My latest post on the credit bubble in Mexico had one especially interesting comment. Jose Manuel asked to consider the loan sizes in the country as a factor that might explain the prevalence of multiple borrowing.
The comment is highly relevant. What Jose Manuel suggests is that loans in Mexico are unusually small. And in a way, he is right. On a per capita GNI basis, Mexico's loans are smaller than in any other country. By contrast, India's loans are nearly three times larger. This has two potential implications: first, small microfinance loans put less of a burden on Mexican borrower incomes, and second, their inadequate size encourages clients to borrow from multiple lenders in order to meet their requirements. And yet, I find that both implications are incorrect and that multiple borrowing levels in Mexico continue to point to a very large bubble.
Multiple borrowing as sign of success
It has been often noted by practitioners that microfinance loans are by design insufficient to meet the needs of some microentepreneurs. Indeed, SKS founder Vikram Akula argued this very point in his letter to the Wall Street Journal in August 2009 – a year before the Andhra Pradesh crisis. In addition to this well-accepted fact, Akula also cited a study conducted in Andhra Pradesh in 2007 that found clients with multiple loans having better repayment rates. Here is how I described this study, when warning of an oncoming crisis in Andhra Pradesh five years ago:
The Krishnaswamy study found that multiple borrowers, representing 7-10% of clients in his sample, consisted primarily of highly motivated entrepreneurs seeking to raise more capital than what was offered by any one MFI. This is unsurprising – due to the nature of their cycle-based lending model, MFIs knowingly underfund their borrowers, thus assembling funds from multiple MFIs is a logical strategy that Krishnaswamy suggests is simply a replacement for the informal funding sources the individuals would have tapped otherwise. This is also consistent with the money management practices documented by Collins et. al. in Portfolios of the Poor. However, as the market heats up and multiple borrowing becomes increasingly widespread, the number of multiple borrowers grows beyond these stand-out individuals…
On this point, I haven't changed my mind – it isn't multiple borrowing itself that concerns me, but rather, the channel through which it occurs. There is a subtle, yet enormous difference between a motivated entrepreneur seeking out additional capital for her business by soliciting multiple MFIs, and a borrower who takes on another loan that's actively – or even aggressively – pushed by a loan officer looking to meet his monthly bonus quota. In Mexico, I suspect there is a lot more of the latter than of the former.
So what about the loan size in Mexico? Recall that main reason why microfinance loans are often too small is that loan size is used as a way of establishing repayment history for clients who don't already have one. With each loan repaid on time, clients become eligible for a larger loan amount. But that figure doesn't increase forever – once credit history is established, repayment capacity becomes the limiting factor, and not necessarily how many prior loans the borrower may have had.
This suggests that, as microfinance markets mature, the mechanism for setting loan amounts begins to look more like traditional retail lending. And so, driven by the laws of supply and demand, the loan size in established microfinance markets should arrive at an equilibrium.
On the supply side, lenders would prefer to lend as much as their clients need, but not more than they are able to repay. After all, it takes roughly the same amount of effort to evaluate a client for a smaller loan as for a larger one, so larger loans should increase profits. For lenders, purposefully lending less than is tantamount to leaving free money on the table – an unlikely outcome for profit-driven institutions.
On the demand side, borrowers face a similar dynamic. Each loan application and repayment process consumes time, expense, or both (e.g. sitting in group meetings, traveling to a branch), which normally borrowers would prefer to minimize. All else equal, borrowers would thus prefer fewer loans.
The intersection of these two drivers – lenders seeking to maximize profits and borrowers seeking to minimize costs – would set the equilibrium loan size. Of course, microfinance isn't quite so simple.
To this basic model, one should add a few adjustments. For some larger amounts, lenders may find funding the full amount to be excessively risky, even if the borrower's repayment capacity is not in doubt. In such cases, the lender may well expect the borrower to supplement the offered loan with loans from competing MFIs. On the borrower side, the inflexibility of microfinance disbursement and repayment cycles may lead clients to seek additional loans on top of their existing ones. And of course, there are the exceptional clients whose risk tolerance and business acumen prompts them to seek out funds than MFIs would normally be unwilling to provide. In short, some level of multiple borrowing is a natural feature in microfinance markets. The question is how much?
It is here that comparisons to other markets are useful. After all, are Mexican borrowers so much more business-savvy or have so much more volatile incomes that they require a larger number of multiple loans to manage their funding needs? Likewise, are Mexican lenders really so extraordinarily risk averse that they make it standard practice to give away the extra profit they could receive by lending larger amounts, and instead expect their clients to go borrow more from competing MFIs? Indeed, what reason is there to think that the market equilibrium for loan sizes in Mexico is governed by rules that are so vastly divergent from everywhere else?
So why are loans in Mexico small?
If the above theory is right, the answer should be that Mexican loans aren't big or small – they're exactly right for their market. So why do they seem small?
Consider first the per capita GDP ratio (or its close cousin, GNI, preferred by MIX Market). At first glance, it seems a reasonable proxy for comparing the levels of client indebtedness. But it's not. First, the per capita GNI comparison is normally used as an indicator for depth of outreach – how poor are the clients of the MFI? By that argument, the figures in Mexico imply that the country's MFIs serve clients who are substantially poorer than the average Mexican.
As it happens, there is some data that helps to get a sense of comparison. According to the Banerjee/Duflo study in Andhra Pradesh, the average household income consumption for MFI clients in 2010 was 11,497 Rs/month, which translates to 68% of India's per capita GNI that year. Meanwhile, the Angelucci/Karlan study of Compartamos cites household income per adult at 1,571 pesos/month, which translates to 17% of Mexico's per capita GNI. Thus, in the relative terms of per capita GNI, the typical microfinance borrower in Mexico is four times poorer than her counterpart in Andhra Pradesh.
Clearly, using per capita GDP (or GNI) as a proxy for loan size relative to borrower incomes vastly overstates the incomes of typical microfinance clients in Mexico.
Interest rates, again…
The other major reason why Mexican loans may seem small is that what's being measured is the loan amount, which is not at all the same as the total obligation undertaken by the borrower. With an average portfolio yield of 82.3%, Mexican loans are unusually expensive – more than three times the average yield of 26.6% for MFIs worldwide in 2012 or the yields in India and Bosnia during their market peaks (25% and 24%, respectively).
The impact of such differences on client cash flows is quite dramatic – a client holding 4 loans with an 80% interest rate faces a total payment obligation that's nearly the same as a client with 6 loans at 25% interest. The stress on borrower cash flows in Mexico is thus much larger than the loan amount alone would indicate.
Back to the bubble
Thus far, the high prevalence of multiple loans in Mexico has been my primary indicator of a microfinance bubble in Mexico. I recognize that for any single client, multiple loans are a poor indicator of repayment capacity – exceptional entrepreneurs or individuals with particularly volatile incomes may well need multiple loans to meet their needs. The question is not whether any one client has multiple loans, but whether their overall prevalence in the market indicates overheating.
I've tried to show that the size of the loans in Mexico does not imply less financial stress on borrowers or a higher propensity to take on multiple loans. Ultimately, their size is governed by the same laws of supply and demand, whose equilibrium is unlikely to differ much from other markets.
If we accept that the figures for multiple borrowing at market peaks in Bosnia and Andhra Pradesh reflect the limits of those markets' credit capacity, and we observe that multiple lending in Mexico is far above those levels, then we should also recognize that Mexican microfinance has likewise exceeded its credit capacity, and what we're in fact seeing is a very large bubble.
 Weighted average loan size = 3.9% of per capita GNI in Mexico, and 11.0% in India. For each MFI, loans weighted by number of active borrowers. Source: MIX Market 2012 data.
 For India, average number of adults in survey household was 4, so assuming that all income is consumed, per adult income is calculated at 2,874 Rs/month. GNI per capita is expressed in current USD (Atlas method), i.e. not adjusted for purchasing power parity, so as to be consistent with loan sizes, which likewise aren't adjusted by PPP. There may be some differences in foreign exchange rates used to express loan size in USD (Oanda.com, 12/31/2010 for both countries), which may not fully correspond to the World Bank Atlas method. The samples in the Hyderabad and Compartamos studies should not be taken as representative of microfinance clients in either country (for example, Hyderabad is relatively wealthy, as are the areas surveyed in Mexico); however, the data is sufficient to demonstrate that Mexican client incomes are substantially smaller than the country's per capita GNI than is the case for Indian clients.
 MIX Market 2012