Embedding Climate Resilience Into Credit Origination Models for Small Businesses in Southeast Asia

The MSME credit gap in ASEAN remains considerable. But MSMEs may be facing a new enemy. The threat lies in weather patterns and whether we can succeed in keeping the temperature rise at acceptable levels. Event-based vulnerability has always been in evidence anyway. It will just get more accentuated. 

Core issues, as many practitioners rightly point out,  have been the lack of formal data and the difficulties banks face in servicing low-return prospects. I suggest these issues will be made more complicated by the uncertainities around weather.  MSMEs are very vulnerable to extreme weather events in any market, anyway. They ply their trade in heavily populated habitats, often in close proximity to the sea or river systems, and have little by way of insurance or reserve cash. Their outlets or establishments might be structurally fragile. Much of their business runs on supply chain credit and informal borrowing. Another category of MSMEs consists of start-ups, consultants and free lancers. Many of these firms are home-based. Their ability to function can be severely disrupted by weather events as they may not be able to travel, could fall ill or do not have manpower located in different places. Also, their uncertain incomes can mean inadequate insurance and medical cover. A third category are those who are related to farming, tourism and local services in the hinterland. They face typhoons, floods and landslides every year. 

In this age, financial inclusion must mean giving people and businesses the resources they need and helping them build resilience. This calls for a more proactive, forward-looking process that can get small businesses set up and fully prepared. Such methods must include climate resilience as a key deliverable. 

Step 1- Give the business a basic cash line. This puts a small amount of money in the hands of a small business which can produce at least one valid ID or utility bill. Once the money has been lent and is expected to be repaid, a file is open and behaviour can be tracked. If the money is not repaid or delayed inordinately, this business is not right for further lending. However, one can consider weightage for what we may call as climate empathy. If a small business is in a vulnerable zone and appears to be bonafide enough, it may be given a one-time basic loan, regardless. 

Step 2-Score the business upon getting the first repayment. This can be a very basic but effective scoring model where weights are provided for three things-the credential provided; whether the loan was repaid; whether the loan was repaid on time. A highly configurable and flexible decisioning engine would enable this. 

Step 3-Add a macro-environmental variable which takes into account the context of operations of that business. This can be key to determining how much to cap his borrowing at and what kind of interest rates will be levied on a continuous basis.

-Nature of trade(add weight)

-Location of business(add weight)

-Kind of establishment(add weight)

-Mobile phone usage(add weight)

-Transaction volumes(add weight)

Step 4-Add a weather variable based on recent weather events in the area and their impact. 

-Weightage for rainfall or flooding index

-Weightage for power shortage index

-Weightage for weather related illnesses. 

-Weightage of weather impact on product/service sold/traded

Step 5- Offer weather-related insurance or provide incentives(such as rate discounts) if such insurance is already taken. 

This approach accounts for the current situation of the business from a multivariate approach as well as  climate context and then makes room for some reasonable resilience to be built(or provides a reward if such resilience is indeed already present). What the actual algorithm is may be determined by the coders of the software. It could be linear regression or supervised machine learning where logistic regression is applied. Unlike variables used for a FICO score, some of the above variables may not have a linear relationship with the likelihood of a MSME repaying a loan. For instance, the monthly income has a linear relationship; track record of paying a mortgage has too; location of trade may have a non-linear relationship. Hence, use of supervised machine learning may be a better option. Should a basic score be already available via a credit bureau(which often may not be the case here), that should be treated further by use of the non-linear variables or alternative variables(the ones mentioned above).

Now, we are left with the basic question of user interaction. Onboarding MSMEs is a key point of concern. The process has to be short and simple in order to not lose the borrower, especially if the loan in question is relatively small. The first part where a small loan is provided and a single credential is shared is simple enough. Below is a possible sequence which may be turned into figma flows:-

1. Upon repayment, a score is provided.
2. Max borrowing limit is flashed. 
3. Offer is made-and can continue to be made each time. 
4. Resilience score is also provided with embedded offers of insurance or other services. 
5. Score keeps changing as borrowing and repayment go on. 

 While it is not likely to have occured yet in many cases, it should be possible for the decisioning engine to connect with not only local credit bureaus, tax offices and government agencies, but also news wire services, local met office, local police feeds(if available) and digital publications. The management of APIs, therefore, becomes quite important. The inception process of the scoring schema needs a team and manager which is quite holistic in approach while being detailed-oriented. This is not a hygiene requirement but something essential. An important output to the borrower should be the degree of resilience he/she has against contextual extreme weather events. 

With the above approach, it may be possible to set up a working model for unsecured lending that deals with credentials challenges, fragmented data and non-existent or very thin files and the absence of weather/climate related criteria in evaluating borrowers. It also provides borrowers data-based assessment of their credit and climate resilience and empowers them to initiate steps towards strengthening their situation.