Floods and Poverty Traps: Evidence from Bangladesh
There are large negative effects of floods on the spatial incidence of poverty. These effects are especially strong in the short term in the immediate aftermath of major floods though there also appear to be longer-term negative effects. However, normal flooding is necessary and beneficial for agriculture, transport and fisheries. The probability of catastrophic flooding in Bangladesh is about once in a decade, and micro-flood-insurance is one important policy tool to mitigate the effects of catastrophic flooding on the poor.
AMRITA DASGUPTA
W
Bangladesh is unusually disaster prone, with natural disasters claiming the bulk of human losses (Table 1). Floods are the second most frequent occurrence after windstorms and cause the greatest losses. Bangladesh lies in a deltaic floodplain, draining three major rivers – the Ganges, Brahmaputra and Meghna – and is crisscrossed by numerous channels and small rivers. The hydrometeorological characteristics of the three basins are unique and their combined discharge is among the highest in the world, at 1,214 × 109 cu m. The peak run-off depth, at 8.5 m, is about the highest in the world.
With such high numbers of run-off discharge and depth concentrated in the peak monsoon season (80 per cent annual precipitation occurs between June and September), the occurrence of severe flooding is especially likely. More than 90 per cent of the land is made up by fertile alluvial plains less than 10 m above sea level. The low-lying land is also prone to catastrophic floods from cyclones and storms during the monsoon season. In terms of total land area subject to risk of flooding, Bangladesh is, in fact, the most flood-prone country in the world. Normal flooding (borsha) affects roughly 25 per cent of the land every year. Abnormal flooding (bonya) has been known to submerge more than 50 per cent of the total land area [Food and Agriculture Organisation 2004]. Roughly 92.5 per cent of the source of drainage lies outside the boundaries of the country [NASA 2005].
Floods and poverty are inextricably linked. Over millennia, people have naturally settled in flood plains and river basins. The nature of river basins is such that their land is highly fertile, perfect for intensive agricultural settlements. Indeed, many sites of civilisation have grown in river basins. The Nile in Egypt, the Ganges in India, the Huang He in China, the Mekong in south-east Asia, and, in modern times, the Mississippi of the US are examples. Over time, many of these river basins have developed some of the highest population densities in the world, with delayed demographic transition consequent to poverty [Wright 2005: 153] trapping many such countries in the middle phases. With these growing population densities, the carrying capacities of the land become strained. Poverty and low education levels then become prevalent and poverty, environmental effects and high fertility rates may become locked in a selfperpetuating cycle [ibid: 157]. We show some of these comparative characteristics of population, population density and poverty in the five major river basins (mentioned above) and in Bangladesh in Table 2.
In people’s attempts to control the rivers on which their livelihoods are based, the natural channels of water flow are often disrupted, and the resulting silt levels in rivers have become a problem. The Brahmaputra and Ganges, for example, each carry suspended particle loads in excess of 700 million tonnes per year, more than thrice the mean annual particle discharge from the Mississippi (200 million tonnes per year), while the sediment load of the Huang He is the highest in the world. In attempts to resolve this, levees and dikes have been built with varying success, so that floods and their adverse effects on the poor have become endemic.
Geography has emerged as a key underpinning of poverty in many parts of the world. Jeffrey Sachs (2000) has argued that geography can create poverty traps. Periodic disasters such as epidemics and flooding, he suggests, can create conditions where it is incredibly difficult for the poor to escape from poverty through normal routes. The existence of spatial poverty traps has been widely ignored in the economic development literature. “Until very recently, the outpouring of econometric studies of cross-country economic growth neglected physical geography as a relevant dimension of analysis”, notes Sachs (2000). Floods can devastate the physical and social capital of societies and destroy whatever tiny amounts of savings poor households have and when these floods are as frequent and catastrophic as they are in Bangladesh and similar flood basins, the effects can be ruinous.
I The Research Question
The paper examines statistical evidence to establish the significance (or otherwise) of the impact of floods as a major cause of persistent poverty in Bangladesh. This research question is a subset of the expanding class of recent inter-disciplinary research that seeks to understand enduring poverty in poor or “backward” regions around the world in terms of causes that now range from tropical diseases [Sachs 2000], to ecology, transport and communications costs, inland isolation, wars
Economic and Political Weekly July 28, 2007
and conflict, demography, biology and other social and cultural factors.
Poverty is primarily a household phenomenon. As Amartya Sen (1981) notes, “the family rather than the individual is the natural unit as far as consumption behaviour is concerned”. Procedures to ascertain in which welfare category, poor or nonpoor, families fall typically identify a minimum requirement of food and non-food consumption based on national norms, and measure individual household consumption against these defined values (termed as “poverty lines” or monetary equivalents). The aggregation of the group of the poor in a community then involves moving to an overall measure, most commonly the “headcount” measure, which is expressed as the ratio of the total number of poor as a proportion of the total numbers in a given community. As a result, studies on the causes of poverty have traditionally tended to focus on the household as the unit of analysis. They have primarily emphasised household attributes, such as age, sex, size, education, health, occupation, and other characteristics of primary income earners and members as the main determinants of poverty. In the literature, the main geographic element commonly referred to is the distinction between rural and urban households, where typically the sample is partitioned into these two categories. Separate poverty lines are defined for the two, since minimum standards vary.
Floods, however, are a geographic- or location-specific occurrence that in principle affect all households in a given area without any necessary distinction in their effects between the poor and non-poor. Yet, we propose that the effects of flooding are asymmetric in their impact on poor versus non-poor households. There are at least four possible ways in which poor households might be affected more than non-poor households: (i) if the numbers of poor households are unusually concentrated in lowlying flood-prone lands because, for example, land is cheaper, and are unable to migrate out of the areas due to social and economic constraints and costs; (ii) if poor households have some economic incentives to remain in flood-prone areas, such as the positive effects of non-catastrophic floods on their sources of incomes; (iii) if catastrophic floods are infrequent and relatively random or unpredictable events such that poor households living in flood-prone lands may be myopic; and (iv) if such catastrophic floods have inter-generational or longer-term negative effects from destruction of physical capital assets and human capital of the poor over time. All four factors can individually and cumulatively lead to the flood-induced spatial or geographic poverty trap referred to earlier, as opposed to non-location specific causes of household poverty.
The distinction between geography-specific and householdspecific causes of poverty is likely to be crucial for the design of public policy to address poverty. For instance, two very different questions have emerged in respect of the recent floods and their effects on the poor in New Orleans, USA in the wake of the recent Katrina disaster: (i) Did the disaster primarily highlight the problems of poverty and race among households in inner-cities, common to New Orleans and other US cities (whether flood-prone or not)? Or, (ii) Did it highlight the effects on poor households of living in flood-prone lands in the Mississippi Delta, independent of community and household characteristics? If it is the former, then public policy interventions should aim primarily at assisting disadvantaged households; and if it is the latter, then interventions targeted at mitigating the impact of floods become necessary for households living in flood-prone lands.
II Testing for the Effects of Floods on Poverty in Bangladesh
We begin with a basic model of how floods (F) might affect poverty rates (P) at the household level. Since poverty is generally affected by a number of non-flood (N) factors as well as floods
Table 1: Occurrence of Natural Disasters in Bangladesh 1904-2005
No of Events Killed Injured Homeless Affected Total Affected Damage $ (000’s)
Drought 5 18 0 0 25002000 25002000 0 Average per event 4 0 0 5000400 5000400 0 Earthquake 6 34 625 15000 3500 19125 0 Average per event 6 104 2500 583 3188 0 Epidemic 28 403103 0 0 2757519 2757519 0 Average per event 14397 0 0 98483 98483 0 Extreme temperature 15 1891 2000 0 84000 86000 0 Average per event 126 133 0 5600 5733 0 Famine 11900000 0 0 0 0 0 Average per event 1900000 0 0 0 0 0 Flood 65 50082 102220 32853724 296714773 329670717 15665100 Average per event 771 1573 505442 4564843 5017857 241002 Wave/surge 2 3 10 12000 0 12010 0 Average per event 2 5 6000 0 6005 0 Windstorm 144 614132 874633 9961443 53009561 63845637 3008880 Average per event 4265 6074 69177 368122 443373 20895
Source: ‘EM-DAT: The OFDA/CRED International Disaster Database’, www.em-dat. net - Université catholique de Louvain - Brussels - Belgium.
Table 2: Comparative Flood Plain Statistics, 1990s
Bangladesh Indo-Gangetic Basin Mekong River Basin Yellow River Basin Nile River Basin Mississippi Basin (Delta)
Area(km2) 1,45,000 860,000 795,000 795,000 3.3*106 3.2*106
(94,000) Rainfall (mm) 2030 1160 1250-1875 478 615 835 (1250) Annual discharge (m3) 1214*109 500*109 437*109 58*109 84* 109 500*109 Population statistics (million) 144 440 63 158 591 70 (10 for Delta) Poverty indicators (per cent) 50 15 - 55 28-60 n a 50 24 (50 for African-American)
Source: Author’s compilation, based on various country data sources and from CGIAR Challenge Programme for Water and Food; United States Interagency Flood Plain Management Review Committee, 1994; United States Improvement of Housing and Infrastructure Conditions in the Lower Mississippi Delta, Housing Assistance Council, 2000; USGS Fact Sheet 2005-3020, March 2005.
Economic and Political Weekly July 28, 2007 (F), the basic model has to incorporate both sets of factors. To quantify the impact of floods on poverty in Bangladesh, we apply a multivariate regression analysis [Barrow 1988] and determine the statistical significance of the effects of floods (F) versus nonflood factors (N) on poverty. Measures of poverty thus make up the dependent variable that is proposed in the underlying model to be affected by both non-flood and flood variables.
The Model
According to the Ravallion and Wodon (1997) model of poverty, a household’s consumption is determined as follows:
logCi = a + B’Xi + Δ’Di + εi …(1)
Where Xi and Di are vectors of non-geographic (household demographics, education, occupation, etc) and geographic (urban, rural, district, etc) variables, respectively (and ε is the random error term). Equation 2 shows that poverty is also a linear function of consumption, and hence of geographic and non-geographic factors. More formally, where the nominal per capita consumption for a household i is denoted by Ci, N the population size, wi the weight for household i, and Z the poverty line, the incidence of poverty P is defined as:
P = ΣCi<Z (wi/N)[(Z-Ci)/Z] …(2)
From these two equations, we can deduce that the same variables affecting consumption, therefore, affect poverty in a similarly linear fashion. Thus we propose that a household’s poverty (Pi) is a function of non-geographic (X) and geographic (D) variables. By derivation we assume the same model holds true with respect to other measures of poverty. Since our interest primarily focuses on floods, we distinguish basically between two types of factors, flood and non-flood variables, and our general model follows:
Pi = a + bNi + cFi + εi …(3)
where F is a flood-specific geographic variable primarily measured by the extent of vulnerability to flooding (flood-prone districts), and N is a vector of non-flood geographic and nongeographic variables (population density, rural, urban, presence of major cities, etc).
Dependent Variables
To ensure robustness, three different poverty measures are used in our regression analysis: (a) poverty headcount ratios (P0)
(b) poverty gap ratio (P1) (c) human poverty index (HPI).
Poverty headcount ratio is as defined earlier, basically an aggregation of numbers of households below a given poverty line as a ratio of the total population in any given spatial area. The poverty gap measures the extent by which the average household’s mean consumption falls below the defined poverty line, thereby measuring the depth of poverty in a given region. Since the poverty gap increases with the distance of the poor below the poverty line, it is a better measure of the depth of poverty than the headcount ratio. The sources of data for the poverty ratios used in this paper are the Bangladesh Bureau of Statistics [BBS 1991; BBS 1995a; BBS 2002] and Ravallion and Wodon (1997) for some specific geographic data simulations on poverty in Bangladesh that are already conditioned on nongeographic household characteristics. The last data source is especially important because it isolates the pure
Table 3: Bangladesh – Floods and Poverty Determinants at Greater District Level, 1988-99
| Dependent Variables Poverty Measure Year | Intercept | Flood-Prone | Independent Variables Non-Flood | N | R2 | ||||
|---|---|---|---|---|---|---|---|---|---|
| Districts | Rural | Literacy | Major City | Pop Density | |||||
| P0 = headcount | 1989 | 12.83 | 17.4*** | 16.11*** | 1.28** | -17.98*** | -0.01** | 34 | 0.58 |
| t-Stat | 0.96 | 2.6 | 4.01 | 2.06 | -3.75 | -2.11 | |||
| 1991 | 10.89 | 22.45*** | 15.12*** | 1.64*** | ‘-18.3*** | -0.01*** | 34 | 0.72 | |
| t-Stat | 1.09 | 4.44 | 4.97 | 3.47 | -5.05 | -3.67 | |||
| Geographic conditional1 | 1989 | 10.75 | 9.44* | 7.93*** | 0.95* | -16.92*** | -0.01* | 34 | 0.53 |
| t-Stat | 1.04 | 1.82 | 2.54 | 1.97 | -4.54 | -1.8 | |||
| Geographic conditional1 | 1991 | 19.52 | 23.23*** | 5.59 | 1.64*** | -19.06*** | -0.02*** | 34 | 0.56 |
| t-Stat | 1.54 | 3.63 | 1.46 | 2.75 | -4.16 | -3.49 | |||
| Change in ΔP0 Geographic conditional1 | 1989-91 t-Stat | 8.78 0.85 | 13.79*** 2.64 | -2.33 -0.74 | 0.68 1.41 | -2.14 -0.57 | -0.01*** -2.48 | 34 | 0.25 |
| P0 = headcount | 1999 | 35.81*** | 8.48** | 1.25 | 0.55* | -3.52 | -0.01* | 34 | 0.25 |
| t-Stat | 5.37 | 2.51 | 0.62 | 1.76 | -1.46 | -1.88 | |||
| P1= poverty gap | 1999 | 7.4*** | 2.71** | -1.13 | 0.28** | -3.06*** | -0.001 | 34 | 0.38 |
| t-Stat | 2.86 | 2.08 | 1.45 | 2.37 | -3.28 | -1.5 |
Notes: *** Denotes coefficient statistically significant at 99 per cent confidence level. ** Denotes coefficient statistically significant at 95 per cent confidence level.
* Denotes coefficient statistically significant at 90 per cent confidence level. 1 Geographic conditional denotes poverty headcount ratios conditioned on household characteristics [Ravallion et al 1997].
Table 4: Bangladesh – Floods and Poverty Determinants at Lower District Level, 1995-2000
| Dependent Variable | Independent Variables | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Poverty Measure | Year | Intercept | Flood-Prone | Coastal | Tribal | Non-Flood Major City | Pop Dens | Wage Rate | N | R2 |
| (HPI) | 1995 t-Stat 2000 t-Stat | 45.36*** 16.89 42.28*** 16.37 | 2.31** 2.22 1.78** 1.98 | -2.26 -1.5 -2.78** -2.41 | 11.05*** 4.99 6.56*** 3.78 | -3.49*** -3.15 -2.52*** -2.66 | 2.67E-5* 1.74 1.38E-05 1.05 | -0.4*** -2.94 -0.32*** -3.26 | 64 64 | 0.49 0.44 |
Notes: *** Denotes coefficient statistically significant at 99 per cent confidence level. ** Denotes coefficient statistically significant at 95 per cent confidence level.
* Denotes coefficient statistically significant at 90 per cent confidence level.
Economic and Political Weekly July 28, 2007
geographic variations in poverty, which we can then use to test uniquely for the relative importance of floods as a determinant of such poverty, which has not been done before to our knowledge.
The HPI is a non-income dimension of lack of well-being or deprivation proposed by the United Nations Development Programme (UNDP). It is a broader measure which takes into account life expectancy, education, and standard of living [Wright 2005].1 The HPI district data source for Bangladesh is Binayak Sen and David Hulme (2005).
Independent Variables
Focusing on floods, a spatial independent variable, we investigate these effects at three different levels, starting with the older 20 greater districts (distinguished by rural and urban areas). Next we move to a much more disaggregated level, into the 64 administrative districts of current-day usage in Bangladesh. We conclude with a finer analysis, still using data at the village level.
The flood variable (F) is a dummy variable, taking the value of 1 in flood-prone districts and 0 in non-flood-prone districts. Areas were designated as flood-prone based on flood depth and inundation levels, as well as proximity to areas subject to major river flooding. Data concerning these variables were taken from the Food and Agriculture Organisation (FAO), the Bangladesh Agricultural Research Council, and the UNDP [Geographical Information System Project 1999]. The GIS maps for reference are available with the author This was supplemented and corroborated by actual flood incidence maps and data from major occurrences (2004, 1998, 1987).
Other non-flood independent variables (N) used are a rural dummy (distinguishing between rural and urban areas, and taking the value of one when poverty is measured in a rural area, zero otherwise in urban areas), literacy rates, population density, presence of major city dummy (five major identified cities in Bangladesh), and tribal/coastal dummies as appropriate. The data source for these values was the BBS (1995b).
III The Results and Analysis
Poverty Headcount/Poverty Gap
The results of the regression analysis at the greater district level are first reported in Table 3. The most powerful and consistent result is that the coefficient of the flood-prone districts’ dummy variable is highly significant, statistically, and carries the expected positive sign in all estimated equations. This indicates that flood-prone districts tend to have consistently greater headcount ratios of poverty, confirming the geographic trap thesis in flood-prone lands.
The second important finding is that the quantitative effects vary significantly between years. For example, in the 1988-91 period, the headcount poverty ratios are the highest and jump in all districts and much of this jump is attributable to the sharp rise in both the size and significance of the flood-prone variable. This corroborates well with our hypothesis that floods have much larger short-term effects. The period 1988-91 is especially important in highlighting the short-term effects for Bangladesh, due to the catastrophic flooding that took place in 1988 [Brammer 1989].
The third finding is that not only do floods affect headcount ratios, they also affect the poverty gap ratio, and therefore the depth of poverty. The conditional geographic equations also point to the same effects. The overall results suggest that floods are inextricably linked to greater poverty incidence in the most floodprone lands and areas in Bangladesh.
Of the other variables in the equations, the location of households in a rural area significantly raises the probability of being poor. This is consistent with expected results since rural poverty is the dominant phenomenon in an agricultural setting such as Bangladesh. The presence of a major city positively reduces the poverty incidence in districts by a factor that is as quantitatively significant as that for flood-prone areas. The explanation for such an occurrence is widely known, in that cities offer an escape from rural poverty and a widening of opportunities for earnings. Migration to cities would appear to be one of the major escape routes out of poverty [Wright 2005: 132].
Two puzzling results are also evident from the regressions. First, population density does not appear to be negatively associated with higher poverty, in contrast to usual expectations that higher population density should result in greater poverty. In fact, the statistical significance and quantitative impact remains low. Second, and more puzzling, is the apparent positive relationship between literacy and higher rates of poverty. Again, the statistical significance and size ofthecoefficient isrelatively low. We suspect that since the effects of literacy are being captured at the household level already, and because the presence of the major city variable is also inevitably further capturing the effect of education in lowering poverty, that the remaining spatial effects are probably indicative of some areas in Bangladesh with relatively high literacy as well as high poverty. Typically, this might be seen in the coastal and tribal hill tract areas of Bangladesh, which we shall explore in the next section, covering HPI.
Table 6: Bangladesh – Occurrence of Floodsby Intensity, 1960-2005
Affecting >10million >15million >30million All Floods Catastrophic>10m Catastrophic>15m Catastrophic>30m
1960-653 0 0 0 1966-703 2 1 0 1971-753 1 1 1 1976-804 1 0 0 1981-854 1 1 1 1986-905 2 2 1 1990-955 2 0 0 1996-2000 5 1 1 0 2001-055 1 1 1
Source: Author’s estimates, derived from EM-DAT: The OFDA/CRED International Disaster Database, www.em-dat.net – Université catholique de Louvain-Brussels-Belgium.
Table 5: Bangladesh – Relationship between Floods and Wages at the Village Level
| Dependent Variable | Year | Intercept | Flood Depth | Flood Depth2 | Independent Variables Flood Depth3 Duration | N | R2 | |
|---|---|---|---|---|---|---|---|---|
| Daily wages | 1998 | 45.5*** | -1.6 | 3.6* | -0.3** | -0.1 | 30 | 0.52 |
| t-Stat | 6.8 | -0.2 | 1.9 | -2.4 | -0.9 |
Notes: *** Denotes coefficient statistically significant at 99 per cent confidence level. ** Denotes coefficient statistically significant at 95 per cent confidence level.
* Denotes coefficient statistically significant at 90 per cent confidence level.
Economic and Political Weekly July 28, 2007
HPI
The results using HPI as the dependent variable are reported in Table 4. Although HPI is a much broader variable of human deprivation, floods continue to have a major effect in raising this deprivation index. Note that neither 1995 nor 2000 were flood years, so that we are picking up in the size of the coefficient relatively longer-term effects of floods on the poor in flood-prone districts. As we discussed earlier, in a major flood occurrence, it is very likely that much stronger short-term effects would come into play. The presence of a major city continues to reduce deprivation levels, again consistent with our previous findings. The population density variable is no longer significant. Tribal areas have markedly higher deprivation indices, and corroborate our inference earlier. Living in coastal areas, however, appears to provide a lower HPI, denoting better living conditions. Finally, higher wage rates naturally lead to lower HPI, as expected.
Flood Depth and Duration
The literature suggests that not all floods have detrimental effects. Small annual floods are beneficial and in fact necessary for agriculture, such as crop cultivation, fisheries, transports, and other trades. Small-scale flooding fertilises the soil and replenishes water bodies. Therefore, the relationship between flooding and poverty is likely to be more complex than a simple linear relationship that we have so far alluded to.
In order to understand this complexity, we have to get much closer to an understanding of the impact of floods. To do this, we use a unique data set at the lowest possible spatial level, that of a village level, where data on extent of normal flooding and daily agricultural wage rates were collected for 1998-99 (primary data from the special FMRSP survey were provided by International Food Policy Research Institute 1999). We regress flood depth, the square of flood depth, and the cube of flood depth, as well as flood duration (measured in days), to gain this understanding of the effects on wages. The results are shown in Table 5.
The results clearly demonstrate our hypothesis on the complexity and non-linearity of the relationship. The data suggests that in place of a simple linear relationship, the relationship is strongly positive, with rising wages associated with higher normal floods up to a certain turning point of around five feet. Beyond this amount, the relationship reverses, turning negative. Flood duration has the expected negative sign, as the longer flooding remains, the worse for land and human life. Crops cannot grow as well with excess water. It is not, however, statistically significant.
IV Conclusions and Policy Implications
The broad results of this paper are important in highlighting the persistently negative effects that large-scale floods have on creating poverty in populations of countries such as Bangladesh, where people’s lives are tied to the vicissitudes of flood plains. We believe this is the first paper to examine the quantitative size of the effects of floods on poverty incidence after controlling for the impact of other variables, as well as the variation of this effect across time and using different measures
Economic and Political Weekly July 28, 2007
of poverty. These negative effects of floods are especially strong in the short term, the immediate after effects of major floods, as we see in the case of the 1988-91 episode. There also appears to be clearly persistent longer-term effects, because catastrophic flooding depletes savings, capital, and assets, both physical and human.
Our results, however, also suggest that normal flooding is also necessary and beneficial. For this reason, policies need to focus not on preventing flooding, but on addressing specifically catastrophic flooding. Systems such as levees are useful only in preventing smaller flooding, and may prove unable to handle large catastrophic floods, often making the effects worse. So the search for a solution for river floods that takes into account normal, beneficial flooding as well as catastrophic flooding, has to be built on different foundations, including housing strategies, early flood warning systems, pre-flood evacuation, raised evacuation roads and bridges, and increasingly important, household flood insurance and financial support for the poor.
To be viable, the last suggestions must hold economic interest for both parties involved: the affected as well as the insurers. The frequency of catastrophic flooding is a very important policy question in countries such as Bangladesh. Complex river models and probabilities are often used to derive estimates of the probabilities of such catastrophic floods. Unfortunately, they have often proven far too optimistic, with standards such as once in a hundred years. In Bangladesh, we show instead (Table 6) that using a very simple breakdown of flood occurrence by subperiods between 1960 and 2005, and depending on the tolerable threshold of their impact on the population, catastrophic floods can be expected to occur roughly once in 10 years. Our estimates are consistent with the results of a much more careful and longer work, where Messerli et al (2000) estimate from reconstructed flood data going back to 1870 that 12 major floods are recorded over 128 years in Bangladesh, with alternating phases of more frequent floods (1902-22) and less frequent ones (1923-54).
As we calculate in Table 6, the frequency of floods that affect greater than 10 million people (roughly 7 per cent of the population) are relatively normal in Bangladesh. Therefore, people are likely to have developed coping strategies for such events. It is only when we move into the category of floods that affect more than 30 million people that these become truly catastrophic events, occurring roughly once every decade. These events in the recent past have included the 2004, 1998, and the 1988 floods.
It is possible to find innovative insurance schemes for such infrequent yet catastrophic events at the very micro level for poor households. Peter Hazell (2004) outlines the requirements for any insurance system to be viable. Micro-insurance, managed by non-governmental organisations, is a realistic and practical option. Hazell poses that for micro-insurance to be a viable option, insurance products need to be both affordable, and equally importantly, accessible. Such plans have been suggested for “rainfall or drought” insurance, where insurance payments are indexed against average annual rainfalls and awarded to policy holders where a drop in rainfall occurs below a certain level. This is being explored in Nicaragua by the World Bank, and also in India. Similarly, flood insurance might be implemented along the same guidelines, where insurance payments are awarded to households when river flood levels exceed a certain threshold.
Problems with micro-insurance strategies also exist. Smallerscale systems with low capitalisation might be inadequate for dealing with such large-scale disasters such as flooding in Bangladesh: i e, if it concurrently affects the entire community.
Additionally, with such schemes as insurance, well-regulated and competitive markets become necessary. Generally, insurance has had little entrance into developing societies such as Bangladesh and India. Alternatively, economists have suggested remedies ranging from diversification of income to public works. It is clear that a multi-pronged approach is often necessary to address the effects of natural disasters and their impacts on the poor, but financial assistance and risk-insurance schemes must play a
EPW
Email: amritadg@gwu.edu
Note
[I am very grateful to Binayak Sen (World Bank), and Tarik Yousef (Georgetown University) for their guidance, comments and suggestions on the paper. I am also grateful to them for referring me to some of the data sources used in the paper, and to Bejoy Dasgupta (IIF) for his helpful suggestions and guidance on the statistical analysis. The research was conducted as part of my environmental studies work for a science project.]
1 HPI is defined as a weighted average of three factors. The first factor, P1, is measured by the percentage of people not expected to survive to the age of 40. The second factor, P2, represents the percentage of illiterate adults in the population. The final factor, P3, representing the standard of living factor, is itself dependent on three other factors: percentage of the population without access to safe drinking water, percentage of the population without access to health services, and percentage of severely malnourished children under the age of five. P3 places equal weight on each of the three factors, quite simply an average of these indicators. Thus, with geometric weighting, HPI is defined as:
333)] 1/3
HPI = [1/3(P1 + P2 + P3
References
BBS (1991): Household Expenditure Survey 1989, BBS, Dhaka.
Brammer, H (1989): ‘Floods in Bangladesh: Geographical Background to the 1987 and 1988 Floods’, The Geographical Journal, Vol 156, No 1, March 1990, pp 12-22.
Food and Agricultural Organisation (2004): ‘Gateway to Land and Water Information: Bangladesh National Report’, Internet http://www.fao.org/ ag/agl/swlwpnr/reports/y_sa/z_bd/bd.htm. Accessed: October 10, 2004.
Hazell, Peter (2004): Potential Role for Insurance in Managing Catastrophic Risks in Developing Countries, International Food Policy Research Institute, Washington DC, http://www.ifpri.org/media/issue brief/ ib_markets.htm. Accessed December 8, 2004.
IFPRI (2000): Bangladesh Coping Strategies FMRSP-IFPRI Household Survey, 1998-99, International Food Policy Research Institute, Washington DC.
Messerli, Bruno, Martin Grosjean, Thomas Hofer, Lautaro Nunez, Christian Pfister (2000): ‘From Nature Dominated to Human-Dominated Environmental Changes’, Quanternary Science Reviews, 19, pp 459-79.
NASA Visible Earth Satellite Images (2005): (Jacques Descloitres, MODIS Rapid Response Team, NASA/GSFC). Internet: http:// visibleearth.nasa.gov/view_rec.php?id=4578. Accessed 10/14/05.
Ravallion, Martin and Quentin Wodon (1997): ‘Poor Areas, or Only Poor People?’, World Bank Policy Research Working Paper No 1798, World Bank, Washington DC.
Sachs, Jeffrey (2000): Tropical Underdevelopment, CID Working Paper No 57, December, X Centre for International Development at Harvard University, Cambridge.
– (2005): ‘Can Extreme Poverty Be Eliminated?’, Scientific American, August. Sen, Amartya (1981): Poverty and Famines: An Essay on Entitlement and Deprivation, Oxford University Press, Oxford.
Sen, Binayak and David Hulme (ed) (2005): ‘Chronic Poverty in Bangladesh: Tales of Ascent, Descent, Marginality and Persistence’, The State of the Poorest in Bangladesh 2004-05, Bangladesh Institute of Development Studies and CPRC, Dhaka.
Wright, Richard T (2005): Environmental Science, Pearson Prentice Hall, New Jersey.
Economic and Political Weekly July 28, 2007
