1.2.1 Elite Motivation

I begin from the expectation that countries which have faced greater hazards in the past will prepare more in the future. This idea is not new (Fox and Weelden Reference Fox and Van Weelden2015), but I offer a more nuanced and comprehensive explanation for why this is the case. Existing work highlights an empirical correlation between past hazards and preparation but does not sufficiently identify the causal mechanism linking exposure to greater preparedness (Keefer et al. Reference Keefer, Neumayer and Plumper2011; Hsiang and Narita Reference Hsiang and Narita2012). Do incumbent governments view natural hazards as a threat to their economic performance (World Bank 2009; Hsiang and Jina Reference Hsiang and Jina2014)? Do they perceive natural hazards as a potential cause of conflict (Hsiang et al. Reference Hsiang, Burke and Miguel2013)?

I suggest instead that ruling elites anticipate a potential threat to their incumbency when citizens can observe the preparedness counterfactual. If an individual has not faced a hazard in the past, as in places that experience these events rarely, they will have little ability to separate the quality of preparedness efforts from the intensity of the hazard itself. In other words, because the effects of a hazard are endogenous to preparedness efforts, someone without past experience of a similar event will be unable to gauge the value of existing investments to their or the broader welfare. These individuals cannot conceive of the preparedness counterfactual at the time of a natural shock: “what would have been the impact of a disaster in the absence of preparedness spending?” (Healy and Malhotra Reference Healy and Malhotra2009: 389).

In contrast, those individuals who have previously been affected by a natural hazard are able to evaluate the quality of preparedness efforts at the time of a future hazard – the counterfactual to what would have happened without preparedness – and can use this information to evaluate the quality of incumbent politician performance. This expectation contrasts with the idea that individuals cannot observe and evaluate preparedness programs, relative to disaster response (Healy and Malhotra Reference Healy and Malhotra2009; Gailmard and Patty Reference Gailmard and Patty2019): where individuals have previously experienced a failure to prepare, I posit that they will be able to evaluate the quality of preparedness efforts at the time of a future hazard.

Even if we assume that individuals have no ability to observe the actual policy changes that have occurred, in the form of disaster preparedness, otherwise similar individuals may still differ in the interpretation of their current welfare based on a comparison between the effects of past and current natural hazards. An individual who has been exposed to a natural disaster in the past, particularly one in which the government was ill-prepared, will have different expectations for the likely outcomes of a subsequent hazard. If the results of that hazard are better than in the past, then they are likely to have a higher estimate of their current welfare than someone who had not been exposed to a disaster in the past and so has much different expectations about the effect that a hazard would have on their wellbeing.

This supposition is in line with qualitative evidence regarding the behavior of individuals who were and were not exposed to a significant previous hazard at the time of a new hazard. In India’s Odisha state, which has a history of tropical storms, a severe cyclone in 2013 affected areas that had and had not been exposed to a similarly intense cyclone in 1999. Officials involved in evacuation efforts at the time noted that those individuals living in areas that had previously been affected were easily evacuated multiple days in advance, while those who had not been hit by earlier cyclones were considerably more reluctant to evacuate and, in the end, some forced evacuations were necessary in these regions (Odisha State Government Official #2 2014). If behavior at the time of the cyclone is associated with differing past exposures to natural hazards, then we might believe that other kinds of behaviors, such as voting, may also exhibit these differences.

Politicians who are knowledgeable about past hazard exposure can then anticipate this future evaluation and incorporate it into their policymaking decision calculus (Bussell Reference Bussell2017). The opportunity to observe multiple shocks over time implies that once a natural hazard has occurred, politicians may assume that at least some proportion of voters will be able to estimate the preparedness counterfactual based on their experience with this shock. In this sense, past experience of disasters changes the information available to a segment of voters, making them more competent at evaluating the performance of the incumbent than their non-disaster-exposed peers. As a result, the perceived electoral value of preparedness may increase, making preparedness activities more compelling as an electoral strategy than was previously the case. Investment in disaster preparedness, then, becomes a strategic electoral move on the part of incumbent elites, to increase the chances of retaining power in the face of a future hazard.

The relevance of this calculation to preparedness policies, however, depends on whether ruling elites perceive a general threat to their incumbency. This is a relatively straightforward expectation that politicians in electorally competitive democratic regimes or authoritarian regimes with a viable opposition will be willing to implement new policies that they expect to be received well by the public. Even if ruling elites anticipate that citizens can evaluate their preparedness efforts, these evaluations will only matter in contexts where those citizens can pose some sort of political threat, through voting, support for an opposition party, riots, or otherwise.

Combining these two sub-variables of elite motivation results in four general expectations for the character of incentives regarding disaster preparedness, summarized in Table 2. We should observe the strongest political incentives for preparedness where both past exposure and opposition threat are high. Where past exposure is high, but opposition threat is low, we should still expect to see medium to high incentives, given the risk of instability and negative public response in the face of a future hazard. Where there is a clear opposition threat but low past exposure, politicians will have only low to medium incentives to prepare, as the costs of preparation will compete with other policy areas that may have more obvious benefits to voters. Finally, where both past exposure and opposition threat are low, we should see the weakest political incentives for reform.

The discussion to this point thus suggests that where ruling elites have an expectation that disaster preparedness efforts are likely to result in electoral support at the time of a future hazard, and where they face a competitive environment in which they would benefit from additional popular support, we should expect to see political interest in disaster preparedness initiatives. To be successful in these initiatives, however, also requires the capacity to coordinate relevant actors who are capable of designing and implementing such reforms. I now explore this aspect of the argument in greater detail.

1.2.2 Capacity

The second key variable in my argument is the capacity of the state to coordinate disaster preparedness efforts. Following existing research, I conceptualize “capacity,” applied to state or non-state actors, as the ability of an organization to achieve the goals they set for themselves (Centeno et al. Reference Centeno, Kohli and Yashar2017; O’Reilly and Murphy Reference O‘Reilly and Murphy2022: 1). For a government organization, this typically involves “the ability to raise revenue efficiently, the ability to enforce a monopoly on violence within its territory, and the provision of public goods in such a way that supports the functioning of markets, especially the legal capacity to attain the rule of law” (O’Reilly and Murphy Reference O‘Reilly and Murphy2022: 1). For non-state actors, this implies parallel abilities to raise and manage financial resources, run a functional organization, and deliver relevant goods to target clients. Such a conceptualization implicitly assumes consistent underlying institutional structures that support and enable this capacity. This is, then, an argument about institutional capacity to support preparedness initiatives.

Substantial work in the social sciences has highlighted the importance of bureaucratic capacity for the effective functioning of government institutions. In their agenda-setting work, Evans and Rauch evaluate the degree to which state institutions exhibit “Weberianness” as a measure of state capacity, considering the presence of competitive salaries, internal promotion and career stability, and meritocratic recruitment (Evans and Rauch Reference Evans and Rauch1999). This capacity, they argue, is then linked to the ability of states to achieve economic growth (Evans and Rauch Reference Evans and Rauch1999). More generally, the quality of institutions came to be seen as a key factor affecting outcomes of both economic and social welfare (Rothstein and Teorell Reference Rothstein and Teorell2008, see also Levitsky and Murillo Reference Levitsky and Victoria Murillo2009).

Related work focuses on the frequent gaps in, or limitations of, government capacity and looks outside the state to identify alternative providers of public goods. In the case of Kenya, Brass (Reference Brass2012, Reference Brass2016) shows how nongovernmental organizations (NGOs) increasingly play both direct and indirect roles in government service provision, through their participation in governance and in the direct delivery of services. More generally, Cammett and MacLean argue that “In many parts of Africa, Asia, the Middle East, and Latin America, states are unable to provide extensive social welfare services, but a diversity of non-state actors such as multinational corporations, ethnic and sectarian organizations, NGOs, community-based organizations, and families provide and facilitate access to much of the welfare that exists on the ground” (Cammett and MacLean Reference Cammett and MacLean2011: 1–2). In addition, the role of non-state actors in service provision is seen only to be increasing (Cammett and MacLean Reference Cammett and MacLean2011: 1–2). These arguments imply that we may observe cases of non-state actors filling in where the state is unable to implement programming directly.

At the same time, it is insufficient to assume that non-state actors have consistent capacity to provide these services across countries. Non-state actors differ dramatically in their size, resources, modes of work, and connections to international affiliates or sponsors.Footnote ⁹ Relatedly, the relationship between non-state providers and state actors themselves is likely to have an important effect on outcomes (Cammett and MacLean Reference Cammett and MacLean2011), particularly in those countries with significant restrictions on civic freedoms. Thus, variation in the capacity of non-state actors is also likely to affect their ability to play a role in disaster preparedness.

My first expectation related to capacity, summarized in Table 3, is that the governance of disaster preparedness efforts, and specifically the character of roles for the state and civil society actors, will depend on the relative capacities of these groups. Where both the state and civil society are of high capacity, the state should lead disaster preparedness efforts while effectively drawing on civil society actors to support its mission, which I characterize as “State-Led.” In contrast, where only the state has high levels of capacity, disaster preparedness efforts are likely to rely predominantly on the bureaucracy in a “State-Dominant” model. Here, when the government draws on non-state actors, it is likely to be only for additional support during times of hazard response, rather than in planning and preparedness exercises. In those cases where civil society has more dominant capacity, a “Society-Reliant” model may arise, with government delegating as many preparedness activities as possible to non-state actors. Finally, in areas with low capacity in both the state and civil society, what disaster preparedness activities we observe are likely to be haphazard and uncoordinated, given the minimal ability of local actors to successfully implement reforms.

Overall, I argue that we should see the quality of disaster preparedness initiatives increase in line with levels of state capacity, where the government is able to direct the activities of state and/or non-state actors in developing and implementing preparedness initiatives. Because the government itself is the primary focus for most disaster preparedness programming, I expect outcomes to be better where the state has high capacity versus cases where civil society has high capacity. High state capacity is also important to ensure that diverse actors engaged in preparedness efforts do not, at best, duplicate work and, at worst, engage in activities that conflict with each other.

1.2.3 Theoretical Implications

The two primary variables in my argument combine to produce a set of expectations about the overall character and quality of disaster preparedness that I expect in each country or subnational region, as summarized in Table 4. A key implication of the argument is that disaster preparedness efforts will be insubstantial where ruling elites do not have incentives to support them, even in the context of high state capacity. Given this, we should observe the strongest performance in those places where there are considerable political incentives to prepare and there is overall high capacity to do so. We should also observe considerable efforts to prepare in those places with high political incentives but where there is more limited capacity. In these contexts, I expect to see evidence of governments making creative efforts to engage in preparedness within the kinds of diverse governance structures outlined in Table 3. In contrast, where there is higher capacity but only limited political incentives, we may see efforts that look like preparedness, but these will be more “window dressing” than deep and substantial interventions. This is because the government has the resources available to put forward basic policies and/or structures of preparedness, but ruling elites do not have the incentives to implement these initiatives in a comprehensive way. Finally, where there are lower electoral incentives and lower capacity, we should see the weakest efforts to invest in preparedness. In these cases, we may observe individual and disconnected programs, but no widespread and thorough disaster preparedness effort.

2.2.1 Component 1 – Making Disaster Risk Reduction a (Political) Priority

In the best cases, political attention to disaster risk reduction is reflected in the presence of a central government body that is established through formal policies, active in disaster-related planning and implementation, and that incorporates both public and private agencies into its processes. This body should be linked to, and actively engaged with, lower-level bodies that are also officially responsible for disaster-related planning. Government bodies are also actively engaged with community actors.

2.2.2 Component 2 – Assessing Risks and Enhancing Early Warning

The second component of disaster preparedness concerns the ability of national governments to anticipate future hazards through active risk assessment, monitoring of potential hazards, and programs for early warning. While there remain limitations on risk assessment, monitoring, and early warning practices in all the countries considered here, those performing well on this measure exhibit a range of tools and techniques for collecting and managing information on natural hazards. In general, these countries have established national systems in place and, often, are partnering with international actors to manage comprehensive programs that account for a range of different hazards in their area.

2.2.3 Component 3 – Building a Culture of Safety and Resilience

Developing a culture of safety is at the heart of the third disaster preparedness component. In this evaluation, I focus primarily on information sharing within government at the central level, as well as educational and awareness programs.

2.2.4 Component 4 – Reducing Underlying Risk Factors

The fourth component focuses on mitigation and risk reduction, particularly in the context of climate change. The emphasis here is on policies and practices to reduce the overall risk that a natural hazard will evolve into a disaster. Thus, these are efforts that go beyond preparing for what to do when a hazard occurs and instead emphasize ways to change and improve practices to limit the threats associated with hazards. In practice, these are the types of activities that were rarely observed in fieldwork, as most countries are still focused on immediate response activities and, at best, efforts to prepare for hazards. Nonetheless, there are some concerted efforts, especially in those countries scoring relatively high in this area.

2.2.5 Component 5 – Preparedness for Response to Natural Hazards

The final disaster preparedness component concerns a country’s overall approach to preparedness. In those countries exhibiting relatively high performance, the authority(ies) for organizing and implementing disaster management protocols is(are) clear, there are funds allocated to these activities, and there are subnational programs in place related to disaster response.

2.3.1 General Lack of Preparedness

The first argument, that governments will not prepare for natural hazards, is established in theoretical work on natural hazards (Healy and Malhotra Reference Healy and Malhotra2009; Gailmard and Patty Reference Gailmard and Patty2019). I provided empirical evidence in this and the previous section that invalidates this expectation, and will only offer a brief elaboration here. The evidence presented in this section on the thirteen country cases and subnational states in India, clearly shows that nearly all of these cases have engaged in at least a minimum of disaster preparedness. In only one country, Togo, and two states, Chhattisgarh and Jharkhand, are scores low on all measures. While these are relative scores, they still indicate attention to the question of disaster preparedness, in the manner emphasized by international standards, across a range of issue areas and policy techniques.

2.3.2 Past Exposure to Natural Hazards

One established explanation for disaster preparedness where it does occur rests on past exposure to natural hazards (World Bank 2009; Keefer et al. Reference Keefer, Neumayer and Plumper2011; Hsiang and Narita Reference Hsiang and Narita2012; Hsiang et al. Reference Hsiang, Burke and Miguel2013; Hsiang and Jina Reference Hsiang and Jina2014; Fox and Weelden Reference Fox and Van Weelden2015). While my argument incorporates past exposure into a broader theoretical logic, the existing expectation focuses solely on past exposure. Those countries that have experienced hazards, and resulting disasters, in the past should be more likely to invest in preparedness for similar events in the future, holding constant other characteristics.

I test this hypothesis using measures of past exposure from the EM-DAT global database on hazard events. Figure 2A plots the relationship between the share of the population affected by natural hazards over a sixteen-year period and each country’s performance on the international standards for disaster preparedness, as discussed in this section.Footnote ¹⁷ Figure 2B plots the relationship between the number of events per million people and disaster preparedness performance for India’s eighteen largest states. In neither case does the graph display a clear correlation between these two measures, suggesting that past exposure of a population to natural hazards does not on its own predict future investments in preparedness.

Figure 2A Past exposure is not directly associated with preparedness outcomes – country cases

Figure 2B Past exposure is not directly associated with preparedness outcomes – Indian states

2.3.3 Economic Conditions

If past exposure does not explain preparedness, then perhaps it is those countries with better economic conditions who are better able to prepare (Cohen and Werker Reference Cohen and Werker2008). In other words, is there a positive correlation between national wealth and disaster preparedness? While there may be some relationship between economic conditions and preparedness, this is by no means clear or consistent, when considering gross domestic product per capita in purchasing power parity or nominal terms, as shown in Figures 3A and 3B. The variation in outcomes is most obvious for those countries with overall scores of 4 and 5 on disaster preparedness. While India is the wealthiest country and also performs at the highest level on preparedness, Bangladesh, Ethiopia, and Malawi all scored four overall and are substantially less wealthy than India on a per capita basis. This distinction is most stark for the case of Mozambique, which also scored at the highest level, but is one of the two poorest countries overall in the study. In Figure 3C, I plot the relationship between gross state domestic product per capita and disaster preparedness again for India’s eighteen largest states. There is also no clear relationship between economic strength and preparedness outcomes in India’s states.

Figure 3A GDP per capita (PPP) displays no clear association with preparedness – Country cases

Figure 3B GDP per capita (nominal) displays no clear association with preparedness – Country cases

Figure 3C GSDP per capita (USD) displays no clear association with preparedness – India’s states

2.3.4 Moral Hazard

A final possible predictor of government investments in preparedness is the degree to which country leaders expect external actors, such as international aid agencies, to provide relief at the time of a hazard (Cohen and Werker Reference Cohen and Werker2008). If this type of moral hazard is at play, then those countries most dependent on external aid should prepare the least and those least able to rely on outside actors should prepare the most.

The most straightforward measure of moral hazard is arguably past receipt of international aid. Those governments that are accustomed to receiving aid should be more likely to expect aid in the future. To measure past receipt of aid, I draw on AidData measures of international aid receipts in 2009, standardized by population. As I do not have similar subnational measures for India, I do not test this hypothesis in the Indian states. As shown for the country cases in Figure 4, there is no clear relationship between aid receipt and performance on the disaster preparedness.

Figure 4 Past international aid is unassociated with disaster preparedness

Each of these descriptive analyses suggests that there is little evidence to support existing arguments highlighting individual country characteristics such as past exposure, economic conditions, or moral hazard as likely determinants of government preparedness for natural hazards. What this does not imply is that there is no association at all between these characteristics and preparedness outcomes. Instead, my argument highlights the important ways in which both past exposure and access to international aid may contribute in important, but under theorized, ways in combination with other state characteristics. I explore these relationships in greater detail in the following sections.

3.2.1 Measuring Past Exposure

To measure past exposure, my politically oriented argument implies that only certain types of exposure to natural hazards should be relevant to preparedness investments. What I expect to matter to individuals’ perceptions of disaster intensity and damages, and thus voting behavior, is the degree to which they have personally been affected by past hazards and can then use these experiences to gauge the relative effects of future events. Thus, only those indicators measuring the effects of past events on current voters should be relevant to political calculations over preparedness. This is not to say that voters care only about the effects of disasters on their own welfare, but rather that it is only those individuals who have been affected in the past, and who are able to vote in the future, who are likely to be able to estimate the effects of later preparedness efforts and vote on that basis.

Given this argument, I develop a measure of past exposure based on the number of individuals affected by a hazard. Specifically, I draw on estimates from CRED’s Emergency Events Database (EM-DAT) for the total individuals affected by natural disasters, normalized to a per capita estimate. To account for year-to-year variability in natural hazard exposure, I generate an overtime measure of the indicator, based on sixteen-year totals and the sixteen-year average, during the period 1995 to 2010.Footnote ¹⁹ This period accounts for the ten years prior to the adoption of international standards under the Hyogo Protocol and the first six years of the implementation period. The logic here is to consider any hazards that viably affected adults who would be politically relevant during the policymaking period under consideration. My preferred measure is the total affected individuals over the sixteen-year period (per capita), as this is the best indicator of the total number of individuals who have recently been affected (even if some are doubled counted, as these individuals may be even more responsive to disaster-related policies).

I also use a secondary measure to evaluate exposure at the subnational level in India. Unfortunately, because many events cross state boundaries, per state data on the number of affected individuals for each hazard event is not available in the EM-DAT database. As an alternative, I consider the absolute number of events that occurred in a state over the same period, and develop a population normalized measure (events per one million people) to compare past exposure in the Indian states.

3.2.2 Measuring Opposition Threat

Effective measurement of opposition threat can depend on the nature of the regime. Given the range of regime types included in this study, it is necessary to consider the differing types of opposition faced by ruling elites in electoral democracies versus autocracies. To do so, I divide the case countries first based on regime type and then apply relevant measures of opposition threat to each subset of countries.

There are a range of techniques for measuring the level of democracy, or lack thereof. Here, I am most concerned with whether a country operates as a viable electoral democracy or not, where electoral democracy is understood as a country with a free and fair electoral system, political competition, and protection of basic civil liberties. This is a “procedural minimum” conceptualization of democracy (Collier & Levitsky Reference Collier and Levitsky1997: 434). The Freedom House Electoral Democracy indicator maps well to this conceptualization, as it incorporates measures of civil liberties, political rights, and the character of the political process. I combine this measure with information from a prominent dataset on autocratic regime transitions, which codes periods of autocratic and democratic rule (Geddes, Wright, and Frantz Reference Geddes, Wright and Frantz2014). A regime is coded as autocratic if the leader(s) came to power by undemocratic means, if the leader(s) came to power by democratic means but then changed institutional rules to limit future electoral competition, or the military effectively interfered in democratic elections (Geddes, Wright, and Frantz Reference Geddes, Wright and Frantz1994: 317).

For those cases coded as democracies, I measure opposition threat as a function of electoral competition. Where electoral competition is higher, there is a greater threat to the ruling regime from the opposition. I use the Legislative Index of Electoral Competitiveness indicator from the Database of Political Institutions, which codes the number of seats held in the national legislature by the largest party relative to other parties. For democracies, the relative scores are 5: only one party won seats, 6: multiple parties won seats, but one party received more than 75 percent of seats; and 7: the largest party won fewer than 75 percent of seats. I consider a score of 5 low competition, 6 medium competition, and 7 high competition.

For subnational units in India, the same measures of electoral competition and capacity are unavailable or less informative in their application. For electoral competition, all 18 major states are coded as 7, highly competitive, during the electoral terms from approximately 1995–2010 (51 election cycles),Footnote ²⁰ other than one term in one state. Thus, a better measure would be whether the ruling government changed during the same period, which encompassed the opportunity for up to three changes in ruling government. Here, in two states (including one that held only two elections) there was no change in government, in five there was one change (including the other two newly formed states that held two elections each), in five the ruling government changed twice, and in the remaining six the government changed in all three elections. I code as highly competitive those states where the ruling party(ies) changed at least twice and medium competitive the remaining states, with the recognition that a reasonably high level of electoral competition appears in every state during this period.

An additional note on electoral competition, perhaps better described as political alliance, concerns the role of party politics at the state and central levels. There is evidence that central governments in India privilege the state governments led by their party (Khemani Reference Khemani2003; Panda Reference Panda2016). Thus, state governments might be expected to be more likely to implement policies put forward by their party at the national level, to maximize potential benefits from that specific policy. For that reason, I also draw attention in the case discussion to whether the party that implemented a disaster preparedness program in a state was the same as the party that promoted the initiative in the central government.

For autocracies, it is somewhat less straightforward to measure opposition strength. Here, I draw on a new measure of Autocratic Party Strength developed by Self (2022), which considers the degree to which the party is in control at the local level as well as centrally. Self’s primary measure is an overall indicator of the strength of all parties in a country. To estimate the relative strength of the opposition, he draws on a second measure of constraints on the opposition, based on the assumption that where parties are strong in general but there are significant constraints on the opposition, the measure of party strength should be generally measuring strength of the ruling party, not the opposition. I use the same combination of these two measures to estimate opposition threat, as shown in Table 9. The logic of these predictions is as follows: where party strength is high and constraints are high, there is a strong ruling party with little viable competition. Where party strength is high and constraints are low, there is significant risk of competition from strong opposition parties. Where overall party strength is low, but constraints on the opposition are high, there should also be minimal opposition threat. Finally, where overall party strength is low, but constraints are also low, there is at least some chance an opposition might emerge to threaten ruling elites.

3.2.3 Application of the Measurement Framework

By applying the past natural hazard exposure measure to the country cases, I estimate the affected share of population summed over a sixteen-year period to range from 7 percent to 125 percent. The annual average affected share of the population ranges from 0.4 percent to 8 percent. For the purposes of categorization, I categorize countries as having low past exposure if their total affected population is <25 percent and/or their average annual affected population is <4 percent. The categorization of country cases based on this measure is shown in Table 10A.

Within India, states ranged from having experienced between 0.2 hazard events per million people and 1.8 events. I categorize states as having high past exposure if they experienced 0.5 or more events per million inhabitants. The coding of Indian states according to this measure is shown in Table 10B.

For opposition threat, four countries among the cases are clear electoral democracies – Ghana, India, Malawi, and Senegal – while eight others are classified as autocracies – Bangladesh, Ethiopia, Gambia, Mozambique, Pakistan, Togo, Zambia, and Zimbabwe. There is one intermediate case, Kenya, which is classified as a democracy in the Geddes et al. data but not as an electoral democracy by Freedom House. I include Kenya within the set of electoral democracies based on the country receiving a score of 7 on the measure of legislative electoral competitiveness in all the years under consideration. The Indian states are all classified as democratic.

The application of the opposition threat measures to the country and state cases results in the categorizations shown in Tables 11A and 11B.

Combining these measures of past natural hazard exposure and opposition threat produces the categorization of cases shown in Tables 12A and 12B. My argument, as summarized in Table 2, predicts that the greatest elite motivation to prepare for natural hazards should appear in those countries and states in the lower right quadrant of the tables. Elite motivation to prepare should also be present, though somewhat more muted, in the countries and states in the lower left quadrants. Cases in the upper right quadrant should demonstrate a low to medium level of incentives, while the lowest motivation for preparedness should occur in cases occupying the upper left quadrant. For all these cases, it is useful to remember that there are no countries coded as having low potential threat from the opposition, so we might expect to see some evidence of related dynamics in all cases, depending on past exposure. In the following section, I explore evidence from the country cases to examine the extent to which observed political behavior is consistent with these expectations.

3.3.1 High Elite Motivation for Preparedness – the Case of India

As in the previous section, I begin with an in-depth discussion of India, to detail the ways in which the combination of past exposure and substantial electoral competition has motivated political actors to prepare. India is a federal democracy with parliamentary legislatures at the central and state levels. Significant policymaking authority and implementation power is delegated to the states, and this includes primary responsibility for disaster management.Footnote ²⁴ The bureaucracy, in the form of the Indian Administrative Service (IAS) and State Administrative Services, also plays an important role in policy design and implementation, as discussed in greater detail in Section 4. However, India’s elected officials retain significant leverage over the bureaucracy and should be seen as the primary actors in determining whether new policies are put in place (Bussell Reference Bussell2012, Reference Bussell2019).

In the years leading up to and including this study, India’s electoral democracy witnessed robust and highly competitive elections, both nationally and at the state level. The 1990s saw the central government under the power of coalition governments led first by the Indian National Congress (INC) and then the Bharatiya Janata Party (BJP), followed by a return of the Indian National Congress in 2004 and continuing through Reference Bussell2014, when the BJP returned to power. These coalitions required support from multiple smaller parties for the ruling government to retain power. Similar electoral dynamics were at play in many, if not most, of India’s states, and the states can provide initial evidence of a policy shift toward disaster preparedness.

It is common for Indian policy innovation and evolution to begin at the state level.Footnote ²⁵ The era of disaster preparedness in India can be traced to the October 1999 super cyclone 05B that devastated the coastal state of Odisha. While cyclones are a regular occurrence on the eastern Indian coast, this super cyclone brought with it damages of a new magnitude, killing more than 10,000 people (Mohanty Reference Mohanty2021) and directly affecting more than 1.2 million, or 3 percent of the state’s population. This was a shock to both the state’s citizens and its leaders. By the end of the year, the state government established India’s first disaster preparedness body: the Odisha State Disaster Mitigation Authority (now the Odisha Disaster Management Authority). The agency was tasked with coordinating all disaster-related activities, including preparation, hazard response, and reconstruction. The design of this agency was a model for the national-level policy put in place six years later.

Though the 1999 super cyclone was clearly a motivating force, it was not necessarily the case that the state government would respond in this way, particularly given that disaster management is not included explicitly in the Indian Constitution and, thus, at this time fell under the primary ambit of the Union (central) government, not the states. Odisha was also not a particularly wealthy state at the time, having a below-average state domestic product per capita in the year of the cyclone.Footnote ²⁶ What else, then, might have contributed to the state embarking on a substantial institutional innovation at this moment?

I suggest that the Indian National Congress government in power in the state saw this cyclone as a threat to its incumbency. The INC was in its first term leading the state since the emergence of the Janata Dal (JD) as a viable electoral threat. While the previous Janata Dal Chief Minister and former INC member Biju Patnaik had passed away in 1997, the opposition was now led by his son, Naveen Patnaik, in the form of a Janata Dal breakaway party, the Biju Janata Dal (BJD). The BJD was set to face Congress in its first state election in February 2000, just months after the cyclone hit. And while the 3 percent of the population affected by the cyclone might not seem substantial in electoral terms, its electoral impact could be substantial. The ten districts affected by the cyclone contain nearly half of Odisha’s state assembly constituencies and India’s first-past-the-post, single-member district electoral system frequently sees seats won by a small percentage of votes.Footnote ²⁷ Thus, affected voters in these areas could feasibly have a significant effect on election outcomes.

Importantly, the cyclone was also seen as a major failure of state institutions. As one journalist recounting the event put it, “The 1999 cyclone remains etched in our memories as an apocalyptic failure of the administration” (Mohanty Reference Mohanty2021). At the time, disaster management activities, such as they existed, were managed at the district level but relied on state- or national-level resources, such as the telecommunications infrastructure and weather monitoring. In 1999, these institutions failed, with dire results. “[T]he Odisha government learned about the change of path of the super cyclone pretty late and soon after the state’s communication networks collapsed and the situation could no longer be monitored” (Santosh Kumar, as quoted in Mohanty Reference Mohanty2021). Even if there had been more advance warning, there were not evacuation procedures and structures in place and the lack of oversight from the state level also meant that there was minimal coordination across districts. This range of administrative failures is seen by many as causes of the disaster (Ray-Bennett Reference Ray-Bennett2016).

The incumbent government did not have sufficient time to implement programs whose relevance would be immediately obvious, as I argue this is only possible at the time of a future hazard. Nonetheless, it could do something that would feasibly be understood by survivors to reflect an understanding of where the state had failed. This is the context in which it rapidly implemented a new state-level authority with explicit responsibility for coordinating natural hazard-related activities of all forms, including preemptive disaster management activities, coordination of all actors involved in disaster mitigation and response, and reconstruction after a hazard occurs. These characteristics reflected emerging best practices internationally at that time but were unique in India and also echoed the character of failures in the face of the super cyclone.

Nonetheless, when the 2000 election took place, the Congress government was punished by state voters, and particularly those in the areas most affected by the cyclone. A coalition of the BJD and the Bharatiya Janata Party won a combined 75 percent of the assembly seats in the ten districts affected by the 1999 cyclone, while only winning 71 percent of seats in non-affected districts (for an average of 72 percent of all seats).

Erramilli (Reference Erramilli2008) has described the 1999 cyclone as a “focusing event” for the state government, in that the policy changes subsequently taking place reflected a shift not only in policy, but in mindset, from a focus on disaster response to preparedness. He also posits that this shift was in line with the dynamics of party institutions at the time, which were consistent with the provision of public goods in the form of disaster planning (Erramilli Reference Erramilli2008. See also Chhibber and Nooruddin 2004). My argument is different, in that he does not focus on how the cyclone generated electoral relevance for disaster preparedness substantially after the focusing event.

The continued electoral relevance of disaster preparedness is evident in the attention to policy implementation and improvement after the BJD coalition came to power. Rather than assume the Congress policy was a failure due to their losing the election, the new government saw it as an opportunity to further efforts supporting recently affected areas. Not only did the BJD-led government implement the disaster management authority, but it also initiated in 2001 a first-of-its-kind Disaster Rapid Action Force as a specialized unit of the Police to be trained and supplied to act immediately in the face of a natural or industrial hazard. This was followed by a formal disaster management policy in 2005 (before the national policy was passed) and continued support for disaster preparedness activities throughout the period under consideration. These efforts are viewed as highly successful, with the state effectively managing large evacuations and other preparedness actions during subsequent cyclones that have resulted in substantially fewer deaths during major hazards (Harriman Reference Harriman2013).

The second state to implement a disaster management initiative with an emphasis on preparedness was Gujarat. As in Odisha, this occurred after a major disaster, in the form of the 2001 Bhuj earthquake. But understanding the policy change process in Gujarat requires attention not just to this precipitating event, but also to an earlier natural disaster, Cyclone 03A, that hit the Gujarat coast in mid-1998. Why did the state government launch its major disaster preparedness efforts after the earthquake, but not this cyclone?

Cyclone 03 A caused substantial destruction along the Gujarat coast in June 1998. Across 12 of the state’s districts, an estimated 3,000 people were killed and 4.6 million were affected (9 percent of the state’s population) (EM-DAT). The state government was critiqued for its failure to notify the public of the impending storm, amid substantial structural destruction across the coast (Rawat Reference Rawat2019).

What was the state’s response? High-level officials note that there was an initial exercise to develop a cyclone-oriented policy in the wake of the storm. Yet, that effort dwindled and produced no tangible results (Erramilli Reference Erramilli2008: 185). It was not until three years later, in the aftermath of a massive earthquake, that the state government made a clear policy shift.

The 7.6 magnitude Bhuj earthquake is estimated to have killed 20,000 people and affected more than 6 million, or 12 percent of the state’s population. Twenty-one of the state’s thirty-three districts were affected. The broader infrastructural destruction was also significant, with major damage to roads and buildings in the affected regions.

In this case, the policy response was rapid and substantial. In addition to a major relief effort, the state government of Gujarat launched the Gujarat State Disaster Management Authority (GSDMA) less than two weeks after the quake occurred. As in Odisha, the agency was charged with overseeing all hazard-related activities, with an eye not just to response, but also to mitigation. This was the beginning of a significant investment in preparedness that can be seen today in the state’s disaster preparedness programs.

The state government that launched the GSDMA was the same one that failed to engage substantially after the cyclone, so what was the difference? It is not sufficient to say simply that that the earthquake was a larger disaster, even though it arguably was. The 1998 cyclone had similar costs and affected individuals as the 1999 cyclone in Odisha, and yet it did not result in the same policy reforms.

I posit that significantly different electoral conditions related to the earthquake increased the political relevance of natural disasters to a degree that pressed the government’s hand regarding the necessity of this policy move. This was the case due to the timing of elections, the character of BJP support in the affected regions, and the electoral risks posed by voters affected by natural disasters repeatedly in a single electoral term.

First, the 1998 cyclone occurred just months after the BJP successfully won reelection in the state and was not expecting another assembly contest for five years. The earthquake, in contrast, happened two years before the next planned election. While still not as proximate to an election as the 1999 cyclone in Odisha, reelection prospects would have been closer to the front of mind for politicians than previously.

Second, the districts affected by the cyclone were in a region of particularly strong support for the BJP, with the party winning 83 percent of the assembly seats there, in comparison to 64 percent overall in the state. Contrast this with the districts affected by the earthquake in 2001, but not the 1998 cyclone. Here, the BJP had won only 57 percent of the assembly seats. This region, then, posed a much higher risk of retribution in the upcoming election in the face of limited policy response. In addition, with more time leading up to the election, there was a chance that initial preparedness efforts could pay off in the face of any additional hazards prior to the election.

Finally, it is significant that all the districts affected by the 1998 cyclone were also affected by the 2001 earthquake.Footnote ²⁸ This suggests that voters in these areas, though previously strong supporters of the BJP, would be able to perceive that very little had been done in the intervening years to address the threats posed by natural hazards. This, combined with lower support in the other earthquake-hit regions, meant that the BJP faced a potential natural hazard-related electoral threat in more than two-thirds of the state’s districts. A lackluster response was no longer a viable electoral option.

The next major policy shift was at the national level. The central Indian government introduced two major policy reforms, the Disaster Management Act of 2005 and the National Policy on Disaster Management in 2009, during the terms of the Indian National Congress party-led United Progressive Alliance. The first policy set out plans for institutions at the central, state, and district levels to provide response to natural hazards. The second policy offered a more comprehensive framework for disaster preparedness and greater incorporation of local and community-based actors.

Why did the national government choose this moment to implement reforms? The perceived need for a disaster policy was not new to experts in the field or national leaders. After the 1999 cyclone in Odisha, the central government established a High-Powered Committee to review the need for disaster management in the country. A highly detailed report of the committee was released in 2001 and laid out recommendations that covered nearly or all of international standards adopted in Hyogo, four years before the agreement was initiated.

When the report was released, however, the constitutional authority for disaster preparedness remained unclear and the states of Odisha and Gujarat were already mobilizing to address disasters at the subnational level. The central government’s response at the launch of the report, then, was to delegate responsibility to the states. There was no clear economic or other reason not to implement a national policy, the government simply did not perceive a sufficient incentive to do so. This perspective changed dramatically after the 2004 Indian Ocean tsunami.

The 2004 tsunami, caused by the Sumatra-Nadaman earthquake, is estimated to have killed more than 16,000 people in India and affected nearly 700,000 across the states of Tamil Nadu, Andhra Pradesh, and Kerala, and the union territories of Puducherry and the Andaman and Nicobar Islands. This event is viewed as one of the worst natural disasters in modern Indian history. It was also a turning point for India’s disaster management strategy.

The central government’s Disaster Management Act was passed on December 26, 2005, one year to the day after the 2004 tsunami. In a government Task Force review of the Act conducted in 2010, the authors note that initially the GoI decided that policies should be only at state level, despite the recommendation of HPC for both central and state policies. But “in the aftermath of the 2004 Asian tsunami, the GoI decided that a central law on disaster management was essential” (Ministry of Home Affairs, 2013: xvi).

A key difference between the tsunami and major previous major disasters was that it severely affected more than one state. As a result, the varying levels of preparedness across the states could have substantial effects on how different citizens were affected. Yet, there was no central government plan in place for attempting to assist and resolve these differences. The tsunami made clear the substantial potential costs to human lives and infrastructure of such an approach.

The costs were also made evident in three electorally important states. The Congress party was in power in all the major states affected by the tsunami, Andhra Pradesh, Kerala, and Tamil Nadu, either on its own or in a coalition government. At the same time, it also faced substantial competition in each place, with strong opposition in the form of the Telegu Desam Party, Left Democratic Front, and Dravida Munnetra Kazhagam–led coalition, respectively. Andhra Pradesh and Tamil Nadu had also both played important roles in the Congress’ success in the national parliament. A failure to respond to a disaster affecting such a large segment of the population across multiple politically relevant states could pose severe electoral threats.

The 2005 Act and the subsequent Disaster Management Policy were also designed in ways that suggest a response to subnational demands for increased disaster preparedness support, in the wake of diverse hazard exposure across the country.Footnote ²⁹ Rather than developing programs resting solely within the central government, in both cases the policies provided resources to actors at the subnational level, in addition to formalizing central bodies, in a manner that appealed to regionally based political parties in the coalition. Thus, the central government retained overall control of the policies while allowing for regionally specific programming both to address substantial variation in hazards and to enable local actors within the overall framework.

After the introduction of the 2005 national act, the considerations for state level policies changed. All states were mandated to implement their own disaster management authorities to oversee preparedness and response efforts. Generally speaking, states have “an incentive, in the form of federal grants and fiscal transfers, to adopt and implement central guidelines” (Erramilli Reference Erramilli2008: 33). At the same time, evidence suggests that such transfers can be influenced by the political alignment between central and state authorities (Khemani Reference Khemani2003; Panda Reference Panda2016). This implies potentially varying incentives at the state level for implementation. In practice, we see that such national mandates frequently do generate staggered responses in the states, such as with the adoption of the constitutional amendments mandating new decentralized political institutions at the local level (Bolhken 2015). Similar choices appear in the case of state disaster authorities, with a pattern that reflects an electoral logic.

In the two years following the national government’s new Act, nine states introduced disaster management authorities, five of which were led at the time by the Congress government. Some of the quickly adopting states had relatively high numbers of previously exposed individuals, such as Assam and Bihar, and the national policy may have reduced barriers to the development of policies to serve these voters. The first group to adopt after the national policy also included those with low levels of past exposure, but these were nearly all states led at the time by the Congress party. It is this electoral alignment with the central government that provides a logic for implementation in these cases, including Andhra Pradesh and Karnataka.

Of the six states that delayed the launch of their authority by multiple years, in contrast, none were led by the Congress in the two years following the national act. Multiple of these states subsequently introduced a disaster management authority when leadership of the state government switched hands to the Congress, such as Chhattisgarh and Punjab, or, when led by the BJP after the national government switched power to a BJP-led coalition in 2014, such as Rajasthan. This suggests that an alliance with the central government was key to decisions to implement new policy, particularly in those cases that had not experienced substantial recent hazards affecting large portions of their populations.

Overall, India’s strategy toward disaster-related policies reflects attention to the electoral risks associated with diverse natural hazards in a competitive electoral environment. Regular experience with hazards in certain parts of the country, and past voter response to these hazards, provides substantial incentive for the central government to provide not only a comprehensive national-level institutional structure, but also to enable state-level policies in line with the specific incentives that exist for those subnational governments. In turn, state governments’ actions indicate a similar set of dynamics, in interaction with incentives generated by the federalist system.

3.3.2 High Electoral Incentives for Reform

The remaining countries where I expect to see a strong electoral incentive for reform are Bangladesh, Kenya, and Malawi. Bangladesh offers a particularly interesting case. Within the central government, politicians appear to perceive a relationship between risk of future hazards and the need for preparedness efforts, which may in part be due to the experiences of past leaders. Governments in Bangladesh have faced popular unrest due to severe natural disasters, such as the flood-related famine of 1974 (Hossain Reference Hossain2017; Shabhanaz and Bussell Reference Shabhanaz and Bussell2017: 13). Considering both earthquakes and cyclones, compounded risks due to increased frequency and climate change, respectively, have been met with new efforts to invest in preparedness.

Subnationally, however, the character of disaster preparedness at the local level seems to suffer from a bureaucracy-heavy implementation model. Local bureaucrats have primary responsibility for implementing preparedness programs, leaving electorally minded local politicians with little opportunity to influence or claim credit for preparedness efforts. “This effectively disincentivizes elected local public officials from following through [on] their commitment to work with communities to tackle disasters, participate in training programs, and understand the catastrophic impact of natural disasters on the most vulnerable communities” (Shabhanaz and Bussell Reference Shabhanaz and Bussell2017: 30). In addition, the lack of coordination between the local bureaucrats and politicians “creates an information gap whereby the central government fails to maximize the use of local knowledge in mapping and managing disasters, and relies on institutional data rather than high-resolution, qualitative data to develop and implement their plans” (Shabhanaz and Bussell Reference Shabhanaz and Bussell2017: 31).

Overall, these dynamics suggest that political incentives from both past exposure and electoral competition can encourage disaster preparedness efforts in high-risk regions. Yet, elected politicians must be in a position to play a role in designing and/or implementing these policies for such motivations to generate positive outcomes. Where politicians see no path to benefiting from preparedness, they are unlikely to exert effort and, in so doing, may diminish the effects of otherwise promising policies.

3.3.3 Medium-High Political Incentives for Preparedness

Countries that I expect to have medium to high levels of political incentives for disaster preparedness derive these incentives from high levels of past exposure and low to medium levels of opposition threat.

These countries are Ethiopia, Mozambique, Zambia, and Zimbabwe. In line with these expectations, the incentives for preparedness in Ethiopia derive more from past exposure than electoral competition. At the time of this study, the coalition government led by the Ethiopian People’s Revolutionary Democratic Front (EPRFD) was highly entrenched with little viable opposition and was placing pressure on pro-democracy advocates (Reimer et al. Reference Reimer, Tripson, Ven Johnson and Ven Johnson2014: 117). Thus, the government faced little direct electoral pressure that might result in preparedness initiatives. Yet, as shown in Section 2, Ethiopia is one of the higher-performing countries in this study on disaster preparedness initiatives.

While electoral incentives from opposition competition are not a major factor here, there is potential for public dissatisfaction due to a failure to prepare for natural hazards. The most common and substantial natural hazard in Ethiopia is drought, and the risks of food security and malnourishment associated with drought have only increased in recent years. In addition, “Poor land and environmental management and dramatic population growth in recent decades have increased food insecurity and heightened demands for food aid and emergency relief during drought crises” (Reimer et al. Reference Reimer, Tripson, Ven Johnson and Ven Johnson2014: 97). Even without these recent changes, the fact that there are “many people living in high-risk areas makes them highly vulnerable to natural hazards regardless of whether shocks are frequent or not” (Reimer et al. Reference Reimer, Tripson, Ven Johnson and Ven Johnson2014: 97).

This widespread exposure to past natural hazards across a significant portion of the population increases the salience of such issues within the government. As multiple interviewees noted, “Ethiopia has experienced natural hazards and disasters in the past, so the government is currently investing in preparedness” (Reimer et al. Reference Reimer, Tripson, Ven Johnson and Ven Johnson2014: 117). That this response is due to concerns about broader perceptions among the public, rather than electoral competition, was also made clear in interviews. “The government does not rely on electoral success to stay in power, so it is unlikely that they are improving DRM for electoral benefit. But the government may be spending more on preparedness to improve their legitimacy in the eyes of their people and the international community” (Reimer et al. Reference Reimer, Tripson, Ven Johnson and Ven Johnson2014: 117). In sum, the Ethiopian government is concerned with the public perceptions of state engagement in preparedness efforts, rather than any specific threat from other political parties, and this has implications for its support of preparedness initiatives.

3.3.4 Low-Medium Political Incentives for Preparedness

In the third set of countries, Gambia, Ghana, Pakistan, and Togo, high levels of potential threat from political opposition combined with lower historical levels of exposure to natural hazards suggest lower incentives to promote policies targeted toward disaster preparedness.

Ghana offers an interesting case in which observers note a government interest in disaster preparedness alongside the public’s unwillingness to adjust to natural hazard threats. A review of the country’s progress on disaster preparedness during the five-year period noted that the national government had “successfully strengthened human capacity through training volunteers, raising awareness of DRR in the public sphere through the media, and establishing regional DRR platforms” (DeCuir et al. Reference DeCuir, Fuerst, Iannuzzi and Leaver2014: 80, FN 295). At the same time, the national disaster management agency, NADMO, faced resistance to more substantial programs for individuals in high-risk areas, such that “when the government attempts to relocate populations to less vulnerable locations, they clash with government officials and often return to vulnerable areas” (DeCuir et al. Reference DeCuir, Fuerst, Iannuzzi and Leaver2014: 80). This suggests that while the government perceives preparedness programming to be politically relevant, individual citizens do not recognize sufficient risk to make significant changes to their living environments, even with government assistance.

This contrast between policy aims and public perceptions of risk is consistent with a context in which there is substantial electoral competition, but where most citizens have not been directly affected by natural hazards in the past. While Ghana’s electoral democracy is not without some issues (DeCuir et al. Reference DeCuir, Fuerst, Iannuzzi and Leaver2014: 79), it has experienced competitive elections for multiple decades, with regular alternation in political power at the national level. Thus, there may reasonably be an incentive for elected officials to demonstrate their effectiveness to citizens. And in practice, observers note that “Ghanaians expect NADMO and the government to care for affected communities after a natural disaster” (DeCuir et al. Reference DeCuir, Fuerst, Iannuzzi and Leaver2014: 80). But these same individuals are not so directly affected by these hazards that they are willing to support government investments in preparedness, making it difficult for the state to move forward effectively with preparedness initiatives.

3.3.5 Low Political Incentives for Preparedness

The category for country cases with low electoral incentives to invest in preparedness includes only Senegal. I expect to see relatively little evidence of political efforts to promote disaster preparedness.

Senegal demonstrates how low levels of political incentives can result in a focus on disaster response, rather than preparedness. In those cases where the government has attempted to focus on preparedness, such as in efforts to move residents from high-risk to low-risk neighborhoods, individuals tend not to be ready to make substantial changes to reduce their risk profiles (Agnihotri et al. Reference Agnihotri, Child, Koob and Wald2014: 46). On the other hand, the government seems to perceive that disaster response is a clear area of voter interest and that this can be utilized to electoral ends. “The Senegalese government appears to be responsive to the public’s calls for disaster response when the failure to provide assistance affects the government’s political capital” (Agnihotri et al. Reference Agnihotri, Child, Koob and Wald2014: 46) and at the same time, “Evidence from the interviews suggests that flooding has provided an opportunity for the ruling party to diminish the popularity of opposition members in office by withholding or delaying resources after natural hazards” (Agnihotri et al. Reference Agnihotri, Child, Koob and Wald2014: 47). This latter strategy appeared to be the case in the flood-prone city of Saint Louis, where flood assistance declined in the year after an opposition mayor took power in the city (Agnihotri et al. Reference Agnihotri, Child, Koob and Wald2014: 45).

This evidence suggests that the preparedness efforts observed in Senegal are only minimally driven by political incentives. To the extent that preparedness goals are aligned with strategies to support disaster response, these efforts are likely to gain political support, but broader, programmatic changes are unlikely to emerge from direct political initiative in this environment.

4.1.1 State Capacity

My country-level measure of state capacity draws from the O’Reilly and Murphy (Reference O‘Reilly and Murphy2022) Index of State Capacity, which is an index of measures from the V-Dem dataset. The baseline measure includes V-Dem scores for “the rule of law, the authority of the state over its territory, the rigorousness and impartiality of public administration, whether public expenditures are on particularistic or public goods, the modernity of the state’s source of its revenue, and the universality of the provision of education” (O’Reilly and Murphy Reference O‘Reilly and Murphy2022: 713). I begin with the baseline capacity measure for the five years prior to the start of the study in each region, 2007–2011 for African countries and 2010–2014 for South Asian countries. I then generate an average capacity measure for each country across the five-year window. The Index of State Capacity baseline measure ranges from −4 to 4 and the average scores for countries included in this study range from −1 to 1. I classify those countries receiving a capacity score of −1 or 0 as lower capacity and those scoring a 1 as higher capacity, as shown in Table 14A. That said, these groupings indicate capacity levels relative to each other, not the global set of countries, given that these countries all sit at the middle of the range of scores and none are toward the extremes.

For the Indian states, there are no established measures of subnational capacity in the literature. Here, I use an alternative measure of corruption in the delivery of public services. This measure is intended to capture the quality of the bureaucracy via the degree to which the state apparatus can direct the behavior of its individual bureaucrats toward the successful and efficient implementation of government programs.Footnote ³¹ The measure comes from a citizen survey conducted in 2004–2005 by Transparency International India and the Centre for Media Studies. Respondents were asked about their experiences with eleven different citizen-facing departments and an overall corruption score was generated for each state (Transparency International India 2005).

Among the eighteen major Indian states, the corruption scores range from 240 (lowest corruption) in Kerala to 695 (highest corruption) in Bihar. Because there is no clear “break” between higher- and lower-corruption states, I normalized the scores on a scale from 0 to 10 and reversed the order, so that the lowest scoring on the corruption index have the highest capacity scores. I then coded those states scoring 5 or higher on the indicator as high capacity and those scoring below 5 as lower capacity. I use this measure to estimate the levels of state capacity in these states.

4.1.2 Civil Society Capacity

Measurement of civil society capacity is not as well established in the existing literature as state capacity. I aim to generate a measure that considers both the presence of non-state organizations and the ability of these organizations to operate within a given country context. In addition, I consider historical measures of international aid.

I first generate a measure of local nongovernmental organization presence based on data from the World Organization of Non-Governmental Organizations (WANGO). WANGO tracks currently active NGOs, so these data are more recent than the other measures used here, but are the most comprehensive available across countries. The per capita measure is based on the total number of domestic NGOs listed for each country and current (2022) population data. The case countries fell into three general categories, which I categorize as having an NGO presence that is Low (<4/million people), Medium (>4 and <8/million people), or High (>8/million people), as shown in Table 15.

The second measure of non-state capacity is based on the presence of international aid in a country. While some aid typically goes directly to governments, the ability of domestic NGOs to operate also often rests on financial and other inputs from international actors.Footnote ³² Here, I use measures from AidData of overall bilateral and multilateral commitments to each country in 2009, which was the most recent year for which there was data for all countries in the study. I then generated a per capita measure based on populations in 2009. Again, I subsequently place the countries into three groups, following a general sorting in the data that uses the following rubric: Low (<$40/person), Medium (>$40 and <$90/person), and High (>$90/person), shown in Table 16.

In theory, I would prefer a measure of domestic non-state capacity that accounts for the status of civic freedoms in each country. This would help to ensure that any NGOs that are listed as being active are actually able to work successfully. In practice, inconsistent availability of data on civic freedom across the study countries means that it is not possible to develop a measure that takes this into account in a consistent manner.Footnote ³³

My alternative, to combine measures of domestic NGO presence and international aid, while limited in its direct measurement of civic freedom, also has benefits. The presence of domestic NGOs may be not only a direct measure of NGO presence, but also an indirect measure of civic freedom, as implied by the ability of individuals to form and maintain these organizations. This measurement strategy also, as noted earlier, accounts for the frequent relationship between international aid and domestic NGOs.

To facilitate a more general discussion, it is useful to collapse the capacity measures into two categories. I collapse the multifaceted civil society capacity measure into a single dimension by grouping those countries that received a “high” on either measure with those receiving a “medium” on both measures as higher capacity and the remaining countries as lower capacity, shown in Table 17.

For India, I use the same world database of NGOs as for the cross-national comparison. These data include the state where an NGO is based and so can account for state-level presence of NGOs. I then generate a normalized measure based on the number of NGOs per million people in the state, as shown in Table 18. While there is some NGO presence in all states, a few states have a substantially higher presence than the others. I do not have a comparable measure of international aid commitments at the subnational level in India; thus, I rely here solely on the measure of NGO presence.

4.1.3 Overall Capacity Profiles

The set of government and civil society capacity measures enables an overall categorization of countries into two-dimensional capacity profiles, with predictions in each category for the likely governance style of disaster preparedness management, shown in Tables 19 A and 19B.

What are the implications of these profiles for an argument about the relevance of latent capacity to disaster preparedness? My primary supposition is that latent capacities should set the stage for particular models of disaster preparedness. Specifically, I expect differing alignments of state and non-state capacity to result in differing governance strategies for preparedness, where preparedness exists at all. In other words, latent capacity does not on its own determine whether a country will exhibit disaster preparedness. Rather, capacity serves as a foundation for preparedness efforts and shapes the character of that preparedness where it occurs. In the context of disaster preparedness, it also matters who has the capacity. I anticipate that the most comprehensive disaster preparedness initiatives will appear in those contexts where the state is high capacity and has an electoral incentive to apply this capacity to preparedness efforts, as elaborated in the previous section.

4.2.1 India – a High-Capacity State, with Subnational Variation

In India overall, my capacity measures suggest an outsized role for the state in preparedness, with a minimal and subsidiary role for non-state actors. It is useful to provide additional background on the Indian administrative set-up as context for the specifics of the case discussion. The Indian government operates through multiple layers of administrative hierarchy, with leadership resting in the central services. These administrative groups include the Indian Administrative Services (IAS), a generalist branch that runs most departments; the Indian Police Service (IPS); Indian Foreign Service (IFS), and others. Officers in these services are hired through a highly competitive, country-wide merit-based process, and after joining a service are typically posted to a specific state cadre for their career. Once an individual reaches a certain level of tenure and experience, they may be eligible for positions in central government departments, such as the Department of Road Transport and Highways, the Central Bureau of Investigation, or the National Disaster Management Authority. State governments also have their own administrative services, to fill positions at lower levels within government departments. Though not without critique, these bodies provide the foundation upon and through which all government programs are implemented in India.

In practice, for disaster management, there is a clearly delineated set of central government organizations with responsibility for leading disaster-related efforts, as well as delegation of responsibilities to related bodies at the subnational level. Three departments – the National Disaster Management Authority (NDMA), National Disaster Response Force (NDRF), and National Institute for Disaster Management (NIDM) – are responsible for, respectively, policy guidelines related to disaster preparedness and response, active response to natural hazards, and providing disaster training programs.

The functioning of these organizations reflects the substantial administrative capacity of the central Indian state. Take, for example, the NDMA. In most high-level bureaucratic positions in India, members of the administrative services are placed in positions based on their generalist training, not due to specific areas of expertise. In the NDMA, however, all senior officials in the secretariat have previous experience with disaster-related issues, such as having previously worked in a disaster-prone area (NDMA Official, March 24, 2014). At the same time, the generalist training of senior bureaucrats can be particularly useful in a cross-cutting issue area such as disaster preparedness. A key role for the NDMA is to generate guidelines for mainstreaming disaster preparedness in both cross-departmental and regular activities of line departments. Having administrative leadership with experience across multiple departments should enable these activities.

The NIDM, charged with capacity building, takes an active role in developing overall training programs and specialized initiatives to respond to new hazards. When unanticipated landslides and floods hit the state of Uttarakhand in 2014, the institute sent a team to conduct site visits and interviews as the basis of a report on the disaster (NIDM Official, March 25, 2014). Reports of this kind are then used to disseminate information and contribute to future training programs. The training programs themselves are implemented based on requests from states and other organizations, including NGOs. But the programming is offered by the NIDM itself, not through partnerships with other organizations.

In these ways, the disaster management and preparedness efforts of the Indian government are conducted largely without the support of, or coordination with, civil society actors. While local NGOs might participate in specific training activities or assist in response, these efforts are clearly designed and implemented by the state in line with plans and procedures set out in government policy. Thus, India offers strong support for a state-dominant model in a case where state capacity outweighs that of the non-state sector.

Within the state cases, there should be slight variations on this model. In Gujarat, high state capacity with low presence of civil society suggests a dominant role for the government in disaster preparedness initiatives, with minimal engagement of non-state actors in implementation of state policies. Odisha, in contrast with Gujarat, ranks relatively lower on state capacity – scoring near the midpoint of the states – but has a higher presence of non-state actors to potentially participate in preparedness initiatives. Thus, state-led initiatives that actively engage with local nongovernmental organizations are more likely there. Andhra Pradesh is a high-government capacity state, scoring nearly the same as Gujarat on this indicator. In contrast with the other cases, however, Andhra also has a relatively high presence of NGOs. To the extent that there are disaster preparedness programs in place, they should be state-led initiatives that leverage the availability of non-state actors. The final case, Karnataka, represents low levels of both state and civil society capacity. Given these institutional conditions, there should be more haphazard management of any existing preparedness activities.

In Gujarat, there is a highly state-driven preparedness model, as anticipated by my argument. The introduction of the state disaster management authority was done in a manner that focused explicitly on leadership of, and execution by, the state. Indeed, as one expert notes, at the level of state policy, “Gujarat seemed to have dispensed with an institutionalized role for NGOs during non-disaster times” (Erramilli Reference Erramilli2008: 88). This approach also prevailed locally, such that Erramilli observed, in Gujarat’s village-level preparedness and response activities, “unlike in Tamil Nadu or Orissa [Odisha], NGOs did not play a major role” (Erramilli Reference Erramilli2008: 84).

In contrast with Gujarat, Odisha’s preparedness model reflects a state-driven approach with greater reliance on non-state partners. Development of the state disaster management agency was done with substantial input from the UNDP and local non-state actors were a focus of training initiatives (Odisha Government Official, March 28, 2014). Recounting interviews with Odisha state officials, Erramilli also noted that “respondents, by and large, repeated the official policy position that underscored the value of non-official participation. They appeared to have recognized the severe limitations of state abilities, which was evident in their cooptation of NGOs” (Erramilli Reference Erramilli2008: 81).

The Andhra Pradesh government has also played a leading role in its preparedness efforts. At the same time, it has drawn on multiple academic partnerships, including inputs from the Central Research Institute for Dryland Agriculture (ICAR-CRIDA) for its drought forecasting efforts (Tejaswi and Kumar Reference Tejaswi and Kumar2011: 447). There is less direct evidence of NGO activity in disaster preparedness efforts, and fewer preparedness activities in general than in Gujarat or Odisha. Nonetheless, what programs exist are clearly led by the state, in line with my expectations for a high-capacity state.

Finally, in Karnataka, one of the most successful preparedness initiatives has been implemented by an autonomous government body, the Karnataka State National Disaster Monitoring Centre (KSNDMC). This administrative setup allows the organization to hire specialized individuals from outside the traditional administrative hierarchy. The team is made up of scientists and support staff, rather than bureaucrats from Karnataka’s lower-capacity administration (Chakrabarti Reference Chakrabarti2019: 137). Thus, here there is an effort to draw from alternative non-state sources, academic rather than NGO, to build new disaster-specific capacity within the state. This is consistent with expectations for a state with lower levels of both government and NGO capacity.

Overall, these state cases reflect both the general tendency for Indian preparedness efforts to privilege state actors in developing and implementing programs with outside participation only where there are gaps in state capacity and available resources in the non-state arena.

4.2.2 State-Dominant Capacity

Similar to India, in case countries where the state has significant capacity relative to civil society, such as in Malawi and Zambia, I expect to see disaster preparedness efforts driven almost entirely by government, with little obvious presence of civil society participation.

Malawi is a useful example of a high-capacity state. Overall, interviews support the general perspective that bureaucratic capacity is relatively high in Malawi and particularly regarding levels of corruption, which are seen to be “vastly improved over many other African countries” (Bussell and Malcomb Reference Bussell and Malcomb2014: 155). This suggests the potential for a strong state role in disaster preparedness.

In practice, the development of disaster preparedness programs in Malawi has been tightly tied to the efforts of international donors and nongovernmental organizations. First, regarding donors, interviews highlighted that “Much of the recent efforts by donors have been concentrated on building capacity of Malawi’s government to better respond to natural disasters, climate change, and other risk factors” (Bussell and Malcomb Reference Bussell and Malcomb2014: 115). Second, while the overall presence of domestic civil society organizations is relatively low, those that exist “are becoming increasingly connected and more influential in the stakeholder process on disaster planning” (Bussell and Malcomb Reference Bussell and Malcomb2014: 156). Thus, external groups have been actively working to support the government in its disaster preparedness efforts and “most government organizations appeared to be working with select donor organizations in an effective and cooperative manner that benefits vulnerable populations and results in increased investment” (Bussell and Malcomb Reference Bussell and Malcomb2014: 156). This offers a nice illustration of the ways in which effective state capacity can be used to leverage even seemingly minimal non-state capacity to improve preparedness. While the government is a primary actor in preparedness efforts, rather than rely only on government institutions, Malawi’s bureaucracy has engaged with civil society actors to enable them to play a more prominent role in Malawian preparedness efforts than otherwise expected, in concert with the government.

4.2.3 State-Led Capacity

Among the thirteen country cases, two exhibit high levels of both state and civil society capacity: Ghana and Senegal. I highlight here the approach taken in Ghana as an example of outcomes in this category.

Administratively, Ghana’s national disasters-related body, NADMO, is primarily responsible for coordinating both state and nongovernmental actors in preparedness and response efforts, as expected by my argument. However, the effectiveness of NADMO is limited by its institutional development and relative place in the government hierarchy. For example, “NADMO was unable to forecast and provide early warnings to affected areas so that stakeholders could prepare and respond to the emergency early. In addition, NADMO lacks an emergency operations center where all NGOs and stakeholders can meet and receive updated and accurate disaster assessments” (De Cuir et al. Reference DeCuir, Fuerst, Iannuzzi and Leaver2014: 70). Government policies on disaster preparedness and disaster risk reduction have also given a primary planning role to district authorities, rather than NADMO. “The district assemblies are the ultimate planning authorities and NADMO is not party to their deliberations or plans. Thus, the agency lacks the authority to force these assemblies to implement their DRR plans” (DeCuir et al Reference DeCuir, Fuerst, Iannuzzi and Leaver2014: 80–81).

Given these constraints, NADMO officials have leveraged their coordination role to push preparedness efforts. As one interviewee noted, “local civil society actors, in coordination with NADMO, have organized themselves into Disaster Volunteer Groups that focus on DRR and preparedness efforts through sustainable farming and agricultural practices to combat potential natural disasters such as flooding. In addition, these groups are trained by NADMO officials on preparedness measures to reduce the effects felt when natural disasters occur” (DeCuir et al. Reference DeCuir, Fuerst, Iannuzzi and Leaver2014: 82). NADMO officials are also known to participate in training activities offered within civil society (DeCuir et al. Reference DeCuir, Fuerst, Iannuzzi and Leaver2014: 82).

In addition, civil society actors are pushing forward preparedness on their own. Smaller NGOs have become particularly active in this regard. “Rural Integrated Relief Services, for example, records natural disaster shocks, develops assessments of these shocks, educates individuals on issues concerning climate change, and offers training programs for disaster management. ABANTU for Development, another local NGO, raises awareness of gender issues in complex emergencies by bringing women into the mainstream of disaster prevention in Ghana” (DeCuir et al. Reference DeCuir, Fuerst, Iannuzzi and Leaver2014: 83). Large organizations are playing their own role, with “UNDP, United Nations Children’s Fund (UNICEF), and other organizations … pushing the preparedness agenda and helping NADMO and other government institutions to focus their efforts on these types of activities” (DeCuir et al. Reference DeCuir, Fuerst, Iannuzzi and Leaver2014: 81).

Ghana thus offers a model of state–society coordination in a country where national government policy has not effectively supported a fully state-led model. Decentralization of responsibilities to the district level makes it more difficult for a national-level body to coordinate state activities. Instead, NADMO relies on its ability to coordinate civil society actors to push forward the disaster preparedness agenda, while also allowing for outside efforts by other civil society actors.

4.2.4 Society-Reliant Capacity

Where the state has lower capacity, but civil society is relatively strong, I expect to see programs that are heavily reliant on the participation of these non-state actors. The countries falling into this category are the Gambia, Kenya, Mozambique, Togo, and Zimbabwe. In these contexts, a lack of internal state bureaucratic capacity leads the government to rely on non-state actors for the provision of even very basic public services (Brass Reference Brass2012). This model may take multiple forms, depending on historical dynamics affecting the character of government institutions and the nature of civil society’s strengths and weaknesses. Nonetheless, the anticipated result is a distribution of preparedness efforts that lies primarily within civil society organizations, not within the state.

Mozambique offers a particularly nice example for how such a capacity dynamic can lead to relatively successful preparedness outcomes in the context of a motivated state. Here, a state disaster management agency – the INGC – governs and directs all activities, of both government and civil society actors, across each country. This model may often involve collaboration in the planning of disaster preparedness programs, but the final say on what programs are implemented, how, and by whom, remains with the state, not with civil society actors.

The INGC plays the role of convener, having set up the CTGC (the disaster management technical council), which serves to guide all disaster preparedness and management activities. As a part of this, NGOs are incorporated into disaster management activities in multiple ways.

As a result, the general sense among NGO representatives interviewed for this project was that disaster preparedness and response were both much better than they had been before the development of the INGC. As this NGO representative noted,

When there are specific preparedness programs that require activities on the ground in the periods between cyclone seasons, the government also leads the efforts to select and coordinate civil society actors to achieve these goals. This strategy serves to alleviate the capacity and human resources constraints of the government itself, by substituting NGO representatives for government employees in the field. “When an NGO starts working in Mozambique, they go to the government and tell them that they want to do work here and then the government helps allocate the NGO to areas of need. This is done through the Ministry of Foreign Affairs – they check with other ministries and look at the needs of the country and the skills of the NGO and then tell them where they can be of most use,” noted the director of an international NGO operating in Mozambique (Bussell and Malcomb Reference Bussell and Malcomb2014: 162). What this looks like in practice, for example, is evident in the implementation of village disaster preparedness committees, which are used to train and mobilize leaders in rural communities:

Thus, this society-reliant model for disaster preparedness allows the Mozambique government to take advantage of civil society where it exists, particularly in areas where the state’s own capacities may be lacking. While bureaucrats within the INGC have reasonable freedom from political involvement in their daily activities,

4.2.5 Uncoordinated Capacity

In the final set of cases, Bangladesh, Ethiopia, and Pakistan, neither the government nor civil society has high levels of latent capacity upon which to draw for developing a disaster preparedness program. This is not to say that there will be no preparedness programs present, but where they do exist, I expect them again to reflect these similar (low) levels of capacity. I examine in greater detail an example of clearly uncoordinated activities in Bangladesh.

The dynamics of absence have clear implications in Bangladesh. Here, there is a centralized Ministry of Disaster Management and Relief that “also organizes the National Disaster Management Council that has representatives from the various institutions at national and local levels to tackle natural disasters in the country” (Shabhanaz and Bussell Reference Shabhanaz and Bussell2017: 16). In addition, “The National Disaster Management Regulatory Framework (NDMRF) provides a set of guidelines for disaster preparedness, risk reduction, and response.” The framework was developed with the intent to mainstream risk reduction efforts within government, NGO, and private sector activities (Shabhanaz and Bussell Reference Shabhanaz and Bussell2017: 17).

However, this strategy does not effectively “account for government bodies outside of the traditional hierarchy of institutions. For example, almost all cyclone shelters in Bangladesh are primary schools built on raised ground, [which] effectively makes the shelters as much a budgetary concern for Ministry of Education as they are for Ministry of Disaster Management and Relief” (Shabhanaz and Bussell Reference Shabhanaz and Bussell2017: 19). This has led to significant problems in disaster response, let alone preparedness, such as “when 48,000 families were displaced after Cyclone Aila, and spent over 14 months on the streets. This was because of lag time between constructing new houses, roads, and schools as each project falls under a different ministry in the government” (Shabhanaz and Bussell Reference Shabhanaz and Bussell2017: 20).

This lack of coordination extends to the government’s relationship with civil society actors.

Yet, as has been the case more generally since the aftermath of the 1974 drought (Hossain Reference Hossain2017), the “government relies on local NGOs, civil society organizations, and international agencies in building capacity of communities and providing emergency relief” (Shabhanaz & Bussell Reference Shabhanaz and Bussell2017: 25). This may take the form of relatively small NGOs – such as Muslim Aid – who “build capacity at the lowest level – in villages and urban wards – on cyclone and flood preparedness using locally sourced and trained volunteers” (Shabhanaz & Bussell Reference Shabhanaz and Bussell2017: 25). In contrast, substantially larger NGOs – such as BRAC – “provide disaster preparedness training through their numerous education programs and village-level meetings” (Shabhanaz & Bussell Reference Shabhanaz and Bussell2017: 25). In some cases, the government – via the Department of Disaster Management – has been able to partner with these larger organizations. This has resulted in “the development and implementation of several training modules to educate and prepare local communities against floods, cyclones and major natural disasters” (Shabhanaz & Bussell Reference Shabhanaz and Bussell2017: 25) Nonetheless, these efforts can only go so far given the limited resources available to state disaster staff at the local level.

These dynamics suggest that a lack of capacity in the government led Bangladesh to mobilize and organize its disaster-related activities insufficiently. Civil society actors themselves have mobilized related to natural disasters, but often after the fact, as a part of response efforts and in efforts to pressure the government to improve its disaster preparedness efforts (Shabhanaz and Bussell Reference Shabhanaz and Bussell2017: 32).

5.1.1 Disaster Preparedness Performance in India

As in previous sections, I discuss India first, and it is the case that offers perhaps the best overall example of how political incentives combined with state capacity can produce strong outcomes. At the same time, I also draw attention to how the subnational cases reflect differing expectations and outcomes according to my argument.

Within India’s states, there are clear variations in the relationship between electoral incentives, capacity, and preparedness outcomes. The strongest performer on preparedness overall is Gujarat, and the character of its performance is directly in line with the expectations of my argument. While Gujarat is also a relatively wealthy state, it did not invest substantially in preparedness policies until electoral conditions combined with repeated serious disasters to generate a political incentive to do so. Once this came to be, the state was able to move quickly and substantially toward a robust preparedness model due to significant internal bureaucratic capacity. In the context of relatively low civil society capacity, the state itself took charge of preparedness efforts and has been recognized for the substantial reductions in human and capital losses these improvements have produced during recent hazards (Rawat Reference Rawat2019).

Odisha is a lower-capacity state, but in a context where electoral incentives and leveraging of civil society could potentially result in considerable efforts to prepare. This is precisely the observed outcome, as a devastating cyclone combined with a highly competitive electoral environment to motivate the state government toward a radically different model for disaster preparedness. In implementing this model, however, the state has faced limitations due to somewhat lower levels of state capacity than in the cases of Andhra Pradesh and Gujarat.

In Andhra Pradesh, the state seems poised to engage in substantial efforts to prepare for natural disasters, particularly given its high levels of state capacity, but has underperformed relative to states such as Gujarat and Odisha. While the state introduced a disaster management authority in line with national policy and has made progress in areas such as risk assessment and monitoring, it has not made preparedness a political priority in the same way as some other states. This is in line with the expectations of my argument for a state that has lower electoral incentives to prepare. It has engaged in the minimum activities that increase basic preparedness, but not in ways that dramatically reduce the risk of disasters in the state.

Karnataka displays the lowest preparedness of the Indian cases considered in detail here. While it has implemented programs in a few areas, such as hazard monitoring, the overall preparedness model is uneven and lacks considerable government attention. This is consistent with my expectations in a state that faces minimal electoral incentives to invest in preparedness and is not equipped with substantial state or civil society capacity to mobilize preparedness efforts otherwise.

In India as a whole, there is a substantial central government effort to invest in institutions of preparedness, intended to support state efforts. Yet, despite high levels of wealth relative to the other country cases, these efforts did not emerge in a comprehensive manner until a substantial disaster, in the form of the 2004 tsunami, imposed significant damages across multiple electorally important states. It was only then that the central government began to leverage its substantial bureaucratic capacity to develop resources that could not only assist in hazard response, but also supply the policy guidelines and training needed for states to improve their performance. This resulted in an overall strong model for disaster preparedness that, while not without areas for improvement, provides a substantial foundation for minimizing disaster risk, even in the absence of substantial participation from civil society actors.

5.1.2 Anticipated Strong Performance

I continue the evaluation with the other countries exhibiting both high levels of electoral incentives and high capacity, where I would expect to see the strongest performance on disaster preparedness. These countries are Malawi and Zambia, in addition to India.

In Malawi, outcomes are in line with my expectations. The country does well on the disaster preparedness measures overall but is still somewhat limited in its performance relative to places like India. While it was difficult to assess political incentives in this case, the high capacity of state actors, combined with a small number of highly active donor agencies and NGOs, has allowed the state to push forward on its preparedness efforts in a substantial way.

For Zambia, we see medium level performance on the components of disaster preparedness. The state clearly coordinates disaster preparedness activities by both the bureaucracy and civil society actors and these efforts are directly motivated by past exposure to natural hazards. At the same time, there is little evidence that politicians themselves take seriously disaster preparedness as a core element of their political agenda. In this context, I would expect performance on preparedness and related efforts to continue to improve alongside increased electoral competition.

5.1.3 Anticipated Substantial Effort

In a second set of cases, Bangladesh, Ethiopia, Kenya, Mozambique, and Zimbabwe, my argument anticipates medium levels of disaster preparedness, which reflect political incentives to prepare crossed with limitations from lower levels of capacity, often within the state. Thus, there should be efforts to invest in preparedness, even if the goals of these efforts are difficult to meet in the context of lower capacity.

Bangladesh exhibits strong political incentives to invest in preparedness and these dynamics are clearly evident in overall preparedness programming. Interestingly, the degree to which Bangladesh outperforms my expectations seems to be due to the role of civil society actors, despite limitations on the state’s ability to coordinate these activities effectively. Here, it is worth noting that my overall strategy for measuring non-state capacity may not fully account for the predominance of natural hazard-related civil society activity, relative to other non-state actors. Thus, there is a more substantial level of preparedness programming than anticipated. The risk over time, however, is that if the government is unable to coordinate ongoing activities effectively, there may be considerable duplication of effort and confusion in response at the time of hazard events.

In Ethiopia, preparedness efforts reflect political incentives combined with an empowered, if lower capacity, bureaucracy. The long history of past exposure in this case means that politicians are well-acquainted with the risks of natural hazards. These state actors have enabled a disaster management body to design and implement its programs and given it the power to coordinate efforts with civil society actors in a highly structured manner. Thus, while there are still gaps in Ethiopia’s overall preparedness profile, the country has displayed considerable effort to overcome limitations to capacity both inside and outside the state, enabling it to achieve relatively strong preparedness outcomes.

Mozambique also demonstrates the relatively long-term effects of high impact disasters. High intensity cyclones in Mozambique have increased the political salience of natural hazards for Mozambique’s politicians, motivating investments in preparedness even multiple years after these events. As in Ethiopia, Mozambique has a highly empowered disaster management body with the purview to coordinate activities with civil society actors. The effect of substantial natural hazards has been to incentivize politicians to support development of a strong disaster management agency, again serving to demonstrate an increase in disaster-specific capacity within the state. This body is then able to mobilize civil society actors in ways that increase overall capacity. There are still issues in the Mozambiquan model, but the outcomes observed here are largely consistent with an argument that emphasizes the importance of political incentives and capacity in generating substantial efforts to increase preparedness.

Kenya, in contrast, somewhat under performs my expectations, due in large part to the lack of significant direct government efforts toward disaster preparedness. Voter interest in issues other than past exposure has affected the perceptions of politicians that disaster preparedness can be used to build political advantage in Kenya’s competitive electoral environment. In addition, the low levels of state capacity in Kenya mean that there is little opportunity for internal program development without substantial political interest. As a result, the efforts that do exist are left largely to civil society, which is limited in its capacity to pursue the range of initiatives laid out in standards for disaster preparedness.

The final case in this category is Zimbabwe. Here, there are some preparedness initiatives in place, but the state is limited in its capacity and places its own constraints on the activities of civil society, in ways that curb more substantial progress. This means that even while there are long-term incentives to invest due to past exposure, capacity is so limited as to place barriers on any major political efforts. While I would expect to see more effort in this case, these outcomes are generally consistent with the argument.

5.1.4 Anticipated “Window Dressing”

In two countries, Ghana and Senegal, the alignment of political incentives and capacity is such that I expect to see some evidence of disaster preparedness efforts, but in ways that reflect limited political incentives to prepare. In other words, these are states that have the capacity to prepare, but lack the political incentive to do so.

For Ghana, the overall outcomes are in line with the predictions of my argument. A disaster management agency was put in place, but it was situated institutionally in a manner that gives it limited authority to act relative to other government institutions. State bureaucrats demonstrate their capacity to coordinate with civil society actors to push forward preparedness efforts but must act within the constraints of the national disaster management agency’s placement within the government. And, whereas politicians are acting in a competitive electoral environment, the general lack of exposure within the population limits citizen support for preparedness efforts. Thus, there is a medium level of performance that is in line with more limited electoral incentives in a higher-capacity state.

In the Senegalese case, there is substantial activity in disaster preparedness by state actors, but these actions are spread across a variety of politicized agencies and not well coordinated. This reflects a lack of overall electoral incentives to coordinate government actors within a preparedness program. There are also many preparedness efforts among civil society actors, but again, without coordination by a central agency that could ensure combined benefits of this work and minimize risks of duplicative effort. As a result, while preparedness activities are clearly visible, their character reflects my expectations in that they lack the political oversight and organization that could lead to substantial preparedness.

5.1.5 Anticipated Low Performance

The remaining country cases – the Gambia, Pakistan, and Togo – exhibit levels of electoral incentives and capacity suggesting relatively low outcomes on preparedness.

The Gambia’s overall performance on preparedness is in line with my theoretical expectations. Low levels of past exposure in the population and constraints on the opposition limit political incentives to prepare. Even so, the government has implemented a viable disaster management body with the potential to push forward preparedness initiatives. Where these efforts fail is in the government’s lack of coordination, and even competition, with civil society actors in the preparedness space. While civil society actors have relatively higher capacity in Gambia than the state, the government’s failure to coordinate has resulted in substantially limited progress on preparedness.

Pakistan’s disaster preparedness performance is somewhat better than I would expect, but the character of these efforts is in line with the expectations of my argument. Political actors face electoral concerns but are not highly motivated by past exposure. In addition, complicated institutional dynamics related to authority over disaster preparedness programming mean that national, more than local, politicians can claim credit for preparedness outcomes. As a result, preparedness initiatives are primarily the purview of local bureaucrats who are highly resource constrained. Thus, much of what is observed in practice are high-level program initiatives with variable implementation at the local level.

For Togo, disaster preparedness is minimal, at best. In this case, it seems that even where there was some knowledge of needs for general disaster preparedness, the lack of substantial past exposure to direct these efforts and lack of funds to support general development, let alone preparedness efforts, stymied whatever motivations existed for developing a preparedness agenda. This is a clear example where the combination of limited electoral incentives and limited capacity results in highly limited preparedness outcomes.

Overall, with some interesting nuances, this discussion suggests that the argument laid out in this volume performs remarkably well in predicting three core dynamics of disaster preparedness: the character of elite motivation for preparedness, the governance of preparedness initiatives, and overall levels of success in achieving preparedness, as measured by international standards. At a minimum, this argument provides a substantially stronger explanation for these outcomes than those laid out in previous research.

5.2.1 Generalizability

The case studies evaluated here constitute a significant portion of countries in Africa and South Asia. As noted in Section 1, I expect the argument to hold in many other countries in these regions and, quite plausibly, other areas of the Global South. In addition, the dynamics of political incentives and capacity are likely to be just as relevant in the Global North, where much research on natural hazards has focused. In these countries, generally higher levels of capacity should mean that the dominant factor is political incentives, as highlighted in existing work. What must be considered to understand preparedness, then, is whether past exposure and electoral competition generate conditions in which previously exposed voters play a dominant role in political decision-making. This is likely to become even more relevant as shifts in climate conditions result in changing patterns of hazards around the world.

5.2.2 Disaster Preparedness and Climate Change

The hydro-meteorological character of the primary hazards considered in this study is particularly important when evaluating how to apply these lessons to our expectations about climate change-related hazards. I earlier provided evidence that natural hazards are increasing in their frequency globally and this is occurring alongside, and is plausibly related to, changes in global air and water temperatures. This suggests not only that there will be more hazards, but that they are likely to occur with differing intensities and in different places than in the past. What does this mean for how governments are likely to respond to such hazards in the future?

One possible descriptive outcome would be the merging of government efforts to prepare for, and reduce the risks of, disasters with climate change-related initiatives. Over the course of this study, it became clear that governments were beginning to consider links between disaster-related programming and climate change adaptation, for both substantive and financial reasons. While in some cases this led to difficulties with coordination across departments, the general sense was that a unified approach could maximize overall resources, particularly around risk reduction.

Yet, which countries should be successful in these types of coordinated efforts? I anticipate that whether countries can move forward on a climate change-informed natural hazards policy agenda depends on quite similar factors to those considered here. Political elites will evaluate the relative implications of climate adaptations and the strength of their political position. The factors that affect these dynamics may differ from those central to disaster preparedness, however, given significant and often more visible investments associated with strategies such as decarbonization (Guy ND).

In practice, this could also mean that countries with the capacity to prepare for hazards – states with stronger internal bureaucracies – may lag in their responsiveness due to a lack of prior exposure. For example, countries with moderate average temperatures, for which small changes do not have as drastic of consequences as in higher average temperature countries, may fail to address the risks of these temperature changes despite having the capacity to do so. These political actors are only likely to engage in significant efforts when the threats of these changes become obvious to the voting population through direct exposure to a hazard. As a result, it may require experience with significant new climate-induced hazard events to motivate political actors to prepare appropriately.

In contrast, governments in areas with a history of past exposure, where hazards may change but are unlikely to disappear, could be better poised to prepare for such events than might be expected based on economic resources. External actors with an interest in supporting climate change adaptation could benefit from approaching these policy actors from the perspective of reducing risks with which governments are already familiar, rather than developing entirely new frameworks for fundamentally related policy areas. This would enable motivated actors to act, by providing additional resources and external capacity to effectively compliment ongoing internal efforts.

Overall, these dynamics suggest that substantial climate change adaptation, and the implications of these policies for natural disaster risks, is not necessarily going to appear in the richest countries first. Countries facing repeated, and potentially worsening, threats from climate-induced hazards may instead be at the forefront of techniques and strategies to reduce these risks. Yet, these countries will not all have the capacity to implement substantial programs and they may also face increased risks due to failures of richer nations to invest in climate change-minimizing reforms. Our global interest may then be best served by efforts to support risk-reducing climate change adaptation in the Global South. Insights and innovations from these efforts can then be leveraged to inform policies in the Global North as political incentives inevitably evolve in the wake of changing hazard profiles.