Planning for Disaster Recovery
Planning for Disaster Recovery
Changing the default setting
Abstract and Keywords
Rules can promote decisive, timely action, but a rules-based system is only as good as the data that drive it. Because no rule is perfect, there should be some discretion to deal with situations in which the rules fail. The challenge is to allow discretion without allowing people to game the system. The data need to be resistant to manipulation and strike the right balance between cost, speed, and reliability. Any data that could trigger action will depend on investments before a disaster in design of the data-collection system, including an audit function, and in the human and technological capacity to collect data in a timely manner. Three types of data could be used to trigger action: ground data on the damage to or losses of people and buildings, area average index data on damage and losses, or parametric indexes.
Ato Samual Mengisha speaks proudly about the Gebrale Iddir, the funeral society of which he is the secretary, in the village of Bukicho in the southern highlands of Ethiopia. The society was founded about twenty years ago, and his family has been a member ever since. Life is still hard for the families in this village, which is dependent on rain-fed agriculture, growing enset (false banana), coffee, and maize. In the area, funerals are expensive and rather too common.
Although conditions have definitely improved in recent years, more than a third of the families are classified as not having enough resources for their basic needs, and in a village like this, about one in ten children does not survive past the age of 5. Just as everywhere across the world, providing a decent burial for a deceased family member is important. Relatives and friends are invited to share their grief. Hospitality in the form of food and drink plays a big part in a dignified burial. Community cohesion is such that relatives of the deceased can rely on support from others in this time of need, but, still, not being able to pay for a funeral when required is sad and shameful.
This iddir has fifty-four members. Every year, members pay a membership fee of about US$5, either in instalments or in one go; new members pay an entry fee of $4 as well. When a member or the spouse of a member dies, the society pays out $12 to the family in cash—a smaller sum if another close relative dies. Contributions are (p.52) predictable, and so members can plan carefully to ensure they have the cash rather than being surprised when a sudden death requires extra cash. Ways to raise extra emergency cash—again before the pot is empty—are specified. At the time of the conversation with Ato Samual, the iddir had cash reserves of about $290. It also owned a canvas tent, plates, and glasses worth about $200, which members brought along to a funeral.
Ethiopia has tens of thousands of iddirs that operate like this. Typically, they have fewer than 100 members. Probably originating in the Gurage communities, the iddir model can now be found across Ethiopia, even in urban areas and among civil servants, teachers, and World Bank employees in their Addis Ababa office. Contributions vary, but everywhere the rules are very similar. Membership is selective, and an entry fee is payable. A group usually holds the reserves in cash or buys some assets. Iddirs are considered an essential part of the Ethiopian cultural fabric. Many of these funeral societies in Ethiopia are expanding to take on other risks such livestock death or illness. Essentially, they can be thought of as mutual societies, historically at the root of some of the largest insurance companies in the world.1
As noted in Chapter 2, similar groups can be found elsewhere—for example, in India, Tanzania, and South Africa.2 They offer a sensible solution to a problem. A delay in finding the resources for an uncertain event such as a funeral is stressful and wastes time, and so organizing financing beforehand is one part of the solution. It also helps that the response by the group is well defined: a pre-determined cash transfer plus some practical assistance to help with the funeral. And the simple and clear decision rules for implementation help to make it functional. In fact, this is small-scale risk management that works for rich and poor alike: those wanting bigger funerals either set up groups with higher fees and payouts, or, more commonly, join multiple groups so they can receive multiple payouts to finance a more expensive funeral.
(p.53) Meanwhile, there are no opportunities for procrastination, for strategic behaviour, or for benefactor behaviour. In fact, iddirs have group incentives to ensure that disasters are avoided—that is, for risk reduction. Because they can force all members to attend meetings or take part in specific activities as part of their membership, and because it is in their collective interest to limit mortality, they are proving to be excellent vehicles for spreading health education and other developmental messages.
Why aren’t similar principles applied to managing disasters across the world? Of course, the problems that arise when a sudden death occurs in a village are far smaller than when a disaster strikes. Lots of people are affected at the same time, and planning for large-scale disasters and the decision making required for preparation and responses are far more difficult. But some of the principles are very similar, whether one is dealing with a community planning sensibly for burial support or a national government or international organization preparing a credible response to a possible disaster.
The Flaws of Human Decision Making
At the heart of any credible pre-disaster plan is a system for post-disaster decision making. Decision making is hard, even if the political will is there. However, if the system for decision making is well thought through, it will serve as a basis for a viable alternative to discretionary begging-bowl financing. If it is not, it is back to the begging bowls.
In recent decades, research in psychology and behavioural economics has illuminated the flaws of human decision making, but also what to do about them. Three related lessons from this behavioural science research can help in the design of better decision-making systems to cope with disasters. The first lesson is that people tend to place too little value on the future; they have a bias towards the present day instead. The result is that they are not inclined to act now for future benefit, even if there is plenty of evidence of regret afterwards. An (p.54) example is not saving enough for retirement—and, indeed, not protecting oneself against potential future harm.3 To overcome this bias, people can use commitment devices to lock themselves into certain actions. Odysseus understood that, and he found a way to pre-commit to certain actions by tying himself to the mast and avoiding the song of the Sirens. Poor Ethiopians have clearly found another way by paying a regular contribution to a fund that can be used to pay for funerals.4
The second lesson is that people tend not to change what they are doing unless the incentive to do so is strong—the ‘status quo bias’.5 Inertia and procrastination go hand in hand with this bias: inaction tends to be preferred over action. Businesses and governments have begun to discover they can exploit this human tendency by making the choice for clients or constituents (such as enrolment in a pension scheme) and then forcing them to opt out if they disagree with that choice. There is strong evidence that if action has to be taken to opt out rather than opt in, enrolment in pension schemes increases dramatically.6 In general, changes in the default setting has proved to be an effective strategy for nudging people towards better outcomes, and it is now recognized as one of the primary routes through which behavioural science is informing public policy.7
The third lesson builds on the finding that people tend to have limited information-processing capacities, which affects the quality of their decision making, even when the stakes are high.8 Behavioural scientists have suggested that much can be gained from keeping things simple: ensure that plans and any decision making they entail are kept as clear and simple as possible. Well-designed, intelligent decision rules and triggers will make the difference.9 So for funerals, whereas different people may have different financial needs, different mortality risks, or different desires for simple or extravagant funerals, the iddirs have only one fixed contribution for all members and one fixed payout when a member dies. This is not necessarily the best possible outcome for all members, but a simple, clear set of decision rules allows this group to function and to function well.
So, what does this mean for decision making around disasters? It suggests the need for a clear-cut, straightforward disaster response plan that can be easily implemented. Strong commitment devices should be found to ensure that the tendency to resort to the default setting of inaction is changed to one of action. Clear-cut triggers and decision rules will aid decision making. It is not rocket science—in fact, it is probably more difficult than that; it is about human behaviour, not Newton’s laws, and there are far more unknowns. Any extreme event that may lead to a disaster will happen in very specific circumstances involving many people whose actions and reactions are not easily predicted. How then does one make effective decisions on launching a successful response?
No one has either the time or the imagination to plan fully for everything that would have to happen after any disaster, so some post-disaster discretion will always be needed. The questions to be answered here are these: What should be decided before a disaster, and what should be decided afterwards? What should be on standby, ready for take-off, at all times, and what can be decided later?
One option would be to prepare rough plans for the people or buildings that will be protected in a disaster, how they will be protected, and who will pay for it. Then, after the disaster, a group of experts, bureaucrats, or political leaders would quickly work out the details of the response. The problem with this approach is that in a highly charged post-disaster environment, there will be negotiation over the details of how the money will flow and to whom, and there will be procrastination while waiting for more information. All this will take time. Strategic discretion is the enemy of speed, particularly after a big disaster, and particularly in an environment rife with benefactors, begging bowls, and bureaucracies. Around the world, there is no shortage of pre-disaster plans laying out the principles for who is supposed to make what decisions when, but again and again that information does not lead to timely action—it just provides ammunition (p.56) for long, drawn-out negotiations. It will not get the response off the ground in time.
Instead of governments and their partners making approximate commitments, roughly agreeing to everything that will happen in the months after a disaster, they could take some lessons again from the Gebrale Iddir. It does not offer a full response plan for everything that happens after the death of a member. Rather, it provides precisely defined support in cash and in kind under specific circumstances, with all the financing sorted out beforehand. In general, it is more useful for governments and their partners to make precise, focused commitments to a credible set of minimal actions that pre-identified implementers have the authority and financing to undertake immediately. These actions would be financed and implemented based on pre-agreed rules and triggers, without the need for any begging bowls or any further green light from political masters. This precise, immediate action plan would define what data would be collected on day one and what financing and actions they would automatically trigger; what data would be collected on day two and how they could accelerate or decelerate actions, etc. The idea is similar to a launch sequence: over time the data force a (minimal) sequence of actions taken in line with a predefined timetable.
The plan could be as simple as providing fast cash or food transfers to drought-affected people. Such a response would require some investments beforehand in identifying the vulnerable households and categorizing them by level of vulnerability, and it would require agreeing beforehand to the rainfall patterns or satellite data that would automatically trigger food or cash support to households. If a drought is mild, perhaps only the most vulnerable households would receive support, but for a more extreme drought support could automatically be extended to less vulnerable households as well. Government or its partners could always provide top-up support over and above this minimal response on a discretionary basis, but the minimal response would be planned for and financed in advance of any possible drought and would happen automatically.
(p.57) Or the plan could simply apply to the reconstruction of government-owned infrastructure following an earthquake. Before any earthquake, the ‘lifeline’ infrastructure (such as key roads, bridges, or hospitals) could be identified and prioritized as part of the pre-disaster plan, and the implementing agency responsible for post-disaster reconstruction of this infrastructure would be identified. Immediately after an earthquake of a given magnitude and epicentre, a pre-agreed budget would automatically be made available to the implementing agency for reconstruction of this lifeline infrastructure. The implementing agency would be responsible for immediately launching reconstruction of the lifeline infrastructure, using the initial budget and applying the pre-agreed principles for prioritizing reconstruction to the specific details of the earthquake.10
An earthquake could also trigger an additional budget for a detailed, objective assessment of the full extent of the damage for all lifeline infrastructure. Such an assessment would automatically unlock additional budgetary funds if the initial automatic budget turned out to be insufficient, so that all lifeline infrastructure is in full working order within six months of any earthquake. Planning for a disaster might also include provision for the implementing agency responsible for post-disaster reconstruction to invest in reconstruction capabilities, develop and implement procurement pre-qualification criteria, and sign retainer agreements with construction companies to ensure that the country has the capacity within the construction sector to respond adequately to an earthquake.
This does not mean delegating all post-disaster decision making to technical agencies and replacing all post-disaster discretion with pre-agreed rules and objective triggers. We are merely proposing that a credible, rules-based plan that everyone knows will be implemented be in place, based on an agreement made before the disaster about who or what will be protected, how the protection will work, and who will pay.
In countries that have taken this approach, a concise, credible plan changes the default setting for responding to disasters from ‘wait and (p.58) see’ to ‘implement what is in the plan’. This approach may not be perfect, but it sidesteps the delays and warped incentives of the begging bowl. And if this immediate response plan is well constructed, it will allow timely, sensible actions.
All Systems Go
A rocket launch sequence needs to culminate in take-off, and that is possible with a precise countdown and a set of predefined systems that ensures that all systems are go. In the same way, certain decision-making and implementation systems need to be in place before a disaster to ensure a response plan can be implemented.
A clear command and control system for strong leadership of the post-disaster response is required for plan implementation. All too often, poor coordination post-disaster hampers responses. Implementers can provide helpful input on how command and control of implementation should work, and how, at the technical level, different implementing agencies can work together towards common goals. All parties should agree on a coordination system, and that can only be done beforehand.
Information systems that can be scaled up quickly during crises are essential to the implementation of response plans. They can then relay information to the coordinators and allow plans to be adapted to actual need. But this will only work and will not unravel carefully prepared and balanced plans if clear decision rules are made beforehand on when and how to use the information. During the 2014 Ebola outbreak in West Africa, coordination as well as good data from the ground were crucial for an effective response: without good knowledge of where the pandemic had spread and who had been in contact with the sick and the dead, no effective response was possible. Similarly, without good data on where communities are cut off and what is required to re-establish infrastructure after flooding or an earthquake, those implementing a response plan will move resources to the wrong places.
(p.59) Scalable delivery systems to reach poor and vulnerable populations after a disaster are also essential. Setting up the logistics for delivering food or health care after disasters requires considerable planning beforehand. Instead of establishing new delivery structures, governments might find that a much more promising route is to build a disaster response into the existing systems. Increasingly, even poor countries have developed rather well-functioning social-protection mechanisms, targeting poor and vulnerable households. The idea is to have these systems ready at all times to expand quickly to reach more people or to implement higher levels of transfers. Such shock-responsive social protection could provide the institutional structures needed to reach the poor and vulnerable quickly and at scale.11
During ‘normal’ circumstances, these schemes tend to be relatively well defined—for example, Ethiopia’s Productive Safety Net Programme (PSNP) covers many millions of food-insecure people in normal circumstances. As part of the disaster-preparedness plan, one could take all the steps needed beforehand to allow this scheme to expand at scale—both in terms of giving higher payouts to already enrolled people and in terms of expanding the scheme to other pre-defined groups that now are vulnerable. This possibility is present in the PSNP. This is also the principle underlying Kenya’s Hunger Safety Net Programme: it can expand when a drought occurs, using a pre-defined trigger. Similar arrangements could be made for other essential services during a crisis, such as expanding public health or WASH (water, sanitation, and hygiene) interventions during crisis situations, not in an improvised way but in one planned ahead of time. And there are alternatives or additions to cash transfer schemes if the circumstances allow: for example, populations may be covered by subsidized agricultural or homeowners’ disaster insurance.
After a disaster strikes, much humanitarian support is often delivered in kind, such as food, but evidence suggests that this is frequently not the best response.12 Trying to deliver food directly to people is costly and logistically demanding. Often during disasters, even when there are droughts or harvest failures in particular areas, (p.60) the overall food supplies per se are not the problem. The more cost-effective, more transparent, and faster alternative is to ensure that vulnerable populations are offered income support—cash—so they can afford to buy the food and other essentials they need. However, markets must also be monitored to ensure that they are functioning and well stocked. With today’s technology, cash can be sent to vast numbers of people via mobile phones and other methods, which makes cash transfers even faster and more transparent than people procuring and delivering goods. The typical current default in delivering support is to deliver it in kind. Changing the default setting by always delivering cash unless there is an explicit rationale not to do so would be consistent with the evidence.13
Lessons from Insurance
At this point, a plan is in place, and systems are ready to deliver at scale when required. But how will those implementing the plan ensure it will happen—that is, how will they make sure there is take-off when the launch sequence is completed? Is it possible to design objective triggers that are both reliable and trustworthy? And, above all, can the data system that drives decisions be both difficult to ‘game’ and sensible? These requirements raise many additional questions. As for gaming, is it possible to know that the data really are objective and trustworthy? Can people manipulate the data system in their favour? As for being sensible, does the system strike an appropriate balance between accuracy, timeliness, and cost? That question, in turn, raises other considerations: Is the system accurate enough to be used to trigger post-disaster action? What is the likelihood of it misfiring, either triggering when it should not or not triggering when it should? Is the system timely enough? Does it strike the right balance between early, approximately targeted action and later, more precisely targeted action? Finally, how much will the data system itself cost?
These are the very questions that insurers grapple with when they sign disaster insurance contracts. Therefore, those designing good (p.61) triggers to induce action by governments or international organizations could learn a few lessons from them. As insurers know all too well, as soon as the data begin to drive decisions and financial flows, there will be attempts to manipulate or falsify the data—for example, to make a fraudulent claim that a home was damaged by a flood even though it was not. Or a poorly designed system could incite some people to become negligent about reducing their risk to disasters—an issue known as ‘moral hazard’. For example, they might not adopt basic flood protection measures because they know their insurance will cover any loss. Moral hazard, as characterized by economists, is not a moral judgement on the behaviour of an individual in a system; it is a judgement on the system itself. If farmers can receive more money from insurers, government, or donors when they underinvest in risk reduction or adaptation to climate change, there is something wrong with the system. If governments can get more money from donors or the international humanitarian system or indeed insurance companies if they underinvest in resilience, there is also something wrong with the system.
A credible system, then, has to be based on credible data. After a flood, how does the national government know whether a school has actually been damaged, whether a farmer has actually lost her harvest, or whether a homeowner has actually lost his home? How does the government ensure that people are not fiddling with the data to get more support than they should? If the data that drive post-disaster decisions are discretionary, then the system is discretionary. Credible data are needed on damage and loss, or at least on credible proxies for these, and that means investing in data systems, people, and processes before a disaster to ensure that after the disaster the money flows to where it is needed. In short, the data have to be objective and protected from meddling, and the system must be structured in a way that does not give people incentives to change their behaviour pre-disaster to game the system.
(p.63) immediate aftermath of a disaster it may be that only quite crude information is available to drive decisions, but over time more accurate information can be usefully collected to refine and better target later, complementary responses.
Insurance companies have struggled with this problem for decades and have come up with three rules-based approaches for trying to capture whether a person or building has suffered damage from a disaster while also trying to economize on the cost of information. Described in Box 4.1, these approaches are the individual loss assessment, the area average index, and the parametric index.
How do these three types of data rank in terms of cost, accuracy, and speed? Typically, the individual loss adjustment is the most (p.64) expensive and the slowest; the area average index is in the middle on both counts; and the parametric is the least expensive and fastest. However, the order is reversed when it comes to accuracy, with individual loss adjustment on top, followed by the area average index and then the parametric index.
Applying Rules-Based Approaches to Disaster Planning
The three approaches described in Box 4.1 are also the right ones to consider when thinking about developing data systems that will provide a foundation for automated specific response plans. In some cases, a crude individual loss assessment, such as targeting all those displaced or all those with a collapsed roof or flooded home, may be quite effective. The data system underlying this approach would have to operate much like loss adjustment works for individual indemnity insurance.
This approach requires a system for individual loss adjustment and auditing, as well as trained adjusters. Otherwise, the response might be too slow, or it might be manipulated by those on the ground. Some conditions may also have to be set for protection, such as requiring that a building satisfy a certain building code. In most post-disaster situations, however, the default setting should not be requiring implementing agencies to wait for a full comprehensive assessment of damage or loss. Rather, the response should be designed to evolve as new data become available. Where early action is important, as it usually is, triggers should be based on index insurance products (area average or parametric indexes, depending on the relative cost, speed, and accuracy). The use of clear and transparent triggers based on the information that is immediately available (and not with long delays on losses linked to data collection) could serve as a kind of triage for prioritizing response. Actions, then, would be based on forecast rainfall or harvest failures, the distance to the epicentres of earthquakes and their magnitude, or the number of cases of a serious infectious disease.
(p.65) These triggers could build on the data generated by the existing early warning systems, which are much better than they used to be. For example, better information is available on hydro-meteorological systems and their likely consequences, including their impacts on harvests, flooding, or cyclones; slow-onset disasters can be better predicted, including links to famine or public-health problems; and even early warning systems for rapid-onset disasters such as earthquakes and tsunamis are in place in some countries and are getting better.14
It is important to continue to improve early warning systems, but by themselves they are by no means sufficient for good decision making. An early warning is of limited use if the main response, especially support for reconstruction and for protecting livelihoods, is always late. The longer the wait after a disaster, the easier it is to justify a need, but acting early is typically much more cost-effective. For example, reports of Ebola circulated for many months before any kind of serious international action was taken.15 Similarly, drought in some areas in the Horn of Africa and East Africa was reported in 2011, many months before the world took notice and began to respond.16 To work as an index insurance product, a trigger should not lead to a set of options for a decision-making body; it should result in an automatic decision. In other words, a defined set of indicators reaching particular pre-agreed values should lead to a defined action, as in insurance. Early warning systems would turn into early action systems.
Other monitoring data can also be useful when contemplating action. Some observers have claimed that the Ebola outbreak was subdued not only by social mobilization but also by careful information gathering and processing (such as incidents of unsafe burial or rumours of hidden cases) that received an immediate response. Similarly, there is good evidence that in the 2005 Kashmir earthquake systematic information platforms such as RISEPAK offered live updates on where the needs were throughout a long period, thereby improving the effectiveness of the response.17 Thanks to the advances (p.66) in satellite imaging and social media and in other digital areas, systematic information gathering is now considerably easier, but planning beforehand on how to use it and what information will trigger specific action is important. And some data systems, such as those for national health and nutrition surveillance data, are still quite expensive and will take many years to build.
User-generated data are, of course, susceptible to fraudulent manipulation, and so by themselves are not useful as a trigger for financing. However, they could still be used as part of a rules-based system for action, in particular for guiding when to collect the data needed for more accurate but expensive back-up triggers. For example, suppose a government wants to protect farmers against a severe loss in crop production and chooses to do so using a combination of a trigger based on rainfall and a back-up trigger based on an area average yield index. If the area average yield index is quite expensive to calculate, the government could decide whether to collect the data for a back-up trigger using a cellphone survey that indicated it had been a very bad crop year—a situation not being picked up by the rainfall trigger.
Rules and triggers also can be applied to early action, recognizing that no early-action rule can be perfect. It is inevitable that a response will be too late for some droughts and too early for others. But a rule for early action does not have to be perfect to be better than waiting—it just has to be good enough. And the science of droughts is certainly good enough to ensure that early action is better than waiting.
In fact, providing cash or food to households early in the face of an ensuing drought seems to be much more cost-effective for reducing food insecurity than waiting until the drought is in full swing. In the Horn of Africa and pastoral areas of East Africa, the rainy season in early 2011 failed, and the rains the previous year had also been poor, so substantial hardship could be expected. Nevertheless, several months passed before a response began, leading to delays in reaching people.18 A more appropriate response would have been to trigger at least some actions as soon as it became clear that the rains were poor, several (p.67) months earlier, even though farmers still had reserves by then. In that case, help could have been on the ground much more quickly, pre-empting difficulties (and benefiting from much cheaper operations).
Reconstruction of damaged lifeline infrastructure such as hospitals and key roads after a large earthquake is another area that could benefit from agreement before a disaster on an objective, transparent, independent, manipulation-resistant procedure for determining the damage and rules for determining who will pay for reconstruction. And yet this crucial work is often delayed because after a disaster different parts of government are negotiating over the total cost of the damage and who will pay.
Just as it is for disaster insurance, the heart of a minimal response plan will always be the data. If the data are too easy to manipulate, too costly, too slow, or too unreliable, the system will not be politically sustainable. And just as for insurance, getting it right will require investing in systems and people before disasters, so that the data-collection process can run smoothly during and in the aftermath of a disaster. Good data will be at the core of any attempt to change the default setting from inaction to action because they will give credibility to the decision rules required.
For Benefactors, a New Strategy
Benefactors who care about the impact of their financial support should be willing to settle for a new decision-making system. Instead of waiting for the appearance of begging bowls after a disaster, benefactors could agree to commit the majority of their funds to financing the planning, preparation, and implementation of the coordinated default response to a disaster as described in the response and recovery plan. It would not be a vague, unwieldy plan, but a pre-agreed, coordinated plan with a specific, defined set of actions to which they would commit.
To make this work, discretion needs to be replaced by rules to guide decision making—that is, triggers for action and a credible commitment. (p.68) It will help to overcome procrastination and inertia, the enemies of fast decision making. In the process, many of the bureaucratic or political incentives for inaction will disappear. And moving from post-disaster discretion towards pre-disaster rules can help to clarify ownership of the risk—who is responsible for what—and this can promote good incentives for investments in preparedness and risk reduction and avoid regret afterwards.
Even when decision making is governed by decision rules and algorithms and early warning leads automatically to early action, there is still ample room for political leadership. It is not about the computers taking over, generating triggers to which one needs to respond without judgement. It is about changing the default setting. Instead of inaction and the status quo being the default, triggers will start actions, and policymakers will need to act to stop action.
Leaders of national governments and international organizations would then have to justify why they stopped or changed their early action systems from being implemented in, say, a drought. In the recent Ebola outbreak, for example, slow decision making and delayed declaration of an emergency were generally acknowledged to have caused loss of life and hardship for many. The approach to a disaster response we propose would have changed the dynamic of the decision making, putting a focus on stopping action if leaders had so wished. During a drought, a response would also be triggered automatically—for example, based on harvest forecasts from weather data. Leaders could stop the response, but they would no doubt be cautious in doing so.
As plans are implemented, new data become available, and further actions and course corrections are triggered, leaders will need to keep everyone on board and hold the course. They will need to show concern and commitment to all those affected, communicate what is being done to all involved, and coalesce all in the mission to deliver. No pre-agreed plan, no rules will be perfect. There will be times when triggers fail or are imprecise. Leaders will need to judge information on unforeseen matters, justify any deviations from plans, and act accordingly, without allowing special interests to take over and game the system.
(p.69) The result of this new approach will be the emergence of a new accountability during disasters. Meanwhile, leaders can take political credit for implementing the plans to which they previously agreed. Yes, they can overrule systems, but they can expect to be judged for deviations from agreed plans.
1. By ensuring that as little as possible must be decided by stakeholders when a disaster strikes, rules can promote decisive, timely action.
2. A rules-based system is only as good as the data that drive it. The data need to be resistant to manipulation and strike the right balance between cost, speed, and reliability.
3. Any data that could trigger action will depend on investments before a disaster in design of the data-collection system, including an audit function, and in the human and technological capacity to collect data in a timely manner.
4. Three types of data are particularly useful for triggering post-disaster action: ground data on the damage to or losses of people and buildings, area average index data on damage and losses, and parametric indexes.
5. No rule is perfect, and so there should be some discretionary back-up system to deal with situations in which the rules fail.
6. Benefactors should channel their financial support into precise sets of plans in which it is clear who exactly is being protected, how, and who is paying.
A Snapshot of the Literature
Early behavioural economics literature provides strong motivation for considering simple plans. Simon’s Rational Choice and the Structure of the Environment (1956) hypothesized that human beings have limited cognitive capacities, and therefore simple approaches may be more (p.70) effective than technically first best approaches that are not sufficiently well understood. One implication is that using simple algorithms and clear decision rules will improve decision making (Gigerenzer and Goldstein 1996). For example, in medicine the use of checklists in diagnosis has contributed to much better diagnosis and treatment (Pronovost et al. 2006; Gawande 2010). Clear heuristics also help to overcome confirmation bias when relying on experts. This bias is the tendency to search for or interpret information in a way that confirms one’s preconceptions, leading to statistical errors.
Rational decision making is also affected by other cognitive biases (for an introduction, see Kahneman 2011). Present bias is the tendency to over-value immediate rewards at the expense of long-term intentions. It can lead to procrastination—situations in which the current self defers decisions towards some long-term goal while planning for some future self to pursue it. Hyperbolic preferences can be used to model this behaviour, which is called time-inconsistent and leads to regret (as the future self will have wanted the current self to have acted differently). O’Donoghue and Rabin (1999) discuss such preferences in more detail. Present bias may be overcome using commitment devices and other ways to increase self-control. Thaler and Bernatzi (2004) show how ‘Save More Tomorrow’ can overcome lower pension savings because of the present bias caused by limited self-control. The essence of the programme is straightforward: people commit in advance to allocating a portion of their future salary increases towards retirement savings. Ashraf et al. (2006) found that commitment savings products can increase savings for people who may display present bias in the Philippines.
Another bias affecting decision making is status quo bias, a preference for the current state of affairs, even if there would be gains from change (Samuelson and Zeckhauser 1988). Loss aversion is a possible explanation (Kahneman et al. 1991). Overcoming status quo bias is possible through changes in the default option, such as automatically enrolling workers in pension schemes and making them act to leave the scheme rather than making them act to join. Madrian and Shea (2001) show how (p.71) changing this default increases the uptake of retirement savings products in a context in which too little retirement savings take place.
At the level of governments, time-inconsistent preferences have also been observed. It is well recognized that governments are often tempted not to do what they said they would do, or do things now that later on they will regret. This time inconsistency can often benefit from institutional designs that make government’s announcements credible. For example, a credible commitment to a monetary policy rule, such as the Taylor rule (Taylor 1993), could perform substantially better than a discretionary approach in which government is given full discretion at every moment in time (Lucas 1976; Kydland and Prescott 1977). This insight led to central banks being given some degree of political independence over monetary policy, so that they can operate under clear rules and not the discretion of political leaders.
Time inconsistency problems in public policy appear in a range of situations. For example, in developing countries a government is unable to make a commitment to private firms that, if they invest in public infrastructure such as a water or electricity distribution network, they will be allowed to recoup their investment (Estache and Wren-Lewis 2009). A government that cannot commit to sticking to its original contract—that is, it will not renegotiate a contract as soon as the private investment in infrastructure is made—will not find any private firms willing to invest in infrastructure, leading to an infrastructure deficit.
Buzzacchi and Turati (2014) find that in a situation in which a beneficiary is able to make investments in risk reduction and a benefactor is unable to observe these investments, a decision by the benefactor to cap its total budget for discretionary relief—a commitment device—can improve welfare by reducing the incentive for the beneficiary to underinvest in risk reduction.
The economics of insurance fraud, or misreporting of information, offers some insight into insurance contracting. Townsend (1979) considers a model in which verifying claims is costly, and Lacker and Weinberg (1989) consider an alternative formulation in which (p.72) verifying claims carries no cost, but the beneficiary is able to falsify claims at some cost. Both models find deductibles to be part of the optimal solution—for losses below a threshold there will be no insurance claim payment—and Lacker and Weinberg’s model recognizes the use of underinsurance for extreme, falsifiable perils. Both findings are observed in insurance markets.
The idea of an indexed approach to financial protection probably began with the book by Chakravarti (1920) outlining a detailed proposal for the sale of rainfall-indexed insurance across India. Since then, the idea has gained momentum, particularly for agricultural insurance in developing countries—the traditional farm-based loss adjustment was too expensive and exposed the insurer to moral hazard (Hazell 1992; Skees et al. 1999; Hess et al. 2005)—and for quick post-disaster liquidity at the sovereign level (Ghesquiere and Mahul 2007), where a detailed assessment of losses is too slow for immediate post-disaster needs.
The downside of indexed protection is that the index may not accurately capture the actual situation on the ground, thereby underestimating a severe loss or overestimating a minor one. Unsurprisingly, the more inaccurate the index, the less useful it is as a risk management tool (Clarke 2016) and the lower the demand for indexed protection (Mobarak and Rosenzweig 2012). One might reasonably believe that an inaccurate index would be little used and therefore would do no damage. However, Morsink (2015) found that Ethiopian farmers who were offered indexed protection but did not take it up received significantly lower discretionary post-disaster transfers from other farmers in their community than farmers who were not offered the indexed protection. Low quality indexed protection, it seems, can crowd out informal risk sharing.
How challenging is it to develop accurate indexes, particularly parametric indexes? Jensen et al. (2014) found the accuracy of a parametric index designed by a top academic team to capture the drought-induced mortality of livestock in northern Kenya quite poor. Clarke et al. (2012) conducted a similar but somewhat crude analysis of (p.73) weather index insurance for Indian farmers. They found that the indexes miss an average of one of every three catastrophe years. Various authors have argued for long-term public investments in accurate indexes (Verdin et al. 2005; Carter et al. 2007; Chantarat et al. 2007; 2009).
Indexed social protection has been implemented by the government of Mexico through its CADENA programme, which provides farmers with free state-level insurance against drought. It has been shown to induce positive risk-management responses demonstrated by the higher yields observed where coverage is available (Fuchs and Wolff 2011). A range of authors have considered the opportunities for and practical challenges of using indexes to trigger shock-responsive social protection (Alderman and Haque 2007; Barnett et al. 2008; Bastagli and Hardman 2015). Pelham et al. (2011) make the case for the importance of shock-responsive social safety nets as a tool for managing disaster risks. In particular, the authors show that safety nets can be useful both pre-disaster, to prevent and mitigate disaster risks, and post-disaster, to cope with the effects of natural disaster shocks. Hobson and Campbell (2012) have investigated the conditions needed to achieve a successful response to shocks through a social safety net.
Building on Hobson and Campbell (2012), Slater and Bhuvanendra (2014) argue that a successful shock response through a safety net programme requires a wide range of institutional and financial arrangements to be in place. Bastagli and Holmes (2014) analyse the criteria necessary to determine whether social protection is effective in responding to shocks, while Grosh et al. (2011) look into what can be learned from previous crises about what may constitute an appropriate crisis response. (p.74)
(10.) An alternative approach would be to use rules to trigger financing but leave the prioritization until after the disaster—for example, to be decided under a (p.118) clear accountability and decision-making framework using information from a Post Disaster Needs Assessment (PDNA).
(13.) Some issues have to be monitored when moving to cash. One is that cash transfers retain their value. Therefore, if prices increase, the value of the transfers needs to be adjusted in a reasonable way to keep purchasing power constant. Also, food supplies need to be guaranteed, but this is about monitoring markets and potentially promoting or supporting imports, not necessarily delivering food to people directly.
(1) There are exceptions—for example, for rain gauges or river flow stations that have to be read manually. However, even these are increasingly automated.
(14.) For example, the Japanese earthquake and tsunami early warning system saved thousands of lives in the aftermath of the 2011 Tōhoku earthquake and tsunami by enabling high-speed trains to be slowed so that they did not derail, dangerous machinery to be shut off so it did not cause damage, and people to find cover or move to high ground so they were not injured or killed. See ‘How Japan’s Rail Network Survived the Earthquake’, Railway Technology, <http://www.railway-technology.com/features/feature122751/>.