https://orcid.org/0000-0001-9359-8117
![Sample our Environment & Agriculture journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5934/TOC-Subject-Sample-2021-Environment-Agriculture.jpg"})
ABSTRACT
Pacific island countries are highly vulnerable to destructive and unpredictable disasters. This paper identifies the intensity of disasters for each country in the Pacific based on the distribution of economic damages and population affected by disasters, and estimates the impact of disasters on economic growth and international trade using a panel regression method. The results show that ‘severe’ disasters have a negative and significant impact on economic growth and lead to a deterioration of the trade balance. Going further, this paper proposes a simple and consistent approach to incorporate disaster risks for economic planning. Better incorporating the economic impact of disasters in the medium- and long-term economic planning would help policy-makers improve policy decisions and to be better prepared for disasters.
Acknowledgments
An earlier version of this paper has been circulated as an IMF working paper under the title of ‘The Economic Impact of Natural Disasters in Pacific Island Countries: Adaptation and Preparedness.’ The working paper was written when the authors were working at the International Monetary Fund. We are thankful to Nitya Aasaavari, Giovanni Melina, Saad Quayyum, Scott Roger, Alison Stuart, and Marina Mendes Tavares for their thoughtful comments. Chau Nguyen provided research assistance. I am also particularly thankful to two anonymous referees for their thoughtful comments and suggestions. The views expressed in this paper are those of the authors and do not necessarily represent those of Bank of Korea or International Monetary Fund. Any remaining errors are the authors' sole responsibility.
Disclosure statement
No potential conflict of interest was reported by the authors.
Correction Statement
This article has been republished with minor changes. These changes do not impact the academic content of the article.
Notes
1 See, for example, Dell et al. (Citation2014), Felbermayr and Grőschl (Citation2014), and Cavallo and Noy (Citation2011) for a detailed survey on the existing empirical literature on the growth impact of disasters and climate shocks.
2 Felbermayr and Grőschl (Citation2014), however, find significant adverse growth effects for all types of disasters.
3 Jones and Olken (Citation2010) find that climate shocks have a negative and significant impact on the exports of poor countries and agricultural products or light manufacturing industries, with a less severe impact on heavy manufacturing industries or raw materials.
4 Li et al. (Citation2015) also show that poorer, warmer, and non-coastal cities in China – which might be less prepared for adaptation to climate change – suffer more from temperature warming.
5 The database is available at http://www.emdat.be/.
6 We use observation-level natural disaster data from the EM-DAT which represent a certain type of disaster that happened in the year. For instance, if each type of disaster happened twice in the same year, it is counted as one observation but with two occurrences. Our analyses in this and later sections are based on the observations instead of the occurrences (event-level) because macroeconomic variables used in the analysis are observed on an annual basis.
7 The country codes for PICs are as follows: Fiji (FJI), Kiribati (KIR), Marshall Islands (MHL), Micronesia (FSM), Palau (PLW), Papua New Guinea (PNG), Samoa (WSM), Solomon Islands (SLB), Timor-Leste (TLS), Tonga (TON), Tuvalu (TUV), and Vanuatu (VUT).
8 It is notable that the likelihood of disasters is computed as the probability that at least one disaster occurs in a given year in Table . For instance, the cases that disasters occur once or twice in a year are considered as the same when computing the likelihood.
9 We truncated the impact on a country's population to 100% in cases which reported a population effect of greater than the country's population.
10 This finding may reflect the improvement in recording disasters. However, we cannot conclude that the upward trend in the occurrence of disasters in PICs is mainly attributable to improved data recording owing to: (i) a clear increasing trend since 2010 which is quite recent year and (ii) a decreasing trend in the occurrence of world-wide disasters since 2000.
11 Estimated damages include both direct damages and indirect losses by disasters.
12 Thus, our proposed method to incorporate the economic impact of disasters in economic projections considers both the probability and the intensity of disasters (Section 5).
13 Specifically, Fomby et al. (Citation2013) measure the natural disaster intensity by the sum of the fatalities and 30% of the total people affected. A disaster is considered as ‘moderate’ if the intensity measure is greater than 0.01% of the population and as ‘severe’ if it is above 1%. This practice was first used by Becker and Mauro (Citation2006) as a way to quantify country-specific external shocks.
14 There are many missing observations in the economic damage of each natural disaster in the EM-DAT.
15 75th percentiles in the damage-to-GDP ratio and in the population affected-to-total population ratio correspond to 7.0% and 7.6%, respectively. Meanwhile, 25th and 50th percentiles correspond to 0.3% and 1.2% for the damage-to-GDP ratio and 0.3% and 1.3% for the population affected-to-total population ratio.
16 In addition to the impact on growth and trade, disasters are likely to affect the fiscal balance via a decrease in government revenue and an increase in government expenditure. Government revenue tends to decrease in the aftermath of severe disasters due to lower GDP growth especially from the disruptions in the agriculture and tourism sectors, while government expenditure may increase in the disaster year to support the economic recovery from the damages of disasters. However, the impacts of disasters on the fiscal balance, government revenue and expenditure are estimated to be insignificant. The results are not reported, but they can be provided upon request.
17 The panel data are unbalanced since some variables are not available for the whole period in all 12 countries.
18 Due to data limitation in PICs, we could not include richer set of control variables on structural, domestic policy, and external factors which may also relate to GDP growth and other dependent variables. Instead country and year fixed effects are expected to capture some of cross-country and time-variant heterogeneity.
19 The threshold effect has also been found in the effect of public debt on output growth (e.g. Cecchetti et al., Citation2011; Lee et al., Citation2017) and the effect of working hours on productivity growth (e.g. Lee & Lim, Citation2017).
20 The results are broadly consistent with a different specification for the natural disaster intensity: e.g. the intensity measure based on higher percentiles between damages and population affected if both data are available.
21 The broadly consistent results can be obtained by using the measure of natural disaster intensity itself, constructed from damages and population affected, instead of dummy variables (see Section 4.3 and Table ).
22 It is commonly known that the system GMM method is appropriate for data that constitute a short time series and a large number of individuals (Blundell & Bond, Citation1998).
23 More rigorously, we can adopt a threshold regression method proposed by Hansen (Citation1999) to determine a threshold level. However, we instead chose the way above as the focus of this paper is to measure the average size of economic impact of severe disasters instead of identifying the threshold level itself.
24 Note that there exist large variations in the economic loss by disasters across countries: i.e. 6.6% and 4.4% of GDP for Vanuatu and Tonga, respectively, but only 0.2% of GDP for Kiribati. In addition, the country's own loss estimates for Fiji are also similar levels, corresponding to 1.6% of GDP (Government of Fiji, Citation2017).
25 For robustness, we also estimated a separate regression by including lagged disaster dummy variables as additional explanatory variables to capture the lagged growth and trade effects by disasters. However, the estimated lagged effects were found to be small and insignificant (see Section 4.3 and Table ), which is consistent with González, London, et al. (Citation2021)
26 We do not include financial supports (grants or loans) from donors as separate explanatory variables, although these variables may have impacts on imports and the trade balance, because the separate inclusion of grants did not affect the results significantly.
27 See Osberghaus (Citation2019) for a comprehensive literature survey on the effect of disasters and weather variation on international trade.
28 Some existing studies found that different types of disasters have distinct effects on economic growth (e.g. Fomby et al., Citation2013; Loayza et al., Citation2012). However, the dummy variable for floods is not included in our estimation because no flood event above 75th percentile occurred over the sample period in the Pacific region.
29 We can also include more lags for natural disaster dummy variables as a robustness check. The results confirm that the effects on growth and trade for two and three years lagged disaster variables are similar to the effects for one-year lagged disaster variables.
30 However, one should note that this result may also relate to the measurement issue of a natural disaster intensity as a large proportion of the highest percentile mostly belongs to the disaster events only with the data on population affected in which the intensity measure is less likely to capture the actual economic impact.
31 In this specification, we include only the sample of disaster events to estimate the differentiated effects by the natural disaster intensity.
32 IMF (Citation2017) proposes that the ‘large disaster’ scenario should consider a one-off shock to public debt, GDP growth, and export growth in the year of disaster. It also allows for tailoring the scenario to country-specific circumstances.
33 For the calculation of the macroeconomic impacts of Cyclone Pam in Vanuatu by the difference of economic variables between 2014 (pre-cyclone) and 2015 (cyclone-year), the impacts on GDP growth can be underestimated while that on the fiscal balance and the trade balance can be overestimated since the government's several large infrastructure projects started in 2015. Countries tend to record relatively higher GDP growth along with larger fiscal deficit and trade deficit during the infrastructure boom. This can offset some of adverse growth impacts of the cyclone and it likely leads to deterioration in the fiscal and trade balance (along with an increase in government expenditure and imports).
34 The adjustment for the fiscal balance against disasters should also be considered. Given difficulty in measuring the impact of disasters on fiscal variables, this paper does not consider the direct impact but it instead consider indirect adjustment through the impact on GDP growth and the trade balance (see Footnote 16).
35 By doing this, we can avoid a double adjustment for natural disaster impact in the year of one-time adjustment.
Additional information
Notes on contributors
Dongyeol Lee
Dongyeol Lee is a Deputy Director at Bank of Korea in Republic of Korea. He obtained Ph.D. in Economics at Michigan State University. His research interests are in economic growth & development, industrial organization, applied econometrics and international economics.
Huan Zhang
Huan (Irene) Zhang is a Senior Data Scientist at Convoy in the United States. She received MA in School of Advanced International Studies (SAIS) at Johns Hopkins University. Her researches have focused on the growth and development in low-income countries.