A BACKGROUND PAPER >> MACRO IMPACTS OF SHOCKS This work is a product of the staff of The World Bank and the Global Facility for Disaster Reduction and Recovery (GFDRR) with external contributions. The sole responsibility of this publications lies with the authors. The findings, analysis and conclusions expressed in this document do not necessarily reflect the views of any individual partner organization of The World Bank (including the European Union), its Board of Directors, or the governments they represent, and therefore they are not responsible for any use that may be made of the information contained therein. Although the World Bank and GFDRR make reasonable efforts to ensure all the information presented in this document is correct, its accuracy and integrity cannot be guaranteed. Use of any data or information from this document is at the user’s own risk and under no circumstances shall the World Bank, GFDRR or any of its partners be liable for any loss, damage, liability or expense incurred or suffered which is claimed to result from reliance on the data contained in this document. The boundaries, colors, denomination, and other information shown in any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. The Macroeconomic Impacts of Natural Disasters in the Caribbean Background Paper Eric Strobl (University of Bern) January 27, 2020 1 Abstract This paper provides a review and assessment of the current literature on the macro- economic impact of natural disasters in the Caribbean, including other non-Caribbean studies that may have implications for it. It also discusses, in view of existing studies, what factors may make Caribbean economies more resilient to these extreme events, as well as whether there are damage thresholds beyond which recovery will be more difficult. Finally, recommendations are provided for future data collection and research that might provide further light on the issues. 2 Contents 1 Introduction 4 2 Impacts 5 2.1 GDP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.1.1 Theoretical Predictions . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.1.2 Direct vs. Indirect Losses . . . . . . . . . . . . . . . . . . . . . . . . 8 2.1.3 Existing Evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2.1.4 Possible Reasons for Lack of Long-Term Impact . . . . . . . . . . . . 13 2.2 Foreign Reserves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 2.3 Debt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 2.4 Other Macroeconomic Indicators . . . . . . . . . . . . . . . . . . . . . . . . 19 3 Natural Disasters and Economic Resilience 19 4 Natural Disasters, Thresholds, and Economic Recovery 21 5 Discussion and Recommendations for Future Research and Data Collec- tion 25 6 Concluding Remarks 29 References 31 3 1 Introduction The Caribbean is one of the most disaster prone regions of the world. Moreover, there is also increasing concern that with climate change at least the number of climate related natural disaster may rise. For instance, in terms of tropical storms Knutson et al. (2019) note that current evidence from state of the art models seems to suggest that in the North Atlantic Ocean Basin, where the Caribbean is located, the average intensity of storms and the proportion of higher intensity (Category 4 and 5) storms will increase, although it is not clear whether there will also be a greater frequency of tropical storms in general or the more intense ones. Additionally, extreme precipitation events are expected to rise, both due to tropical storm activity (Knutson et al. (2019)) and non-storm related rainfall (McLean, Stephenson, Taylor, and Campbell (2015)), leading to increased flooding and leading slides. While it difficult to make similar predictions with regard to non-climate related events, such as earthquakes and volcano outbreaks, the region is certainly susceptible to these. Us- ing available data since 2000 (EMDAT1 ) Figures 1, 2, and 3 depict the distribution of the number of events, persons affected, and damages by biological (epidemic, insect infestation), climatological (drought, wildfire), geophysical (earthquake, mass movement, volcanic activ- ity), hydrological (flood, landslide, wave action), and meteorological (extreme temperature, storm) types. Accordingly, while geophysical events constitute only 3% of all events, they constitute nearly 10% of persons affected and close to 18% of damages. They thus also play, and are likely to continue to do so, an important role in the region. Consisting mostly of small open, highly indebted economies reliant on a few sectors, these natural events are likely to have large impacts in the Caribbean. Moreover, given the region’s high indebtedness it may also a place where recovery is most challenging after a natural dis- aster strikes. Thus understanding the actual macroeconomic impacts of natural disasters in 1 See https://www.emdat.be/ 4 the Caribbean is arguably crucial for policymakers. Nevertheless, while there is a sizeable literature examining various aspects of the macro-economic impact of natural disasters in the region, there appears to be no explicit conclusions or summary in this regard that could act as a guidance for policy makers or researchers. The purpose of this background paper is to synthesize the literature, identify gaps that need to be addressed and provide some direction in terms of how this might be possible addressed with existing database and methodologies or by collecting necessary data. The rest of the paper is organized as follows. In the next section we will discuss and summarize the existing literature that explicitly examines the macro-economic impact of natural disasters on the Caribbean, as well as results from other non-Caribbean studies that might be relevant. This will include general impacts on GDP, as well as other other macro-economic indicators particularly important for the region. Section 3 will discuss what factors are likely to make Caribbean economies more resilient, as suggested by existing ev- idence. The subsequent section (Section 4) will provide a discussion and a review of the small literature regarding the existence of possible damage threshold beyond which recovery may be more difficult, as well as provide some own estimates in this regard. In lieu of the gaps and weaknesses of the current literature Section 5 will provide a discussion and recom- mendations for future research and data collection in order to increase our understanding of the macro-economic impact of natural disasters in the Caribbean. Finally, Section provides some concluding remarks. 2 Impacts . 5 Figure 1: Number of Events Figure 2: Total Persons Affected 2.1 GDP 2.1.1 Theoretical Predictions Hypothetically natural disasters could have either a negative or positive impact on affected 6 and Botzen, Deschenes, and Sanders (2019). economies; see Hallegatte and Przyluski (2010) Figure 3: Total Damages With regard to the former, the underlying intuition is straightforward, destruction of human and/or physical capital, if large enough, will lead to pushing economies away from their growth path. The uncertainty regarding such a negative impact is whether this will just manifest itself over the short-term, so that economies recover quickly back to their equilib- rium growth trajectory, or if recovery will be only gradual. In contrast, a positive effect is feasible if the loss of capital results in the replacement of more productive and modern technologies that will push the economy to a newer, higher growth path. From a theoreti- cal modeling perspective, the predictions in this regard rest in part crucially on the choice of economic growth model. Models based on neo-classical growth theory general are only able to predict a negative impact since they they assume rather then try to model technical change. Whether then this effect is short or long term will depend on if the disaster event permanently changes savings, depreciation, or productivity growth. While endogenizing productivity growth may still lead to predictions of negative effects, as in the AK models, in others, such as the vintage capital models, an accelerated deprecia- 7 tion of capital may result in higher technology growth if technology is updated.2 Similarly, in learning models the destruction of capital or labor could simulate learning and productivity growth during construction. Since clearly the predicted macroeconomic impact of natural disasters depends on the choice of theoretical model describing economic growth, the impact of natural disasters is largely an empirical question. 2.1.2 Direct vs. Indirect Losses Important in understanding and measuring the economic impact of natural disasters is that losses due to these events can be both direct and indirect (Hallegatte and Przyluski (2010)). Direct losses refer to the destruction in physical (capital stock, buildings, infrastructure, etc.) and human (deaths and injuries) capital that are directly attributable to the physical forces of the event and these tend to be almost immediate. These direct losses can then lead to interruptions, i.e., indirect losses, to the normal functions of the economy. CACDN (1999) note that such indirect effects can take a number of forms. Firstly, there may be losses in sales, wages, and/or profits due to interruptions caused by direct physical damage to commercial structures or from infrastructure failure for damaged firms. There may also be reductions in output by not directly damaged firms that provide inputs to the damaged and/or affected firms via lower demand. Similarly, the loss in output by directly affected firms could lead to reducing output in other firms that depend on this output as inputs into their production process. Finally, firm closures and reduced production and sales could result in lower wages or loss of employment, in turn leading to loss of consumer expenditure and hence lower demand for firms’ output. Compared to direct losses, indirect effects are likely to extend to the medium- and even long-term. But indirect effects may also lead to possible gains to the economy. For instance, changes in future production, employment, and income and/or changes in these flows outside the damaged area may compensate for initial 2 This positive effect is known as the ’build-back-better’ hypothesis in the literature 8 disaster-induced losses. Related to this, there may be boosts to production, and hence prof- its, outside the impact area to owners of commodities inflated in price by disaster-induced shortages, as for example in the construction sector. Also there may be positive economic stimuli of jobs and production generated from cleaning up and rebuilding activities for some parts of the economy inside and outside the affected area. Finally, it is crucial to also highlight the relationship between direct and indirect losses and GDP. In this regard one needs to remind oneself that GDP is the total monetary or market value of all the finished goods and services produced within a country’s (or region’s) borders in a specific time period, and as such is a flow rather than a stock measure. Thus destruction in capital, physical or human, will not itself be captured in measures of GDP. Rather it is only through their indirect effects, such as reductions in output due to business interruptions and/or gains in output due to restructuring, relief, or investment to replace destroyed capital, that GDP can be affected. If each dollar in direct damages does not translate into a dollar in indirect losses (gains) within the period that GDP is measured, then direct damages are unlikley to be equivalent in value or even conceptually. 2.1.3 Existing Evidence With regard to the empirical perspective on the economic effect of natural disasters, since the first known study in 1993 by Albala-Bertrand (1993), there have been a large number of researchers have set out to quantify the macroeconomic impact of natural disasters. As a result there are now several studies that review and try to draw conclusions from the liter- ature, in particular Cavallo, Noy, et al. (2011), Klomp and Valckx (2014), AG van Bergeijk and Lazzaroni (2015), Noy, duPont IV, et al. (2018), and Botzen et al. (2019). While all of these reviews emphasize the large differences in data and methods across existing studies, they all also seem to reach the consensus that the literature points to a negative rather than a positive impact of natural disasters on economic wealth and/or growth. Importantly, most 9 of the studies also seem to suggest that this negative effect is relatively short-lived, lasting at most a few years, although this could possibly be in part due to some methodological and data shortcomings (Noy et al. (2018)).3 Most of the studies on the macroeconomic impact of the natural disasters have used data that pools a large number of countries, either developing or developed, or both, across time. Given that most Caribbean economies are small open economies, typified by relatively non diversified economic structures, it is questionable whether one can directly extend most of the findings from these directly to the Caribbean. However, there have also been a handful of studies specifically focusing on the Caribbean, or at least for which the majority of the sample consisted of the Caribbean. One should note in this respect that most of these have focused on the impact of hurricanes. This may not be surprising given that such tropi- cal storms are relatively frequent in the region, as compared to, for example, earthquakes or volcanoes), and also are easier to model meteorologically capture than, for instance, floods. The earliest published study to examine the Caribbean is Hsiang (2010), who, using a sample of 28 Caribbean-basin countries (which also includes several Central American na- tions) over the 1970 to 2006 period modeled hurricane destruction using a tropical storm wind field model.4 His results showed no aggregate effect on total country level production. In contrast, also using a sample of Caribbean and Central American nations, but over the 1950 to 2004 period and a locally (within country) constructed destruction proxy derived from a tropical wind field model5 , Strobl (2012) found an overall negative impact on GDP 3 One exception is the widely cited, but still unpublished, study by Hsiang and Jina (2014) who find a very long-term effect looking at tropical cyclones. However, a closer examination paper of the study reveals two aspects that should lead one to question the general findings. Firstly, the authors do not take into account exposure in their destruction index, which could lead to systematic measurement error bias if the more affected coastal populations are changing over time. Secondly, the derived long-term effect is based on a standard joint-significance on a large number of lagged cyclone destruction indicators. This for one hides the many insignificant marginal effects the authors find within these lags, but also ignores any possible correlation across lags due to the temporal link between storms during La Nina and El Nino phenomenia. 4 This approach was first employed by Yang (2008). 5 One should that, in contrast to Hsiang (2010) and Yang (2008), the study by Strobl (2012) explicitly 10 growth. This effect, however, was short-lived, only lasting up to a year after the storm. Quantitatively, his results suggested that the impact could nevertheless be large, where the average hurricane reduced the GDP growth rate by about 0.8, but the largest event pushed economies by up to 8.5 percentage points off their equilibrium growth path. The short-lived negative effect of tropical storms was echoed by Bello (2017) with a sample also including Latin American countries, as well as by Cashin and Sosa (2013) for the Eastern Caribbean. Focusing only on Caribbean countries, Moore, Elliott, and Lorde (2017) proxied total hur- ricane destruction by using a storm (but not location) specific power dissipation index and also find a negative, but a slightly larger effect than Strobl (2012). However, the authors did not explore any more long term impacts. Finally, Mejia (2014) also focuses on Caribbean economies only, and finds a negative but insignificant effect on GDP. There are also a handful of studies that have looked at the impact of other natural dis- asters on the Caribbean apart from tropical storms. In particular, using synthetic control methods, Best and Burke (2019) investigated the impact of the 2010 earthquake on macroe- conomic losses in Haiti. In this regard, the authors found that even up to 6 years after the event (the 6 year cut-off was limited by the end of their sample period) losses still amounted to an average of 12% of Haiti’s annual GDP. For floods, droughts, and extreme temperatures Bello (2017) also discovered a small short-lived negative impact for his Latin American and Caribbean sample, but no implications following geological disasters. Finally, as with hurri- canes, Mejia (2014) found no macroeconomic impact of floods in Caribbean countries. Aggregate effects may of course hide considerable heterogeneity. One suspect in this regard is at the sectoral level, where different sectors may react differently to the direct or indirect costs. For instance, given the low level of physical capital agriculture may be more likely to recover quickly than other industries; see, for instance, Blanc and Strobl (2016). took into account local population exposure in constructing the destruction index. As indicated earlier, not doing so may not only induce attenuation bias but also possibly systematic measurement error. 11 Other sectors, in contrast, like construction may actually receive a boost in demand after a natural disaster, as was found by Strobl and Walsh (2009) the US. Indeed sectoral differences in impacts seems to be characteristic for the Caribbean, at least in terms of the impact of hurricanes.6 In this regard, Hsiang (2010) found that there were negative impacts on the wholesale, retail, restaurants and hotels, the agriculture, hunting, and fishing, the transport and communication, the mining and utilities, and the tourism sectors, but a positive effect on the construction sector, and no implications for manufacturing. Importantly, for none of these industries was there an effect beyond four years of the event, with most only experi- encing significant consequences up to 1-2 years later. Looking explicitly at agriculture in the Caribbean, as measured by agricultural exports, P. Mohan (2016b) also finds a negative im- pact that lasts only a year, and a similar result was found by Strobl (2011) when measuring agricultural production in the region with satellite data. This short-lived fall in agricultural output also seems to hold when one examines individual crops, as shown for example for bananas in the Caribbean by P. Mohan (2017) and for a number of different crops in Jamaica in Spencer and Polachek (2015). A short-term negative impact on the tourism industry, like in Hsiang (2010), was also found by the study Granvorka and Strobl (2013). One can also think of aggregate GDP figures potentially masking more complicated dy- namics in terms of its components. To this end, P. S. Mohan, Ouattara, and Strobl (2018) ex- amined the impact of hurricanes on national accounting components for a set of 21 Caribbean countries over 60 years. Their results showed that the impact differed widely across these factors, where there was a fall in exports and private consumption, but an increase in im- ports, government consumption, and investment. Moreover the timing of the effects was not perfectly aligned. Another reason why studies using aggregate macroeconomic data may be masking any 6 All existing sectoral studies seem to focus again only on tropical storms. 12 longer term effects is that most natural disasters can be considered relatively localized events, where even within small economies only parts of the region will be affected, or at least there is considerable heterogeneity across spaces. Nevertheless, the existing sub-national studies indicate a similar lack of any long-term implications of natural disasters, at least in terms of hurricanes. For example, while Bertinelli and Strobl (2013) using satellite derived nightlight intensity as a proxy of localized economic activity, find that aggregate data will underesti- mate the total impact of hurricanes, their data did not show any longer term effect of the storms at a localized (1km) level. One may want to note in this regard, that both evidence in Strobl (2011) and Spencer and Polachek (2015), in examining the impact on agricultural on intra-national regional units, suggested that the same is true for the agricultural sector. Even at the household level, the impact of hurricanes in the Caribbean appears to only last a short time, as shown for Jamaican households in terms of per capita consumption by Henry, Spencer, and Strobl (2019). 2.1.4 Possible Reasons for Lack of Long-Term Impact Overall, one can thus reasonably conclude that the empirical evidence for the Caribbean appears to largely provide evidence that the impact of natural disasters is negative but does not reach beyond the short-term, at least for climatic disasters. There are a number of possible, not necessarily mutually exclusive, reasons for this, which we discuss below. The easiest answer is that Caribbean economies indeed manage to readjust fairly quickly even after a very large negative shock. In this regard, one has to remember that although natural disasters are rare events by definition, their occurrence is not new to the region. Thus there may already be buffering or adjustment mechanisms in place. One can think of a number of channels through which this may occur. An obvious candidate are remit- tances. In this regard, Yang (2008) found that after hurricanes the remittances to developing countries constituted in monetary value nearly three quarters of the total damages. Also, 13 Ebeke and Combes (2013) show that remittances have a non-linear buffering effect, where above a certain level (17% of GDP) they further destabilize countries affected by natural disasters. Evidence specifically for the Caribbean is unfortunately scarce, but for the few existing studies remittances do seem to play a non-neglible role. For instance, Clarke and Wallsten (2003), examining the case of Jamaica after Hurricane Gilbert in 1988, found that remittances covered 25% of losses, while Henry et al. (2019) show that these can reduce the fall in consumption expenditure in Jamaica after a hurricane by about 75%.7 Households in the Caribbean may also be preemptively saving in order to deal with nat- ural disasters when they occur, although this is likely limited for the poorest among them (Carter, Little, Mogues, and Negatu (2007)). In this respect, Henry et al. (2019) found evi- dence that Jamaican households that had savings used these to buffer the losses associated with hurricanes. Finally, migration may also relieve some of the pressure on economies after natural disasters in the Caribbean. For example, Spencer and Urquhart (2018) found that damaging hurricanes increased migration to the US by about 6% in Central America and the Caribbean. Nevertheless, given the scarcity of evidence, it is difficult to conclude that these informal buffering mechanisms are the primary reason as to why one does not find any long-lasting effects of natural disasters in the Caribbean at the macro-level. Methodologically speaking, another possible reason is that the damages due to natural disasters are poorly modelled, thus leading to attenuation bias - in other words under- estimating the effect of natural disasters in the econometric analysis.8 . There are two issues in this regard: (i) direct versus indirect effects and (ii) choice of damage function. In terms of (i) it is the extent of the indirect effects that by definition will determine any long-term 7 One may want to note in this regard that a significant proportion of remittances received in the Caribbean is in kind through shipment of barrels, and thus may not be capture by official monetary sources (Kirton (2005)) 8 Attenuation bias due to (non-systematic) measurement error pushes the estimate on a regressor towards zero; see Woolridge (2009). 14 effect, and arguably probably constitute the larger part of the total impact; see Hallegatte and Przyluski (2010). However, in essentially all existing studies proxies of the natural disaster damage are built on proxies of the direct losses due to the events9 , and thus im- plicitly assume that both channeles can be proxied by these on a one for one basis. This is in large part because there is a general lack of understanding of how the direct losses translate, and thus can be modelled, into indirect losses, at least quantitatively.10 The mea- surement error involved in using direct loss measures as proxies for indirect losses is likely to become more important as one considers the longer term, and additionally may also be very different across countries, depending on their economic, political, and institutional structure. There may also be a problem in how the economic impact is generally measured in the empirical literature, namely in terms of the impact on GDP growth, which implicitly iso- lates the marginal effect of natural disasters. However, Hallegatte and Vogt-Schilb (2019) argue that examining the impact on the marginal productivity may substantially underesti- mate the true output losses, and that instead one should focus on consequences for average productivity. Empirically, however, doing so would pose a likely insurmountable challenge. More specifically, this would mean looking at the impact of GDP levels rather than growth rates. However, it is well known that GDP in levels is generally non-stationary, and the standard approach is to transform the data into first differences, hence only allowing one to isolate the marginal effect.11 9 There is of course also the issue of whether the proxies used to model the direct losses are adequate enough to avoid attenuation bias even for these. In this respect, the literature has evolved considerable from using measurement ridden, and possibly endogenous, ex-post estimates of direct damages to using the physical characteristics of the event and simplified damage functions, taking local exposure into account, as proxies; see Felbermayr and Gr¨ oschl (2014). Nevertheless, there still remain questions about how accurate these damage functions are. For instance, for hurricanes most researchers use wind speed as a input in the damage function, but ignore damages due to storm surge or rainfall. Moreover, exposure is generally measured via proxies of local population or income, and thus inherently ignores ex-ante disaster mitigation measures, such as building type. 10 One approach has been to use input-output tables to simulate the indirect losses, such as in Pan (2010). However, these implicitly assume that input-output coefficients do not change as a result of a disaster. Moreover, they only capture one aspect, namely the effect through provision of inputs of goods, on economies. For a further discussion see Noy et al. (2018). 11 In some circumstances GDP may co-integrated in levels with the explanatory variable of interest, but 15 A final, more conceptual, but related to the modeling challenges, reason was put forth by Noy et al. (2018). In particular, the authors argue that economies are not generally stagnant over long periods and thus may not return to their pre-event trajectories. However, in order to identify any long term impact of natural disasters, it is necessary to predict what the effect would have been had the event not occurred. This would require the explicit modeling of the long-term dynamics, which is likely to be challenging.12 Typically, researchers simply use other non-affected islands as control groups, albeit with different econometric strategies.13 2.2 Foreign Reserves Holding foreign reserves has been shown to be an important buffering mechanism for large ere, Cheng, Chinn, and Lisack (2015). Moreover, negative shocks; see, for instance, Bussi` Moore and Glean (2016) show that the optimal foreign reserve adequacy may be double for small states, as essentially are Caribbean economies are, compared to other countries. While there are no explicit studies on how foreign reserves may help mitigate the negative impacts of the natural disaster shocks, Kaplan and Strobl (2020) examine how these respond to hurricane shocks in Caribbean islands. They find that for the higher income islands there is an increase in reserves a month after the shock, which they argue is likely due to foreign aid and remittance inflows. In contrast, the lower income islands experience a fall in reserves up to one month after the shock, followed by a small increase a month later. this is unlikely to be the case with natural disaster loss data. 12 Noy et al. (2018) suggests that this could be done via the input-output model approach, but concedes that it would not account for changes in the use inputs due to the natural disaster, as noted above. 13 One appraoch has been to use synthetic control methods where one examines the impact of a specific event on a particular country relatively to a synthetically created similar sample of ’control’ countries; see, for example, Best and Burke (2019) for the case of the 2010 earthquake in Haiti. However such studies usually require the context of a long time period for a single country that was hit by only one large event, and a set of control countries that were not affected by this or any other similar events during that time period. Given the frequency of natural disasters in the Caribbean it is difficult to find such conditions. 16 2.3 Debt Post-disaster management, including reconstruction, can put a considerable financial bur- den on the economies affected and require large amounts of government expenditures. The recurrent use of government expenditure to absorb the impact of natural disasters then is likely to lead to higher levels of debt and higher interest rates accompanied by lower credit scores, which results in higher budget deficits, and causes debt to further increase, thus creating a vicious cycle threatening debt sustainability ((Koetsier (2017) and (Borensztein, Cavallo, and Valenzuela (2009)). High debt also keeps borrowing costs high, which discour- ages private investment, and constrains fiscal flexibility. Worryingly, there is also emerging evidence that high levels of debt negatively affect economic growth (Reinhart and Rogoff (2010)). In this regard, Greenidge, Drakes, and Craigwell (2010) show for the Caribbean that as debt rises beyond 55 percent of GDP there is a downward impact on economic growth. A summary of the empirical literature suggests broadly that government debt increases after a natural disaster in developing countries; see, for instance, Lis and Nickel (2009) and Melecky and Raddatz (2011). More specifically looking at debt default, Klomp and de Haan (2015) provides evidence that investors perceive natural disasters as an adverse shock that makes government debt less sustainable which may eventually trigger a sovereign default. as a matter of fact Klomp (2017) shows that the probability of a sovereign debt default increases by about three percentage points after major earthquakes and storms. However, Noy and Nualsri (2011) find that developing countries conducting pro-cyclical fiscal policy as a result of natural disaster can effectively decreasing debt. There are also a number of papers specifically focusing on the Caribbean debt response to natural disasters. More specifically, Rasmussen (2004) in a study of the Eastern Caribbean found that the median public debt increases by 6.5 percentage points following a disaster, mainly because of an increase in spending and a small reduction in revenue. The result of 17 a rise in public debt after natural disasts is also echoed in the Eastern Caribbean study by Lugay and Ronald (2014). In contrast Acevedo (2016) investigates the effect of storms and floods on public debt for the region and finds that debt only increases with floods. Similarly, in a study of hurricanes, Ouattara and Strobl (2013) demonstrate that hurricane strikes cause no significant increase in government deficit. There are two other noteworthy aspects that arise from the literature specifically looking at the impact on debt for the Caribbean. Firstly, the effect seems relatively short-lived, although not all studies explicitly investigate medium or long-term effects. For instance, in Rasmussen (2004)’s study the impact last only up to three years. In contrast, Ouattara and Strobl (2013) only find an impact in the year of the hurricanes. As matter of fact, in a follow up study Ouattara, Strobl, Vermeiren, and Yearwood (2018) using monthly data demon- strate that any fiscal deficit only occurs in the month of the event. This short-term effect is also echoed in the recent study by P. Mohan and Strobl (2020) for the Eastern Caribbean, who find that the debt increases only up to six months after an island suffers damage from a hurricane. Secondly, there is mixed evidence as to what is driving the increase in debt. More specifically, while for Ouattara and Strobl (2013) rise is due to more fiscal spending, but the monthly data instead suggest that it is a fall in revenue that is driving it, Rasmussen (2004) shows that both play a role. Of course, part of the short-lived effect of debt may be because foreign aid allows countries to avoid borrowing as much as they otherwise would to finance post disaster management. While there is no study that explicitly investigated this, for their finding that debt decreases after natural disasters in the Caribbean Heger, Julca, and Paddison (2008) suggest that the inflow of foreign aid is a likely explanation. 18 2.4 Other Macroeconomic Indicators There are unfortunately few other macroeconomic indicator’s response to natural disasters that have been explicitly investigated for the Caribbean. One exception, Heinen, Khadan, and Strobl (2018) find there is a short-term impact of hurricanes and flooding on inflation lasting up to month after the event. Another, Strobl, Kablan, et al. (2017), examines the impact of tropical cyclone on small island developing states, many of which in the sample are Caribbean, and show that there is real exchange rate depreciation up to two months after. Nevertheless, this paucity of literature on other macroeconomic aspects thus leaves a large gap in understanding how natural disasters affect factors such the labour market and financial markets. 3 Natural Disasters and Economic Resilience An important, largely still unresolved, question that remains about natural disasters is what makes countries more resilient than others. More specifically, while a number of studies show that developing countries are more affected than developed ones - see the reviews by Klomp and Valckx (2014), Lazzaroni and van Bergeijk (2014), and Botzen et al. (2019) - this finding provides little insight as to what part about being developed makes countries more resilient to the large negative shock associated with such events. Nevertheless, there are a number of studies that have explored some of the possible channels. For instance, Felbermayr and oschl (2014) find that disasters have smaller impacts in more democratic countries and Gr¨ in countries that are more open to trade and have better-developed financial markets. The latter channel was also found to be an important factor in reducing the negative effects of natural disasters by McDermott, Barry, and Tol (2013). Lack of corruption was also found to be an important factor in reducing the impact specifically for earthquakes by Escaleras, Anbarci, and Register (2007). 19 While the aforementioned channels provide first insights into on certain characteristics of an economy that are likely to be helpful in mitigating natural disaster effects, one large gap in the literature is still what role actual disaster mitigation measures, such insurance coverage, government contingency funds, or resilient infrastructure and buildings, could play. And, without a further understanding of this it is arguably difficult to formulate actual dis- aster mitigation policies. As a matter of fact the few studies that exist suggest that specific mitigation policies may actually play a crucial role. For example, Anbarci, Escaleras, and Register (2005) demonstrate that ex-ante preparedness for eartquakes in terms of building design and seismic code implementation can play an important role in reducing the number of subsequent deaths. Also, Henry et al. (2019) show that more resistant housing can help Jamaican households reduce the impact of hurricanes. Although explicit empirical evidence in so far how much insurance coverage might act as a buffer currently is unavailable, Von Pe- ter, Von Dahlen, and Saxena (2012) do show that it is the uninsured losses that drive most of the subsequent negative macroeconomic impact in countries. This finding may be partic- ularly alarming for the Caribbean, where estimates suggest that at least two thirds of losses are not insured (IMF (2019)). Of course it is also helpful to look at other literatures that have explored what deter- mines resilience to large negative shocks as a guideline as to what could also be important for natural disasters in the Caribbean. Unfortunately, most of the studies on this topic anchez, Rasmussen, and R¨ examine developed countries; see Caldera S´ ohn (2016) for a re- view. Nevertheless a few papers provide insights into what aspects might provide resilience to large negative shocks for the developing world. For instance, Edwards (2006) shows for a set of Latin American nations that a floating exchange rate can help buffer a negative shock, a result also echoed in Cerra, Panizza, and Saxena (2013). Cerra et al. (2013) also show that expansionary fiscal policy and automatic fiscal stabilizers could play a mitigating role. In their study of resilience to financial contagion, Ahrend and Goujard (2015) find that 20 stringent domestic capital adequacy requirements for banks, greater reliance of a domestic banking system on deposits, controls on credit market inflows, and openness to foreign bank entry can play a role. 4 Natural Disasters, Thresholds, and Economic Recov- ery There may damage thresholds beyond which it would be harder for economies to recover from natural disasters, an argument put forward, for instance, by Renaud, Birkmann, Damm, ın (2010). More specifically, the authors argue that external shocks like natural and Gallop´ disasters, can induce an economy to move from one regime (social, economic and ecological) to a very different other one. They, however, also concede that it is difficult to identify such thresholds.14 While there is a small economic literature investigating tipping points in general, usually from a theoretical perspective and focusing on the climate change context (see, for instance, Lemoine and Traeger (2014)), there is as far as we are aware no explicit theoretical or empir- ical study for the case of natural disasters. The nature of natural disaster damages, however, is suggestive of the possibility of tipping points. Firstly, since most natural disasters mea- sured in their physical units (wind speed, peak ground acceleration, etc.) tend to follow a fat-tailed distribution, so do the resultant damages; see Blackwell (2014), Pisarenko and a and Pac´ Rodkin (2014), and Jindrov´ a (2016). Intuitively this means that most events akov´ tend to cause mostly relatively small damages, with a few resulting in very large losses. However, estimates from a standard linear econometric model portray the ’average’ effect, and thus are more likely reflective of the impact of smaller damages in such a distribution. Secondly, and related to this, a linear econometric model, as most empirical applications 14 For a case study of tipping points in the context of coastal systems and tropical storms see ? (?) 21 employ, assumes a linear relationships between the dependent and explanatory variables. There is of course no reason to a priori assume that relationship between losses and macroe- conomic outcomes would be linear. As a matter of fact the few studies that have specifically modelled this do in fact suggest that there is a non-linear relationship. More specifically, oschl (2014) show that a disasters in the top 1 percent of the distribution Felbermayr and Gr¨ cause a country’s GDP growth rate to fall by 7%, compared to 0.5% from those in the top 5%. Similarly, in their meta-analysis of empirical studies of the macroeconomic impact of natural disasters, Klomp and Valckx (2014) show that those that focused only on larger disasters found on average larger negative impacts. One may want to also note that a poor modeling of a possibly non-linear link between losses and macroeconomic variable could also be a reason behind the reason for not being able to identify a significant long-term relation- ship. Whether such thresholds exist for the Caribbean is as of date unknown. Certainly, data on losses due to natural disasters suggest a very fat-tailed like distribution. To illustrate this we use the loss data generated from the 2G CCRIF Loss Model for hurricanes - which incorporates both losses due to wind and storm surge, but not precipitation, damage - for all storms in the North Atlantic Basin back to 1870 until 2014, but assuming 2014 exposure for the Caribbean, as a percentage of total assets.15 Their distribution is shown in 4. As can be seen, most losses due to hurricanes are well below 2% of capital assets, but a few of the historical storms, if they had occurred in the present setting, would have caused destruction of capital assets of over 10%, with the maximum nearly 40%. It seems not unreasonable that there is some, or even several, threshold of losses that could constitute some tipping point beyond which the macroeconomic implications are very different from the low loss, more common, scenario for the region. 15 See Ouattara et al. (2018) for details on the data. 22 Figure 4: Caribbean Hurricane Damage Ratio (%) Distribution In order to gain more explicit insight into the possibility of the existence of thresholds for the Caribbean we also use the data to investigate non-linear effects of hurricane losses on GDP growth rates in the Caribbean. Given that there is incomplete GDP data for many of the Caribbean islands we as in Bertinelli and Strobl (2013) instead use nightlights as proxy for GDP. These data allow us to construct an annual panel of GDP growth (as proxied by nightlight intensity growth) and hurricane loss ratios over the period 1993-2013. To analyze their relationship we estimate the following growth equation: L ∆log (N Li,t−1→t ) = α + βLOSSt−l LOSSi,t−l + πt + µi + it (1) l=0 where N L is the sum of nightlight intensity in territory/country i in year t, LOSS is the share of capital asset losses due to hurricanes, µ is a unit specific fixed effect, and π are yearly indicators. The results of estimating equation 1 are provided in Table 1. As can be seen from the first column, hurricane losses have an immediate negative ef- fect and statistically significant on the GDP growth rate. Taking at face value, the average 23 Table 1: Regression Results (1) (2) (3) LOSSt -1.165* -1.164* -1.112* (0.503) (0.510) (0.504) LOSSt−1 0.047 0.067 (0.598) (0.595) LOSSt−2 0.590 (0.408) Observations 567 567 567 Nr. territories/countries 27 27 27 Notes: (a) Robust standard errors in parentheses; (b) ** and * indicate 1 and 5 per cent significance levels, respectively; (c) All specifications include year dummies. non-zero loss ratio observed in our sample period (0.007) suggests a 0.8% reduction in the GDP growth rate, whereas the highest observed value would imply a reduction of 13%. The inclusion of further lags of LOSS shows that these are insignificant. One may want to note that this is in congruence with a large part of the literature globally and for the Caribbean, as discussed earlier. To investigate possible thresholds we employ the estimator by Baltagi, Li, et al. (2002), which fits partially linear panel data models with fixed effects based on an Epanechnikov- kernel-weighted local polynomial fit. In our context, this estimator fits: ∆log (N Li,t−1→t ) = α + γlog (N Lt−1 ) + g (LOSSi,t ) + πt + µi + it (2) where g () is an unknown, possibly non-linear, function describing the relationship be- 24 tween ∆log (N Li,t−1→t ) and LOSSt . The linear prediction of ∆log (N Li,t−1→t ) is depicted in Figure 5, along with its 95% confidence bands and a scatter plot of the ∆log (N Li,t−1→t ) and LOSSi,t data points. Accordingly, there is a very slight decrease in the estimated negative effect of the losses on the GDP growth rate as losses become larger, although this levels off at around a 4.5% loss ratio. If anything the impact on growth becomes slightly positive at the very end of the observed range of losses, although this is based on a single data point and hence the confidence band is very wide. We also re-estimated equation 2, but modeling LOSS at time t as linear but also including LOSS at time t − 1 as possibly non-linear. The results, shown in Figure 6, show a less negative effect than for LOSS at time t, and while the effect in contrast seems to fall at the end of the range of observed losses, again this is estimated on account of one observation. Thus it seems, based on this preliminary exercise, there is no obvious first hint of any threshold beyond which the impact on a Caribbean economy will be substantially greater than below it. However, at the same time Figures 5 and 6 also exemplifiy the basic challenge in trying to find such thresholds if they indeed exist. More specifically, the availability of data points at higher loss ratios with which one might want to estimate such non-linearity is extremely limited - a feature, as mentioned above, that is typical of data on extreme events. 5 Discussion and Recommendations for Future Research and Data Collection As noted above, one finding that stands out from the empirical literature in general, but also specifically for the Caribbean, is the, although potentially large, ultimately short-term nature of the macroeconomic impact of natural disasters. However, even if one were to assumes that the data and methods employed in current empirical studies are appropriate, one still has to realize that the empirically isolated impact from most of these studies is a ’net’ effect, that 25 Figure 5: Non-Linear Estimate of βLOSSt Figure 6: Non-Linear Estimate of βLOSSt−1 is net of all buffering mechanisms for instance. To this end it would be useful to explicitly investigate how much these help mitigate the actual short-term, or perhaps even suppress any long-term, macro-economic effect of natural disasters in the region. As a starting point 26 there are two obvious candidates for which data could be collected: • Foreign Aid : Although existing evidence suggests that foreign aid inflows after natural disaster are not large relative to the damages (Becerra, Cavallo, and Noy (2014)), they may still play a role in mitigating the negative effects. A useful data source to capture this financing and link it specifically to natural disaster events would be the Financial Tracking Service (FTS) managed by the UN Office for the Coordination of Humanitar- ian Affairs, which reports humanitarian aid flows from government donors, multilateral organizations, NGOs, private foundations, and the private sector. Importantly these aid data can also distinguish between aid committed and pledged, as well as the source country/organization. While they are actually available since 1992, they are believed ¨ to provide reliable coverage only from about the year 2000 (Fuchs and Ohler (2019)). • Remittances : Remittances, as noted above, have shown to potentially play an impor- tant role as a buffering mechanism in the aftermath of losses caused by natural disasters. Moreover, remittances have constituted a crucial influx into Caribbean economies; see Orozco (2017) and Ramcharran (2019). A standard source for these data are the In- ternational Monetary Fund (IMF) Balance of Payments Yearbooks. Such information may, however, also be collected nationally in the Caribbean, as for example in the Re- mittance Reports in Jamaica, and thus be of better quality, or at least more complete. One caveat with measuring these solely in financial flow terms, regardless of the source, is however that in kind remittances, such as through the shipment of barrels, play an important role in the Caribbean and may not be captured in official sources (Kirton (2005)). Combining these data with other macroeconomic indicators and natural disaster data and using a Panel VAR-X model, where the losses are treated as the exogenous variable, could then help identify the role of these other mitigation factors in reducing any impacts. 27 One should note that given that most studies involve estimations on pooled data across countries, the identified impact is necessarily the average effect. However, there may very well be heterogeneity across countries - in other words economies may differ in their vulner- ability for a given shock. The discussion above suggests a number of potential important heterogeneities within the Caribbean: • Sectoral Structure/Diversification : Caribbean economies vary considerable in their in- dustrial structure, with tourism, agriculture, financial services, and some rudimentary manufacturing, dominating most. Mostly, however, as small open economies they tend to be non-diversified (P. Mohan (2016a)), and there is some evidence that diversifica- tion tends to amplify the effects of natural disasters (Botzen et al. (2019)). Challenges in constructing indices of diversification for Caribbean economies may arise because output figures for sectors other than manufacturing and agriculture are difficult to quantify in a similar manner. One possibility is to use population employed in these sectors, although this will overestimate the importance of relatively more labour inten- sive aspects of the economy. One may also want to note that sectors such as agricultural crops and tourism tend to be seasonal, and not taking account whether their seasons coincide further complicates the matter of adequately measuring diversification. • Credit Constraints Credit constraints have been shown to be an important determinant in terms of the extent to which natural disasters affect economies; see McDermott et al. (2013). To create a measure of differences across economies and time in the Caribbean there are a number of potential proxies available, including the level of credit to the private sector as a proportion of GDP or total banking assets. Importantly, Svirydzenka (2016) has recently created an annual database with 9 different proxies of financial market development, covering most of the world, which covers at least all sovereign nations of the Caribbean. These could serve as an initial proxy for the region. A natural empirical tool to investigate how differences across the Caribbean may play a 28 role in the macroeconomic impact with such data is the panel threshold model developed by Hansen (1999). This would allow one to explicitly identify thresholds in the characteristic as it is important to the relationship. How direct losses to human or physical capital feed into the economy is poorly under- stood in the literature in general, and even more so for the Caribbean. More precisely, there is no single study in the region that focuses on trying to understand how such damages perpetuate throughout the economy once they take place. While it would be difficult to quantify this with any precision, there is nevertheless some data available that could provide some rudimentary insight. Firstly, for many disaster incidences in the past in the Caribbean, ECLAC has implemented the DaLA methodology methodology that includes calculating in- direct losses, as for example with Jamaica after Hurricane Ivan (ECLAC (2004)). While certainly some caution should be taken in interpreting the data (Moore and Phillips (2014)), compiling these allows some cross-country, cross-incident comparisons in terms of direct ver- sus indirect losses and effects. Another approach would be to try to use input-output tables to roughly predict indirect losses, keeping mind the drawbacks of doing so (Galbusera and Giannopoulos (2018)). Some headway has been made in this regard - see, for instance, the broad sector Eora National IO Tables, or more detailed attempts for parts of the Caribbean na and Horridge (2015). There are likely also to be locally constructed tables from by Lude˜ national sources that could prove helpful and are not yet in the public domain. 6 Concluding Remarks This paper provided an overview of the existing literature on the macro-economic impact of natural disasters in the Caribbean. Accordingly, while there are a non-negligible number of studies on various aspects of the topic, and this number is growing, there is still much to be 29 discovered in this regard. While there seems to be some consistency in some of the findings, like the potentially large but short-lived impact of natural disasters, given the current data and methodological weaknesses it would be too premature to describe these as stylized facts within any real confidence. Moreover, there are large gaps that simply have not been yet addressed due to a lack of at least publicly available data, as for instance, the indirect impacts of such extreme events. 30 References Acevedo, S. (2016). Gone with the wind: Estimating hurricane and climate change costs in the caribbean. International Monetary Fund. AG van Bergeijk, P., & Lazzaroni, S. (2015). Macroeconomics of natural disasters: Strengths and weaknesses of meta-analysis versus review of literature. Risk analysis , 35 (6), 1050– 1072. Ahrend, R., & Goujard, A. (2015). Global banking, global crises? the role of the bank balance-sheet channel for the transmission of financial crises. European Economic Review , 80 , 253–279. Albala-Bertrand, J.-M. (1993). Natural disaster situations and growth: A macroeconomic model for sudden disaster impacts. World Development , 21 (9), 1417–1434. Anbarci, N., Escaleras, M., & Register, C. A. (2005). Earthquake fatalities: the interaction of nature and political economy. Journal of Public Economics , 89 (9-10), 1907–1933. Baltagi, B. H., Li, D., et al. (2002). Series estimation of partially linear panel data models with fixed effects. Annals of economics and finance , 3 (1), 103–116. Becerra, O., Cavallo, E., & Noy, I. (2014). Foreign aid in the aftermath of large natural disasters. Review of Development Economics , 18 (3), 445–460. Bello, O. (2017). Disasters, economic growth and fiscal response in the countries of latin america and the caribbean, 1972-2010. CEPAL Review . Bertinelli, L., & Strobl, E. (2013). Quantifying the local economic growth impact of hurricane strikes: an analysis from outer space for the caribbean. Journal of Applied Meteorology and Climatology , 52 (8), 1688–1697. Best, R., & Burke, P. J. (2019). Macroeconomic impacts of the 2010 earthquake in haiti. Empirical Economics , 56 (5), 1647–1681. Blackwell, C. (2014). Power law or lognormal? distribution of normalized hurricane damages in the united states, 1900–2005. Natural Hazards Review , 16 (3), 04014024. Blanc, E., & Strobl, E. (2016). Assessing the impact of typhoons on rice production in the philippines. Journal of Applied Meteorology and Climatology , 55 (4), 993–1007. Borensztein, E., Cavallo, E., & Valenzuela, P. (2009). Debt sustainability under catastrophic risk: The case for government budget insurance. Risk Management and Insurance Review , 12 (2), 273-294. doi: j.1540-6296.2009.01168.x Botzen, W., Deschenes, O., & Sanders, M. (2019). The economic impacts of natural disasters: A review of models and empirical studies. Review of Environmental Economics and Policy . ere, M., Cheng, G., Chinn, M. D., & Lisack, N. (2015). For a few dollars more: Bussi` Reserves and growth in times of crises. Journal of International Money and Finance , 52 , 127–145. CACDN. (1999). The impacts of natural disasters: A framework for loss estimation. National Academies Press. Caldera S´ anchez, A., Rasmussen, M., & R¨ ohn, O. (2016). Economic resilience: What role for policies? Journal of International Commerce, Economics and Policy , 7 (02), 1650009. Carter, M. R., Little, P. D., Mogues, T., & Negatu, W. (2007). Poverty traps and natural disasters in ethiopia and honduras. World development , 35 (5), 835–856. 31 Cashin, P., & Sosa, S. (2013). Macroeconomic fluctuations in the eastern caribbean: The role of climatic and external shocks. The Journal of International Trade & Economic Development , 22 (5), 729–748. Cavallo, E., Noy, I., et al. (2011). Natural disasters and the economy—a survey. International Review of Environmental and Resource Economics , 5 (1), 63–102. Cerra, V., Panizza, U., & Saxena, S. C. (2013). International evidence on recovery from recessions. Contemporary Economic Policy , 31 (2), 424–439. Clarke, G. R., & Wallsten, S. (2003). Do remittances act like insurance? evidence from a natural disaster in jamaica. Evidence from a Natural Disaster in Jamaica (January 2003) . Ebeke, C., & Combes, J.-L. (2013). Do remittances dampen the effect of natural disasters on output growth volatility in developing countries? Applied Economics , 45 (16), 2241–2254. ECLAC, U. (2004). Assessment of the socioeconomic and environmental impact of hurricane ivan on jamaica. ECLAC, Port of Spain, Trinidad and Tobago. See http://www. eclac. org/publicaciones/xml/1/20501/L636-completo. pdf (accessed 17/01/2013) . Edwards, S. (2006). Monetary unions, external shocks and economic performance: A latin american perspective. International Economics and Economic Policy , 3 (3-4), 225–247. Escaleras, M., Anbarci, N., & Register, C. A. (2007). Public sector corruption and major earthquakes: A potentially deadly interaction. Public Choice , 132 (1-2), 209–230. Felbermayr, G., & Gr¨ oschl, J. (2014). Naturally negative: The growth effects of natural disasters. Journal of development economics , 111 , 92–106. ¨ Fuchs, A., & Ohler, H. (2019). Does private aid follow the flag? an empirical analysis of humanitarian assistance (Tech. Rep.). Kiel Working Paper. Galbusera, L., & Giannopoulos, G. (2018). On input-output economic models in disaster impact assessment. International journal of disaster risk reduction , 30 , 186–198. Granvorka, C., & Strobl, E. (2013). The impact of hurricane strikes on tourist arrivals in the caribbean. Tourism Economics , 19 (6), 1401–1409. Greenidge, K., Drakes, L., & Craigwell, R. (2010). The external public debt in the caribbean community. Journal of Policy Modeling , 32 (3), 418-431. doi: https://doi.org/10.1016/ j.jpolmod.2010.02.004 Hallegatte, S., & Przyluski, V. (2010). The economics of natural disasters: concepts and methods. The World Bank. Hallegatte, S., & Vogt-Schilb, A. (2019). Are losses from natural disasters more than just asset losses? In Advances in spatial and economic modeling of disaster impacts (pp. 15–42). Springer. Hansen, B. E. (1999). Threshold effects in non-dynamic panels: Estimation, testing, and inference. Journal of econometrics , 93 (2), 345–368. Heger, M., Julca, A., & Paddison, O. (2008). Analysing the impact of natural hazards in small economies: The caribbean case [WIDER Working Paper Series]. (025). Heinen, A., Khadan, J., & Strobl, E. (2018). The price impact of extreme weather in developing countries. The Economic Journal , 129 (619), 1327–1342. Henry, M., Spencer, N., & Strobl, E. (2019). The impact of tropical storms on house- holds: Evidence from panel data on consumption. Oxford Bulletin of Economics and Statistics . 32 Hsiang, S. M. (2010). Temperatures and cyclones strongly associated with economic pro- duction in the caribbean and central america. Proceedings of the National Academy of sciences , 107 (35), 15367–15372. Hsiang, S. M., & Jina, A. S. (2014). The causal effect of environmental catastrophe on long- run economic growth: Evidence from 6,700 cyclones (Tech. Rep.). National Bureau of Economic Research. IMF. (2019). Building resilience in developing countries vulnerable to large natural disasters. IMF Policy Paper . Jindrov´ a, P., & Pac´akov´a, V. (2016). Modelling of extreme losses in natural disasters. In- ternational Journal of Mathematical Models and Methods in Applied Sciences, volume 10, issue: 2016 . Kaplan, O.-O., Sandrine, & Strobl, E. (2020). Impact of hurricanes strikes on international reserves in the caribbean. Journal of International Money and Finance , forthcoming. Kirton, C. D. (2005). Remittances: The experience of the english-speaking caribbean. Beyond small change: Making migrant remittances count , 261–294. Klomp, J. (2017). Flooded with debt. Journal of International Money and Finance , 73 , 93-103. doi: 10.1016/j.jimonfin.2017.0 Klomp, J., & de Haan, J. (2015). Bank regulation and financial fragility in developing countries: Does bank structure matter? Review of Development Finance , 5 (2), 82-90. doi: https://doi.org/10.1016/j.rdf.2015.11.001 Klomp, J., & Valckx, K. (2014). Natural disasters and economic growth: A meta-analysis. Global Environmental Change , 26 , 183–195. Knutson, T., Camargo, S. J., Chan, J. C., Emanuel, K., Ho, C.-H., Kossin, J., . . . others (2019). Tropical cyclones and climate change assessment: Part ii. projected response to anthropogenic warming. Bulletin of the American Meteorological Society (2019). Koetsier, I. (2017). Natural disasters and (future) government debt. , 48-74. Lazzaroni, S., & van Bergeijk, P. A. (2014). Natural disasters’ impact, factors of resilience and development: A meta-analysis of the macroeconomic literature. Ecological Eco- nomics , 107 , 333–346. Lemoine, D., & Traeger, C. (2014). Watch your step: optimal policy in a tipping climate. American Economic Journal: Economic Policy , 6 (1), 137–66. Lis, E., & Nickel, C. (2009). The impact of extreme weather events on budget balances and implications for fiscal policy [Working Paper Series]. (1055). Lude˜na, C. E., & Horridge, M. (2015). Gtap regional input-output data for jamaica, trinidad & tobago, puerto rico and the dominican republic (Tech. Rep.). Inter-American Devel- opment Bank. Lugay, B., & Ronald, J. (2014). The impact of natural disasters on public debt accumulation in selected eastern caribbean countries [Working Paper Series]. (14/6). McDermott, T. K., Barry, F., & Tol, R. S. (2013). Disasters and development: natural disasters, credit constraints, and economic growth. Oxford Economic Papers , 66 (3), 750–773. McLean, N. M., Stephenson, T. S., Taylor, M. A., & Campbell, J. D. (2015). Characterization of future caribbean rainfall and temperature extremes across rainfall zones. Advances in Meteorology , 2015 . 33 Mejia, S. A. (2014). Debt, growth and natural disasters a caribbean trilogy (No. 14-125). International Monetary Fund. Melecky, M., & Raddatz, C. (2011). How do governments respond after catastrophes? natural-disaster shocks and the fiscal stance. Mohan, P. (2016a). Caribbean diversification and development. The World Economy , 39 (9), 1434–1453. Mohan, P. (2016b). Impact of hurricanes on agriculture: evidence from the caribbean. Natural Hazards Review , 18 (3), 04016012. Mohan, P. (2017). The economic impact of hurricanes on bananas: A case study of dominica using synthetic control methods. Food policy , 68 , 21–30. Mohan, P., & Strobl, E. (2020). Hurricane losses and government debt: evidence from the eastern caribbean. unpublished mimeograph . Mohan, P. S., Ouattara, B., & Strobl, E. (2018). Decomposing the macroeconomic effects of natural disasters: A national income accounting perspective. Ecological economics , 146 , 1–9. Moore, W., Elliott, W., & Lorde, T. (2017). Climate change, atlantic storm activity and the regional socio-economic impacts on the caribbean. Environment, development and sustainability , 19 (2), 707–726. Moore, W., & Glean, A. (2016). Foreign exchange reserve adequacy and exogenous shocks. Applied Economics , 48 (6), 490–501. Moore, W., & Phillips, W. (2014). Review of eclac damage and loss assessments in the caribbean. Noy, I., duPont IV, W., et al. (2018). The long-term consequences of disasters: What do we know, and what we still don’t. International Review of Environmental and Resource Economics , 12 (4), 325–354. Noy, I., & Nualsri, A. (2011). Fiscal storms: public spending and revenues in the aftermath of natural disasters. Environment and Development Economics , 16 (1), 113-128. Orozco, M. (2017). Remittances to latin america and the caribbean in 2016. The Dialogue. Avaliable online: https://www. thedialogue. org/wpcontent/uploads/2018/01/Remittances-2017-1. pdf (2019-02-15) . Ouattara, B., & Strobl, E. (2013, 01). The fiscal implications of hurricane strikes in the caribbean. Ecological Economics , 85 , 105–115. doi: 10.1016/j.ecolecon.2012.10.002 Ouattara, B., Strobl, E., Vermeiren, J., & Yearwood, S. (2018). Fiscal shortage risk and the potential role for tropical storm insurance: evidence from the caribbean. Environment and Development Economics , 23 (6), 702–720. Pan, Q. (2010). Economic losses from a hypothetical hurricane event in the houston-galveston area. Natural Hazards Review , 12 (3), 146–155. Pisarenko, V., & Rodkin, M. V. (2014). Statistical analysis of natural disasters and related losses. Springer. Ramcharran, H. (2019). Analyzing the impact of workers’ remittances on household con- sumption in latin american and caribbean countries. Journal of Economics and Fi- nance , 1–19. Rasmussen, T. N. (2004). Macroeconomic implications of natural disasters in the caribbean [IMF Working Papers]. (04/224). 34 Reinhart, C. M., & Rogoff, K. S. (2010). Growth in a time of debt. American economic review , 100 (2), 573–78. Renaud, F. G., Birkmann, J., Damm, M., & Gallop´ ın, G. C. (2010). Understanding multiple thresholds of coupled social–ecological systems exposed to natural hazards as external shocks. Natural Hazards , 55 (3), 749–763. Spencer, N., & Polachek, S. (2015). Hurricane watch: Battening down the effects of the storm on local crop production. Ecological Economics , 120 , 234–240. Spencer, N., & Urquhart, M.-A. (2018). Hurricane strikes and migration: Evidence from storms in central america and the caribbean. Weather, climate, and society , 10 (3), 569–577. Strobl, E. (2011). Impact of hurricane strikes on local cropland productivity: Evidence from the caribbean. Natural Hazards Review , 13 (2), 132–138. Strobl, E. (2012). The economic growth impact of natural disasters in developing countries: Evidence from hurricane strikes in the central american and caribbean regions. Journal of Development economics , 97 (1), 130–141. Strobl, E., Kablan, S., et al. (2017). How do natural disasters impact the exchange rate: an investigation through small island developing states (sids)? Economics Bulletin , 37 (3), 2274–2281. Strobl, E., & Walsh, F. (2009). The re-building effect of hurricanes: evidence from employ- ment in the us construction industry. Economics Bulletin , 29 (4), 3059–3066. Svirydzenka, K. (2016). Introducing a new broad-based index of financial development. International Monetary Fund. Von Peter, G., Von Dahlen, S., & Saxena, S. C. (2012). Unmitigated disasters? new evidence on the macroeconomic cost of natural catastrophes. Woolridge, J. M. (2009). Introductory econometrics: A modern approach,(5e). South- Western, Michigan State University , 378 , 57. Yang, D. (2008). Coping with disaster: The impact of hurricanes on international financial flows, 1970-2002. The BE Journal of Economic Analysis & Policy , 8 (1). 35