Our starting point was that it is important to come to an agreement on what the problems are, in order to find solutions to them. We remain, nonetheless, open to hear your alternative proposals. In any case, we wish to still leave you time to reflect until the end of this month. Beyond that, we will be forced to launch actions in each of our countries to make this conflict, which we do not know how to resolve together, public.
We will do what is necessary with partner NGOs around the world to give these actions the greatest visibility possible. We therefore ask you to consider this letter as a forewarning. Do know that we sincerely regret to have reached this point. In order to find sustainable and definitive solutions together, we would be honoured to count on your personal attention to this matter so that we may advance in the road to dialogue.
In the hope of hearing from you, please accept our distinguished greetings. Original source: ReAct. It was desired to include practical methodologies that took a broad perspective on disaster risk, vulnerability and resilience. The following criteria for the review were selected:. Composite indicator including those on a spatial basis or scorecard approach. These are the methodologies that are being studied — more model based approaches which produce results in terms of a 'real' number, such as average annual losses, were excluded.
Studies focussed on post-disaster recovery are excluded. This was chosen to ensure relevance to disasters. This was chosen to ensure the review remained practically small as there are a large number of single hazard risk index methodologies and as single hazard risk indices tend to be much more focussed on physical risk and less inclusive of social or economic aspects.
Communities or governments are targeted national or sub-national in scope , not households or individuals nor single sectors. Full methodology is published or otherwise publicly available. This was considered important to enable analysis of the practices used in the methodology. Focus is on present day — climate vulnerability studies of the future are excluded. An extensive search of the academic and grey literature was undertaken that used the VuWiki, 15 Scopus, Web of Knowledge and Google Scholar, as well as forward and reverse citation searching utilising a snowball approach.
A Google Web search was also conducted to capture methodologies reported in the grey literature. The search sought articles published between 1 January to 31 March The review found documents of potential interest which were, upon further review against the criteria, narrowed to documents detailing methodologies or implementations.
These are listed at Annex 1. A list of methods initially captured that were subsequently excluded is at Annex 2. Figure 1 displays the countries where the lead authors or their institution of these methodologies are based. A large number have been developed by researchers in the United States and Western Europe, although academics in China have been very active in index development and a small number of researchers in developing countries have also developed indices.
Data was gathered on the title, author, year of first implementation, geographic areas used, variable selection, data collection, imputation, normalisation and weighting, index structure, variables used, outputs and the use of any sensitivity or uncertainty analyses based on the review by da Silva and Morera, 14 and the index construction practices in the OECD's Handbook on Constructing Composite Indicators: Methodology and User Guide.
The variables used in each index were recorded and grouped into sub-indicators, indicators, categories and environments based on the phenomena each variable was measuring. This classification hierarchy is illustrated in Figure 2. These were analysed to determine the frequency of use of different concepts across different methodologies as well as the composition of each methodology. The methodologies analysed can be divided into five groups based on similar approaches to purpose, data gathering and index construction:.
This category contains the most variation, ranging from simple indices using a few equally weighted variables to complex, multi-layered hierarchies with weights selected through more rigorous methodologies. The majority have been similar implementations of SoVI in jurisdictions outside of the USA using alternative variables, but some have used PCA in alternative contexts, including with data gathered by community survey. These methods are focussed at the sub-national level, as PCA typically requires a large number of study units to produce reliable results.
As it is a data reduction technique it is also suitable for the data rich environments of developed countries where large statistical agencies collect comparable data across many small areas. PCA can be implemented in a range of different ways, with a key choice being the rotation method used in constructing the principal components.
Where the rotation method used was listed they have all used varimax rotation as a means of minimising the number of factors, due a desire to attempt to explain the conceptual significance of each factor. However it is also likely that the use of a varimax rotation in Cutter's SoVI was also influential. These methods have been mostly developed for the use of communities or governments as a self-assessment tool and as such focus on explicit elicitation of disaster preparedness and risk reduction outputs. These methods generate an index based on analysing the relationship between vulnerability inputs and disaster impacts using either simple or multiple linear regression or Data Envelopment Analysis DEA.
Four methods in the literature used more advanced construction methods, which have not been broadly deployed. They feature novel use of statistical methods and simulation to produce the index result, which potentially makes them more difficult to understand and less transparent to end users.
They are:. The Local Disaster Index. Produced as part of the IADB suite of disaster indices the Local Disaster Index attempts to identify the impact of small-scale disasters on national and local development in a country. It utilises localised historical disaster data and a series of equations to produce an index that indicates how widespread and persistent small and moderate disasters are in a country's territory.
Geoscience Australia's Social Vulnerability. This approach developed by Geoscience Australia synthesises a vulnerability index for individual households in small areas, which is then summed to produce the area index. It uses synthetic micro-estimation with census data to produce simulated households in the area. Scenario analysis with complex decision trees, which were constructed based on population survey, was then utilised to determine individual household vulnerability.
These household values were then statistically analysed to produce vulnerability values for each area. To develop a risk index in China Jin et. This method also enabled the provision of confidence estimates on the resulting vulnerability values. Ge produced a social vulnerability index using the Projection Pursuit Cluster PPC method, which is a data reduction and rotation methodology similar to Principal Component's Analysis. Figure 3 shows the year of development or submission for publication of each methodology.
In the several years after Briguglio published their index in a small number of composite indicator methodologies were published. The rate of publication increased through the middle s, however during the past five years there has been a large increase in the rate of publication with close to two-thirds of composite indicator methodologies developed since Table 1 shows the geographic level at which each methodology has been applied.
Those at the national level have studied nations or been completed on a global basis. Sub-national administrative units are studies completed in well-defined sub-national areas. Community level refers to studies in locations where the geographic boundary is not well defined or doesn't correspond to an administrative area. Multiple levels refer to studies which have applied the same methodology at both national and sub-national levels. Despite the high profile given to many indices that compare nations, they only comprise a fifth of the total. Three quarters are focussed on sub-national settings, with the majority of these based on well-defined administrative units.
This is consistent with the large number of data driven methodologies, as statistical data is typically provided on the basis of these territorial units. Only a small number of methodologies have been developed for application at multiple levels. This suggests that most authors are attempting to tailor their approach to a particular level and at least in this sense are not in danger of committing the ecological fallacy, a concern that has been raised in the literature. Of the 25 national or multiple level methodologies only 23 directly compare nations, with two of the multiple level methodologies taking a gridded approach to mapping the index value.
Of these 23 methodologies that compared nations only eight were global in scope, with the remainder prepared for particular regions or some other sub-set of countries such as developing nations or Small Island Developing States. The majority of sub-national methodologies have been applied in single countries with only nine methodologies applied in a cross-national context.
Only five of these are applied in more than two countries. The small number of methodologies that are applied across multiple countries is consistent with the literature, which has identified this approach as both difficult and a key gap in the field. These are displayed in Table 2 and Figure 4. In the vast majority of methodologies 90 the variables were chosen by expert judgement relying on the literature, theory models and stakeholder knowledge. A smaller number 9 utilised statistical analysis such as examining correlations to exclude redundant variables to assist in variable selection.
Three methods directly used stakeholders through workshops to select variables, whilst the remaining four used other means or a mix of approaches. In many cases developers cited variable inclusion in other indices in the literature as justification for inclusion, rather than empirical evidence or theoretical vulnerability and resilience frameworks. A variety of methods were used to collect data which are summarised in Table 3.
The majority 76 utilised existing data collected by national statistical agencies and other government or non-government organisations that gather sociological and economic data. For sub-national methods the data in many cases came from national censuses to enable the use of small areas. Due to the use of census data which is collected relatively infrequently, the data used in a number of the indices was up to 10 years old at time of publication. A smaller number of methodologies 10 used household surveys to gather relevant data. The more explicitly stakeholder focussed methodologies utilised either workshops 4 or surveys of relevant stakeholders 4 to collect information.
Expert advice was relied on in only two methods. A further ten studies adopted a mixed approach to data collection utilising data from statistical agencies, expert advice and stakeholder responses. The majority of methods 96 did not perform any imputation with a small number using either case deletion 2 or some form of single imputation 8 to deal with missing data. A more important factor in dealing with missing data is that, as discussed in the literature, a number of authors acknowledged that data availability influenced variable selection.
Many methodologies applied no normalisation to the data, either because it was not relevant to the aggregation method or because the data types were already consistent. Where normalisation was applied Min-Max methods were the most popular followed by standardisation, categorical scales and ranking. A range of other methods or mixed methods were used in other cases. The results are summarised in Table 4. A broad variety of methods for weighting variables in index construction have been deployed, including a number of bespoke methods.
The results are summarised in Table 5. Equal weighting of variables and indicators appears to be the default setting in index construction, having been used by 44 index methodologies. Where equal weights were used indices were constructed to either weight each variable equally or in some multi-level hierarchies to weight each branch equally.
When we reached there Colin had some breathing, he was making sounds. I'm from England keppra cijena Ellis joins some 2, other people who over the course of three days line up in the wee hours of the morning in the hopes of getting free medical care. Gently fry the pork for four or five minutes each side. The group gives no detail on its financial arrangements. He was real cool, and then after that we had a workout.
Two of the scorecard methods did not use aggregation of the variables and don't have any weighting. In some other cases 13 the study authors chose weights based on expert experience and the literature, but did not employ any rigorous participatory or statistical method for selecting weights. Nineteen indices used a participatory method of selecting weights. The most popular participatory method was the Analytic Hierarchy Process, used in 8 indices.
Eleven studies used other participatory methods included ranking, 30 , 31 DEMATEL analysis, 32 , 33 , 34 the Delphi method, 35 the Budget Allocation Process, 36 and workshop or other interview based assignment. Principal Components Analysis is the most popular statistical weighting method, used by 17 methods and typically implemented using the procedure developed for Cutter's Social Vulnerability Index.
In most cases 10 PCA factors were aggregated with equal factor weights. Three methods used the factor score which indicates the percent of variance explained as the factor weightings, three did not aggregate the factors and one was unclear. Other statistical methods employed include triangular fuzzy numbers and stochastic simulation, 28 data envelopment analysis, 28 , 41 , 42 multivariate regression against other indices, 28 , 43 multivariate regression against outcome measures, 44 , 45 , 46 microsimulation and decision tree analysis, 27 the projection pursuit cluster model, 29 the procedure developed for the Disaster Deficit Index, 47 and the procedure developed for the Local Disaster Index.
A variety of inductive and deductive approaches were used to construct the indices. Three approaches used no aggregation, three of the principal components approaches just provided the factors and used no further aggregation, 74 used a hierarchical approach, 14 combined factors from principal components analysis, whilst the remainder employed other mostly relational approaches.
Of the hierarchical indices the majority 59 used a simple structure with aggregation at each level by arithmetic or geometric weighted mean. In the remaining 15, the theoretical models incorporated a risk, vulnerability or adaptive capacity equation which was typically at the top of the hierarchy, the equation's components being constructed as the weighted mean of the indicator variables.
Some of the methodologies based on PCA also employed a hierarchical approach in the combination of the factors, once they had been calculated. Numbers of variables in deductive methodologies with different numbers of intermediate levels between the variables and the reported index. Almost all methodologies provided some display of the results with maps and tables being the most popular as summarised in Table 7.
A small number of methods published no output whatsoever. Although many papers discussed the potential limitations of the methodology developed, only twenty have any explicit analysis of uncertainty or sensitivity. Three methodologies explored the sensitivity to alternative aggregation methods. Two assessed the use of different groups of variables. Only two methodologies provided estimated errors on the resultant vulnerability scores.
Of the methodologies included in this study only one comprehensive sensitivity analysis was undertaken, for Cutter's SoVI 50 which incorporated an investigation of different geographic levels and 54 unique variations on the index construction. However other work that failed to meet the inclusion criteria has also examined global approaches to sensitivity and uncertainty analysis. One other compared the results against ethnographic assessment and one community survey methodology validated the resulting index against a broader set of survey questions. There was some variation in the findings of the sensitivity analyses that were undertaken, with some finding low sensitivity to changes to the methodology and others finding high sensitivity to change.
There was a large variation in the number of variables each methodology used with the minimum being 2 and the maximum being , however most methodologies used relatively few with two thirds using less than The distribution of variables is illustrated in Figure 5. Most of the methodologies employed mining of national and international statistical databases.
Methodologies using this data collection technique used the fewest variables on average median 18, minimum 2, maximum Some of these techniques included a stage of variable exclusion based on correlation or other statistical analysis. Those using community surveys collected more variables median However the stakeholder focussed methods using surveys or workshops relied on substantially more variables median The methodologies used variables of which were unique.
An analysis of their frequency of occurrence found that the most used variables were dominated by common statistical indicators. This is consistent with the dominant data collection methodology. The 11 most common variables are shown in Table 8. As per the classification hierarchy shown in Figure 2 , the variables were grouped under Sub-Indicators, an average of 6.
The most common Sub-Indicators were strongly influenced by the most common variables with some Sub-Indicators, such as Household Water Access appearing despite not having component variables in the top This appears to be due to the larger number of ways of measuring certain sub-indicators. The 10 most common sub-indicators are shown in Table 9. The Sub-Indicators were grouped under Indicators, an average of 3.
At the Indicator level in the classification hierarchy the significant use of various variables describing age distribution in communities and properties of housing stock was demonstrated. The 10 most common indicators are shown in Table The Indicators were grouped under 15 categories. The number of methodologies that included variables from each of the categories is shown in Table This demonstrates that a majority of the methodologies included some measure of demographics, education and health, with existing indices and measurement of aspects of government and the environment being used the least.
Looking at the number of Indicators, Sub-indicators and Variables in each category and the proportion of variables that are used more than once it is possible to better understand how much variety within the literature there is in terms variables to represent these different concepts. The proportion of variables used more than once is indicative of the level of agreement in the literature on what variables to measure to understand the properties of that category. This is shown in Table Number of Indicators, Sub-Indicators and Variables in each category and the proportion of variables in each category that are used in more than one methodology.
The 15 categories were grouped into 6 environments, to better enable visual analysis of the composition of each index. The use of variables in these 6 different environments in the different methodologies is summarised in Table The most common variables are related to various social aspects of communities especially demographics, education and health. Respectively population density, number of doctors and literacy rate were the three most common variables in these categories.
Variables representing various economic aspects of communities: livelihoods, labour market and economy were the next most common. The number of renters and access to clean water were, respectively, the most common variables in these two categories. Existing indices were used in only 21 of the methodologies, with most relying instead on directly collected data. However despite the risk of double-weighting, by also including variables that are already present in an included index this appears to have only occurred in two methodologies.
Although the prevalence of different variables provides some insight into their popularity in disaster risk, vulnerability and resilience indices it does not reveal the make-up of the individual indices. To better understand their composition the proportion of the variables classified into each environment was calculated for each index. The results of this classification are displayed in Figure 6 and the averages across all the indices displayed in Table These show that most indices are dominated by variables related to the social environment, with a much smaller number using high proportions of variables from the disaster environment.
However when disaster resilience variables are specifically examined i. Variables that compose each index classified into one of six environments as a proportion of the number of variables in each index. Proportion of variables from each environment that comprise each index, on average, for all methodologies and for methodologies that only include variables in that environment.
Examining the correlations between the proportions of variables included Table 15 shows the strongest relationship between the social and disaster environments; the more social variables that are included, the fewer disaster variables that are included. It is also possible to compare how the proportion of variables in each index varies according to the type of methodology which is shown in Table Only the Deductive, PCA and Stakeholder-based methodologies have been included due to the small number of methodologies in the Relational and Novel techniques categories.
This demonstrates that there is some variation in the type of variables used in these methodologies depending on the approach used. Although Deductive and PCA approaches appear to be broadly similar in the proportions of variables included, PCA approaches feature far fewer disaster related variables. Stakeholder based approaches include many more disaster related variables, making up approximately half of the variables in these methods with a much lower focus on economic variables.
Proportion of variables from each environment present in methods using the three most common construction approaches. It is desirable to know whether the large number of composite indicator methodologies is actually adding new explanatory power to understanding of vulnerability, risk or resilience or whether they are repeatedly using the similar sets of variables and only varying the construction method.
Ideally this would be tested by comparing index values for the methodologies in the same area, but with little geographic overlap and data unavailability this would not be practical. To gauge the amount of variation in the choice of variables across all the methodologies a custom measure, the 'Overlapping Score' was created to measure the proportion of elements in common at each level of the classification hierarchy variable, sub-indicator, indicator, category, environment.
This index has been constructed such that a set of methods using identical variables, sub-indicators etc. These results show that there is relatively low internal consistency between the methodologies when measured at the variable level this increases substantially when considering the indicator level and above. This suggests that many of these methodologies may not offer substantially different results in presenting an understanding of risk, vulnerability or resilience. By examining the average of the absolute values of the correlation of each method against all others it is possible to uncover those methods that include a more unique set of variables — this was applied at the indicator level.
The four most unique methodologies are:.
Predictive Indicators of Vulnerability Communities Advancing Resilience Toolkit Local Disaster Index Disaster Deficit Index These methodologies had fewer indicators in common with the rest of the set, partly due to a focus on more unique concepts and partly because they use a relatively small number of variables. Similarly the least unique methodologies with the highest proportion of elements in common were the two versions of Joerin's Climate Disaster Resilience Index. This review has revealed a broad range of practice in the development of composite indicators for the measurement of disaster risk, vulnerability and resilience.
There is substantial diversity in the literature, with a range of variable selection approaches, data collection methods, normalisation methods, weighting methods, aggregation approaches and variables being used. However this review has also identified a number of trends which may limit the utility of composite indices in improving the understanding of these concepts. Although the review found considerable diversity in the methodologies of index construction the majority take a fairly standard deductive or hierarchical approach with a weighted sum of the variables included in the index.
In most cases the main point of difference in index construction was the choice of variables for inclusion. Hierarchical approaches are easy to construct and are relatively simple to understand which may largely explain their prevalence. Principal Components Analysis was also commonly employed, with many cases being strongly influenced by the publication of Cutter's Social Vulnerability Index SoVI.
In a number of instances its use has extended beyond addressing some of the problems associated with collinearity in deductive indices to more detailed analysis of the principal components and their spatial variation, thus taking advantage of PCA as a data reduction tool. PCA has only been applied at a sub-national level, however it is likely that comparing nations would not offer significant advantages over other methods as for it to be statistically valid only a small number of variables could be included.
Stakeholder based methods were less popular than the deductive methods or PCA, however appeared to be more targeted with more variables directly related to disaster resilience. The large number of variables gathered in the stakeholder focussed methodologies could be problematic. Although a number of these are checklist-based self assessments, the volume of questions could lead to little attention being paid to responses and overall disengagement from the process. The literature generally agrees that resilience is not a 'check the box' approach but is related to systemic performance, 55 which these methods may not focus on.
Because many stakeholder based approaches are self assessment and others are being driven by a single small research group it is difficult to ascertain their full geographical coverage. Aside from pilot locations, their implementation is often not reported. This makes it difficult to assess implementation difficulties or conduct reliability analysis to identify a shorter list of questions. Relational techniques, such as linear regression and data envelopment analysis have received some attention for both the comparison of nations and sub-national areas. Although the indices that use relational methods are internally validated they have a number of potential problems that could limit their application.
Data on disaster impacts may be of poor quality and have limited time coverage to adequately reveal the relationship with variables. Indices produced with these methods should be accompanied by robust uncertainty estimates that incorporate both aleatoric and epistemic uncertainty to better enable assessment of their reliability. A small number of composite indices have taken more advanced approaches, some of which are closer to more model based assessments of disaster risk.
Greater sophistication in this area may produce indices that better reflect theory models. However these 'black box' methods may be more difficult to understand for end-users and promote a less critical acceptance of the results. The review of the literature found 24 methodologies that were not sufficiently well described to include in the analysis. Although these included a number of proprietary rankings, many methods published in the academic literature also failed to provide sufficient detail to enable analysis, let alone replication. Even methods included in the review failed to include sufficient detail that would enable reconstruction of the index.
Where information was sought from statistical databases, some methodologies are unclear about the agency from which the data was sourced and the years to which variables refer to. Some studies did note the year from which the data was sourced, which was up to 10 years prior to publication potentially making the index out-of-date. This has implications for decision makers using the index results. Although demographic variables may change relatively slowly this lack of currency was often not well communicated in the results of these studies, nor were attempts made to 'nowcast' the index values.
Those based on community or stakeholder survey often did not include the questions asked, making comparison with other surveys difficult. Numerous researchers have been pointing out flaws in index construction and calling for greater use of sensitivity and uncertainty analysis for quite some time as outlined in the introduction. This study has found that these calls have been largely unheeded. Despite the large increase in number of disaster risk, resilience and vulnerability composite indicators being developed, there has not been an increase in the use of sensitivity and uncertainty analysis.
Few methodologies are undertaking explicit sensitivity and uncertainty analyses and where it is undertaken it is typically limited to one or two aspects of the methodology. This makes it difficult to assess quality, especially when many choices in index construction appear to have been made arbitrarily or with limited justification. Some sensitivity analyses have found significant impact of methodological choices on the resulting index values which has implications on the broader use of these indices by policy-makers.
Without appropriate caveats or the provision of uncertainty estimates, decision makers may believe the index results to be much more reliable than they actually are. Most studies communicated results for example by using maps and summary tables. However many did not provide full numerical results. Although full reporting on index results for large numbers of study units is difficult in the academic literature more effort needs to be made to make these available to enable improved review of these studies, for example by comparing different index results for the same set of territories.
Furthermore only 4 methodologies provide interactive portals to access and visualise the data in graphical, map or table form. Interactive options may be more preferred by policy makers, the community and others interested in the results of these studies which suggests that many authors may not be making them useful for end-users. This study aimed to review composite indices that claim to measure disaster risk, vulnerability and resilience. The limited use of disaster related variables is likely to be due to limited availability.
This is supported by the fact that those methodologies that directly elicited information from stakeholders included substantially more of these variables than those that depended more on data gathered from statistical databases. However without access to the results of all the indices in this review and a social disadvantage index for comparison, this is difficult to assess. Analysis of the inclusion of the variables found a strong negative correlation between use of social variables and use of disaster variables.
This suggests that social variables are the proxy of choice when access to more disaster related data is difficult, with demographic variables being very common. A large number of these variables are collected through workshops and stakeholder and community surveys. Alternative formulations of questions on disaster resilience, for example customised to the terminology and context of a particular jurisdiction, may limit the broader applicability of these methodologies. Work surrounding the implementation of the post development agenda, such as the development of a new "10 Essentials" may assist in addressing this gap.
This highlights the difficulty in adapting tools across national contexts, particularly where those tools utilise data from statistical agencies which may not be available for similar scales, forms and time periods across multiple nations. An updated tool, consistent with the new 10 Essentials may be beneficial in improving cross-national coverage although the differing contexts of nations may prove too difficult for this tool to enable meaningful comparison. Two key motivations have emerged from this analysis of composite index and dashboard methodologies of disaster risk, vulnerability and resilience.
Many methods use primarily statistical data to compare large numbers of study units. Consistent comparison of different jurisdictions may be desired by national and international organisations seeking to inform decisions on resource allocation. Another group of methods seek to provide a tool, primarily to sub-national authorities, for self-assessment of disaster vulnerability through the asking of targeted questions.
Self-assessment may be desired by national and local governments and communities seeking to improve their performance. These two groupings are consistent with the three key motivations found by da Silva and Morera of ranking relative performance and diagnosing performance and influencing change. There does not appear to have been any efforts to make the results of self-assessments more comparable.
This could be achieved by establishing quantitative measures of performance where each assessing authority would choose their own benchmarks. However this may not produce desirable results; the use of self-assessments for resource allocation could potentially bias them towards assessing their performance as lower than reality to enable greater access to resources. The number and variety of composite indicator methodologies that have been developed clearly indicate their potential end use for decision makers working in disaster risk reduction, humanitarian and emergency response, civil protection or other fields related to disaster resilience.
However the limitations of the present literature have a number of implications for end users and there is a risk that biases and uncertainty may lead to inappropriate decisions. To counter this risk end users should consider multiple techniques when attempting to understand community vulnerability and resilience. Stay ahead with Tip Sheet! Free newsletter: the hottest new books, features and more. Parts of this site are only available to paying PW subscribers. Thank you for visiting Publishers Weekly. There are 3 possible reasons you were unable to login and get access our premium online pages.
You may cancel at any time with no questions asked. You are a subscriber but you have not yet set up your account for premium online access. Add your preferred email address and password to your account.
Dementia Rate In England & Wales Drops 24% From To (OPEN MINDS Weekly News Wire Book ) eBook: Terry Griffin, Monica Oss, Sarah. Between there were a total of 1, unintentional fatal . From to , the standardized mortality rate of drowning in China In fact, 75% of all drowning deaths occur in open fresh water, places On average people drown in the UK each year. A 12 week social media strategy.
You forgot your password and you need to retrieve it. Click here to access the password we have on file for you.