ASSESSMENT OF THE ECONOMIC IMPACT OF INVESTMENTS IN DISASTER RISK REDUCTION AND PREVENTION: THE CASE OF TABASCO Oscar A. Ishizawa, Juan José Miranda and Itzel de Haro Contributions: Xijie Lv (World Bank), Adrián Pedrozo (UNAM) and CONAGUA officials Funded by Authors: Oscar A. Ishizawa, Juan José Miranda and Itzel de Haro Contributions: Xijie Lv (World Bank), Adrián Pedrozo (UNAM) and CONAGUA officials © 2017 The World Bank. 1818 H St. NW Washington, DC, 20433 USA Telephone number: 202-473-1000 Web site: www.worldbank.org This work has been carried out by World Bank staff, with external contributions. The opinions, interpretations, and conclusions expressed herein do not necessarily reflect the views of the World Bank, its Board of Executive Directors or the countries represented by it. The World Bank does not guarantee the accuracy of the data contained in this publication. The boundaries, colors, names, and any other information shown in the maps of this document do not imply the expression of any opinion whatsoever on the part of the World Bank concerning the legal status of any territories, nor the endorsement or acceptance of such boundaries. Rights and Permissions The material contained in this work is copyrighted. The World Bank encourages the dissemination of its knowledge, and authorizes the partial or total reproduction of this report for non-profit purposes, provided the source is cited. All queries on rights and licenses, including subsidiary rights, should be addressed to: World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax number: 202-522-2625; e-mail address: pubrights@worldbank.org. 3 EXECUTIVE SUMMARY............................................................6 1. INTRODUCTION.........................................................................8 2. HYDROMETEOROLOGICAL CHARACTERIZATION OF THE 2007 AND 2010 EVENTS........................................... 12 2.1. ESTIMATED DAMAGES AND LOSSES: 2007 AND 2010 EVENTS ................................................................................ 18 2.2. COMPREHENSIVE WATER PLAN OF TABASCO (PHIT)...... 19 3. DATABASES................................................................................20 3.1. SCHOOL DROPOUT RATE........................................................21 3.2. NIGHTTIME LIGHTS (NTL)...................................................... 24 3.3. FLUVIAL FLOODING MODEL (GLOFRIS)........................... 26 3.4. INVESTMENTS IN PROTECTION INFRASTRUCTURE ........................................................................... 27 4. METHODOLOGY .......................................................................30 4.1. METHODOLOGY FOR THE ESTIMATION OF THE EFFECT ON SCHOOL DROPOUT ................................. 34 4.2. METHODOLOGY FOR THE ESTIMATION OF THE IMPACT ON NIGHTTIME LIGHTS................................. 35 5. RESULTS .....................................................................................36 5.1. RESULTS RELATED TO SCHOOL DROPOUT ......................................................................... 37 5.2. RESULTS RELATED TO NIGHTTIME LIGHTS ...................40 TABLE OF 6. CONCLUSIONS..........................................................................44 CONTENTS BIBLIOGRAPHY ........................................................................48 ANNEXES................................................................................... 50 5 EXECUTIVE SUMMARY This study assesses the role of the investments made For that purpose, estimations were made of the impact in disaster risk reduction (DRR) and prevention in the on the school dropout rate for Basic Education students actual reduction of the impact of floods in the state of (primary and secondary levels), and of the effect on Tabasco between 2007 and 2010. the economic activity of the state (measured through changes in the intensity of nighttime lights [or NTL]). This, Based on a previous study conducted within the together with georeferenced information on the location framework of the Technical Assistance Agreement of the investments in DRR and the flood level present between the Mexican Government and the World Bank, during each of the events—estimated through a fluvial it was found that the intense rainfall events occurring flooding model called GLOFRIS (GLObal Flood Risks in Tabasco in 2007 and 2010 were similar in terms of with IMAGE Scenarios)—, made it possible to obtain their hydrometeorological characteristics (rainfall) and estimates with a spatial disaggregation level of 1 km2. spatial distribution (rainfall fields). When analyzing the accumulated rainfall recorded for both events, it was On the one hand, the results show that the fluvial found that the magnitude of the 2010 floods was greater flooding level (GLOFRIS) indeed has a negative effect than that of the 2007 floods. However, damages and on the level of school dropout and on the changes in losses estimations recorded by the National Center NTL. Specifically, a higher negative impact is observed for Disaster Prevention (CENAPRED) show an 80% in 2007 compared to 2010—in the latter year, the reduction in 2010, compared to 2007 (World Bank, 2014). estimation is controlled for the presence of investments The similarity between these two events provides an in protection infrastructure. In 2007, it was found that ideal “natural experiment” scenario to evaluate the one additional decimeter of flood (1 dm) generated an benefits of the investments in DRR made between increase in the school dropout rate of almost 0.1%, while 2008 and 2010. the impact in 2010 barely amounted to 0.023%. Likewise, NTL show a decrease of 30% in light intensity after the Motivated by the devastating floods that occurred 2007 floods, compared to 5.4% in 2010. in October 2007, the Mexican Federal Government, together with the Government of the State of Tabasco, On the other hand, no robust results were found for the designed the Comprehensive Water Plan of Tabasco effect of the investments in DRR, even after correcting (PHIT, Plan Hídrico Integral de Tabasco) and began its a possible bias resulting from their location. Therefore, implementation. The aim of the PHIT was to come up our estimations may lack accuracy due to the lack with a set of solutions that guaranteed the population’s of information concerning the exact extent of the safety, as well as the performance and continuity of investments, as well as because of problems related economic activities and the balance of ecosystems to the time frame of school dropout data available and in the occurrence of floods. The PHIT provides for the lack of NTL information for the months before the the implementation of structural measures (physical 2010 floods. investments, such as embankments and reinforcements) and non-structural measures (non-engineering In conclusion, the results obtained in this study measures, such as workshops, trainings, development of reinforce those found in the previous one (World Bank, early warning systems and drawing of risk maps, among 2014): the impact of the 2007 floods was greater than others). As part of the PHIT, a total of USD 753,613,871 that of the 2010 floods, even though the latter year was (MXN 9,518 million)1 were invested in DRR measures characterized by rainfall events of a greater magnitude. between 2008 and 2010, which amounts to 2.18% of As regards investments in DRR, the assessment of the state’s GDP in 2010. About 84% of these resources their impact would require more accurate spatial and were invested in infrastructure for the protection of the time frame data. population and productive areas (amounting to 56% of the overall investment), as well as in the reconstruction of protective hydraulic infrastructure (approximately 27% of the overall investment) (World Bank, 2014). The aim of this analysis is to assess the impact caused by the floods and the investments in DRR on the socioeconomic wellbeing of the population of Tabasco. 1 For the purpose of this document, all figures in United States dollars (USD) have been calculated at the average exchange rate for 2010 according to the Banco de México, which equals to 12.62981 Mexican pesos (MXN) per US dollar. (Source: http://www.banxico.org.mx/ SieInternet/consultarDirectorioInternetAction.do?accion=consultarCuadro&idCuadro=CF373&locale=en, checked on August 1st 2017). 7 1 INTRODUCTION Consisting of 17 municipalities and 1,456 localities, the After the devastating floods that occurred in October state of Tabasco is situated in Southeastern Mexico. and November 2007, the Mexican Federal Government, According to the National Institute of Statistics and together with the Government of the State of Tabasco, Geography (INEGI), 42.6% of the population lives in decided to dramatically increase the investments in rural areas2. Out of the economically active population, DRR at the state through the implementation of the 19.65% work in the primary sector, while the largest Comprehensive Water Plan of Tabasco (PHIT). Given proportion of the population works in the secondary that most of the investments in DRR were made sector and accounts for 59% of the state’s GDP (mostly between 2008 and 2010, and that the floods that took coming from oil production). place on September 2010 had similar characteristics to those of the 2007 floods —both in hydrological terms Due to its location and hydrological characteristics, and in terms of flooded areas—Tabasco offers a unique Tabasco constantly sustains floods. Only between opportunity to analyze the impact of the investments in 2003 and 2013, 33 hydrological disasters, such as DRR on different socioeconomic variables. torrential rains, floods, and tropical storms, hit the state (CENAPRED, 2013). The purpose of this analysis is to Floods occurring in 2007 affected slightly more than assess the impact of both floods and investments in 75% of the state’s population (approximately 1.5 million disaster risk reduction (DRR) and prevention on the people), 73% of the road infrastructure (6.500 km), socioeconomic wellbeing of the population of Tabasco almost 570,000 hectares of agricultural land, and with the greatest degree of disaggregation possible. 123,000 households (CENAPRED, 2013). As a result, the In order to measure the impact on the population, 17 municipalities that make up Tabasco declared a state this study takes into consideration floods occurred of emergency (DOF, 2007). According to calculations in Tabasco in 2007 and 2010, given the fact that they made by the National Center for Disaster Prevention are similar events, and that a great proportion of the (CENAPRED), damages and losses caused by the investments in DRR were made after the 2007 floods 2007 floods were estimated above USD 2,517,852,604 and before the rainy season of 2010. (MXN 31.8 billion3), amount only exceeded by the 2 The INEGI defines as “rural” all those localities with less than 2,500 inhabitants. 9 earthquake that hit Mexico City in 1985 and by the Comprehensive Water Plan of Tabasco (PHIT), Then, hurricanes Wilma and Stan in 2005 (ECLAC, 2008). a description of the databases (Chapter 3), and the methodology used (Chapter 4) are introduced. Lastly, After the 2007 floods, the Federal Government allocated Chapter 5 shows the results obtained regarding the USD 752,188,671 (MXN 9.5 billion4) to investments in risk impact of floods and investments in DRR on the school reduction and prevention, corresponding to 2.18% of the dropout rate and on the economic activity (measured GDP of Tabasco in 2010. About 84% of these resources through the changes in NTL). were invested in infrastructure for the protection of the population and productive areas. Additionally, the Mexican Fund for Natural Disasters (FONDEN) allocated USD 950,133,058 (MXN 12 billion), equivalent to 3.13% of the 2010 GDP of Tabasco, to the reconstruction of public infrastructure, such as federal and state roads, schools, and hospitals (World Bank, 2014). This study specifically aims at analyzing the impact of floods and investments in DRR on two social aspects: (i) the school dropout rate, and (ii) the economic activity of the state, for which changes in Nighttime Lights (NTL) recorded before and after the event are used as proxy. In order to estimate the accumulated annual rainfall in flooded areas of Tabasco, the fluvial flooding global model developed by Deltares (Global Flood Risk with IMAGE Scenarios, GLOFRIS) was used, which makes it possible to obtain information with a resolution of 1 km2. This, added to monthly NTL data obtained from the National Oceanic and Atmospheric Administration (NOAA), allowed for the estimation of the impact of floods on the local economy based on evidence that suggests that NTL are a good predictor of the economic activity (Henderson et al., 2011, 2012; De Janvry et al., 2016; Bundervoet et al., 2015). Additionally, to estimate the impact on the school dropout levels of the state, administrative information regarding the number of enrolled students and the number of students who finished school, gathered by the Mexican Ministry of Public Education (SEP), was used. So as to estimate the effect of the investments in DRR made between 2008 and 2010, the estimated effect of the 2007 floods in the affected areas was compared to the impact on the protected and flooded areas in 2010. After controlling for various factors, such as distance from the investment and flood level, the difference in the estimated effects was attributed to the implementation of investments in DRR. The information in this document is arranged as follows: Chapter 2 includes a hydrometeorological analysis of the floods that occurred between 2007 and 2011, with the purpose of justifying why the 2007 and 2010 events are comparable both in extent and in magnitude. This chapter also includes a description of 3 Constant pesos as of 2010. 4 Constant pesos as of 2010. 10 |ASSESSMENT OF THE ECONOMIC IMPACT OF INVESTMENTS IN DISASTER RISK REDUCTION AND PREVENTION: THE CASE OF TABASCO 11 2 HYDROMETEOROLOGICAL CHARACTERIZATION OF THE 2007 AND 2010 EVENTS In a previous study conducted within the framework of a 2007 cycle. The results are presented graphically in Technical Assistance Agreement between the Mexican Figure 2, where the three panels show the results Government and the World Bank, it was found that throughout the year for the three hydrological areas. The the 2007 and 2010 events were similar in terms of their graphs show that, for the three regions under study, the hydrometeorological characteristics (rainfall) and spatial year 2009 had less rainfall compared to the year 2007. In distribution (rainfall fields). A summary of the analysis contrast, in 2008 there was a higher amount of rainfall conducted in that study is presented below. For further compared to the amount observed in 2007. information, refer to the World Bank publication (2014). In 2010, the three regions had an increased volume of Tabasco consists of three hydrological regions. The rainfall compared to 2007, especially at the beginning first one comprises the eastern part of the state and of the season. Despite this, the final value of the covers the Tonalá river area on the border with Veracruz. accumulated rainfall difference between 2010 and 2007 The second covers the area of Bajo Grijalva, located at was less than 200 mm of rain in all three cases. This the central area of the state. Lastly, the third region is suggests that, although in 2010 rainfall volume was located at the western part of the state, and covers higher than in 2007, its levels were in the same range the Usumacinta river area. Figure 1 shows the location statewide, unlike in 2008 and 2009. of these three regions, as well as the capital city of Tabasco (Villahermosa) and the most important dam of As regards the results for 2011, it was found that the the state, Peñitas. accumulated rainfall for such year was more similar to the one recorded in 2007. However, it is worth mentioning The hydrometeorological analysis of the three that its behavior was different across the different hydrological regions consisted in comparing the regions. While in the regions of Bajo Grijalva and Tonalá accumulated rainfall curves (mass curve) during the there was a higher level of accumulated rainfall in rainy season (June-December) for the period 2007‑2011. 2011 compared to 2007, the area surrounding the river For this purpose, the mass curves for every year were Usumacinta showed a decrease in rainfall volume. In view first estimated, and then, the mass curve for the year of the above, it is concluded that in the events of 2010 2007 was subtracted from them. This operation made and 2011 there was a smaller variation for each of the it possible to identify which of the cycles was more regions as regards the accumulated rainfall of 2007, in similar—from the hydrological point of view—to the comparison to the mass curves for 2008 and 2009. 13 FIGURE 1 HYDROLOGICAL SUBREGIONS OF THE STATE OF TABASCO Source: World Bank, 2014 Moreover, when comparing the areas that concentrated Using meteorological information on daily rainfall from the largest amount of rainfall during the events of 2007, the CLICOM (CLImate COMputing Project) System of 2010 and 2011, it was found that the 2007 and 2010 the Center for Scientific Research and Higher Education events show a significant overlapping in the central area of Ensenada (CICESE) in Mexico5, it was possible to of the state, which concentrates approximately 80% estimate the accumulated rainfall for 2007 and 2010. As of the population and is where the city of Villahermosa shown in Figure 4, the accumulated flood level in 2010 is located (see Figure 3), whereas the 2011 floods were was higher than that of 2007 and, in both years, it was mainly concentrated in the southeastern part of the higher than the historical mean (1990-2006). The same state, therefore such event would not be comparable to behavior was observed in 10 of the 17 municipalities that the 2007 floods. make up the state of Tabasco (see Annex 1). In short, when comparing the amount of rainfall and the The same conclusion was reached in the previous spatial distribution of the rainfall events over the three study when analyzing the accumulated rainfall curves hydrological regions that make up the state of Tabasco, (mass curves) of the runoff of rivers Carrizal, Samaria, it was found that the 2010 floods were similar to the ones Grijalva and La Sierra, which are adjacent to the city in 2007. of Villahermosa, as well as the variation in the water storage level (baseflow6) in the basins of the rivers In order to specify the magnitude and intensity of the Usumacinta and La Sierra. In all cases, it was found that 2007 and 2010 floods, the accumulated rainfall, as well the mass curves and the baseflow of the basins were as the recorded flood level, was analyzed in both events. higher in 2010 than in 2007 (World Bank, 2014). Also, 5 The information used was obtained from 18 weather stations located in the state of Tabasco for which there was complete information for the 2007-2011 period. 6 The baseflow is estimated as the daily mean variation of the water level, which is measured in millimeters (mm). 14 |ASSESSMENT OF THE ECONOMIC IMPACT OF INVESTMENTS IN DISASTER RISK REDUCTION AND PREVENTION: THE CASE OF TABASCO FIGURE 2 ACCUMULATED RAINFALL DIFFERENCES FOR THE REGIONS OF TONALÁ, BAJO GRIJALVA AND USUMACINTA Western area: Tonalá 1,000 800 600 400 200 0 -200 2008-2007 2009-2007 -400 2010-2007 -600 2011-2007 0 20 40 60 80 100 120 140 160 180 Central area: Grijalva 1,000 800 600 400 200 0 -200 -400 2008-2007 2009-2007 -600 2010-2007 -800 2011-2007 0 20 40 60 80 100 120 140 160 180 Eastern area: Usumacinta 400 200 0 -200 -400 2008-2007 2009-2007 -600 2010-2007 -800 2011-2007 0 20 40 60 80 100 120 140 160 180 Source: World Bank, 2014 15 FIGURE 3 RAINFALL FIELDS COMPARISON FOR THE YEARS 2007, 2010 AND 2011 Source: World Bank, 2014 FIGURE 4 ACCUMULATED RAINFALL IN THE STATE AND MUNICIPAL when studying the accumulated rainfall recorded in DEVIATION FROM THE HISTORICAL MEAN the area of Bajo Grijalva during 2007 and 2010, it was possible to conclude that the rainfall volume in the 2010 event was a 17% higher than that of the 2007 2,321 floods (see Figure 5). However, when analyzing the extension of the flooded areas (see Figure 6), it was found that the 2007 floods 2,151 had a greater impact than the 2010 event, both in terms of accumulated water and affected area. According to 2,060 the results of the previous study, the reduction in the affected areas in 2010 is attributed to the investments in risk reduction and prevention made between 2007 and 2010. 1990-2006 2007 2010 Source: Prepared by the authors based on information from the Mexican National Weather Service (SMN) 16 |ASSESSMENT OF THE ECONOMIC IMPACT OF INVESTMENTS IN DISASTER RISK REDUCTION AND PREVENTION: THE CASE OF TABASCO FIGURE 5 COMPARISON OF 2010 AND 2007 ACCUMULATED RAINFALL IN THE AREA OF BAJO GRIJALVA 2,500 y = 1.0965x + 65.317 R² = 0.94363 2,000 2010 rainfall (mm) 1,500 1,000 500 0 0 500 1,000 1,500 2,000 2,500 2007 rainfall (mm) Source: World Bank, 2014 FIGURE 6 FLOODED AREAS IN 2007 AND 2010 November 2007 September 2010 Source: World Bank, 2014 17 2.1. ESTIMATED DAMAGES AND LOSSES: 2007 AND 2010 EVENTS As already mentioned, when comparing the 2007 and followed by Huimanguillo and Cárdenas (see Annex 2). 2010 floods, it was found that the rainfall level recorded Such decrease coincides with the fact that most of the in 2010 was higher than in 2007. However, the 2007 floods investments in DRR made in Tabasco are located in that had a greater impact in terms of accumulated water area (see section 3.4). level, as well as in flood extent. As part of the previous study, an analysis of the damages and losses reported by Also, the previous study estimated that in 2007 CENAPRED during the 2007 and 2010 events was carried damages and losses amounted to USD 2,902,577,315 out, and it was found that the damages (in heritage and (MXN 36,659 million7), while in 2010 they came to only assets) and the losses (in goods and services' production USD 585,281,964 (MXN 7,392 million) (CENAPRED, workflows) sustained in 2007 were almost four times 2007, 2011). This means the impact was reduced in greater than those reported in 2010. USD 2,317,295,351 (MXN 29,267 million)—80% less damages and losses—, equivalent to the 7% of the GDP According to the statistics from CENAPRED, the of Tabasco in 2010. Such decrease is mainly translated number of people affected as a result of the 2007 into a decrease in the recorded losses in the primary floods was higher than the one reported in 2010 for sector and in the damages sustained by other productive all the municipalities of Tabasco, except for Balacán. sectors8 (see Figure 7). In particular, it was found that the greatest decrease in the number of people affected occurred in the municipality of Centro—the most densely populated municipality, and also where Villahermosa is located— FIGURE 7 TOTAL ESTIMATED DAMAGES AND LOSSES IN 2007 AND 2010 (IN MILLIONS OF MXN AS OF 2010) 40,000 35,000 30,000 25,000 20,000 15,000 10,000 5,000 0 2007 2010 2007 2010 2007 2010 2007 2010 2007 2010 2007 2010 2007 2010 Primary sector Other productive Social sector Infrastructure Emergency Productive General total sectors response sectors total Source: World Bank, 2014 Losses Damages 7 Constant pesos as of 2010. 8 The category other productive sectors includes Commerce, Manufacture, Construction, Services, and Restaurants. 18 |ASSESSMENT OF THE ECONOMIC IMPACT OF INVESTMENTS IN DISASTER RISK REDUCTION AND PREVENTION: THE CASE OF TABASCO 2.2. COMPREHENSIVE WATER PLAN OF TABASCO (PHIT) The 2007 flood event in Tabasco affected more than Between 2008 and 2010, the Federal Government one million people, and the estimated economic losses invested a total amount of USD 753,613,871 in the state amounted to approximately USD 3 billion (MXN 9,518 million) in DRR measures in Tabasco, which (CENAPRED, 2012). This makes the 2007 floods the is equivalent to 2.18% of the state's GDP in 2010. About greatest disaster over the past fifty years in Mexico, both 84% of these resources were invested in infrastructure in extent and in affected population (CENAPRED, 2008; for the protection of the population and productive CONAGUA, 2012). As a result of the losses and damages areas (amounting to 56% of the overall investment), caused by such flooding, the National Water Commission as well as in the reconstruction of protective hydraulic (CONAGUA), the National Autonomous University of infrastructure (approximately 27% of the overall Mexico (UNAM) and the Government of the State of investment) (World Bank, 2014). In turn, only 6.74% of the Tabasco developed the Comprehensive Water Plan of total investment was allocated to the PHIT and to other Tabasco (PHIT) (CONAGUA, 2012). comprehensive water plans, which equals USD 50,832,119 (MXN 642 million). The remaining 11% was allocated The PHIT was launched in 2008 with the objective of to the construction of tunnels for the Grijalva river, to generating a set of solutions that would guarantee the compensatory payments for the regularization of plots of population’s safety, the performance and continuity of land, to the restoration of tourism-related infrastructure, economic activities and the balance of the ecosystems and to the rehabilitation of the basin Frontera Sur as well in the occurrence of this type of events. As a result, a as of several rural paths and roads (see Figure 8)9. series of structural measures (physical investments, such as embankments and reinforcements) and non- According to the white paper on the PHIT, between structural measures (non-engineering measures, such the 2007 and the 2010 floods, 67 investments were as workshops, trainings, development of early warning made, including the construction, reconstruction and systems, drawing of risk maps, among others) were reinforcement of protection embankments, walls and implemented, grouped into urgent and medium-term riverbank protections, desilting of several carrier drains, actions, so as to reduce flood risk mainly in the urban and construction of control structures and traffic area of Villahermosa. One of the most remarkable bridges, among others (see Annex 9). Section 3.4 offers outcomes of the PHIT was the implementation of more detailed information on these investments. territorial management and risk reduction measures, aimed at preventing the establishment of human settlements in high-risk areas (CONAGUA, 2012). FIGURE 8 OVERALL FEDERAL INVESTMENT IN DRR MEASURES BETWEEN 2008 AND 2010 (IN MILLIONS OF MXN AS OF 2010) Rehabilitation and retrofitting in the irrigation districts of the basin Frontera Sur Reconstruction of roads and rural paths infrastructure Compensatory payments for the regularization of plots of land in hydraulic works Restoration of tourism-related infrastructure in Tabasco Comprehensive Water Plan of Tabasco Comprehensive water plans Diversion tunnels for the river Grijalva Reconstruction of protective hydraulic infrastructure Protection infrastructure for population centers and productive areas Source: World Bank, 2014 0 1,000 2,000 3,000 4,000 5,000 6,000 9 It is worth mentioning that it was not possible to obtain information about investments in DRR measures at state level; therefore, the mentioned amounts correspond to a percentage of the overall amount invested in DRR measures. 19 3 DATABASES 3.1. SCHOOL DROPOUT RATE The school dropout rate in the Basic Education system Where Asgt is the number of students at the school s and (primary and secondary levels) was estimated on the in the educational program g at the beginning of the basis of administrative information from the education academic year t, while Asgt+1 is the number of students sector. The Mexican Ministry of Public Education that still attend school at that education level by the end (SEP) keeps a record, at school level, of the number of of the academic year t+1. enrolled students at the beginning of each academic year, as well as of the number of students who actually It is important to note that the 911 Survey only contains complete it. This information is obtained though the 911 statistical information on the number of students Survey, conducted for each school building registered enrolled by gender, age and school grade, so there is no with the SEP (both in public and private institutions) at further information about the specific characteristics of the beginning and at the end of each academic year. each student. However, it was possible to identify certain According to the academic calendar established by the characteristics at school level, such as funding (private SEP, the academic year starts in August and finishes in or public) and the characteristics of the locality where it June of the following year. is situated (urban or rural)10. Based on the information from the surveys conducted The Mexican Basic Education system is divided into at the beginning and at the end of the 2007-2008, preschool, primary and secondary education. This 2008‑2009, 2009-2010 and 2010-2011 academic years, the analysis is only focused on the primary and secondary school dropout rate was estimated taking into account education levels, comprising children from 6 to 15 years the number of students at the beginning and at the end of age. Basic Education students are taken as a of those periods: reference because they are the most likely to abandon Asgt + 1 school in case of a natural disaster. Since the minimum School dropout ratet = Asgt working age in Mexico is 15 years of age, estimations 10 According to the INEGI, a locality is considered rural if it has a population of less than 2,500 inhabitants. 21 of the impact based on the school dropout rate of schools (SEP, 2009). In this regard, the present study only High School Education students may significantly bias includes data from general secondary education and the results. from the three modalities of primary education. In primary education, there are several modalities: (i) Along with public information about the location of general, (ii) communal, and (iii) indigenous. The most schools in the country provided by the National School common of these three is general primary education, Information System (SNIE) of the SEP, it was possible to which serves the majority of the population, whereas obtain the coordinates of the schools for which there was communal and indigenous modalities serve specific information available from the 911 Survey. Overall, there TABLE 1 NUMBER OF SCHOOLS BY ACADEMIC YEAR AND EDUCATION PROGRAM School Primary education Secondary Total period General Communal Indigenous Total education 2007-2008 1,830 207 101 2,138 725 2,863 2008-2009 1,833 207 101 2,141 728 2,869 2009-2010 1,828 208 101 2,137 742 2,879 2010-2011 1,828 199 101 2,128 742 2,870 Source: Prepared by the authors based on information from the SEP sectors. Community education serves the “population was information available for 2,896 schools located in the that lives in small isolated and disperse localities of the state of Tabasco, 2,201 of which were primary schools, and national territory” (SEP, 2009), while primary indigenous 695, secondary schools. Table 1 shows in greater detail education is a bilingual inter-cultural early education the number of schools for which there was information model for indigenous children. available by analyzed year and education program. In turn, secondary education also has different As regards the schools’ characteristics, it was found that modalities: (i) general, (ii) technical, and (iii) tele- approximately 74.5% of primary schools are urban (see secondary. General secondary education is aimed at Table 2). Out of the communal and indigenous primary developing the necessary skills to pursue higher studies, schools registered in Tabasco with the SEP, 98% are whereas technical education intends to do the same located in rural communities, which is in line with their while providing basic tools that enable students to enter education modality. Also, it can be noted that the SEP the labor market after finishing this level of education. record is mostly comprised of public schools: only 5.3% In turn, tele-secondary is an education system through of primary and secondary schools are private. television aimed at serving the adolescent population who lives in disperse communities and lack the Figure 9 shows the average dropout rates in the state of possibility of attending general or technical secondary Tabasco by education program. As it can be observed, TABLE 2 NUMBER OF PRIVATE AND PUBLIC SCHOOLS WITH KNOWN LOCATION, AND SCHOOLS IN RURAL AND URBAN SETTINGS BY EDUCATION PROGRAM Characteristics Primary education Secondary Total of the school General Communal Indigenous Total education Urban 1,041 2 4 1,047 469 1,516 Rural 2,538 329 193 3,060 982 4,042 Private 177 0 0 177 115 292 Public 3,402 331 197 3,930 1,336 5,266 Source: Prepared by the authors based on information from the SEP 22 |ASSESSMENT OF THE ECONOMIC IMPACT OF INVESTMENTS IN DISASTER RISK REDUCTION AND PREVENTION: THE CASE OF TABASCO the school dropout rate for all education systems in FIGURE 9 SCHOOL DROPOUT RATE IN THE STATE OF TABASCO the state was higher in 2007 than in 2010: the average IN 2007 AND 2010 dropout rate in the state for primary and secondary education in 2007 was 5%, whereas in 2010 it was 4.2%. Taking a closer look at the school dropout rate for 10.9% primary and secondary education at municipality level 9.1% (see Figure 10), it can be found that, except for Jalpa de 6.0% Méndez and Centro, all the municipalities of Tabasco 5.0% 5.4% 4.1% 4.2% had higher rates in 2007 than in 2010. Specifically, it 3.4% 2.5% was estimated that the dropout rate in the city of 1.6% Villahermosa was of 5.2% in 2007 and of 4.5% in 2010, which represents a 0.8% decrease. 2007 2010 Rural General Indigenous Secondary Primary and communal primary primary education secondary primary education education education education Source: Prepared by the authors based on information from the 911 Survey FIGURE 10 MUNICIPAL SCHOOL DROPOUT RATE IN 2007 AND 2010 Teapa Emiliano Zapata Huimanguillo Balancán Cárdenas Tenosique Jonuta Cunduacán Centro Comalcalco Jalapa Tacotalpa Centla 2007 Macuspana Paraíso 2010 Nacajuca Jalpa de Méndez 0% 2% 4% 6% 8% 10% Source: Prepared by the authors based on information from the SEP 23 3.2. NIGHTTIME LIGHTS (NTL) Based on satellite images obtained by the U.S. Air Force Corral et al., 2016). In particular, the research conducted Defense Meteorological Satellite Program (DMSP), it by De Janvry et al. (2016) stands out, which assessed is possible to estimate the intensity level of the light the economic impact of the Mexican Fund for Natural emitted by human settlements. Nighttime Lights (NTL) Disasters (FONDEN) on the municipalities and found offer significant advantages in the assessment of policies that NTL make it possible to estimate the level of local and projects. According to Donaldson and Storeygard economic activity in Mexico both from a time frame11 and (2016), some of these advantages are the following: (i) they a spatial perspective. make it possible to estimate data which is either difficult to obtain or nonexistent, (ii) they have a higher spatial In the particular case of Tabasco, the use of NTL resolution compared to traditional indicators, and (iii) they enables to assess the economic impact of floods and have a wide geographical coverage. investments in DRR through the change in light intensity before and after the event. For this study, satellite Seeking to exploit such advantages, an increasing images taken by satellites F16 and F18 during the period number of studies have found that NTL not only enable ranging from January 2004 to December 2012 were to capture the presence of human settlements around used12. Satellites capture images of each and every the world, but also the level of supranational, national, place of the planet twice a day between 8.30 p.m. and and local economic activity. For example, using data 10.00 p.m. (local time). Then, these images are processed for 188 countries, Henderson et al. (2012), show that by scientists at the NOAA in order to filter out those NTL are a reliable predictor of economic growth for the pixels obscured by the presence of clouds, as well as period 1992-2008. Other studies have found a strong other sources of transient lights (i.e., Northern lights, link between NTL and economic activity at subnational forest fires and moonlight). The purpose of this is to and local level (Harari and La Ferrara, 2013; Hodler and produce information about the intensity level of the light Raschky, 2014; Elliott et al., 2015; Alesina et al., 2016; generated by human activities alone13. TABLE 3 NTL DATA FOR TABASCO, 2004-2012 Digital Number (DN) 2004 2005 2006 2007 2008 2009 2010 2011 2012 0 39.8% 34.2% 36.6% 37.7% 35.7% 44.7% 36.3% 39.8% 38.0% (0, 3) 0.6% 1.8% 4.4% 1.3% 1.5% 0.3% 0.1% 0.0% 0.0% [3, 6) 23.3% 27.2% 24.8% 27.0% 28.5% 17.7% 23.9% 21.7% 22.4% [6, 11) 18.3% 16.9% 15.7% 15.7% 15.9% 16.7% 20.0% 20.0% 20.5% [11, 21) 9.8% 10.9% 10.0% 10.3% 10.2% 10.5% 10.2% 10.3% 10.3% [21, 63) 7.9% 8.6% 8.0% 7.7% 7.8% 9.6% 9.1% 7.9% 8.3% 63 0.3% 0.4% 0.5% 0.4% 0.3% 0.7% 0.5% 0.4% 0.6% Average DN 7.00 7.51 7.03 6.96 7.01 7.56 7.83 7.19 7.53 Average DN (excl. 0) 11.62 11.41 11.09 11.17 10.90 13.66 12.29 11.94 12.14 Amount of pixels with value 354,338 354,272 351,668 350,341 344,338 199,158 297,215 298,480 297,873 Source: Prepared by the authors based on data supplied by the NOAA 11 De Janvry et al. (2016) apply the approach used by Henderson et al. (2012) to determine if the fluctuations of nighttime lights in Mexico are a good predictor of the local economic activity from a time frame perspective. Likewise, from a spatial perspective, they verified that NTL are strongly correlated to the economic characteristics of households and municipalities. 12 Each satellite has its own operating cycle: satellite F16 was operative from 2004 to 2009; and satellite F18, from 2010 to 2013. Thus, the selected sample covers the period under analysis in this study, which has been determined by the events that occurred in 2007 and 2010 in the state of Tabasco. 13 For more details about the filtering process, see Elvidge et al. (1997). 24 |ASSESSMENT OF THE ECONOMIC IMPACT OF INVESTMENTS IN DISASTER RISK REDUCTION AND PREVENTION: THE CASE OF TABASCO The information obtained from the NOAA has a spatial Although NTL facilitated by the NOAA provide an resolution of 30 arc seconds per pixel (about 1 km2 at innovative way of assessing the economic impact of the equator). Every pixel measures NTL intensity, which floods and investments in DRR in the state of Tabasco, is standardized for all satellites by means of a digital they are not free from limitations. To begin with, due number (DN) value that stretches from 0 (darkness) to to the differences between the sensors and the lack of 63 (maximum intensity)14. in-flight calibration, monthly lights captured by different satellites are not directly comparable, so it was necessary Table 3 shows the distribution of pixels by luminosity to apply an inter-calibration procedure developed by Wu level in Tabasco for the period 2004-2012. In general, et al. (2013) in order to improve comparability. Also, as a it is observed that between 35% and 40% of the result of the deterioration of the satellite F16 sensor and territory does not emit enough light to be captured the late sunset in summer, there are no data available by the satellites, which indicates the absence of from June to September 2009, and from June to July of economic activity in those areas. In turn, another 40% the period 2010-2012. This affects, to a certain extent, the of the territory emits light with an intensity value that measurement of the impact of the 2010 floods. The way ranges between 3 and 11, whereas only a tiny fraction in which the achieved results are affected by the lack of (less than 1%) reaches the maximum intensity of 63. information is analyzed below. Based on NTL data, those municipalities and localities which concentrate most of the economic activity of Tabasco were identified. As can be seen in Figure 11, more than 50% of NTL come from the municipalities of Huimanguillo (16.8%), Centro (15.1%), Cárdenas (10.5%) and Macuspana (10.1%). FIGURE 11 DISTRIBUTION OF NTL AND TERRITORY BY MUNICIPALITY IN TABASCO, 2004-2012 16.8 Huimanguillo Centro 15.1 10.5 Cárdenas Macuspana 10.1 Cunduacán Comalcalco Nacajuca Paraíso Centla Tenosique Jalpa de Méndez Jalapa Balancán Teapa Nighttime lights Tacotalpa Jonuta Territory Emiliano Zapata 0% 3% 6% 9% 12% 15% 18% Source: Prepared by the authors based on data supplied by the NOAA 14 The digital number does not necessarily reflect the actual luminosity for several reasons. For example, censoring (predefined maximum luminosity) or sensor saturation. See Henderson et al. (2012) for further details. 25 3.3. FLUVIAL FLOODING MODEL (GLOFRIS) In order to estimate the results, a fluvial flooding temperature data. This process is carried out for a period global model (GLOFRIS), developed by Deltares, was of 30 years. The result of this exercise is the maximum used. Such model contains information on the flood daily flood volume, which represents the amount of level in decimeters of rain at a spatial resolution of water (in decimeters) potentially found outside the river 0.00833 decimal degrees (about 1 km2 at the equator). banks. In addition to the flood extent, the layers indicate GLOFRIS was designed by combining a model of the maximum volume of the river outflow, measured in global scale weather databases, a global hydrological decimeters of water. model, a global flood routing model and a downscaled flood model15. When estimating the GLOFRIS model for 2007 and 2010, it was found that the accumulated water level Having a spatial resolution of about 1 km2 at the equator, resulting from the overflow of rivers was higher in 2010 GLOFRIS represents a significant improvement over than in 2007 (see Figure 12), which is consistent with previous river overflow flooding models that, in contrast, the results obtained through the hydrometeorological have a resolution of about 0.5 decimal degrees or 60 km characterization of the events. at the equator. This provides more flexibility and enables the combination of GLOFRIS with other exposure and vulnerability indicators for damage assessment. In order to generate annual information regarding the amount of water, in decimeters, recorded during the largest scale event, hydrology and routing modules are calibrated using daily global precipitation and ambient FIGURE 12 GLOFRIS MODEL FOR 2007 AND 2010 GLOFRIS 2007 GLOFRIS 2010 Source: Prepared by the authors based on information from GLOFRIS 15 The detailed methodology for GLOFRIS flooding maps can be found in Winsemius et al. (2013). GLOFRIS annual flooding maps are available for the period 2000-2010. 26 |ASSESSMENT OF THE ECONOMIC IMPACT OF INVESTMENTS IN DISASTER RISK REDUCTION AND PREVENTION: THE CASE OF TABASCO 3.4. INVESTMENTS IN PROTECTION INFRASTRUCTURE Based on information obtained from the white paper location of these investments. This document includes on the PHIT, it was possible to identify investments in an updated version of the protection infrastructure protection infrastructure and disaster prevention made inventory existing in the state. For each investment in between 2007 and 2010. According to the white paper, record, the document shows its coordinates in UTM 67 investments were made, including the construction (Universal Transverse Mercator) format, which has made and reinforcement of embankments, walls and riverbank it possible to obtain the exact location of each of them. protections, desilting of carrier drains, and construction When combining this with the information recorded on of control structures and traffic bridges. the white paper on the PHIT, we managed to identify those investments made by the PHIT and their year of On the basis of a study conducted by the National completion. Autonomous University of Mexico (UNAM) in 2014 for the hydrological project to protect the population In order to be able to use ArcGIS data, the coordinates against floods and make a better use of water of the investments in DRR were converted to decimal (PROHTAB, Estudio para el proyecto hidrológico para degrees (DD). We managed to locate 57 out of proteger a la población de inundaciones y aprovechar 67 investments in DRR made between 2007 and 2010 mejor el agua), it was possible to determine the exact (see Annex 9); for the remainder, no record was found FIGURE 13 LOCATION OF DRR INFRASTRUCTURE BUILT IN TABASCO BETWEEN 2007 AND 2010 Source: Prepared by the authors based on information from GLOFRIS 27 in the infrastructure inventory (see Annex 10). As FIGURE 14 SATELLITE IMAGES OF THE TINTILLO FLOODWAY shown in Figure 13, 70% of the investments in DRR are concentrated around the city of Villahermosa. Besides, and in order to confirm the location of these investments, each of the coordinates was checked with Google Maps and the existence of the investments was verified through satellite images. An example of the work carried out was the identification of the Tintillo floodway in the municipality of Centro. As it can be observed in the satellite image (left) in Figure 14, the presence of the floodway is not so evident, except for the surrounding vegetation with a slightly different color shade. However, using the street view (right image), the hydro-agricultural infrastructure becomes more visible. Thus, by means of both views, the presence of the mapped investments was confirmed. However, it should be mentioned that there is a limitation to this analysis, which is that the extent of the investments cannot be defined. For example, in the case of the reconstruction of protection embankments, it was possible to locate them and verify their existence, though not the extent of the reconstruction work undertaken. On the other hand, information on the amount invested or the exact date of completion of the works is also unavailable— only the year is known. However, it is considered that investments registered as completed in 2010 preceded the rainy season (June-December) and, therefore, the floods occurred in August-September 2010. Source: Google Maps 28 |ASSESSMENT OF THE ECONOMIC IMPACT OF INVESTMENTS IN DISASTER RISK REDUCTION AND PREVENTION: THE CASE OF TABASCO 29 4 METHODOLOGY With the aim of measuring the impact of floods on the As it was previously mentioned, georeferenced economic development and on the school dropout rate, information on Tabasco’s Basic Education schools and georeferencing of the databases was used. First, data on investments in DRR made between 2008 and 2010 from GLOFRIS and NTL were mapped in ArcGis, in order was used. Figure 15 shows the geographical distribution to generate information at cell level with a resolution of the schools in the city of Villahermosa (black dots) of around 1 km2 at the equator. Subsequently, Basic and of the investments in DRR (red dots). As shown in Education schools located in Tabasco were included in Panel A, many of the investments made in Villahermosa the map, as well as investments in DRR made between are located on the banks of the rivers surrounding the 2007 and 201016. city. This is because most of the investments made consisted in the construction of walls and protection The location of the schools and of the investments embankments. Due to the fact that the information in DRR made it possible to determine each school’s available on them is limited to the general description accumulated water level (GLOFRIS) present in the of the investment, the year of completion and its 2007 and 2010 floods, and if the school was protected coordinates, their extent cannot be determined. To by said investments. At the same time, by combining solve this problem, we used a 1 km radius to define a the databases in ArcGis, it was possible to create a protection area surrounding the investments so as to be database at cell level with information on the level of able to come to a closer estimation of its impact. water accumulated during the floods (GLOFRIS) and the light intensity (NTL) present during the months To determine the radius of protection, various that preceded and followed the event. Based on this, simulations were carried out applying radii with two databases were created: one at school level, to extensions of 500 meters, 1 km, 1.5 km, and 2 km. On assess the effect on school dropout, and another the one hand, we found that by using radii of more at cell level, to measure the effect on the economic than 1 km, a considerable section of Villahermosa activity (based on the change in NTL). Below, there is (which received 70% of the investments) was protected; a more detailed explanation on how these databases with 2 km radii, downtown Villahermosa was almost were created. completely covered. Based on this observation, and 16 For more detailed information on the databases, refer to the previous chapter. 31 with the aim of making a more realistic analysis, the structures were left out of the analysis so as to avoid areas above 1 km were dismissed. On the other hand, bias in the estimations. As regards the traffic bridges we considered that a 500 m radius was a very small of the Tintillo17 and Sabanilla floodways (see Annex 9), area, and that an important share of the schools and these were excluded from the analysis because they protected areas would not be included under this were built as a supplement to the construction of criterion, so it was also dismissed. Panel B shows the floodways to facilitate the passing of vehicles investments located in Villahermosa and their area of across the construction works. In these cases, it is influence in a 1 km radius. considered that such investments provide protection by means of the floodways and not through the traffic In the case of embankments, walls and riverbank bridges themselves. protections, only the areas of influence belonging to the corresponding river banks were considered. Aided by the description of the investments included in the white paper on the PHIT, it was possible to define the areas of influence more accurately. For example, the description of embankments and protections usually included the protected community or its location on the left or right bank of the river. In order to exclude unprotected areas, in ArcGIS, the radius of the protections surrounding the rivers was cut (see Figure 16). The construction of traffic bridges and floodways— structures that help to drain excess water from the basins—was was also excluded from the analysis. Tabasco’s floodways are located in the outskirts of Villahermosa so as to lower the water level of the rivers surrounding the city. Therefore, their area of influence can be the entire city of Villahermosa. Consequently, we cannot know for certain which areas of the city are benefited by these works. For this reason, these FIGURE 15 LOCATION OF INVESTMENTS IN DRR MADE IN VILLAHERMOSA BETWEEN 2007 AND 2010 Panel A Panel B Source: Prepared by the authors based on information from the CONAGUA and the white paper on the PHIT 17 In the second photograph in Figure 14, the traffic bridge built to allow the crossing of cars can be seen. 32 |ASSESSMENT OF THE ECONOMIC IMPACT OF INVESTMENTS IN DISASTER RISK REDUCTION AND PREVENTION: THE CASE OF TABASCO FIGURE 16 1 KM RADIUS OF PROTECTION OF THE INVESTMENTS IN DRR IN VILLAHERMOSA Source: Prepared by the authors based on information from the CONAGUA and the white paper on the PHIT 33 4.1. METHODOLOGY FOR THE ESTIMATION OF THE EFFECT ON SCHOOL DROPOUT Once the investments in DRR—as well as the areas Cárdenas and Cunduacán (see Table 4). Also, it can be affected and the GLOFRIS index—were located in ArcGIS, observed that the highest percentage of protected schools the schools were mapped using the coordinates obtained is located in the city of Villahermosa, which belongs to the from the SEP's SNIE. On this basis, it was possible to municipality of Centro. assign the flood level present during the 2007 and 2010 events to each of the schools, and to identify those school In order to estimate the effect of the 2007 and 2010 floods, buildings located within a 1 km radius of the investments in and of the investments in DRR in 2010, a linear fixed effects protection infrastructure. model was used at municipal level. With this model, it is possible to use potential non-observable effects related To assign the water level, the distance from each school to the school geographical location as control variable. to the closest flooded cell was considered. In order to Likewise, certain characteristics of the schools reported in avoid possible estimation errors caused by a school being the 911 Survey were used as controls: (i) location in a rural located in a 1 km2 cell, the piece of information used was or urban locality, and (ii) public or private funding. the recorded water level of the cell closest to the school in a 1 km radius. In other words, if the distance from a Estimations for different specifications of the model school to the closest flooded cell was shorter than 1 km, were made, using different definitions of flood. The the flood level of that cell was taken as a reference and, estimations made use the GLOFRIS continuous variable, otherwise, it was considered that the school had not been which indicates the level of water accumulated during affected. This adjustment was made in order to avoid the floods, and a dichotomous variable that defines as possible measurement errors in the case of schools whose flooded those cells with a GLOFRIS value higher than 10 dm coordinates locate them at the edge of a cell. There may (i.e. 1 m). The latter is aimed at differentiating the impact be a case of two adjacent cells where one shows a certain on those schools that sustained a considerable level of water level and the other does not. Therefore, when having flooding18. Taking into account that the average height of a a school on the edge of a non-flooded cell, it may be household’s first floor in Mexico is 2 m, it is expected that a mistakenly assumed that it was not affected. When using flooding of 10 dm or more will result in a considerable loss the water level which is closer to the indicated radius, we of the household’s assets. reduce possible errors in the school location due to the lack of accuracy in the coordinates or in the assignment of the water level of the GLOFRIS model. TABLE 4 PERCENTAGE OF PROTECTED SCHOOLS BY MUNICIPALITY As the academic year established by the SEP starts at the beginning of August and finishes in June of the following Municipality Number Protected schools of schools within a 1 km radius year, and given that the 2007 and 2010 floods occurred at the end of August and the beginning of October, it was Centro 482 15.4% possible to estimate the school dropout rate based on Nacajuca 84 7.1% information on the number of students enrolled before Huimanguillo 332 0.9% the floods and those who completed the academic Cárdenas 259 1.9% year. Specifically, dropout rates were obtained based on Cunduacán 152 0.7% the information gathered during the 911 Survey for the Other 1,478 0% 2007‑2008 and 2010-2011 academic years. Villahermosa 194 32% Source: Prepared by the authors based on information from the SNIE, the In Figure 15, Panel B shows the number of schools CONAGUA and the white paper on the PHIT protected by municipality in a 1 km radius. It should be noted that investments in DRR were only concentrated in the municipalities of Centro, Nacajuca, Huimanguillo, 18 Estimations were also made using a threshold of 5 dm. The results of those estimations are qualitatively similar and are not included in the present report, but are available for the reader. 34 |ASSESSMENT OF THE ECONOMIC IMPACT OF INVESTMENTS IN DISASTER RISK REDUCTION AND PREVENTION: THE CASE OF TABASCO 4.2. METHODOLOGY FOR THE ESTIMATION OF THE IMPACT ON NIGHTTIME LIGHTS For the estimation of the impact of the 2007 and 2010 floods school dropout, a dichotomous variable of flooding was on the economic activity of the state, information on NTL included for GLOFRIS values higher than 10 dm. before and after the events was used. NTL information for the months of the floods was excluded from the analysis To define the period during which the floods occurred, the so as to avoid a measurement bias. As it was mentioned months when the Declaration of Disaster was published in in the previous descriptive section about NTL, satellite's the Official Gazette of the Federation (DOF) for 2007 and capture of light is hindered by the presence of clouds. As 2010 were taken as a reference: October for the 2007 floods the floods in Tabasco are a consequence of the excess of and the period of August-September for 2010. water accumulated in the basins during the rainy season, the presence of a greater number of cloudy days may bias Additional estimations were made taking different the results. To reduce this type of variations in the light definitions of the flood period as a reference. For example, intensity recorded, the average NTL for periods of 1 to the impact was estimated considering the rainy season in 12 months before and after the floods were used. the state instead of the month in which the disaster was declared, and also taking as a reference the month with the In order to measure the impact of floods on the economic highest rainfall level in each of the municipalities. Similar development of the state, the effect of GLOFRIS was results were obtained with all the different specifications. estimated on the logarithmic change of the NTL average during the months before and after the event, according to To identify the cells protected by investments in DRR, the following formula: a procedure similar to the school mapping was carried out. In ArcGis, the location of the investments made was superimposed and the 1 km radii affected were drawn. Then, DNTLitm =Log [ m ] [ Rm NTLi(t+m) - Log Rm NTLi(t-m) m ] the sections of the radii that—according to the description of the investments—were not protected were deleted, and the cells included within those radii were identified. Once the protected cells in each radius were identified, it was Where i is the cell identifier, t is the month when the possible to make the estimations. flooding occurred, and m indicates the number of months throughout which the effect needs to be estimated—from 1 to 12 in our analysis. For example, to estimate the change in NTL taking as a reference the 6-month periods preceding and following the 2007 floods, the average NTL recorded from April to September 2007 (6 months before October, when the declaration of disaster was made) and the average from November 2007, to April 2008 (6 months after) were considered. Once this was done, logarithms were applied to each of the average values so as to standardize the results, and the difference was calculated. With this, it was possible to estimate the effect of 1 dm of water in the average change of NTL compared to the previous period with the following regression: DNTLitm =b0 + b1GLOFRISit + f (1) This model is similar to a regression of simple differences, as it measures the effect on the change in NTL. Also, the effect for those cells with high flood levels was estimated. As with the estimations made to measure the impact on 35 5 RESULTS 5.1. RESULTS RELATED TO SCHOOL DROPOUT The first step in the analysis was the estimation of the The results obtained show the existence of a positive effect of floods on the 2007 and 2010 school dropout correlation between the flood level (GLOFRIS) and rates. For this purpose, a fixed effects model was used the school dropout rate (see Table 5). Particularly, it at municipality level, with the GLOFRIS variable and the was found that one additional decimeter of rain had rate of school dropout in Basic Education for the 2007- increased the dropout rate in 0.11% for 2007, while 2008 and 2010-2011 academic years: in 2010 the increase was of only 0.02%. This result remained the same even when it was controlled by the Dropouttsn= b0 + b1GLOFRISts + b2Rurals + b3Publics + characteristics of the school. b4Publics*Rurals + in+ f (2) Dropouttsn= b0 + b1Floodedts + b2Rurals + b3Publics + TABLE 5 RESULTS OF THE EFFECT OF THE FLOODS ON THE SCHOOL b4Publics*Rurals + in+ f (3) DROPOUT RATE USING GLOFRIS Where Dropout is the school dropout rate recorded during the academic year t in school s located in Variables 2007 2010 2007 2010 municipality n, while the GLOFRIS variable indicates GLOFRIS (dm) 0.00110 0.000231*** 0.00111 0.000228*** the flood level in decimeters of rain present during the Rural 0.00751 -0.00287 floods in the cell located at a distance of less than 1 km Public 0.00822** 0.000864 from the school, and the dichotomous variable Flooded Public*Rural -0.0102 0.00390 differentiates those schools with a GLOFRIS value equal Constant 0.0461*** 0.0397*** 0.0402*** 0.0382*** to or greater than 10 dm. Both estimations are controlled Observations 2,750 2,808 2,750 2,808 for the school characteristics: location in an urban or R2 0.001 0.001 0.002 0.001 rural area, and public or private funding. The interactive ***p<0.01, **p<0.05, *p<0.1 variable Public*Rural and the fixed effects at municipal level were also included (in). Source: Prepared by the authors 37 In contrast, when making estimations for those schools Table 7 and Table 8 show the results for the total sample severely affected by the floods (see Table 6), it was found of Basic Education schools (primary and secondary), that the impact was greater in 2010 than in 2007: in 2007, as well as the disaggregated results by school grade. the increase in the school dropout rate was of almost According to the results obtained with GLOFRIS 0.4%, while in 2010 amounted to almost 0.7%. These (see Table 7), it was found that, when controlling for results indicate that for those schools which sustained investments in DRR, the impact on the school dropout a high flood level the impact on school dropout was rate is still smaller for 2010 than for 2007. In 2007, an slightly higher in 2010. It should be noted that, while the increase of one decimeter of rain caused an increase GLOFRIS results show a magnitude difference of more of almost 0.1% in the Basic Education dropout rate, than 10%, in this case, there is only a 0.3% difference. while in 2010 the increase was merely of 0.023%. The same happens when the floods' effect is evaluated for primary and secondary schools; in this case, the biggest difference in impact was recorded at the primary school TABLE 6 RESULTS OF THE EFFECT OF THE FLOODS ON SCHOOL level for the GLOFRIS variable. DROPOUT USING THE DICHOTOMOUS VARIABLE FLOODED (GLOFRIS = 10) As regards the schools benefited by the protection measures and that sustained floods in 2010, the school Variables 2007 2010 2007 2010 dropout rate increased in Basic Education schools, Flooded (=1) 0.00424 0.00691*** 0.00434 0.00701*** particularly, in primary schools. However, the coefficient Rural 0.00612 -0.00130 is only significant for the latter, with an increase of Public 0.00825** 0.00112 0.026% in the dropout rate, which is significant at Public*Rural -0.00867 0.00295 10%. On the other hand, it was possible to observe a Constant 0.0472*** 0.0392*** 0.0411*** 0.0369*** reduction in the impact on secondary schools flooded Observations 2,750 2,808 2,750 2,808 in 2010 and that were located within the radius of R2 0.000 0.001 0.001 0.002 protection of investments. ***p<0.01, **p<0.05, *p<0.1 On the other hand, when estimating the results for schools with a flood level higher than 10 dm (see Source: Prepared by the authors Table 8), we found that there was a greater increase in the dropout rate in 2010 (0.67%, compared to 0.43% in 2007). This seems to indicate that, in comparison In order to measure the effect of the investments in with the previous model, the schools with high flood DRR, the dichotomous variable Protection was added levels were more affected in 2010 than schools with to the 2010 regressions. This variable enables to identify high flood levels in 2007. It should be highlighted that those schools located within a 1 km radius around the 2010 event was greater in terms of magnitude and the investments. This new specification also includes that, when comparing schools with high flood levels, the interactive variables (GLOFRIS*Protection and we are comparing two groups in which the most Flooded*Protection) to estimate the joint effect of the flooded school in 2007 showed lower water levels schools protected by investments in DRR and that, in turn, than the most flooded school in 2010. This fact helps were affected by the 2010 floods. As in the previous model, to understand why, in spite of the investments made, the results were estimated with a fixed effects model at high rainfall levels continue to exert an impact on the municipality level: increase of school dropout. Dropouttsn= b0 + b1GLOFRISts + b2Protections + b3GLOFRIS * In conclusion, a positive correlation was found between Protection + b4Rurals + b5Publics + b6Publics * Rurals + in+ f (4) the flood level (GLOFRIS) and the school dropout rate in 2007 and 2010. Such correlation is even greater in the Dropouttsn= b0 + b1Floodedts + b2Protections + b3Flooded * case of schools with flood levels higher than 10 dm of water. In turn, coefficients for the interactive GLOFRIS Protection + b4Rurals + b5Publics + b6Publics * Rurals + in+ f (5) and Flooded variables show an increase in the school dropout rate—except for the secondary education level Where t corresponds to the academic year 2010-2011 in the GLOFRIS model and for the primary education and s is the unique identifier of the school located in level coefficient using the dichotomous variable Flooded. municipality n. A possible explanation as to why coefficients in general show an increase of the dropout rate, even when they are within the radius of protection, is that the schools located in a radius close to the investments in DRR may 38 |ASSESSMENT OF THE ECONOMIC IMPACT OF INVESTMENTS IN DISASTER RISK REDUCTION AND PREVENTION: THE CASE OF TABASCO have previous characteristics that lead to an increase Another aspect that we should bear in mind is that of school dropout. It is important to bear in mind that the 911 Survey of the SEP only enables to estimate the floods occur in Tabasco every year and that the high annual dropout rate; therefore, we cannot measure the frequency of these events may have long term effects immediate impacts after the floods. Since the 2010 floods on school dropout. For example, students’ performance occurred during the first month of the academic year, it may be poorer in those schools frequently enduring is possible that in the following ten months the affected floods and, even though they may have recently students may have resumed attendance to school. become protected by investments, the fall in students’ In this case, more detailed and frequent information performance may continue to affect dropout statistics. on the students’ attendance may be essential for a To prove such hypothesis, however, a more detailed more thorough analysis of the effect of investments in analysis on the students’ performance level is needed. protection infrastructure on school dropout. TABLE 7 IMPACT OF INVESTMENTS IN DRR ON THE SCHOOL DROPOUT RATE BY EDUCATION LEVEL USING GLOFRIS Basic Education Primary Secondary Variables 2007 2010 2007 2010 2007 2010 GLOFRIS (dm) 0.00111 0.000231*** 0.00124 0.000290*** 0.00111*** 0.000105 Protection (=1) 0.0182*** 0.00935** 0.0370*** GLOFRIS*Protection 0.0000114 0.000266* -0.000496 Rural 0.00751 -0.000768 -0.00882 0.000595 0.0462 0.00149 Public 0.00822** 0.000657 0.00632* -0.00445 0.0158*** 0.0136** Public*Rural -0.0102 0.00292 0.00560 0.00241 -0.0431 0.00280 Constant 0.0402*** 0.0369*** 0.0385*** 0.0364*** 0.0400*** 0.0365*** Observations 2,750 2,808 2,036 2,072 714 736 R 2 0.002 0.004 0.002 0.003 0.009 0.017 *** p<0.01, ** p<0.05, * p<0.1 Source: Prepared by the authors TABLE 8 IMPACT OF INVESTMENTS IN DRR ON THE SCHOOL DROPOUT RATE BY EDUCATION LEVEL USING THE DICHOTOMOUS VARIABLE FLOODED (GLOFRIS ≥ 10) Basic Education Primary Secondary Variables 2007 2010 2007 2010 2007 2010 Flooded (=1) 0.00434 0.00677*** 0.00413 0.0113*** 0.00890** -0.00418 Protection (=1) 0.0124** 0.0175*** 0.00157 Flooded*Protection 0.00941* -0.00613* 0.0433*** Rural 0.00612 0.000654 -0.0104* 0.00354 0.0450 -0.00121 Public 0.00825** 0.000831 0.00636* -0.00411 0.0158*** 0.0120** Public*Rural -0.00867 0.00208 0.00747* 0.000549 -0.0419 0.00590 Constant 0.0411*** 0.0359*** 0.0395*** 0.0337*** 0.0408*** 0.0399*** Observations 2,750 2,808 2,036 2,072 714 736 R2 0.001 0.004 0.001 0.005 0.008 0.022 *** p<0.01, ** p<0.05, * p<0.1 Source: Prepared by the authors 39 5.2. RESULTS RELATED TO NIGHTTIME LIGHTS In order to measure the impact of floods on the FIGURE 17 COEFFICIENTS OF GLOFRIS IMPACT ON NTL change in NTL, a fixed effects model was estimated FOR 2007 AND 2010 at municipality level. For this purpose, the average light intensity level was estimated for periods of 1 to 12 months before and after the floods according to the 3% following equations: 2% 1% DNTLitmn =b0 + b1GLOFRISit + in + f (6) 0% Impact on NTL -1% DNTLitmn =b0 + b1Floodedit + in + f (7) -2% -3% Where i is the cell identifier, GLOFRIS is the level of -4% water recorded in the year t, month m, for the cell -5% i, and in is the fixed effect for the municipality n, -6% while the variable Flooded identifies those cells with 1 2 3 4 5 6 7 8 9 10 11 12 GLOFRIS values greater than or equal to 10 dm of Months before and after the floods accumulated water. 2007 2010 Figure 17 shows the coefficients obtained for GLOFRIS, for each of the NTL specifications. As it may be Note: Red dots indicate a significance level of p<0.01, green dots, p<0.05; observed, the effect on NTL is negative for almost and yellow dots, p<0.1. Original regressions are detailed in Annex 5 and in Annex 6. all coefficients, except for the change in NTL during the first and second month after the 2007 floods. Source: Prepared by the authors It particular, the change in NTL for 2010 remains almost constant, while in 2007 a sharp drop in NTL may be observed. Even 12 months after the 2007 floods, a 3.5% reduction in light intensity could still be In order to measure the effect of the investments on observed, compared to the average light intensity of the NTL, a dichotomous variable was added to the 2010 12 months preceding the event. regression; such variable identifies the cells found within a 1 km radius around the investments made between This is also observed when estimating the effect by 2008 and 2010. The estimation was made using a linear means of the dichotomous variable that differentiates fixed effects model at municipality level, according to those cells showing a GLOFRIS value higher than 10 dm. the following equations: In contrast with the estimations obtained using the GLOFRIS variable, Figure 18 shows an even sharper DNTLimn =b0 + b1GLOFRISi + b2Protectioni + difference between the impact on NTL in 2007 and 2010. b3GLOFRISi * Protectioni + in + f (8) While in Figure 17 the difference between both events in month 12 amounts to only 3.5%, when estimating the DNTLimn =b0 + b1Floodedi + b2Protectioni + effect for the cells with high flood levels, a difference of b3Floodedi * Protectioni + in + f (9) around 25% in NTL is observed. In these estimations, the interactive variables With both definitions, the results show a significantly higher GLOFRIS*Protection and Flooded*Protection were impact in 2007 when compared with 2010, even when included so as to estimate the effect on NTL for the cells the accumulated rainfall in 2010 was higher than in 2007. showing a certain flood level that, at the same time, were Besides, it should be noted that the missing results for the benefited from the investments in DRR. first and second month of the year 2010 are due to the lack of NTL information prior to the flood, as it may be observed When comparing the results obtained for 2007 with in Figure 19. The same graph shows the drop in NTL in the the model estimated for 2010 for the NTL average for months after the 2007 and 2010 floods. Particularly, a larger 6 and 9 months, it was found that GLOFRIS had a higher drop may be observed in 2007 than in 2010. negative impact in 2007 than in 2010 (see Table 9). For 40 |ASSESSMENT OF THE ECONOMIC IMPACT OF INVESTMENTS IN DISASTER RISK REDUCTION AND PREVENTION: THE CASE OF TABASCO FIGURE 18 COEFFICIENTS OF THE IMPACT OF THE DICHOTOMOUS VARIABLE TABLE 9 IMPACT OF GLOFRIS AND OF THE INVESTMENTS IN DRR ON NTL FLOODED ON NTL FOR 2007 AND 2010 2007 Results 2010 Results Variables ∆NTL 6m ∆NTL 9m ∆NTL 6m ∆NTL 9m 20% GLOFRIS (dm) -0.0305** -0.0333*** -0.00260 -0.00241* 10% Protection - - 0.0338 0.0370 0% GLOFRIS*Protection - - 0.00297 0.00112 -10% Impact on NTL Polynomial - - - - -20% Constant -0.198*** -0.120*** -0.138*** -0.0352*** -30% Observations 18,760 18,760 18,978 19,251 -40% R2 0.020 0.020 0.007 0.004 -50% *** p<0.01, ** p<0.05, * p<0.1 -60% 1 2 3 4 5 6 7 8 9 10 11 12 Months before and after the floods Source: Prepared by the authors 2007 2010 Note: Red dots indicate a significance level of p<0.01, green dots, p<0.05; TABLE 10 IMPACT OF INVESTMENTS IN DRR ON NTL FOR HIGHLY and yellow dots, p<0.1. Original regressions are detailed in Annex 5 and FLOODED CELLS in Annex 6. Source: Prepared by the authors Resultados 2007 Resultados 2010 Variables ∆NTL 6m ∆NTL 9m ∆NTL 6m ∆NTL 9m Flooded (=1) -0.284** -0.304** -0.0541 -0.0542** 2007, it was found that one additional decimeter of Protection - - 0.0312 0.0429 rainfall reduced in approximately 3.1% the light intensity Flooded*Protection - - 0.0686 0.0103 recorded via NTL, while, in 2010, this impact was of only Polynomial - - - - 0.26%. Such result is consistent with what was observed Constant -0.212*** -0.135*** -0.139*** -0.0352*** in the simple model (see Figure 17). Besides, it can be Observations 18,760 18,760 18,978 19,251 observed that, for those flooded cells benefited from R2 0.012 0.011 0.005 0.004 investments in protection infrastructure, no reduction in *** p<0.01, ** p<0.05, * p<0.1 NTL was recorded. Source: Prepared by the authors Similar results are found when estimating the effect on highly flooded cells. Table 10 shows that, in 2007, the effect for a 9-month period was a 30% decrease in the recorded NTL value, while in 2010, it only equaled 5.4%. Likewise, it was found that there was an increase in FIGURE 19 NTL DYNAMICS FOR TABASCO, 2004-2012 the NTL level for the highly flooded cells that benefited from the investments in protection infrastructure. For estimations for the 12 months, refer to Annex 7 50% 2007 2010 and Annex 8. October August-October 40% 30% In order to verify that the results are not affected 20% by a possible bias arising from the location of the 10% investments in DRR, a correction was used that included 0% both the longitude and the latitude of the investments -10% (see, for example, Dell 2010). Aimed at correcting such -20% bias, Dell introduces a new variable which consists -30% of a second degree function built on the coordinates. -40% Based on that method, the following models for 2010 2004 2005 2006 2007 2008 2009 2010 2011 2012 were estimated: Source: Prepared by the authors based on data supplied by the NOAA 41 DNTLim =b0 + b1GLOFRIS + b2Protectedi + b3GLOFRISi * of the coordinates. The estimations shown belong Protectedi + b4f(geographical location)+ in + f (10) to the NTL average for 6 and 9 months. As it can be observed, both for the model including the GLOFRIS NTLim =b0 + b1Floodedi + b2Protectedi + b3Floodedi * continuous variable and for the dichotomous variable Protectedi + b4f(geographical location)+ in + f (11) Flooded, there are significant differences in the results obtained in the simple model and those obtained with Where the introduction of the polynomial. Lastly, an attempt to f(geographical location)= x + y + x2 + y2 + xy (12) make estimations using a fuzzy regression discontinuity x=longitude y=latitude model was made. However, due to the reduced number of protected cells, it was not possible to carry out Table 11 and Table 12 show the results obtained for this estimation. the simple model and other two models that include a second and a third degree polynomial for the correction TABLE 11 RESULTS FOR THE GLOFRIS VARIABLE USING 2 ND AND 3RD DEGREE POLYNOMIALS 2010 Results 2nd degree polynomial 3rd degree polynomial Variables ∆NTL 6m ∆NTL 9m ∆NTL 6m ∆NTL 9m ∆NTL 6m ∆NTL 9m GLOFRIS (dm) -0.00260 -0.00241* -0.00256* -0.00241* -0.00258* -0.00241* Protection 0.0338 0.0370 0.0478 0.0350 0.0480 0.0349 GLOFRIS*Protection 0.00297 0.00112 0.00210 0.00124 0.00212 0.00125 Polynomial - - 0.000956 -0.000135 -0.0000070 0.0000011 Constant -0.138*** -0.0352*** -7.040 0.939 -4.773 0.660 Observations 18,978 19,251 18,978 19,251 18,978 19,251 R2 0.007 0.004 0.014 0.004 0.014 0.004 *** p<0.01, ** p<0.05, * p<0.1 Source: Prepared by the authors TABLE 12 RESULTS FOR THE VARIABLE FLOODED USING 2ND AND 3RD DEGREE POLYNOMIALS 2010 Results 2nd degree polynomial 3rd degree polynomial Variables ∆NTL 6m ∆NTL 9m ∆NTL 6m ∆NTL 9m ∆NTL 6m ∆NTL 9m Flooded (=1) -0.0541 -0.0542** -0.0595* -0.0539** -0.0602* Protection 0.0312 0.0429 0.0457 0.0421 0.0460 Flooded*Protection 0.0686 0.0103 0.0493 0.0113 0.0495 Polynomial - - 0.00104 -0.0000555 -0.0000076 Constant -0.139*** -0.0352*** -7.678 0.365 -5.215 Observations 18,978 19,251 18,978 19,251 18,978 R2 0.005 0.004 0.013 0.004 0.013 *** p<0.01, ** p<0.05, * p<0.1 Source: Prepared by the authors 42 |ASSESSMENT OF THE ECONOMIC IMPACT OF INVESTMENTS IN DISASTER RISK REDUCTION AND PREVENTION: THE CASE OF TABASCO 43 6 CONCLUSIONS The present study was aimed at assessing the impact of Some of the problems that may explain the lack of robust floods and of investments in DRR on the socioeconomic results are the following: wellbeing of the population of Tabasco with a greater level of disaggregation. For this purpose, a flood model 1. Time frame of available information. The lack of data and NTL data with a resolution of 1 km2 were used, as concerning school attendance for the months after the well as georeferenced information on schools and on 2007 and 2010 floods significantly affected the analysis investments on protection infrastructure made between carried out. Owing to the fact that we only have annual 2007 and 2010. As previously mentioned, various models information and that floods were registered during the were applied for the estimation of the impacts, and first months of the academic year, it is not possible variations were used for the definition of Flooded and to estimate the impact for the months immediately Protected. On the one hand, different models were following the events. Likewise, the effect estimated estimated using the flood level recorded by GLOFRIS, in this study may be considerably smaller, as several as well as the dichotomous variable Flooded for a flood months had gone by since the events at the time this threshold of more than 10 dm of water. On the other data became available, and the students affected may hand, the effects of investments in DRR on school have returned to school during that time. Having more dropout and on the change in NTL for schools and cells information available concerning school attendance located at a distance of 1 km were analyzed. would make it possible to estimate the impact of the events on temporary school dropout. The results show that, in fact, there was a negative effect of GLOFRIS on school dropout and on the change in 2. Missing information related to the events under NTL. Specifically, a greater negative impact is observed analysis. The lack of NTL information for the months in 2007 than in 2010, even when controlling for the immediately preceding the 2010 floods hindered presence of investments in protection infrastructure. the estimation of short-term effects. Such lack of However, the effect of investments in DRR is not so information may be correlated to the presence clear, as results are not robust. These results reaffirm of clouds blocking light intensity, thus making it those found in the previous study (World Bank, 2014): impossible to compare the immediate effects both in that the impact in 2007 was greater than that of 2010, 2010 and in 2007. even though the 2010 event was of a greater magnitude. 45 3. Lack of accurate information on the extent of the investments. The existence of very limited information on the type of investments made in Tabasco and their location complicated the estimation of precise impacts for the variable Protection. In the first place, the lack of information as regards the extent, height, and structural form of the investments made it impossible to determine the areas protected by them. In turn, the impossibility of estimating the protection area for investments, such as floodways, limited our results significantly. In the specific case of the city of Villahermosa, those structures played a key role in preventing river overflows. As they drain a large amount of water from the rivers that flow though the city, their effect may be greater than that of other investments in protection infrastructure, like embankments and walls. Thus, it is possible that the fact of being unable to include those structures in our analysis may affect the lack of statistical significance that we have found for the coefficients of the variable Protected. In this case, more accurate data on the extent of the investments, together with information on the potential areas benefited from floodways may substantially improve the significance of the estimations. In conclusion, although more detailed information is obtained using a 1 km2 accuracy level, the time frame of the data available continues to be a key factor for the estimation of the impact. Besides, missing information and lack of detail in the definition of the investments may substantially reduce the statistical significance of our estimations. More accurate geographical information on the location of investments would reduce the use of discretionary definitions for the areas of protection and would enable the estimation of the effects with a higher level of detail. 46 |ASSESSMENT OF THE ECONOMIC IMPACT OF INVESTMENTS IN DISASTER RISK REDUCTION AND PREVENTION: THE CASE OF TABASCO 47 BIBLIOGRAPHY Alesina, A., Michalopoulos, S., & Papaioannou, E. (2016). Ethnic Inequality. Journal of Political Economy, 124 (2), 428-488. Bundervoet, T., Maiyo, L., & Sanghi, A. (2015). Bright Lights, Big Cities Measuring National and Subnational Economic Growth in Africa from Outer Space, with an Application to Kenya and Rwanda. CENAPRED. (2005-2013). Características e impacto socioeconómico de los principales desastres ocurridos en la República Mexicana. CONAGUA. (2012). Libro Blanco CONAGUA-01. Plan Hídrico Integral de Tabasco (PHIT). Corral, L., Schling, M., Rogers, C., Cumberbatch, J., Hinds, F., Zhou, N., et al. (2016). The Impact of Coastal Infrastructure Improvements on Economic Growth: Evidence from Barbados. Washington, DC: IDB Working Paper. De Janvry, A., del Valle, A., & Sadoulet, E. (2016). Insuring Growth: The Impact of Disaster Funds on Economic Reconstruction in Mexico. Washington, DC: World Bank Policy Research Working Paper. Dell, M. (2010). The persistent effects of Peru's mining Mita. Econometrica, 78 (6), 1867-1903. DOF. (2017). Declaratoria de Emergencia por la presencia de lluvias e inundaciones atípicas, ocurridas los días 28 y 29 de octubre de 2007, en 17 municipios del Estado de Tabasco. Retrieved from http://dof.gob.mx/nota_ detalle.php?codigo=5005453&fecha=06/11/2007 Donaldson, D., & Storeygard, A. (2016). The View from Above: Applications of Satellite Data in Economics. Journal of Economic Perspectives, 30 (4), 171-198. ECLAC. (2008). Tabasco: características e impacto socioeconómico de las inundaciones provocadas a finales de Octubre y a comienzos de Noviembre de 2007 por el frente frío número 4. Elliott, R., Strobl, E., & Sun, P. (2015). The Local Impact of Typhoons on Economic Activity in China: A View from Outer Space. Journal of Urban Economics (88), 50-66. Elvidge, C. D., Baugh, K. E., Kihn, E. A., Kroehl, H. W., Davis, E. R., & Davis, C. W. (1997). Relation between satellites observed visible-near infrared emissions, population, economic activity and electric power consumption. International Journal of Remote Sensing, 18 (6), 1373-1379. Harari, M., & La Ferrara, E. (2013). Conflict, Climate and Cells: A Disaggregated Analysis. CEPR Discussion Papers. Henderson, V., Storeygard, A., & Weil, D. (2012). Measuring Economic Growth from Outer Space. American Economic Review, 102 (2), 994-1028. Henderson, V., Storeygard, A., & Weil, D. N. (2011). Frontiers of productivity and output: A Bright Idea for Measuring Economic Growth. Hodler, R., & Raschky, P. (2014). Regional Favoritism. Quarterly Journal of Economics (129), 995-1033. SEP. (2009). La estructura del sistema educativo mexicano. Retrieved from http://www.sep.gob.mx/work/models/ sep1/Resource/1447/1/images/sistemaedumex09_01.pdf Winsemius, H. C., Van Beek, L. P., Jongman, B., Ward, P. J., & Bouwman, A. (2013). A framework for global river flood risk assessments. Hydrology and Earth System Sciences, 17 (5), 1871-1892. Wu, J., He, S., Peng, J., Li, W., & Zhong, X. (2013). Intercalibration of DMSP-OLS night-time light data by the invariant region method. International Journal of Remote Sensing, 34 (20), 7356-7368. 49 ANNEXES ANNEX 1 MUNICIPAL DEVIATION FROM THE HISTORICAL MEAN Huimanguillo Cárdenas Tecotalpa Emiliano Zapata Tenosique Teapa Jalapa Centla Balacán Jonuta Centro Macuspana Nacajuca 2010 Comalcalco Jalpa de Méndez 2007 Cunduacán Paraíso -200 0 200 400 600 800 Source: Prepared by the authors based on information from the Mexican National Weather Service (SMN) ANNEX 2 NUMBER OF PEOPLE AFFECTED DURING THE 2007 AND 2010 FLOODS IN TABASCO, BY MUNICIPALITY Tenosique Teapa 2010 Tacotalpa Paraíso 2007 Nacajuca Macuspana Jonuta Jalpa de Mendez Jalapa Huimanguillo Emiliano Zapata Cunducán Comacalco Centro Centla Cárdenas Balancán 0 100,000 200,000 300,000 400,000 Source: World Bank, 2014 51 |ASSESSMENT OF THE ECONOMIC IMPACT OF INVESTMENTS IN DISASTER RISK REDUCTION AND PREVENTION: THE CASE OF TABASCO ANNEX 3 IMPACT OF GLOFRIS ON NTL FOR PERIODS OF 1 TO 12 MONTHS (2007) ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL Variables 1m 2m 3m 4m 5m 6m 7m 8m 9m 10m 11m 12m GLOFRIS (dm) 0.00874 0.00134 -0.0166 -0.0243* -0.0285** -0.0305** -0.0312** -0.0315** -0.0333*** -0.0352*** -0.0348*** -0.0346*** Constant -0.172*** -0.0918*** -0.152*** -0.182*** -0.211*** -0.198*** -0.180*** -0.141*** -0.120*** -0.0680*** -0.0862*** -0.0973*** Observations 17,555 18,134 18,760 18,760 18,760 18,760 18,760 18,760 18,760 19,394 19,394 19,394 R2 0.002 0.000 0.007 0.015 0.019 0.020 0.020 0.020 0.020 0.012 0.015 0.016 *** p<0.01, ** p<0.05, * p<0.1 ANNEX 4 IMPACT OF GLOFRIS ON NTL FOR PERIODS OF 1 TO 12 MONTHS (2010) ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL Variables 1m 2m 3m 4m 5m 6m 7m 8m 9m 10m 11m 12m GLOFRIS (dm) - - -0.000323 -0.00130 -0.00178 -0.00249 -0.00309* -0.00301** -0.00236* -0.00182* -0.00201* -0.00211 Constant - - -0.110*** -0.131*** -0.140*** -0.138*** -0.118*** -0.0604*** -0.0348*** -0.0150** -0.0405*** -0.0565*** Observations - - 18,978 18,978 18,978 18,978 18,978 19,251 19,251 19,251 19,251 19,251 R2 - - 0.000 0.002 0.004 0.006 0.008 0.005 0.004 0.003 0.003 0.003 *** p<0.01, ** p<0.05, * p<0.1 ANNEX 5 IMPACT OF THE DICHOTOMOUS VARIABLE FLOODED (GLOFRIS ≥ 10 DM) ON NTL FOR PERIODS OF 1 TO 12 MONTHS (2007) ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL Variables 1m 2m 3m 4m 5m 6m 7m 8m 9m 10m 11m 12m Flooded (=1) 0.0272 -0.0306 -0.178*** -0.237*** -0.268*** -0.284** -0.283** -0.291** -0.304** -0.304*** -0.296*** -0.291** Constant -0.166*** -0.0902*** -0.159*** -0.193*** -0.224*** -0.212*** -0.194*** -0.155*** -0.135*** -0.0840*** -0.102*** -0.113*** Observations 17,555 18,134 18,760 18,760 18,760 18,760 18,760 18,760 18,760 19,394 19,394 19,394 R2 0.000 0.000 0.006 0.009 0.011 0.012 0.011 0.011 0.011 0.006 0.007 0.008 *** p<0.01, ** p<0.05, * p<0.1 52 ANNEX 6 IMPACT OF THE DICHOTOMOUS VARIABLE FLOODED (GLOFRIS ≥ 10 DM) ON NTL FOR PERIODS OF 1 TO 12 MONTHS (2010) ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL Variables 1m 2m 3m 4m 5m 6m 7m 8m 9m 10m 11m 12m Flooded (=1) - - -0.0121 -0.0296 -0.0303 -0.0526 -0.0634 -0.0626** -0.0537** -0.0466** -0.0479** -0.0427 Constant - - -0.108*** -0.131*** -0.143*** -0.138*** -0.119*** -0.0618*** -0.0347*** -0.0135*** -0.0399*** -0.0578*** Observations - - 18,978 18,978 18,978 18,978 18,978 19,251 19,251 19,251 19,251 19,251 R2 - - 0.000 0.002 0.002 0.005 0.005 0.004 0.003 0.003 0.003 0.002 *** p<0.01, ** p<0.05, * p<0.1 ANNEX 7 IMPACT OF GLOFRIS AND OF INVESTMENTS IN DRR ON NTL FOR PERIODS OF 1 TO 12 MONTHS (2010) ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL Variables 1m 2m 3m 4m 5m 6m 7m 8m 9m 10m 11m 12m GLOFRIS (dm) - - -0.00039 -0.00141 -0.00188 -0.0026 -0.00318* -0.00308** -0.00241* -0.00185* -0.00206* -0.00215 Protection - - -0.0713 -0.0458 -0.00462 0.0338 0.0627 0.0497 0.0370 0.0261 0.0373 0.0422 GLOFRIS*Protection - - 0.00302 0.00400 0.00356 0.00297 0.00227 0.00176 0.00112 0.000762 0.000998 0.00106 Constant - - -0.109*** -0.130*** -0.140*** -0.138*** -0.118*** -0.0609*** -0.0352*** -0.0152** -0.0409*** -0.0570*** Observations - - 18,978 18,978 18,978 18,978 18,978 19,251 19,251 19,251 19,251 19,251 R2 - - 0.001 0.003 0.005 0.007 0.009 0.006 0.004 0.003 0.003 0.003 *** p<0.01, ** p<0.05, * p<0.1 ANNEX 8 IMPACT OF THE DICHOTOMOUS VARIABLE FLOODED (GLOFRIS ≥ 10 DM) AND OF INVESTMENTS IN DRR ON NTL FOR PERIODS OF 1 TO 12 MONTHS (2010) ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL ∆NTL Variables 1m 2m 3m 4m 5m 6m 7m 8m 9m 10m 11m 12m Flooded (=1) - - -0.0133 -0.0312 -0.0319 -0.0541 -0.0645 -0.0633** -0.0542** -0.0470** -0.0484** -0.0433 Protection - - -0.0847 -0.0574 -0.0111 0.0312 0.0682 0.0587 0.0429 0.0286 0.0398 0.0446 Flooded*Protection - - 0.0955 0.112 0.0894 0.0686 0.0352 0.0167 0.0103 0.0106 0.0151 0.0157 Constant - - -0.108*** -0.131*** -0.143*** -0.139*** -0.120*** -0.0625*** -0.0352*** -0.0139*** -0.0403*** -0.0583*** Observations - - 18,978 18,978 18,978 18,978 18,978 19,251 19,251 19,251 19,251 19,251 R2 - - 0.001 0.003 0.003 0.005 0.006 0.004 0.004 0.003 0.003 0.002 *** p<0.01, ** p<0.05, * p<0.1 53 ANNEX 9 LIST OF INVESTMENTS IN DRR MADE BY THE PHIT AFTER THE 2007 FLOODS AND BEFORE THE 2010 EVENT19 Year of ID completion Municipality Investment UTM Latitude UTM Longitude DD Latitude DD Longitude 1* 2007 Centro Reconstruction, reinforcement and protection of the embankment on the 503672.7 1989570.9 17.99444137 -92.96530801 right bank of the river Carrizal, on the boulevard Miguel Alemán, from the community Buenavista to park La Isla, municipality of Centro 2* 2007 Centro Raising of the protection embankment on the right bank of the river Carrizal, 503672.6663 1989570.934 17.99444167 -92.96530833 from the Central Supply Center to the community Nueva Villa de Trabajadores 3* 2007 Centro Raising of the protection embankment on the right bank of 507345.9372 1992670.395 18.02244722 -92.9306 the river Carrizal, from Pemex Administrative Center to the bridge Tierra Colorada III 4* 2007 Centro Raising of the protection embankment on the right side of the river Carrizal, 510104.6133 1992427.915 18.02024444 -92.90453889 from the bridge La Pigua to the private neighborhood La Pigua 5* 2007 Centro Raising of the protection embankment on the right bank of the river Carrizal, 510704.9067 1991023.15 18.00754444 -92.898875 from the road Arcadio Hernández to the bridge Grijalva IV, considering the confluence of rivers Carrizal and Grijalva 6* 2007 Centro Raising of the embankment on the left side of the river Carrizal over a length 510104.6133 1992427.915 18.02024444 -92.90453889 of 300 m, downstream from bridge La Pigua 7* 2007 Centro Raising of the embankment on the left bank of the river Carrizal over a 510552.1983 1991748.044 18.01409722 -92.90031389 length of 1.5 km, downstream from the end of the retaining wall of the community Indeco 8* 2007 Cunduacán Raising of the embankment on the left side of the river Samaria 482028.6278 1994048.847 18.03484444 -93.16979444 9* 2008 Cunduacán Restitution of the embankment Samaria-Zavala with its corresponding bank 479451.1219 1986230.9 17.96415833 -93.19406944 protection in the settlement Plátano y Cocos, 1st Section, on the left bank of river Carrizal 10* 2008 Centro Restitution of the embankments with their corresponding bank 503128.3 1989511.5 17.99390532 -92.97045045 protections in the settlement Emiliano Zapata 11* 2008 Centro Restitution of the embankments and paths with their corresponding bank 516145.5 1990600.3 18.00368882 -92.84748328 protections in the settlement Barranca y Guanal (stretch 1) 12* 2008 Centro Restitution of the embankments and paths with their corresponding bank 516145.4942 1990600.308 18.00368889 -92.84748333 protections in the settlement Barranca y Guanal (stretch 2) 13* 2008 Centro Restitution of the embankments and paths with their corresponding bank 516189.8885 1991323.17 18.01022222 -92.84705833 protections on the right bank of stream El Zapote 14* 2008 Centro Restitution of the embankments and paths with their corresponding bank 518218.1 1992396.9 18.01991114 -92.82788898 protections in the settlement Barranca y Guanal (stretch 4) 15* 2008 Centro Restitution of the embankments with their corresponding bank protections in 508826.9 1993430.4 18.02931109 -92.91660553 the community Asunción Castellanos 19 Investments included in the estimations are marked with an asterisk. 54 ANNEX 9 LIST OF INVESTMENTS IN DRR MADE BY THE PHIT AFTER THE 2007 FLOODS AND BEFORE THE 2010 EVENT (CONT.) Year of ID completion Municipality Investment UTM Latitude UTM Longitude DD Latitude DD Longitude 16* 2008 Centro Restitution of the embankments with their corresponding bank protections 509524.7 1987076.8 17.97188031 -92.91004201 in the community Gaviotas, (stretches 3 and 4), on the right bank of river Carrizal 17* 2008 Centro Restitution of the fill slopes with their corresponding bank protections in 509111.0441 1993153.941 18.02681111 -92.91392222 community Pino Suárez (el Mangal sector, stretches 1 and 2) 18* 2008 Centro Restitution of the fill slopes with their corresponding bank protections 491599.5 1984060.3 17.94461979 -93.0793281 in the settlement Buenavista, 2nd Section, km 14+500, right bank of river Carrizal 19* 2008 Cárdenas Reinforcement of the embankment Samaria Nueva Zelandia and its bank 464872.3167 1985471.538 17.95710833 -93.33174167 protection in settlement Habanero, river Carrizal 20* 2008 Centro Restitution of the embankments with their corresponding bank protections 503200.1218 1989390.452 17.99281111 -92.96977222 and reinforced concrete walls in the community Carrizal, right bank of river Carrizal 21* 2008 Centro Restitution of the embankments with their corresponding bank protections 503065.8 1988763.8 17.98714722 -92.97104192 and reinforced concrete walls at the settlement Anacleto Canabal, 1st Section, right bank of the river Carrizal 22* 2008 Centro Restitution of the embankments with their corresponding bank protections 504931.9903 1992650.6 18.02227514 -92.95340556 and reinforced concrete walls at the settlement Anacleto Canabal, 2nd Section, right bank of the river Carrizal 23* 2008 Centro Restitution of the fill slopes with their corresponding bank protections in 513117.9 1990990.5 18.00723609 -92.87608065 settlement La Manga, 2nd Section (stretches 1 and 2), right bank of river Grijalva 24* 2008 Centro Restitution of the embankments with their corresponding bank protections 507811.7877 1993123.877 18.02654444 -92.92619722 in the community Pino Suárez, on the right bank of river Carrizal 25* 2008 Centro Restitution of the embankments with their corresponding bank protections 510407.8025 1992344.174 18.01948611 -92.901675 in the community Indeco, on the left bank of river Carrizal 26* 2008 Centro Raising of the embankment in community Ciudad Industrial, left bank of river 509278.3206 1993195.201 18.02718333 -92.91234167 Carrizal 27* 2008 Centro Expansion of the hydraulic section of the Medellín drain and raising of the 510477.4 1995851.7 18.05118887 -92.90099977 embankments 28* 2009 Huimanguillo Construction of breakwaters for the protection of the left bank of river 459686.88 1971323.83 17.82914389 -93.38043901 Mezcalapa in the city of Huimanguillo 29* 2009 Centro Construction of protection of floodway El Tintillo (stretches I and II), 519379.5 1993945.1 18.03389459 -92.81690251 municipality of Centro 30* 2009 Cárdenas Desilting of the 1st stage of carrier drains Naranjeño and W-55, District 012, 438747.8803 1996382.105 18.05514167 -93.57877222 Chontalpa 55 ANNEX 9 LIST OF INVESTMENTS IN DRR MADE BY THE PHIT AFTER THE 2007 FLOODS AND BEFORE THE 2010 EVENT (CONT.) Year of ID completion Municipality Investment UTM Latitude UTM Longitude DD Latitude DD Longitude 31* 2009 Cárdenas Desilting of the 2nd stage of carrier drains Naranjeño and W-55, District 012, 456251.1051 2005612.002 18.1389889 -93.41358333 Chontalpa 32* 2009 Cárdenas Desilting of the 2nd stage of carrier drain W-58, District 012, 450602.3655 1997157.179 18.06245 -93.46678056 Chontalpa, municipality of Cárdenas, Huimanguillo 33* 2009 Cárdenas Desilting of the 2nd stage of carrier drain W-61, District 012, 447730.1326 1997100.719 18.06187222 -93.49391944 Chontalpa, municipality of Cárdenas, Huimanguillo 34* 2009 Cárdenas Desilting of carrier drain Zonapa, District 012, Chontalpa, municipality of 438861.2193 1985200.165 17.95408056 -93.57737222 Cárdenas, Huimanguillo 35* 2009 Centro Construction of the second stage of the protection of floodway Sabanilla-El 513861 1977545.1 17.88570299 -92.86915029 Censo, municipality of Centro 36* 2009 Centro Construction of protection walls on the left bank of the river Viejo Mezcalapa, 503809.0075 1985573.624 17.95831111 -92.96402778 stretches I, II, III, IV, V, VI and VII (Curahueso-Puentre Pedrero), municipality of Centro 37* 2009 Centro Construction of protection walls at the riverside promenade Leandro Rovirosa 508891.9121 1988806.141 17.98751389 -92.91601111 Wade (stretches II, III and IV) in the city of Villahermosa 38* 2009 Centro Construction of protection wall at the riverside promenade Carlos A. Madrazo 508212.855 1988143.873 17.98153056 -92.92242778 (stretches I, II and III) in the city of Villahermosa 39* 2009 Centro Construction of the protection embankment and wall at the community 512109.8443 1993463.811 18.02959722 -92.88558889 Indeco (stretches I, II, III and IV), municipality of Centro 40 2009 Centro Construction of traffic bridge over the floodway El Tintillo I, municipality 519430.7074 1993867.075 18.03318889 -92.81641944 of Centro 41 2009 Centro Construction of traffic bridge over the floodway El Tintillo II, municipality 516620.8 1991847.7 18.01495997 -92.84298343 of Centro 42 2009 Centro Construction of traffic bridge over the floodway Sabanilla-El Censo, 513892.4843 1977589.045 17.8861 -92.86885278 municipality of Centro 43* 2009 Centro Construction of the third stage of the riverbank protection of the floodway 519379.4719 1993945.084 18.03389444 -92.81690278 El Tintillo, municipality of Centro 44* 2009 Centro Construction and raising of the protection embankment Acachapan, 1st stage, 516492.0919 1992463.595 18.02052778 -92.84419444 municipality of Centro 45* 2009 Centro Construction of the bank protection on the right bank of the river La Sierra, 510096.0188 1986267.934 17.96456667 -92.90465 Coquitos sector, municipality of Centro 46* 2009 Centro Construction of the bank protection at Corregidora, 5th Section (Zapata 509501.7 1990273.27 18.00077215 -92.91024463 sector), municipality of Centro 47 2009 Centro Completion of the construction of the control structure over the river Carrizal 470182.9252 1985630.267 17.95862222 -93.28159167 48* 2009 Centro Completion of the construction of riverbank protections 470182.9252 1985630.267 17.95862222 -93.28159167 56 ANNEX 9 LIST OF INVESTMENTS IN DRR MADE BY THE PHIT AFTER THE 2007 FLOODS AND BEFORE THE 2010 EVENT (CONT.) Year of ID completion Municipality Investment UTM Latitude UTM Longitude DD Latitude DD Longitude 49* 2010 Centro Construction of protection embankment at the community Casa Blanca, on 510704.9 1991023.15 18.00754444 -92.89887506 the left bank of river Grijalva in Villahermosa, Tabasco 50* 2010 Nacajuca Desilting of pilot bed of river Samaria (stretch I) in the municipalities of Jalpa 507397.3 2014038.06 18.21557915 -92.9300381 de Méndez, Cunduacán and Nacajuca 51* 2010 Nacajuca Desilting of pilot bed of river Samaria (stretch II) in the municipalities of Jalpa 506735.88 2013296.6 18.2088797 -92.93629608 de Méndez, Cunduacán and Nacajuca 52* 2010 Nacajuca Desilting of pilot bed of river Samaria (stretch III) in the municipalities of Jalpa 506557.13 2012496.72 18.20165058 -92.93798915 de Méndez, Cunduacán and Nacajuca 53* 2010 Centro Construction of spillway channel, protection embankments and traffic bridge 512582.84 1979781.2 17.90592219 -92.88120282 of floodway Sabanilla on the right bank of the river La Sierra, municipality of Centro 54* 2010 Centro Construction of stretch of road, left embankment and floodway of lake Los 516235.9 1990828.45 18.0057503 -92.84662754 Zapotes-lake Don Julián, municipality of Centro 55 2010 Centro Construction of channel on the right bank and curtain of control structure 470242.64 1985492.04 17.95737367 -93.28102576 over the river Carrizal, and ancillary works in the basin of the river Grijalva 56* 2010 Centro Construction of protection embankment against flooding, stretch IV; 501691.53 1987112.94 17.97222716 -92.98402396 residential area Islas del Mundo, Santa Elena and Isla in settlement Miguel Hidalgo, municipality of Centro 57* 2010 Centro Construction of pile wall on the left bank of the river Carrizal, stretch Puente 509925.91 1992485.04 18.0207616 -92.90622688 Carrizal 4 and street Emiliano Zapata, as well as downstream from bridge La Pigua, community Indeco, municipality of Centro 57 ANNEX 10 LIST OF INVESTMENTS IN DRR WITH UNKNOWN LOCATION Year of completion Municipality Investment 2008 Nacajuca Restitution of the embankments with their corresponding bank protections in settlement Libertad 2008 Nacajuca Restitution of embankment and bank protection in settlement El Cedro, left bank of river Carrizal 2009 Cárdenas Desilting of carrier drain San Felipe in District 012, Chontalpa, municipality of Cárdenas 2009 Emiliano Zapata Construction of bank protection in the riverside promenade in the city of Emiliano Zapata 2009 Balancán Replacement and construction of riverside promenade in the city of Balancán 2010 Jonuta Construction and rehabilitation of the protection wall and embankment on the right bank of river Usumacinta, city of Jonuta, municipality of Jonuta 2010 Emiliano Zapata Construction and rehabilitation of the protection embankment of the stretch Carretera Federal Villahermosa-Escárcega to the road junction to bridge Trapiche, Villa Chablé, municipality of Emiliano Zapata 2010 Jonuta Construction of protection embankment with coarse gravel at the settlement El Sacrificio, in Jonuta 2010 Balancán, Emiliano Construction of embankment with coarse gravel and of a sandbag protection wall against floods in population centers in various communities of the Zapata, Jonuta and municipalities of Balancán, Emiliano Zapata, Jonuta and Tenosique Tenosique 2010 Balancán Construction of bank protection of the stretch II of the riverside promenade and trail on the right bank of the river Usumacinta, city of Balancán, municipality of Balancán 58 Funded by