An analysis of the 2010 Household Income & Expenditure Survey (HIES) shows that a large proportion of poor households live near areas prone to cyclone surges (Taupo, Cuffe, & Noy, 2016). 8 The survey was conducted using trained interviewers, trained and supervised by one of the authors. As a proxy for cyclone strength, we focus on measuring the distance of each household from the cyclone path.
8 We first asked about household details and characteristics before elaborating on the impact of TC Pam.
Loss and Damage
Based on our totals from the survey responses, household loss and damage in Tuvalu is estimated at AUD, which is almost 4.4% of 2015 GDP.15 However, total loss and damage at the national level is estimated at approximately 10% of GDP GDP.16 From the survey we find that agriculture causes 5.3% of loss and damage, while 14%. Poor households bore the brunt of the cyclone and caused about half of the total loss and damage. A similar storm coming from the east would be significantly more damaging, as about half the country's population and most of the infrastructure is on Funafuti.
Vaitupu Island was affected, but not to the extent of the other five islands, for which damage and loss information is given below in Figures 5 and 6. 13 If two pigs died as a result of the cyclone, we used a value of AUD 200 if both weigh 20 kg at a local price of AUD 10 per kg. For consistency across households, the information obtained along with local fruit market prices was used to calculate the value of lost fruit trees.
The combination of low elevations and the narrowness of the island contribute to the high level of impact experienced there (see maps in Appendix A). As determined by our calculations of the poverty threshold, and on average, the median loss and damage per household is very similar for poor and non-poor households. However, given the different levels of income, both the median and distribution of the relative loss and damage (to income) are much higher for poor households (Figure 7).
Hazard
Exposure
Most affected households, and even most households on atoll islands, live in areas prone to storm surges and flash floods. Given the small size of all the islands, the entire population of Tuvalu lives no further than a kilometer from the sea. On almost all the islands of Tuvalu, the populated areas are on the western side – the direction from which the storm surge came.17 In general, this meant that on most islands (except, more importantly, the capital Funafuti), the population was highly exposed.
The percentage of the population living on the main island, Funafuti, also increased from 32.6% in 1991 to 57% in 2012.18 We produced GIS maps linked to the household surveys and measured exposure by household distance to the using coastline and household elevation (lower elevation). areas are more prone to flooding associated with storm surges). Households living in houses that are less than 5 meters above the low water line are described in panel (a), while households whose residence is less than 100 meters from the coast are described in panel (b). We further distinguished between the poor and non-poor households using the definition of poverty described earlier.
It is clear to note that the inhabitants of Nui and Nukulaelae are the most exposed, with some significant additional exposures at Nanumea. Similarly, Nanumaga, Nanumea, Niutao and Nukulaelae reported and 66%, respectively - again, this clearly corresponds to the exposure data depicted in Figure 8 which identified Nanumea, Nui and Nukulaelae as the most exposed. On islands without lagoons, populations tend to be concentrated on the western (downwind) side of the island, while on islands with lagoons, populations tend to reside on the lagoon side.
Vulnerability
Responsiveness
16 The majority of people received cyclone warnings from the radio broadcast and island community alarms, but 39% did not receive any cyclone warnings. Fifteen percent of households were not aware of any safe shelters available to them during the cyclone. The average time a household had to travel to reach a safe shelter is 13 minutes, but 57% of households never transitioned to these shelters.
After the storm, virtually all families received assistance from the government, NGOs, family, friends, development partners, and remittances. 13% of families received cash assistance, 67% received assistance in kind, and 23% received other types of assistance (Figure 10).
Poverty
In addition to having a lower income, poor households have more household members and are further away from cyclone shelters. They are more responsive to cyclones in terms of preparation, fortifying their houses and moving valuable assets to safety. More received warnings, fewer cyclone response workshops, less money saved and more evacuated to cyclone shelters.
There are more poor households receiving help from family and friends, but they have less access to credit and generally receive less help.
5 Estimation Results
Interestingly, the "decay" of the damage rate with increasing distance from the cyclone is not very large. For every additional kilometer of distance, the reduction in loss and damage from Cyclone Pam is 0.5%. A cyclone passing 1000 km away from the site, as was the case for Pam in Tuvalu, is not usually modeled as causing damage.21.
In terms of the ability to respond to an oncoming cyclone, we find that households that received early warning (at least 12 hours in advance) did manage to reduce their damage and loss.22 Interestingly, the effectiveness of early warning higher for poor. households in reducing loss and damage; although the differences between the two subsamples are not statistically significant. Cyclone response training and pre-strengthening of homes appear to reduce both loss and damage, but these impacts are largely driven by their effectiveness for wealthier households. One can speculate why this might be the case—perhaps it is related to the increased ability of non-poor households to act affectively on the information they receive—but we have no direct evidence to show that it does may be the reason for these observed differences.
Other variables for which data were collected were not statistically associated with the loss and damage in any materially consistent manner. We suspect that this is the case for other risk models, such as RMS. 22 Although we cannot rule out the possibility that the availability of early warning is somehow endogenously determined.
6 Hypothetical Scenarios of DRR Policies
Had the cyclone also hit Funafuti (which would have been the case if the wave surges had come from the East rather than the West), the total loss and damage would have been an estimated AUD 6 million (14% of GDP). Similarly, Scenarios 6-14 show adjusted exposure using the distance to the coast and altitude with different cases, assigning different observed values of minimum, mean and maximum across all households. 2 Adjusted vulnerability, average income pc (poor households only) 69.4 3 Adjusted vulnerability maximum income pc (poor households only) 19.9.
For poor households, we note that if they all had the lowest per capita income, cyclone-related damage would have been almost 40% higher (scenario 1). Conversely, if everyone had the highest income in the full sample, damages would be 80% lower (scenario 3). Likewise, if everyone lived at the minimum height, the damage would be almost twice as high (scenario 9).
In terms of the possibilities of achieving better preparation for and response to the cyclone, the data suggests some potentially useful interventions. Specifically, strengthening all homes before the cyclone would have reduced damage by 8%, while attending cyclone response training for everyone (a representative from each household) would have reduced damage by 17% - scenarios 19 and 19. In this case, the impact would have been reduced by 77% if everyone had received effective early warning (at least 12 hours before the cyclone wave).
7 Conclusions and Policy
It is important to remember that the most populous island, and the island that contains much of the infrastructure, including the international deep-sea port, airport and most of the government facilities, has not been damaged at all. If Funafuti had been as exposed to this cyclone as the outer islands, the damage and losses associated with the cyclone would have been about three times higher, even though Funafuti is home to only about half of the country's population. Under the circumstances, it is not surprising that the Tuvalu government is considering providing relief to affected households by paying a portion of the reconstruction costs associated with the impact of TC Pam.
Questions about the desirability of state assistance after Pam are therefore also related to other questions about the government's fiscal position and the sustainability of its spending. In this context, it seems advisable for policy makers to consider alternative options to provide ex-post assistance to affected households and at the same time protect the government's fiscal stance. The threats of climate change, sea level rise and climate disasters may require relocation to safer places for households living very close to the coast, in narrow parts of the island and low-lying areas.
More extreme climate change scenarios with greater sea-level rise may require large-scale relocation of the country of Tuvalu, although at this point there are few plans for such dire scenarios (Noy, 2016b). As part of our research after Pam, we also asked the inhabitants of the outer islands about their views on migration as an adaptation option. Of the factors influencing the choice of relocation destination, the availability of jobs again appears to be the most important (although, surprisingly, the availability of health facilities also appears to be important).
8 Bibliography
Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, 1-32. A cost-effective solution to reduce disaster losses in developing countries: hydro-meteorological services, early warning and evacuation.
Appendix A: Loss and Damage & Scenario Maps
