Opinion: This data on loan performance in LMICs isn't the full picture

Newly released data on loan recovery rates across 25 multilateral development banks, or MDBs, and development finance institutions, or DFIs, aims to ease investor concerns about the risks of lending in low- and middle-income countries. The Global Emerging Markets, or GEMS, database, covering loans from 1994-2022, shows relatively low loss rates for these institutions' loan portfolios in emerging markets. However, a closer look reveals important limitations in using this data to gauge true investment risks for private capital in developing economies. Investor uncertainty over the potential losses of investing in low- and middle-income countries is an obstacle to increased private capital flows. Investors don’t have enough experience in these markets to even quantify the risks. So showing the track record of the MDBs and DFIs that have actually invested in those markets over many years is a welcome step forward to dispel some of that uncertainty. Hence the excitement in the global development world when the GEMS database, which pools loan recovery data across 25 MDBs and DFIs — disclosed data from 1994-2022, showing low loan losses across a wide range of emerging markets and developing economies, or EMDEs. But while it’s an important additional set of data, it is important to recognize its limitations in dispelling investor uncertainty. The data released so far is limited to aggregated statistics, not the whole dataset. To be useful for risk modeling, investors need to see the data disaggregated by year and country, at a minimum. Unfortunately, when you slice the data like that, the number of data points per slice quickly gets small — so the confidence intervals around the averages become very high. For example, there are 1,756 data points for loans to private borrowers in sub-Saharan Africa — which sounds like a lot, but is only an average of 1.3 data points per country per year. In addition, only 6% of the dataset is for low-income countries, where the lack of investor information is greatest. Furthermore, the data tells us less about the risk of EMDE assets than you might think: • MDBs and DFIs benefit from risk mitigants not available to private investors. As they are among the few — if not the only — long-term investors in these markets, governments and local counterparts attach more importance to repaying their loans, so that they may benefit from future loans. And they have the backing of powerful governments which can influence governments and central banks to allow them to be repaid even during foreign exchange crises, when others may not get paid. This is known as preferred creditor status, and while for most institutions it is only a convention, it has historically been widely respected. • MDBs and DFIs engage in lengthy due diligence and structuring to limit loan risks upfront, sometimes including arranging government guarantees. • MDBs and DFIs have the capacity and willingness to engage in lengthy workouts and refinancings rather than see loans default, because they care about the success of the underlying project or business. So private investors should be skeptical that they can achieve the same loan performance, unless they co-invest with MDBs and DFIs, and benefit from these advantages. The International Finance Corporation has been able to mobilize over $16 billion from 17 institutional investors into its Managed Co-Lending Portfolio Program, based on IFC’s past loan performance. Similarly, private credit fund manager ILX has been able to raise $1 billion from European pension funds to buy MDB/DFI loans, using GEMS data (in confidence) to demonstrate the risk profile of such a loan portfolio. So it appears GEMS data can give private investors comfort about co-investing with MDBs and DFIs, but not necessarily about investing on their own. More fundamentally, the fact that loan recovery rates on loans to private parties are higher for LMIC loans (between 75% and 85%) than high-income country loans (66%) does not mean that lending in LMICs is less risky. All this shows is that MDBs and DFIs are good at selecting, structuring, and supervising loans in these countries to achieve a risk profile in line with their limited risk appetite. The risk profile of a set of potential loans in a country is distributed on a curve from low to high risk. When we talk about the riskiness of a country, we are talking about the average of that distribution. But MDBs and DFIs use rigorous credit risk assessments to select loans at the left-hand (low-risk side of the risk distribution), so the data in the GEMS database is bound to show lower loan losses than the average for the whole country distribution. So you can only expect to achieve the loan recovery rates if you are able to emulate the thorough loan selection processes of MDBs and DFIs — or let them do it for you. And finally, it is not even clear that MDBs and DFIs could sustain these recovery rates if they mobilize a lot more capital from private investors to co-invest, as many aim to do. The GEMS data release is a welcome step toward greater transparency on how effective MDBs and DFIs have been in lending to LMICs, and it should encourage more investors to co-invest with these institutions. But we need much more disaggregated data, including from private loans without MDB/DFI involvement or government guarantees, to dispel the uncertainty deterring private investment in LMICs.