Two years ago, I blogged weightily about the Grameen Bank's deteriorating loan portfolio. In the second half of 2009, its repayment rate---fraction of amounts due actually paid---fell from 97.81% to 96.55%.That was hardly catastrophic in itself, but marked the sharpest drop for that indicator in its publicly recorded history. While hedging my bets, I raised some serious possibilities:
The Grameen Bank, indeed all big microcreditors in Bangladesh, may be finding it harder to collect on loans. As far as the evidence goes, there has been no epidemic of default. But the combination of years of rapid growth and accelerating declines in key indicators of delinquency are so reminiscent of the lead-up to the global financial crisis that the broad implications hardly need explaining. A partial meltdown in the Mecca of microcredit would not sow the same economic destruction—microfinance is not the heart of Bangladesh’s economy in Schumpeter’s sense—but it could have lasting implications for microcredit worldwide.
It's humbling to look back on these words and see how poorly they forecast what followed. There was plenty of microfinance drama in Bangladesh and beyond, but not the kind I insinuated, at least not in Bangladesh. Roodman was no Roubini (or Rozas). Despite the unforeseen trauma of Muhammad Yunus's removal---which makes a move 97.81% to 96.55% seem trivial---the Grameen Bank seems to be operating normally.
For all that, I think the concerns I raised still stand.
Since writing that post, I've gained a better understanding of the role of flexible loans in the Bank's finances. Treating flexible loans more properly in my own analysis yields a picture of a loan portfolio whose quality is still deteriorating. Flexible loans are intended to refinance and rehabilitate borrowers struggling to repay at the standard 50 payment-per-year speed. Muhammad Yunus described the flexible loan as a detour on the microcredit highway.
I think now that the deterioration I blogged in early 2010 was the result of a ratcheting down of flexible loans from early 2008 to mid-2009. Apparently during that period almost no new flexible loans were extended, so that the stock shrank as old ones were paid or written off. That hardening of posture---that blocking of the slow detour---could explain the delinquency spike I noticed. And I see now that Grameen reversed course in July 2009, even before I blogged (though the acceleration of flexible lending began in February, just as my post appeared). That reversal can explain why in the months that followed my post, the Grameen Bank's publicly available repayment indicators sharply improved.
The mechanical quality of the trends suggest that the overall disbursal of flexible loans is controlled from Grameen Bank headquarters. To a significant extent, then, the non-repayment spike I blogged seems to have been an artifact of policy changes at the Grameen Bank. Actual portfolio quality changed less than the way it manifested: delinquencies went up, but the stock of flexible loans, of dubious long-term value to the Bank, shrank.
So was my story a complete non-story, a mountain made out of a statistical molehill? I don't think so. The see-sawing of policy on flexible lending suggests an institution struggling with a delinquency problem of some magnitude.
And if I rerun the numbers in a way that greatly reduces this artifact, then the portfolio quality trend remains negative and the level worrying. As Rich Rosenberg commented back in 2010, when computing "portfolio at risk" (amounts owed by people behind on their payments), conservative accounting calls for every taka of flexible loans to be considered at risk. Every taka of these loans has been rescheduled for repayment later than originally planned. So flexible loans, even when people are current on them, should be treated just like loans officially recognized as overdue and at risk. I did not appreciate this when I wrote that post, so I only counted flexible loans as being at risk when borrowers were behind in repaying them according to their redifined, lenient terms.
The graph below absorbs Rich's advice. The stacked bars show the shares of outstanding loan amounts that are a) owed by people who have missed 5-9 consecutive installments on standard loans; b) owed by people who have missed 10 or more installments on same; or c) have been refinanced as flexible loans. These sum to the PAR30, "portfolio at risk, payments at least 30 days late." (The authoritative MixMarket corroborates these calculations with a PAR30 for Grameen of 6.95% at December 31, 2010.) Overlaid in blue is the non-recovery rate, which is the percent of amounts due in a month that were actually paid (and which, for lack of data, is not adjusted to treat all payments on flexible loans as late):
Calculated this way, the Grameen Bank's PAR30 reached 9.9% last month, the highest level since May 2004, when the Bank was still recovering from the repayment crisis of the late 1990s.
Worthy of comment is how the Grameen Bank recovered from that old crisis. It improved its product offerings, notably by offering better options for savings. As a result, both its savings and credit businesses grew rapidly. Membership shot from 2.5 million people at the end of 2002 to 8 million at the end of 2009. Delinquent borrowers who had left the fold returned to take out new loans and use them to make good on old; that reduced the numerators in the ratios graphed above. Fresh borrowers multiplied the denominators. As a result, old delinquencies shrank in absolute terms and dwindled in relative importance.
But, as I pointed out in 2010, when growth slowed in the late 2000's, as it eventually had to, the delinquency rate resumed its upward creep. This made me wonder to what extent the Grameen Bank was riding a tiger, constantly trying to grow its way out of repayment difficulties, only to create new and larger ones. Was portfolio growth was masking continuing delinquency problems? As I just suggested, it could do so in two ways, which I'll call "dilution" and "refinancing." Bringing in lots of new borrowers whose payment records would start out spotless would dilute the significance with the Bank's portfolio of the delinquencies among older borrowers. And of course new loans could be made to help those struggling older borrowers catch up on their old loans. (The introduction of flexible loans was in part a way to make this practice transparent.)
To explore this question further, I made the next graph. It compares PAR30 in each month to portfolio growth over the previous twelve. The red contour is the just the skyline of the bar graph above. The blue is the growth rate. The antisymmetry between the two is striking:
What explains this pattern? I'm wont to point out that correlation does not prove causation. Perhaps the two mechanisms I just described, which both have higher growth causing lower reported delinquency, are not actually at work. Maybe in good economic times, the portfolio grows faster and delinquency falls, and in bad times, like now, the opposite happens. But as far as I know, 2006--07, a period like now of slowing portfolio growth and rising portfolio-at-risk, was not difficult economically (though it was politically).
So if the causal mechanism does indeed run from growth to reported delinquency, can we distinguish between the two mechanisms I suggested? Maybe so. I think refinancing requires more growth to cut delinquency than growth through dilution, to the point of implausibility. Suppose the outstanding portfolio is 100 million taka, of which 5 million is at risk. Diluting that risk by adding enough new members to expand the portfolio to 101 million reduces the par PAR30 from 5/100=5% to 5/101=4.95%. So a 1% growth increase reduces the PAR by just 0.05%. Yet in the graph, a 1% increase in growth corresponds to far larger reductions in PAR. The ratio looks to me much more like 1% increase in growth for 1⁄3% fall in PAR.
In contrast, suppose the Bank decides to refinance that 5 million taka of portfolio at risk---to make new loans to pay off old. To entice the borrowers back into the fold, it might need to offer bigger loans this time around. If you'd walked away from $100 loan, would you come back just so you could owe $100 again? A new $200 loan, half going to pay off the old loan, half to put cash in your hands, would be more interesting. So assume the Bank extends loans worth 2 million taka in order to refinance 1 million taka of dubious debt. (This seems realistic if you imagine the Bank offering a new $100 loan to refinance a half-repaid old $100 loan.) Now the total portfolio has grown to 101 million (reflecting the 1 million net increase in loans) and portfolio at risk has dropped to 4 mllion, cutting the PAR from 5/100=5% to 4/101 = ~4%. Here, a 1% increase in growth reduces PAR by 1%. That 1-to-1 ratio is much closer to 1-to-3 I see in the graph. If I assumed the new loans were not twice as big, but a third bigger, I'd have a good match.
So I think the possibility that refinancing is obscuring the true quality of the Grameen Bank's portfolio cannot be easily dismissed. The question arises: what is the Bank's "steady state portfolio at risk" (a term I just made up)? If growth (the blue line) fell over the next few years to zero (lending can't grow forever, can it?) what would the red line rise to? To my eye, if the refinancing theory is correct, meaning that declines in the blue line cause the red one to rise, I'd say 14%. Going by the 3-to-1 ratio, a drop from the current growth rate of 12% to 0% would cause PAR to rise another 12/3=4%, from 9.9% to 14%.
So I'm suggesting: if the Grameen Bank stopped expanding its lending today, within a year it would emerge that 14% of its portfolio is owed by people struggling to repay. That amounts to 10.6 billion taka, or $128 million. And it exceeds the Bank's capital base as last reported, 9.35 billion taka ($112.5 million) at the end of 2010. (The Bank may have boosted its capital base with profits in 2011, or reduced it with a loss. Audited financial statements should appear in the next couple of months.)
How big a threat this purported 14% poses to the capital base depends on how much the struggling borrowers would eventually pay back. If they paid back nothing, then the Grameen Bank is approximately broke, and its savers are at some risk. But if they paid back 80% of what they owe, the situation would be much better. The Bank would still need a capital infusion in order to meet minimum standards (as it already does), but savers would be safe.
Of course, the Grameen Bank's portfolio won't stop growing tomorrow. But that doesn't make this analysis irrelevant. There is still a suggestion here that the Grameen Bank faces greater financial challenges than most realize. If this is true---if there is a big, de facto loss destined to appear on the books---that will force some rethinking of the Bank's business model. Perhaps the manifestation will be as severe as the credit crisis of the late 1990s. Perhaps the problem and the response will be as mild as accepting such a high delinquency rate---flexibility is good for clients---and, to compensate, increasing interest rates, which are extremely low.
Check back with me in another two years and we'll see what happens.
All data and graphs in this post are in this spreadsheet. Underlying data come almost entirely from the Grameen Bank.