The pandemic bond took a financial instrument common to the US insurance sector — insurance-linked securities (ILS) — and applied it to disaster response. Investors receive a coupon, but face losing their principal if an event meets pre-agreed conditions. Bondholders are not betting on the credit quality of the bond issuer, but on the likelihood of the event happening.
The pandemic bond had an extra special twist when compared with other ILS products. It was designed not to help deal with the aftermath of a disaster, but to help prevent it occurring, by paying out at the early stages of an outbreak.
The bond was issued in 2017 at a size of $320m. It has two tranches: the less risky one pays 6.5% over Libor and the most risky one, relevant for the Ebola outbreak, pays 11.1% over Libor. The conditions for payout are rather complicated and include speed of contagion, number of confirmed deaths and the disease crossing national borders.
The issuance of the bond has come under heavy criticism, as reported in GlobalCapital’s sister publication GlobalMarkets, because the severe outbreak in DR Congo has not triggered it. This is in spite of 2,169 deaths and 3,243 total cases as of Tuesday. It is reportedly because the outbreak has not killed enough people in any other country.
It is worth noting that the World Bank has put money into fighting Ebola in DR Congo through a cash window that is part of the same programme as the bond but designed to be more flexible.
The criticisms of the pandemic bond can be put into three categories. First, the trigger terms are too onerous and weighted in favour of bondholders at the expense of those who would benefit from a payout. Second, allowing investors to profit from this type of disaster response is inherently bad. Third, ILS is, in its nature, ill-equipped for this type of disaster response.
More ‘automatic’ system
The second criticism rests on the well-founded idea that healthcare should be the preserve of the state, because it is more egalitarian and efficient. Why should this be any different for disaster responses in less developed parts of the world?
As Olga Jonas, senior fellow at the Harvard Global Health Institute, wrote in the journal Nature: “Making the bonds attractive to investors meant designing them to reduce the probability of payout.”
The counter-argument is one of practicalities rather than morals. In an age of nationalism, governments may not put more money into aid budgets. Hard-pressed international aid professionals may not like the financialisation of charity, but be forced to embrace it nevertheless.
The validity of the third criticism is also tricky to determine.
Conceptually, tying payments to pre-agreed conditions makes sense for disaster response funding, because it can make those payments both certain and timely.
Mark Lowcock, of the United Nations Office for the Coordination of Humanitarian Affairs, put it as follows: “Instead of doing what we do at the moment, watch the problem grow and develop and people like me ring up, launch appeals and some months later the money come[s] through, we have a much more automatic system.”
Yet designing those conditions can be tricky, and can backfire. Even in the more established US ILS market, investors have been caught unaware by perils such as wildfires in California. In the poorest parts of the world, where the insurance market is less well-developed, there is not much in the way of modelling or data for those designing products to draw on.
That creates uncertainty, and if the bonds are marketed to private investors, uncertainty adds risk premium. This makes the product less cost efficient for the intended beneficiary.
When the formulae work, there is little outrage. Peru and Mexico have both benefitted in the last few years from catastrophe bonds issued by World Bank. These were triggered after earthquakes. But when those formulae fail, it leaves people yearning for a less cold-hearted approach.
This pandemic bond has not worked, but that does not invalidate the use of ILS tools for humanitarian aid completely. More evidence, data and refinement are needed.