I’m on my way to Mexico, for what I hope to be the start of a deeper exploration of overindebtedness in the country. Data analysis an ocean away can be revealing, but there’s nothing like seeing the numbers come alive when visiting the field. First stop: Tapachula, Chiapas.
Every analyst has his or her own approach. For me, I find it best to come with a number of hypotheses and then see to what extent reality reflects those initial preconceptions. I like to keep an open mind and am always willing to change my view. Still, having a pre-existing framework in mind helps structure field observations, especially when time is short.
I’ve already shared my thoughts on multiple borrowing and overindebtedness in Mexico, but those go back a couple of months. Since then, I’ve spent a fair bit of time digging deeper into the data and comparing Mexico to what I’ve seen elsewhere (including finalizing a study of Moroccan MFIs during 2008-13, including how they dealt with substantial multiple borrowing during 2009). Based on this and earlier work, I’m putting down some of my hypotheses below.
Hypothesis 1: The Mexican microfinance market reflects very deep penetration of a narrow slice of the population
Some numbers: MIX Market has about 6.1 million active borrowers for Mexico in 2012. Adjusting for the multiple borrowing rates in the Finca study, this implies 2.2 million unique borrowers. Of those, about 75% are women. However, the MIX market doesn’t include several large MFIs (such as Banco Azteca) and even an even greater number of small ones. Then again, the Finca study probably excluded many of those small MFIs, because they don’t report to the credit bureau. So there is an undercount – I have no idea how large. Perhaps a factor of 2, perhaps less, probably not more. So I’ll go with 2, and assume there are 4.4 million unique borrowers, of whom 3.2 million are women.
Those women are from among the poorest segments of Mexican society. After all, in Mexico where the average per capita income is over $10,000, a $400 loan just isn’t very much. Seems unlikely that anyone other than somebody very poor would bother with groups and all the baggage that comes with them, just to borrow $400 for 4 months, at a rate of some 100%. And indeed, from the World Bank Findex microdata (available as a Stata download here), we see an anomaly: 9% of women in the bottom quintile are borrowers, which is nearly double the level of women in the quintile directly above theirs, which in turn is higher than the middle quintile. This is in fact very unusual -- the trend in nearly all other countries is the opposite, with wealthier individuals consistently borrowing more than poorer ones.
The average per capita income level in the bottom quintile in Mexico is $173, which is well above the finding from the Angelucci/Karlan study, which cites household income per adult at 1,571 pesos/month ($112). So I think we’re on reasonably firm ground in describing the population of Mexican microfinance clients as being women in the bottom income quintile (though really, this depends on how widely client incomes vary).
And how large is that? Well, the bottom quintile would have 24 million people, of whom about 65% are ages 15-64. So about 16 million people, or 8 million women. From the analysis above, we have roughly 3.2 million unique women borrowers, or 40% of the target population. That’s drastically different from the 9% cited in Findex – a discrepancy that remains to be explained. But it does seem that the segment of microfinance borrowers is narrow indeed.
What other signals might confirm if this hypothesis is true? There are a couple of signs to look for. One is particularly strong geographic concentration in poor regions. Hence the focus on Chiapas, the country’s poorest state. By extension, microfinance penetration in wealthier regions should be smaller. Another sign would be the diversity of products (i.e. how much above the $400 average?) offered by Mexican MFIs and also the diversity of incomes among the clients. If the dispersion of both figures is narrow, it would support the hypothesis. These are the data I’ll be looking into.
Hypothesis 2: Overlending in Mexico can be sustained for a long time
One of the key misunderstandings of bubbles is that they must pop relatively soon. In fact, they can last years. Let’s go back to that FINCA study. One interesting addition to its multiple borrowing figures was the addition of default rates for each category of borrower (Figure 2). The figures are so high as to be unbelievable. So, nearly 9 out of 10 borrowers holding 6+ loans are delinquent? And even 34% of borrowers holding just one loan are likewise delinquent? Were these figures truly reflective of the market in Mexico, we wouldn’t be talking about a credit bubble – this would already be a full-blown crisis. Indeed, the study itself questions the reliability these delinquency figures. As it happens, there’s a simple test for this.
Consider the one-loan borrowers. They fall delinquent at a rate of almost exactly 1/3rd. If one were to throw a simple die, the chance of rolling a 5 or 6 would essentially be equivalent to the delinquency rate of these single-loan borrowers. So what would happen if one were to throw that die two times? Three times? Well, as it happens, figure 2 shows the odds of throwing a 5 or 6 at least once for each of the number of times the die is rolled. If you replace the die roll with the number of multiple loans held by the borrower, you could likewise compute the probability of default at each of the levels of multiple borrowing. It turns out that the probability of borrowers holding 2, 3 or more loans falling delinquent is essentially identical to the delinquency levels reported in the credit bureau data. The implication is that these delinquency figures have nothing at all to do with the added stress of holding multiple loans. Instead, it likely reflects a problem in how delinquency is reported to the credit bureau in the first place. This is a serious problem for the credit bureau and for Mexican MFIs, but it says little about overindebtedness.
So it seems that multiple borrowers in Mexico aren’t really defaulting at a rate that’s any different from other borrowers. This is exactly as it should be. It’s what I found in the credit bureau data of Moroccan MFIs just before and after the 2009 crisis holds (IFC publication forthcoming). Simply put, even heavily indebted borrowers may perform at essentially “normal” levels – at least until the crisis hits. However, when the crisis comes, the whole edifice breaks down – in Morocco in 2009, multiple borrowers defaulted at levels far above those of single-loan borrowers, while borrowers holding 4+ loans defaulted at the rate of 60%. We saw the very same pattern in Bosnia, and for that matter, with US subprime borrowers.
At heart, it is the bubble itself that sustains multiple borrowers. When stressed, they simply borrow more. In some cases they will default, and those defaults essentially serve as the escape valve for deeply overindebted clients. The steadily rising write-off rates in Mexico are likely signaling this very mechanism. These write-offs can be a way out for the most distressed borrowers, but at the broader level, they are the very process that allows the bubble to grow ever larger. Like a volcano, whose mini eruptions precede the big one, these are warning signs that one ignores at one’s own peril.
So what other signs of a building bubble might one look for? The clearest would be a steadily growing loan/income ratio. Loans that steadily consume ever rising incomes is a telltale sign of a developing bubble. Finding such a trend would largely support my argument that loan amounts in Mexico are not too small for borrowers, though the thoughtful comments to the contrary are serious and should be considered. Unfortunately, this trend may not be so easy to capture. Another option is to look at trends in multiple borrowing itself. If my arguments on loan size were incorrect, what the data should show is a relatively clear dispersion, with some borrowers maintaining relatively large number of loans over time, while others hold a smaller number, reflecting their individual funding needs. One should also see changes in multiple borrowing that go both ways – sometimes borrowers would increase the number of outstanding loans, other times they would decrease. On the other hand, a steady increase in multiple borrowing across the board would be a sign of a growing bubble.
Hypothesis 3: A strong shift in the perception of debt among individuals
It’s very common for bubbles to build among people who are relatively new to debt, and social norms governing debt begin to loosen. There has been ample experience with this – in Ireland and Spain, the housing bubble featured a notable shift in home buying patterns. The ready availability of credit encouraged people to buy with far greater leverage and much less saving than had been the case with earlier generations. The same can be seen in the credit card bubble in Korea, which followed a sudden departure from earlier cultural norms that had eschewed borrowing.
It is telling that one of the key warning signs in Bangladesh, where the 4 leading MFIs seemed to have successfully averted a crisis in 2007-08, was the fact that the MFIs were finding clients to be resistant to more loans. These were after all 2nd and 3rd generation clients, who had grown up with essentially the same microcredit products households and were well familiar with both their advantages and their risks.
In Mexico, there’s no such historical depth with microcredit. An interesting indication would thus be to look for differences in how credit is perceived by older generations, and how they view the borrowing of their children and grandchildren.
Hypothesis 4: Competition trumps coordination among MFIs
A key lesson from the Morocco crisis was the rapid reaction of the leading MFIs, who worked together to address common issues, particularly with respect to sharing client data. This was a massive change for a market that featured very aggressive competition, with the two leading MFIs both vying to top the other in outreach and size. In Mexico, there is a credit bureau, but it isn’t required for many market participants, including some very large ones, like Micronegocio Azteca.
It’s also critical that the coordination that does happen be more than just words, as was the case with Indian MFIs in the leadup to the crisis, when the newly-created MFIN was promoting the quite sensible Code of Conduct, while the leading MFIs, including SKS, were focused entirely on pursuing their IPOs. In the process, the focus on growth pushed aside any commitments to avoid client overindebtedness.
The important question in Mexico today is whether the concerns about overindebtedness and long-term market sustainability go beyond words, and whether real, coordinated action among market leaders is taking place. By the same token, if aggressive competition and growth continues to be their primary driver, it would suggest that the sector is likely heading down the same path as the MFIs in India, Bosnia, Nicaragua, and elsewhere.
Hypothesis 5: Regulators are “solving” problems by ignoring them
In most microfinance crises, the response of the regulators came after the crisis had already manifested itself. The sole exception might be Morocco, where the regulators were already becoming involved in the sector for unrelated reasons, with supervisory visits starting to take place about a year prior to the crisis. This is one reasons why the Moroccan crisis was both less intense and shorter than any of the others. By contrast, the Reserve Bank of India was quite actively seeking to avoid any real supervision of Indian MFIs until their hand had been forced by the state government of Andhra Pradesh, whose actions had essentially stopped all microfinance activity in the state. For that matter, the subprime crisis in the US also followed a period of light regulation, with Alan Greenspan refusing to take any action to regulate the mortgage sector.
If the regulators in Mexico start actively addressing the serious risks in the microfinance sector in the country, for example, by raising minimum capital requirements, expanding credit bureau reporting to all significant MFIs, conducting supervisory visits to the sector’s largest MFIs, and generally start looking seriously at a sector that affects Mexico’s poorest citizens, then that’d be a sign that the crisis may be avoided. If, on the other hand, they continue to deal with the problem by ignoring it, this will be yet another signal that the market is heading towards disaster.
Setting the stage
These five hypotheses form the core of my concerns over Mexico. Did I miss anything here? Certainly – this is hardly an exhaustive list. However, this is a way for me to set down a set of expectations in advance. I’ve also tried to highlight the types of factors that might suggest that the microfinance sector in Mexico isn’t overheated or is likely to achieve a “soft landing.” I will keep you posted.