Natural Disaster Impacts and
Fiscal Decentralization
An Empirical Analysis
MSc Thesis International Economics and Business University of Groningen – Faculty of Economics and BusinessAbstract
This thesis focuses on the relationship between the shocks of natural disasters and the governmental style of fiscal decentralization. The modified model contributes to the existing idea that local actors are able to cope superiorly with the impacts of natural disasters. Operating a panel data set of merged disaster data and government fiscal data, the results show that fiscal decentralization indeed does lead to lower levels of disaster-induced deaths, presenting a significant contribution to the previous literature that higher economic development leads to superior mitigation of natural disasters. Hence, this thesis offers a crucial insight in governmental styles and policies worldwide.
Key words: natural disasters, fiscal decentralization, economic development, institutions
Author: J.W. Vogel
Student ID number: 1889680
Contents
1. Introduction ... 5
2. Literature Review ... 7
2.1 Natural disasters ... 7
2.2 Fiscal decentralization ... 9
2.3 Effects of fiscal decentralization on natural disasters ... 10
3. Empirical analysis ... 14
3.1 Data on variables ... 14
3.2 Model and methodology ... 15
1. Introduction
“Communities will always face natural hazards, but today’s disasters are often generated by, or at least exacerbated by, human activities. At no time in human history have so many people lived in cities clustered around seismically active areas. Destitution and demographic pressure have led more people than ever before to live in flood plains or in areas prone to landslides. Poor land-use planning; environmental management; and a lack of regulatory mechanisms both increase the risk and exacerbate the effects of disasters.”
Kofi Annan.
Natural disasters have devastating effects for humankind. Examples such as the tsunami in Indonesia (2004) and hurricane Katrina (2005) strike to notice in the past decade. According to the International Disaster database EM-DAT, natural disasters are classified in five different categories, which are presented in appendix 1. As one can perceive in appendix 2, hydrological types of disasters, such as floods, occur the most worldwide. Within the last decades there seems to be a surge in hydrological disasters. According to a report by Leaning and Guha-Sapir (2013), this is due to the increasing effects of climate changes. Apart from the trail of death and people displacement natural disasters leave behind on humanity, Figure 1 signifies the worldwide rise in estimated damages of natural disasters from 1975 to 2000, indicating an upward trend in estimated damages.
Figure 1: Estimated damage (US$) billion caused by reported natural disasters 1975 – 2010
As such, the effects of natural disasters are important. Although natural disasters are uncontrollable, their impact also depends on the political response of a country. It is generally understood that as a country develops, it devotes greater resources to safety (Skidmore, 2013). Previous research (Kahn, 2005) has shown that there is a distinct and predictable pattern between losses from natural disaster events and economic development. Consequently, the adaptive abilities of a state depends on the quality of the economic circumstances, such as the GDP, and to what extent the available resources in that specific country are used.
However, the actual distribution of these resources is important as well. One could question whether the distribution of resources would work better via a centralized or a decentralized governmental style. Decentralized governments can respond rapidly when a natural disaster would occur, but often lack knowledge and awareness, especially in developing countries. The relationship between fiscal decentralization and the mitigation of natural hazards has been researched to some extent. Nonetheless, these existing studies were largely descriptive and not analytical (Bardham, 2002). As a result, this thesis attempts to bridge the research gap by focusing on the relationship between the impact of natural disasters and fiscal decentralisation. Bridging this gap is of crucial importance. Since hardly any research has been performed on the subject of fiscal decentralization and natural disasters, it might lead to new insights about how to cope superiorly with the shocks of natural disasters.
2. Literature Review
2.1 Natural disasters
Natural disasters are a compound mix of natural hazards and human action (Wisner, 2003). Statistics show an increasing number of disasters within the last decades. Only within the last four decades, natural disasters have caused more than 3.3 million deaths and 2.3 trillion dollars in economic damages. In the last three decades, the 2010 Haiti earthquake and the 2004 Indonesian earthquake and tsunami have caused the highest death toll from natural disasters (Worldbank, 2010). As argued by Kahn (2005), the number of people affected by natural disasters has risen over the past few decades. This can be due to locations of housing and production facilities in hazard zones.
As a country is developing, there is a tendency towards superior resources to provide safety, including precautionary measures to hinder the impact of natural disasters. In many areas the outcome can be related to starvation and disease. Especially in developing countries vulnerable people often suffer repeatedly from shocks that influence their families, settlements and their livelihoods. Wealthier people can defend themselves better against disasters by money, which can buy design and engineering. Furthermore, richer people can choose to live in a hazardous area, whereas poor people often do not have this choice (Wisner, 2003).
Gencer (2013) argues that due to the urban concentration of population, the highest potential for tragedies happens in the most populated cities. In urban areas, there is a strong connection between vulnerability and poverty. Lack of sanitation, clean water and garbage removal add to the vulnerability of disasters, which can result in even more environmental and health problems. Additionally, disasters are a constraint on economic and human development at the household level and at the national level when roads, hospitals etc. are damaged. According to Wisner (2003), it can be discussed that the pressure of population growth and lack of ‘modernization’ of the economy and other institutions drive human conquest of an unforgiving nature. It is true that some groups are more disposed to damage, loss and suffering in the context of differing hazards. Some key explanations include wealth, occupation, caste, ethnicity, gender, disability, health status, age and immigration status, and the nature and extent of social networks (Wisner, 2003). This is generally in line with de Ville de Goyet et al. (2005), who states that vulnerability to disasters is linked to demographic growth, rapid urbanization, settlement in unsafe areas, environmental degradation, climate change and unplanned development.
2.2 Fiscal decentralization
In the literature that exists, there is a strand that focuses on the effectiveness of a decentralized fiscal system (Skidmore, 2013). Although decentralization occurs in a variety of countries around the world, there is a lack of quantitative evidence that can prove this effectiveness. There have been small numbers of so-called ‘before-after frameworks’. For instance, Santos (1998) evaluated the decentralization initiative in Bolivia, finding that access to basic sanitation services and utilization of elementary and secondary schools augmented twofold following decentralization. Furthermore, in the World Development Report on Infrastructure (World Bank, 1994) it is cited that cost savings next to a quality improvement in public infrastructure projects has followed the transfer of management responsibility to local authorities. In this study, it was found out that decentralized governments were more efficient in providing infrastructure roads and water supplies at low costs. Consequently, these studies provide evidence that it has been more efficient for public services to be carried out decentralized. However, there is still a lack of cross-country analyses to evaluate the effectiveness of government decentralization, due to the unavailability of equivalent data on costs and effectiveness of government activities (Skidmore, 2013).
According to Escaleras (2012), there has been a transferral towards forms of decentralization in countries around the world in order to increase overall government performance. Some of these nations have decided to do so while others have been convinced by organizations such as the World Bank that measure for decentralization requirements. The reasoning behind this was to enhance the provision of local public and semi-public goods and services. Moreover, this especially concerned goods and services that targeted unique local or regional needs. This is linked to the idea of carrying decisions closer to the local governmental officials, as they have superior knowledge and understanding of the unique local demands and needs.
2.3 Effects of fiscal decentralization on natural disasters
As discussed, natural disasters are uncertain events with uncertain impacts. Both national and subnational governments play vital roles in the preparation and responding phase of a natural disaster. This gives an opportunity to assess the efficiency of both systems when it comes to protecting human life. More specifically, in this thesis it is assessed whether decentralized governmental systems are more prone to safeguard against natural disaster risks. In research so far there has been a lack of consistency across studies involving the concept of decentralization (Skidmore, 2013). This may simply be because of various samples and a variety of accepted definitions of decentralization.
Pantoja (2002) suggests that the potential of local governments to deal with their own disaster risk management will depend on sector and institutional level factors. These factors include the degree of (in)formality, dependence on government funds, the level of financial and operational sustainability, and the decision to offer subsidized credit or poverty. Commonly, small, locally based institutions are as a result expected to be more vulnerable to natural disasters than bigger, more geographically dispersed institutions (World Bank, 2000). Furthermore, local governments comprising poor members or reaching poor clients will tend to be more exposed to harsh conditions. Characteristically, younger and smaller governments have a more challenging time in responding to disasters; during the emergency period itself, and in implementing effective preparedness and strategies.
Gemcer (2013) argues that local governments are able to be proactive, rather than waiting for hazards and the subsequent disasters to occur. Moreover, there is internationally a shift from recovery and reconstruction to disaster management prevention and mitigation. Many local governments in various countries worldwide have undertaken disaster risk management programs. In addition to local projects, there are several regional and international programs that support local governments initiatives in disaster risk reduction (Gemcer, 2013).
country. For example, in the United States practically no area is free of natural disasters, and often these disasters are unique – i.e., wind storms in the South East, earthquakes in the West Coast. As a result, since these natural disasters occur in a local environment, having local actors would enlarge the need for risk reduction. This is in line with Skidmore (2013), who argues that in the last few years, central governments have received some criticisms due to the fact that they lack preparation and response abilities when a natural disaster occurs. Public responsibilities include manners such as safety, land use and economic development decisions, all of which play important roles in the mitigation of natural disasters. Within a decentralized system, more autonomous subnational authorities typically provide public safety services.
Dayton-Johnson (2006) argues that local know-how and capabilities are very valuable for disaster preparation and coping efforts after the tragedy has hit. Concretely, local community networks might be able to deliver these facilities at a lower average costs than international networks. The author also provides some examples of domestic policies that can reduce vulnerability and increase resilience. For instance, there should be guidelines that oversee urban development so that the building of houses and industries are not as vulnerable in areas that are subject to many natural disasters. According to Dayton-Johnson (2006), whether domestic or foreign aid provides the resources does not matter as long as there is sufficient location-specific knowledge that can be given by local governments. In addition, in a research conducted by Burnside and Dollar (2000) it is stated that foreign aid cautions might differ broadly in their efficiency depending on the country they are carried out at. This is logical reasoning since disaster-prone countries vary widely in quality of governance, and this will have an effect on the efficiency of disaster-related aid.
once there is an inadequate financial support in a central government, they leave the disaster management to local governments, who in their case lack the skills to deal competently with these issues.
On the other hand, a central authority can provide protection across sub-national regions: it can pool risks. Furthermore, Persson and Tabellini (1996) state that decentralized governments that should prepare for disasters are often lacking knowledge and awareness, and thereby oversee some specific characteristics when funding protective infrastructure investments. Moreover, the synergy between local governments and the national governments is seldom realised in practice (Pantoja, 2002). It is argued that the role of decentralization is diverse and that it should be analysed to what extent local institutions have advantages over national institutions. Assistance should be provided for local actors when constructing regulations, as well as providing elaborate strategies for linking disaster prevention and response activities into long-term sustainable rural development strategies. Since many local authorities in developing countries lack knowledge in how to assist and guide the implementations of activities, mobilize local participation and handle the communication between higher and local policy levels, there is still a lot to be gained (Pantoja, 2002).
To conclude, there is interplay between national and subnational types of governments in the mitigation process of natural hazards. The previous literature shows that the level of fiscal decentralization could play a significant role in lowering the quantity of causalities when a natural disaster strikes. As such, the impact of natural disasters will be tested by proxying the amount of disaster-induced deaths. All in all, these considerations lead to the following hypothesis for this study:
3. Empirical analysis
3.1 Data on variables
In order to test the hypothesis a quantitative study will be performed. I merged disaster data with macroeconomic and governmental fiscal data to create my own panel dataset. The unit of analysis is the disaster event, and 3096 natural disasters are examined from 108 countries over the time period 1972 until 2000. The 29 calendar years are selected for this study because these were the only data available from the International Financial Statistics database in regards of fiscal decentralization.
The essential variables in the analysis relate to death due to natural disasters and the degree of fiscal decentralization in a country. As carried out in other research, the most efficient tool of measuring disaster events is calculating the reported disaster-induced deaths and therefore this measure is used as an indicator of the impact of a natural disaster (Skidmore, 2013). Thus, the dependent variable is the Death rate, which represents the ratio of the total number of deaths due to natural disasters within a country during a year in which a disaster struck to the country’s population. The data on natural disasters comes from the OFDA / CRED database. This databank is a consequence of cooperation between the Office of U.S. Foreign Disaster Assistance and the Centre for Research on the Epidemiology of Disasters. They provide information on all sorts of natural disasters. As many nations would like to establish superior preparedness for when natural disasters occur, demand for complete and verified data has been growing. The CRED utilizes specific criteria to determine whether an event is classified as a natural disaster. This involves ten or more fatalities; 100 or more affected, injured or left homeless; significant damages incurred and whether a declaration of a state of emergency for international assistance was issued (EMDAT, 2014).
Furthermore, following the literature review and previous research on the topic of fiscal decentralization and natural disasters (Skidmore (2013); Escaleras (2012) and Yamamura (2011)), a number of control variables are introduced. These are from the Polity IV data series the level of ‘Autocracy’ and ‘Democracy’ (controlling for social and political institutions), while ‘Homeless’ and ‘Affected’ derive from the EM-DAT database (controlling for the impacts of natural disasters on communities). From the World Development Indicators database ‘GDP per capita’ (controlling for economic developments), ‘Population density’, ‘Fertility’ and ‘Urban’ (controlling for demographics) and ‘Land’ (capturing a country’s total area) are utilized. An elaborate explanation of the variables involved in the equation, as well as a summary of the expected effects from previous studies is listed in appendix 3.
3.2 Model and methodology
Following the considerations above, this leads to the following equation:
(log) Death ratei = αi + β1 ‘Fiscal decentralization’i + β2 (log) ‘Homeless’i + β3 (log) ‘Affected’i + β4 ‘Autocracy’i + β5 ‘Democracy’i + β6 (log) ‘GDP per capita’I + β7 (log) ‘Population density’ + β8 ‘Fertility’+ β9 Urban β + β10 (log)‘Land’+ εi
In which ‘Death rate’i is the number of deaths from natural disasters in country i, αi is a constant and ‘Fiscal decentralization’i is the share of sub-national expenditures.‘Homeless’i , ‘Affected’i , ‘Autocracy’i , ‘Democracy’i , ‘GDP per capita’i, ‘Population density’i, ‘Fertility’i, ‘Urban’i and ‘Land’i are control variables over the period 1972-2000, and εi is the error term.
Appendix 5 lists the summary statistics of the dependent variable deaths in detail. There are two outcomes that strike to notice. First, one can perceive that the variance is very large. This is similar to what has been discovered in previous studies on natural disasters (Escaleras, 2012). Due to this wide variation, the log-form of the death toll is taken. Second, roughly one third of the observations record zero fatalities, meaning that the sample is truncated at zero.
In addition, appendix 6 lists the summary statistics of all subtypes of natural disasters. It appeals that the most common subtypes of disasters are earthquakes, floods and windstorms. To further focus on the relationship between decentralization and disaster-induced deaths, dummy variables are created for these three subtypes of natural disasters in order to control for the regression estimations. For instance, the dummy variably earthquake will take the value 1 if an earthquake hits and value 0 otherwise. Naturally, this works similar for the other subtypes of natural disasters.
4. Results
4.1 Baseline model
Summary statistics of the variables involved in the equation are provided in Table 1. Note that the variables are not logged yet. This table portrays that there is a wide variation in observations for the dependent variable deaths: the mean is 16 times smaller than the standard deviation. Furthermore, the minimum and maximum values range from 0 to 300.000. This last value indicates a period of severe drought in Ethiopia in 1983.
Table 1: Summary statistics, cross country data analysis 1972 – 2000
The estimation results of the pooled OLS regressions are presented in Table 2. Column 1 represents a baseline regression in which the dependent variable deaths is estimated by the explanatory variable fiscal decentralization and various control variables, as introduced in the methodology section. Prior to discussing the individual outcomes of the model, it should be noted that in Column 1 the R-square value of 56% suggests of a reasonable fit, indicating that the model explains 56% of the variability of the response data around its mean. In Columns 2 until 4, dummy variables for the three most common subtypes of disasters are presented: earthquakes, floods and windstorms. In Columns 5 until 7 interaction terms are introduced in combination with decentralization: GDP per capita, autocracy and democracy.
Table 2: Number of deaths by natural disasters and fiscal decentralization (Pooled OLS model)
(1) OLS (2) OLS (3) OLS (4) OLS (5) OLS (6) OLS (7) OLS
VARIABLES ldeaths injured
(earthquake)
injured (flood)
injured (windstorm)
ldeaths ldeaths ldeaths
decentralization -0.0329* -545.9*** -6.242* -3.264 0.0533 -0.0343 -0.0423** (0.0171) (131.3) (3.437) (11.60) (0.0802) (0.0224) (0.0213) dec_lgdppercapita -0.0113 (0.0103) lhomeless 0.199*** 77.30 -18.88 53.44 0.197*** 0.200*** 0.200*** (0.0644) (461.3) (24.75) (46.24) (0.0643) (0.0646) (0.0644) laffected 0.163*** -3.720 22.81 44.14 0.163*** 0.162*** 0.161*** (0.0560) (539.4) (18.87) (56.88) (0.0560) (0.0562) (0.0562) autoc 0.108 2,288* 10.45 31.64 0.102 0.109 0.0681 (0.110) (1,386) (35.26) (52.59) (0.111) (0.111) (0.124) democ 0.0799 538.3 6.770 67.89 0.0748 0.0765 0.0849 (0.0867) (996.0) (25.33) (44.67) (0.0874) (0.0988) (0.0878) lgdppercapita 0.246 6,971** -31.46 -146.3 0.646 0.240 0.252 (0.290) (3,545) (56.32) (169.8) (0.469) (0.299) (0.300) lpopdensity 1.085*** 5,238 13.18 83.86 1.040*** 1.089*** 1.082*** (0.248) (3,937) (49.28) (105.0) (0.258) (0.256) (0.258) fertility 0.689*** 4,396* -35.51 138.3 0.700*** 0.691*** 0.667*** (0.190) (2,458) (51.05) (105.9) (0.194) (0.196) (0.198) urban 0.0319* 110.0 1.475 17.41* 0.0262 0.0325 0.0307 (0.0190) (326.4) (3.378) (9.498) (0.0202) (0.0199) (0.0199) lland 0.572*** 3,556* 66.35** 16.39 0.602*** 0.567*** 0.614*** (0.159) (1,929) (33.16) (97.06) (0.164) (0.173) (0.173) dec_democ 0.000270 (0.00308) dec_autoc 0.00290 (0.00392) Constant -17.79*** -136,022*** -640.3 -1,863 -20.69*** -17.73*** -18.08*** (3.859) (46,066) (891.7) (1,843) (4.724) (4.037) (3.999) Observations 171 31 31 55 171 171 171 Number of countryid 34 13 18 17 34 34 34
Subsequently, several additional regressions are now estimated including a dummy variable signifying a subtype of disaster. As previously discussed, earthquakes, floods and windstorms are the most common types of natural disasters and are therefore selected as dummy variables. In order to test these regressions, a new dependent variable will be introduced: injured. This is explained as the number of people that obtained an injury due to the effects of natural disasters. Since the dependent variable ldeaths creates an insufficient amount of observations when I add the dummy variables, I have chosen to alter the dependent variable to this new one. In addition, injury will be controlled by the same variables as in the baseline model, with again the focus on the independent variable decentralization. One can perceive in Table 2 that the outcomes of creating the dummy variables show some noteworthy results. In Column 2 the dummy earthquake is introduced. The independent variable decentralization is significant at p<0.01. The absolute value suggests that a 1% increase in share of sub-national expenditures reduces the amount of injured annually by 545 people. As such, decentralization certainly has a positive impact in case of an earthquake. Flood is presented as a dummy variable in Column 3. Here, the results are less obvious: decentralization is significant at p<0.1 and lowers the injuries with 6 per year. Column 4 represents the dummy windstorm, which, opposed to the other two subtypes, has no significance regarding decentralization. Next, the results of the regression including interaction terms are as follows. The log of deaths is now selected again as the dependent variable, as in the baseline model. First, adding the variable dec_lgdppercapita – by multiplying decentralization and (log) gdppercapita – in Column 5 does not result in a significant effect. Therefore, it can be concluded that the effect of decentralization on the amount of deaths is not different for different values of the effect of GDP per capita. The same is true for Columns 6 and 7, where respectively dec_democ and dec_autoc are added: both interaction terms do not indicate statistical significance.
4.2 Robustness checks
To further assess the stability of the outcomes, four alternative regressions are performed for robustness checks. First, the dependent variable will be altered from deaths to homeless. Second, I will test for a non-linear relationship between decentralization and natural disasters. Furthermore, a Tobit regression will be estimated, as opposed to the Pooled OLS estimations I performed before. Lastly, it will be estimated whether an endogenous effect might be present. Beneath, Table 3 indicates the outcomes of the diverse robustness checks. Column 1 replicates the baseline model as provided in Table 2.
Table 3: Robustness checks (Pooled OLS, Tobit model and IV regressions)
(1) OLS (2) OLS (3) OLS (4) TB (5) IV
VARIABLES ldeaths lhomeless ldeaths ldeaths ldeaths
decentralization -0.0329* -0.0272* -0.0259** -0.0118 (0.0171) (0.0144) (0.0120) (0.0182) lhomeless 0.199*** 0.204*** 0.192*** 0.180*** (0.0644) (0.0645) (0.0618) (0.0626) laffected 0.163*** 0.348*** 0.163*** 0.165*** 0.140*** (0.0560) (0.0560) (0.0563) (0.0507) (0.0517) autoc 0.108 -0.132 0.0977 0.104 0.104 (0.110) (0.108) (0.111) (0.0880) (0.0918) democ 0.0799 -0.0692 0.0754 0.0758 0.0484 (0.0867) (0.0855) (0.0881) (0.0696) (0.0744) lgdppercapita 0.246 -0.676*** 0.251 0.226 0.144 (0.290) (0.219) (0.300) (0.189) (0.201) lpopdensity 1.085*** 0.634*** 1.050*** 1.053*** 1.089*** (0.248) (0.169) (0.255) (0.149) (0.167) fertility 0.689*** 0.250 0.682*** 0.686*** 0.645*** (0.190) (0.158) (0.196) (0.131) (0.152) urban 0.0319* 0.0428*** 0.0291 0.0288** 0.0337** (0.0190) (0.0138) (0.0195) (0.0117) (0.0132) lland 0.572*** 0.172 0.501*** 0.555*** 0.482*** (0.159) (0.127) (0.151) (0.110) (0.166) decentdecent -0.000462 (0.000310) Constant -17.79*** 3.786 -16.94*** -17.20*** -15.60*** (3.859) (2.960) (3.928) (2.560) (3.129) Observations 171 189 171 171 159 Number of countryid 34 35 34 34 37
First, I will estimate whether fiscal decentralization would influence the widely accepted definition of measuring disaster events, ‘deaths’, differently if it were to be estimated with a different measure of a disaster event. As a result, homeless will be used as the dependent variable as opposed to deaths. The estimation results are presented in Column 2. They suggest that fiscal decentralization produces similar results as in Column 1: decentralization has statistical significance and a negative coefficient. Therefore, this reveals that similar results are found when testing for an alternative natural disaster statistic as the dependent variable, indicating the robustness of the results of decentralization lowering disaster-induced deaths.
Second, to this point, a linear relationship between fiscal decentralization and natural disasters has been assessed. Following the literature review, one could argue that there might be an optimal mix between national and subnational governments. Both centralization and decentralization governmental styles have positive and negative consequences. A high level of centralization can lead to a higher death toll because the central government cannot respond rapidly, whereas a high level of a decentralized governmental style can as well lead to a higher death toll due to the lack of a central government ruling. Thanks to the non-linear relation characteristic of this notion, an U-shape parabola might prove to be the most superior governmental style, aiming for the perfect mix between centralization and decentralization. Appendix 8 depicts the relationship between decentralization and natural disasters in a scatterplot diagram. Observing the figure, one might be able to interpret an U-shape parabola in there. Consequently, a new independent variable named decentdecent will be introduced, which is decentralization multiplied by decentralization. The estimation results are presented in Column 3. By adding the same control variables as in the baseline model, the results suggest no significant results for the new independent variable decentdecent. As a result, I can dismiss the notion that there might be an effect of an U-shape parabola, and proclaim that there is a linear relationship between decentralization and disasters.
sorts: limit observations (y=0) and non-limit observations (y>0). This leads to the following expression:
y* is a latent variable, and censoring takes place from below with a lower limit of 0. Therefore, there are either no deaths (y=0) or there is a positive number of deaths (y>0). The estimation outcomes of the Tobit regression are presented in Column 4, suggesting nearly identical results as in the Pooled OLS model. As one can perceive, the independent variable decentralization is still statistically significant and has a negative coefficient, again with a value around -0.03. Once more, this result shows the robustness of the relationship between the two main variables. The other six statistically significant variables are homeless, affected, population density, fertility, urban and land, all containing positive coefficients.
5. Conclusion
Subnational and national governments, private companies and non-profit organizations engage in various activities to reduce the impacts of natural disasters. The ability to cope with natural disasters is a central policy issue for all countries in the world. Natural disasters can be tragic and catastrophic, but they offer a good indication of the effectiveness of several institutions. Although the occurrence of a natural disaster will always be unexpected and uncontrollable, a lack of a proper preparation and response in the aftermath of the hazard may result in various lives lost. This thesis offers an assessment of the relationship between natural disasters and fiscal decentralization. Fiscal decentralization is proxied by the percentage of total government expenditures controlled by sub-national governments. The primary focus of this study is to estimate the effect of a governmental fiscal structure towards the mitigation of natural disasters. Bardham (2002) posits that previous research on this topic has been limited in terms of the identification of causal relationships, as well as that there are not enough studies with sufficient cross-country data analyses.
In the empirical analysis it is found that larger levels of decentralization are associated with lower levels of natural disaster-induced deaths. The main outcomes of the variables deaths and decentralization are statistically significant and consistent for the Pooled OLS estimations and the various other models in the robustness checks. Therefore, this thesis shows evidence that decentralized government systems provide superior safety in the context of natural disasters, no matter how I specified the model. This is in line with previous research (Escaleras, 2012).
assistance to increasing decentralization, at least with respect to natural disaster risk management. All in all, this thesis offers a contribution to both the literature of natural disasters and the literature of fiscal decentralization.
References
Albala-Bertrand, J. (1993). Natural disaster situations and growth: A macroeconomic model for sudden disaster impacts. World Development, 21(9), pp.1417-1434.
Alexander, D. (1993). Etiology of the Thera eruption and its effects. Nat Hazards, 7(2), pp.187-189.
Anderson, J.L. and Decker, R. W. (1992). Volcano risk mitigation through training. In: McCall, G. J. H., Laming, D. J. C., Scott, S. C. (eds.). Geohazards: Natural and Man-Made. Chapman and Hall, London, pp. 7–12
Bardhan, P. (2002). Decentralization of Governance and Development. Journal of Economic Perspectives 16 (4): pp.185-205.
Battista, F. and Baas, S. (2004) The role of local institutions in reducing vulnerability to recurrent natural disaster and in sustainable livelihoods development. Rural Institutions and Participation Service, FAO
Brambor, T. (2005). Understanding Interaction Models: Improving Empirical Analyses. Political Analysis, 14(1), pp.63-82.
Burnside, C. and Dollar, D. (2000). Aid, Policies and Growth, American Economic Review, Vol. 90
Dayton-Johnson, J. (2006). Natural Disaster and Vulnerability. OECD Development Centre, pp.1-28.
EMDAT (2014). The OFDA/CRED International Disaster database. Universite Catholique de Louvain, Brussels
Escaleras, M., and Register, C. (2012). Fiscal Decentralization and Natural Hazard Risks. Public Choice 151 (1): 165-83.
Fisman, R., & Gatti, R. (2002). Decentralization and corruption: evidence across countries. Journal of Public Economics, 83, 325–345.
Gencer, E. (2013). The interplay between urban development, vulnerability, and risk management. Berlin: Springer.
Goodspeed, T. (2013). Decentralization and Natural Disasters, CESifo Working Paper Series 4179, CESifo Group Munich.
International Monetary Fund (IMF). (2014). International Financial Statistics Online. [online] Available at: http://ifs.apdi.net/imf / [Accessed 22 April 2014].
Introduction to SAS. UCLA: Statistical Consulting Group.
from http://www.ats.ucla.edu/stat/sas/notes2/ [Accessed November 24, 2014]
Kahn, M. (2005). The Death Toll from Natural Disasters: The Role of Income, Geography, and Institutions. Review of Economics and Statistics 87 (2): 271-84.
La Porta, R., et al. (1999). The quality of government. Journal of Law, Economics, and Organization, 15, 222–279.
Leaning, J. and Guha-Sapir, D. (2013). Natural Disasters, Armed Conflict, and Public Health. New England Journal of Medicine, 369(19), pp.1836-1842.
Mill, J. (1977). De Tocqueville on democracy in America. In J. Robson (Ed.), Collected works. Toronto: University of Toronto Press.
Pantoja, E. (2002). Microfinance and Disaster Risk Management Experiences and Lessons Learned, World Bank’s Disaster Management Facility, UNDP, July.
Persson, T. and Tabellini, G. (1996). Federal Fiscal Constitutions: Risk Sharing and Moral Hazard. Econometrica, 64(3), p.623.
Santos, B.S. (1998). Participatory Budgeting in Porto Alegre: Toward a Redistributive Democracy. Politics and Society 26 (4): 461-510.
Sargan, J. D. (1958). The estimation of economic relationships using instrumental variables. Econometrica 26: 393–415.
Shea, J. S. (1997). Instrument relevance in multivariate linear models: A simple measure. Review of Economics andStatistics 79: 348–352.
Tanzi, V. (1995). Fiscal federalism and decentralization: a review of some efficiency and macroeconomic aspects. In World Bank, annual World Bank conference on development economics 1995. Washington: The World Bank.
Skidmore, M. and Toya, H. (2013). Replication of 'Economic Development and the Impacts of Natural Disasters' (Economics Letters, 2007). Public Finance Review.
Unisdr.org. (2014). UNISDR. [online] Available at: http://www.unisdr.org/ [Accessed 7 May 2014].
Wildasin, David. (2008). Disaster Policies: Some Implications for Public Finance in the U.S. Federation. Public Finance Review 36 (4): 497-518.
World Bank. (1994). World Development Report on Infrastructure. New York: Oxford University Press.
World Bank. (2010). Disaster risk management in Latin America and the Caribbean region: GFDRR. Global Facility for Disaster Reduction and Recovery (GFDRR)
Appendices
Appendix 1: General classifications of natural disasters
Disaster subtype Definition Disaster Main Type
Biological Disaster caused by the exposure of living organisms to germs and toxic substances
Epidemic, Insect infestation, Animal Stampede
Climatological Events caused by long-lived/meso to macro scale processes (in the spectrum from intra-seasonal to multi-decadal climate variability)
Extreme Temperature, Drought, Wildfire
Geophysical Events originating from solid earth Earthquake,Volcano, Mass Movement (dry)
Hydrological Events caused by deviations in the normal water cycle and/or overflow of bodies of water caused by wind set-up
Flood, Mass Movement (wet)
Meteorological Events caused by short-lived/small to meso scale atmospheric processes (in the spectrum from minutes to days)
Storm
Appendix 2: Natural disasters by type (1983 – 2012)
Appendix 3: Definition of variables and expected effects on disaster-induced deaths
Variables Definition Expected effects on deaths
Deaths Number of deaths from natural disasters (confirmed as dead and persons missing and presumed dead)
---
Decentralization Share of sub-national (state and local) expenditures (% of total expenditures)
Negative (Escaleras 2012)
Homeless Number of people that became homeless due to natural disasters
No estimated effect to date
Affected Number of people that were affected by natural disasters
No estimated effect to date
Autocracy (scaled from 0 – 12)
Higher values indicate more thoroughgoing autocratic institutions
No estimated effect to date
Democracy (scaled from 0-12)
Higher values indicate more thoroughgoing democratic institutions
Positive (Escaleras 2012)
GDP per capita Gross Domestic Product divided by midyear population
Negative (Kahn 2005)
Population density The midyear population divided by land area in square kilometres.
Negative (Kahn 2005)
Fertility Births per woman if she were to live to the end of her childbearing years
No estimated effect to date
Urban People living in urban areas as defined by national statistical offices
Positive for earthquakes, negative for others (Skidmore, 2013)
Appendix 4: List of countries (N = 46) Albania Italy Argentina Kenya Australia Lithuania Austria Malaysia Azerbaijan Mexico Belarus Netherlands Belgium Nicaragua Bolivia Norway Brazil Panama Bulgaria Paraguay Canada Peru Chile Philippines Colombia Poland
Costa Rica Romania
Croatia Slovenia
Denmark South Africa
France Spain
Germany Sri Lanka
Hungary Switzerland
India Thailand
Indonesia United Kingdom
Ireland United States
Appendix 6: Summary subtypes of disasters
Total 1,687 100.00
