic credit appraisal model evaluating individual debtors’ data based on specific parameters set by government organisations, financial institutions and banks. In short, whilst the credit score / appraisal itself is automated, lenders, in line with ULI regulations, will have limited ability to ‘tweak’ the algorithm depending on risk appetite and edge, leading to increased competition. Were the ULI to be applied to Thailand, would likely be used as the base model, with lenders able to change final requirements and thresholds for loan approval. Alternate data sources, through collaborations with FinTech firms, including psychometric analyses, online-retailing activity, internet usage etc. may also be used to try limit asymmetric information and generate a holistic credit score. Such automation is intended to reduce risk-aversion bias and prevent pro-cyclical credit supply shortages or surpluses, reducing Thai credit friction and boosting productivity in the long run.
   
3. Multi-Lender Ecosystem: Finally, the ULI will offer an online loan marketplace for all potential debtors. On the platform, debtors will be able to compare different loans from rival firms, choosing the offerings that offer the most attractive rates and flexibility for them. For example, in my view, KBank may offer a lower interest rate but a longer tenor, whereas.

Risks:

Although the ULI promises to reduce India’s severe credit friction problems, the interface can present risks to all stakeholders involved. The most apparent include data-privacy and credit risk (Sathyanarayana et.al 2024.) Regarding the former, the ULI deploys digital public infrastructure to centralise and share sensitive data between banks, non-bank financial institutions and government agencies. Thus, if proper data security measures are not maintained across all entities accessing the open-API, the likelihood of cyberattacks, information breaches, large scale data leaks and identity theft will be high. The bottom line is, the more complex and integrated a digital platform is, the more weak links will be open for attack. In addition to information breaches, if the government is lax about the procedures and access to the database, private, sensitive data may be misused by firms to exploit consumers, leading to an ‘irreversible erosion of consumer rights’ (Omarova, 2018.) Under such a scenario, fintech firms may ‘double dip’, using data that was solely designated to loan appraisal for ‘unauthorised commercial uses’ including profiling and surveillance instead. This may lead to price discrimination or “free commerce” (using harvested financial information to sell you products) in the medium to long run etc, reducing meaningful consumer choice. Therefore, strong data protection (ie. Chinese Wall legislation) bills and enforcement must be upheld by policymakers to hedge against data privacy risks.

Furthermore, credit risks are another strong consideration which the RBI and, more specifically, financial institutions must take into account. Due to the experimental nature of ‘big-data-backed, non collateralised loans’ that the ULI is, in essence promoting, the model will not be perfect. As the parameters are refined, the ratio of non performing loans (loans paid over 2 months after the deadline) will slowly decline. However, for non-collateralised loans to make sense for private banks, significant regulation must be introduced. Banks approve loans and decide on interest rates based on risk-premiums: higher risk must be accompanied with higher expected returns (Damodaran, 2012.) Accordingly, were banks allowed to set ‘rational’ interest rates for data-collateralised loans in line with their other product portfolios, interest rates for MSMEs and underserved communities would be understandably high, beyond the legal limit of 15% due to high NPL probability. The opportunity cost of such loans would be too high for commercial banks.

Thus, to keep interest rates beneficial for debtors and guarantee satisfactory returns for banks, further research into fiscal stimulus must be considered. For example, if the ideal interest rates for firms is 5%, banks will have no incentive to lend to underserved communities because they can get the same return, with many magnitudes lower risk from lending to higher net worth individuals or firms. To increase incentives for such loans, policy makers or the responsible authorities may have to introduce programmes like subsidising half the NPLs. In this case, we cannot subsidise all NPLs because this would create a moral hazard incentive for firms to accept all loans and shift the penalty to the authorities; however, half would greatly decrease the probability of default risk. Another consideration could be to reduce capital requirements purely for this program, eg. banks will be allowed to reduce their HQLA requirements (requirements to have high quality liquid assets cover all projected outflows for the next 30 days) conditional on using all the increased capital to fund ULI loans. These are two examples of policies that can make data-collateralised loans more palatable to banks whilst keeping rates low enough to be beneficial for debtors; however, further research and modelling will be required to validate the initial suggestions.

Global Fintech Aggregator / Scoring Models:

Globally, there are two main credit aggregator models of note in addition to India’s Proposed Unified Lending Interface. Firstly, the Entrepreneurial Finance Lab at Harvard University developed the EFL Psychometric score to provide an additional alternative metric to evaluate credit default risk and reduce credit frictions (Bailey, 2013.) Through using ‘psychometrics’ as another data point in appraisals, FinTech companies and financial institutions will be able to predict how debtors will react post loan-approval. Will they be willing to pay back the loan?

How honest are they? What is their risk tolerance? How focussed and analytical are they? These tests are similar to ‘pymetrics’ tests used by large multinational firms to screen job applicants

and can aid lenders, especially to MSMEs or agricultural borrowers (who borrow mostly to invest,) in determining how their investment will be managed. Intuitively, if banks can conclude that Person A is more conservative, more methodical, more focussed, and more extroverted than Person B, they will project that Person A’s investment is more likely to generate consistent returns and that they will be more able to payback the loan. Moreover, if Person A is also honest, they will also be more willing. This is a large advantage compared to traditional appraisal processes: without psychometrics, if both people are from the same area, have the same educational background, and similar incomes etc. both would get rejected. However, such information asymmetry can be addressed under the EFL framework, reducing credit frictions and reliance on informal loans.

Statistically, Bailey states that there is a significant correlation between psychometric scores, business performance and credit default risk. The three most noteworthy, conscientiousness, honesty, and education have a relatively high AUC (Area Under The Curve), which is used to measure predictive power of an indicator relative to other measures in developing countries, thus proving that psychometric analyses do add value to loan appraisal processes; however, psychometrics cannot be used alone (AUC ranging from 0.57 – 0.66.) Their main value comes as a supplement to aid and validate alternative data evaluations. According to Sifrain, whilst using psychometrics alone does outperform socio-demographic credit scoring (Sogesol Bank, Haiti case study), it would not be fully efficient in the context of ‘microcredit operations’ (Sifrain, 2020.)

Chinese Zhima Credit

Another successful alternative credit-scoring indicator is AliPay’s (China) Zhima Credit, which has over 100 million registered users (Bei, 2018.) Through utilising innovative technologies like ‘cloud computing’ and ‘machine learning’ to analyse alternate datasets containing internet consumption, internet behaviour, and online payments as new data points in credit appraisal, Zhima has been proven to improve classification performances and default prediction (Wang, et.al, 2022.) According to Wang, the inclusion of Zhima score increases the performance of all 5 classifiers in his study (C4.5, Random Forest, Naive Bayes, K-Nearest Network, Support Vector Machines, and Back Propagation Neural Network.) In addition, Bei concludes that the score compares favourably to traditional appraisal across many different industries, with the most significant decrease in default probability being recorded in telecommunication loans from 20% to 6% (Bei, 2018.) Although Zhima methodologically differs from EFL’s Psychometric Score, the information being obtained is quite similar: Zhima ultimately predicts a potential debtor’s ‘behaviour preference, performance ability, identity traits and personal relationship.’ Therefore, both models realise that psychological traits that affect honesty and competence determine debtors’ willingness and ability to pay, affecting NPL default risk. Thus, the scoring has been utilised to increase loan volume and exempting upfront payments by up to 42 billion RMB. The latter in particular should aid rural communities that rely on seasonal cash flows the most.

Conclusion:

In conclusion, this paper has performed an initial analysis of the causes and effects of Thailand’s informal loan problem, evaluated the Unified Lending Interface, which we believe could be utilised as a partial solution, and explored other successful fintech aggregator / innovative credit-scoring solutions globally.

We firstly conclude the credit friction problem in Thailand is a severe headwind to our economic development. With over 42% of our population engaged in some form of predatory, informal lending, long term investments in personal education, healthcare and capital are impossible.

Illegally high interest rates, enforced by violent collection methods, consign most rural, MSME households to ‘debt traps’ where an excess of total income is required to service existing debt. Thus, Bailey’s ‘virtuous cycle’ cannot be fulfilled, the evolution of local enterprises out of the ‘survivorship stage’ becomes impossible, meaning that the trickle down chain linking higher quality jobs, income, education and living standards does not exist. The bottom line is this: an extension of formal, compassionate credit is required to revitalise rural productivity, living standards, and decrease core-periphery inequality.

Moreover, with relying on the Puey Ungphakorn institute, we diagnosed that the problem stems down to three main factors: the traditional loan appraisal process, employee risk aversion, and influential power-brokers ‘ผู้มޒอޑݵݴޑพล’. Firstly, the traditional loan process creates information asymmetry, where the real financial health of rural, MSME and agricultural debtors cannot be accurately assessed as their ‘formal documents’ do not reflect their informal cash flows and abilities. Secondly, the incentives of the modern banking sector disincentivise both individual bank employees and entire institutions from extending rural credit due to ‘performance evaluations’ and KPI. On the firm level particularly, risk aversion creates a pro-cyclical credit supply pattern. This is especially challenging for underserved communities who require credit the most during economy downturn. Finally, informal loans can only be enforced by local power brokers because there is no legal mechanism to do so otherwise. Thus, a solution which aims to tackle Thailand’s credit frictions must solve all three causes at their source.

We suggest the adoption of a Unified Lending Interface and alternative credit scoring, which can address the first and second problems. Through creating a data aggregator and automated appraisal process, we can reduce the information asymmetry apparent and more accurately represent the financial health of underserved communities. Furthermore, it may also be beneficial to add psychometric measures into the composite indicator (through logistic regression etc.) which can further evaluate individual debtors’ willingness and ability to repay their loans. Automation will also deal with human risk aversion at both an individual and institutional level. However, risks are apparent: further research will have to be done into developing an accurate credit score and fiscal policies will have to be passed to incentivise banks to lend to higher beta clients and stomach their opportunity cost.

Finally, policies will have to be ratified in collaboration with other government agencies especially the one who can curb the power of local influential power brokers. We believe that the actions suggested will truly address the root causes of credit friction to underserved communities, reducing credit friction. And if credit frictions can be reduced, a bottom-up surge in productivity will likely release the untapped potential of the countryside, encouraging development, enhancing living standards, and reducing core-periphery inequality.

Bibliography

Bei, D., Chuang, H., Li, Y., Liu, Y., Luo, Y. and CAFI (2018). Digital Financial Inclusion in China. [online] Chinese Academy of Financial Inclusion Renmin University. Haidan Beijing: Renmin University . Available at: https://www.findevgateway.org/sites/default/files/publications/files/growing_with_pain_digital_f inancial_inclusion_in_china_cafi_report.pdf.

Damodaran, A. (2012). Investment valuation : Tools and techniques for determining the value of any asset. 3rd ed. Hoboken, N.J.: Wiley.

Kim and The World Bank (2020). Digital Financial Inclusion. [online] World Bank. Available at: https://www.worldbank.org/en/topic/financialinclusion/publication/digital-financial-inclusion.

Klinger, B., Asim Ijaz Khwaja, Carlos Del Carpio and Springerlink (Online Service (2013).

Enterprising Psychometrics and Poverty Reduction. New York, Ny: Springer New York. Omarova, S. (2018). Fintech: Examining Digitization, Data, and Technology.

Pinitjitsamut, P. and Suwanprasert, W. (2022). Informal Loans in Thailand: Stylized Facts and Empirical Analysis *. [online] Available at: https://www.pier.or.th/files/dp/pier_dp_173.pdf [Accessed 10 Sep. 2024].

Professor, S. and Bengaluru, M. (2024). The Unified Lending Interface -Innovations in Lending: Conceptual Insights into the Future of Financial Services. International Journal of Business and Management Invention (IJBMI) ISSN, [online] 13(9), pp.35–48. doi:https://doi.org/10.35629/8028-13093548.

Wang, H., Chen, W. and Da, F. (2022). Zhima Credit Score in Default Prediction for Personal Loans. Procedia Computer Science, 199, pp.1478–1482. doi:https://doi.org/10.1016/j.procs.2022.01.188.

Peter Pakdeejit