Measuring vulnerability

Economic vulnerability
The economic vulnerability of SIDS, which is well documented, stems from a number of inherent characteristics of such states, notably:

The small size of SIDS, which limits their ability to reap the benefits of economies of scale and constrains production possibilities.
A high degree of economic openness rendering these states particularly susceptible to economic conditions in the rest of the world.
Dependence on a narrow range of exports, giving rise to the usual risks associated with lack of diversification.
Dependence on imports, in particular energy and industrial supplies, exacerbated by limited import substitution possibilities.
Insularity, peripherality and remoteness, leading to high transport costs and marginalization.

Environmental vulnerability
Small island states tend also to be environmentally vulnerable, mainly due to:

Limited assimilative and carrying capacity, leading to problems associated with waste management, water storage and other factors affected by small territorial size.
A relatively large coastal zone, in relation to the land mass, making these states especially prone to erosion.
Fragile ecosystems, because of low resistance to outside influences, endangering endemic species of flora and fauna.
A proneness to natural disasters, including earthquakes, volcanic eruptions, cyclones, hurricanes, floods, tidal waves and others – these disasters also affect larger territories, but the impact is more devastating on a small island state.
A relatively high proportion of land which could be affected by climate change, and in particular a rise in sea level, possibly resulting in proportionately large land losses, particularly in low-lying SIDS.
The significant impact on the environment of economic development, often leading to a fast depletion of agricultural land and natural resources.

Vulnerability and economic success
In spite of their economic and environmental vulnerability, many small states register relatively high GNP per capita compared with other developing countries. (There are, of course, a number of SIDS which have low GNP per capita and are classified as least developed states.) High GNP per head gives the impression of economic strength, and masks the fact that economic success may be quite fragile and dependent to a high degree on conditions outside the country’s control.

This has led to the development of an economic vulnerability index, the main objective being to highlight the underlying economic and environmental fragility of many small states.

Constructing the vulnerability index
There are three basic methods for computing a vulnerability index:

Normalization procedure.
Mapping on a categorical scale.
Regression method.

Normalization procedure
The method most commonly used is to obtain data for the components of the index, with each component representing a facet of vulnerability. Since the components of the index are often measured in different units, the observations have to be ‘standardized’ or ‘normalized’ to permit averaging, with the average being called a composite index.

The normalization procedure most commonly used is that which adjusts the observation to take a value of between 0 and 1 using the formula: Vij = (Xij - MinXi) / (MaxXi - MinXi), where: Vij stands for the standardized observation associated with the ith component for country j; Xij stands for the value of the ith component in the vulnerability index, for country j; MaxXi and MinXi stand for the maximum and minimum value of the ith component for all countries in the index. Crowards refined the method to reduce the undue impact of outliers on the distribution of the observations, by assigning the value of 1 to the top decile of values in the observations of a particular variable and a value of 0 to the bottom decile.

The averaging procedure can be based on equal or varying weights assigned to each component. Briguglio and Crowards experimented with varying weights for each component, but their preferred method was that involving equal weights.

Composite indices using this methodology were those by Briguglio (1992; 1993; 1995; 1997), Chander (1996) and Wells (1996) (see selected reading). All these studies concluded that small island states tend to be more vulnerable than larger territories.

The most important shortcoming of this method is that the weights for averaging the components of vulnerability are arbitrarily chosen, and that the distribution of the normalized variables are heavily influenced by outlier observations.

Mapping on a categorical scale
This method, suitable for qualitative data, involves mapping the scores on a categorical scale ranging from the lowest possible incidence to the highest. This approach was used in the study by Kaly et al., where the scale set was 1 to 7. The scores for each component of the index were averaged to derive a composite index for each country. Kaly et al. applied the index to three countries only, namely Australia, Fiji and Tuvalu, since the exercise was a preliminary one and constrained by the funding provided. The results show that Tuvalu is the most environmentally vulnerable while Australia is the least environmentally vulnerable. Again, with this method, there is a degree of arbitrariness and subjectivity in assigning scores and in weighting the components of the index.

The regression method
The third method used for constructing the vulnerability index, proposed by Atkins et al. (1998) and Wells (1997), is based on a regression procedure. Wells and Atkins et al. assumed that GDP volatility is a manifestation of vulnerability and can therefore be taken as a proxy of vulnerability. They then regressed GDP volatility on a number of explanatory variables which represented causes of vulnerability. The coefficients on the explanatory variables of estimated equation were then taken as weights for averaging the three vulnerability components.

This method lets the data produce the weights and does not require the ‘normalization’ of the observations. However it has a number of methodological defects, which limit the operationality and the reliability of the index. The most important defect is that the authors had to assume that the dependent variable (namely GDP volatility) is a proxy for vulnerability, and therefore they had no need to go through a cumbersome regression procedure to compute the vulnerability index. An additional problem with the Atkins et al. method is that the predictive ability of the model is poor.

Components of the index
The vulnerability indices developed so far differ also in terms of complexity. The economic vulnerability indices generally include a relatively small number of variables, often limited to three or four. One reason for this is that many economic variables are correlated with each other and one variable could be used to represent others. Other reasons are that many variables complicate the procedure and data for certain variables are not available across countries.

The most frequent variables used in the economic vulnerability indices relate to economic openness, export concentration, dependence on imports of energy and peripherality. Another variable used is proneness to natural disasters, which is grouped here with environmental vulnerability.

The environmental vulnerability index developed by Kaly et al. utilized a large number of variables (57 in all) since, as argued by the authors, a large number of indicators are required for complex ecological systems. Two other attempts at constructing an environmental vulnerability index were those by Pantin (1997) and by Atkins et al. (1998). Pantin’s work is essentially an attempt to measure the effects of natural disaster on the GDP of the different countries. The Atkins et al. study was tentative and only introduced as an footnote to their study on economic vulnerability, but it indicated that small island developing states tend to be more economically vulnerable than larger territories.

Kaly et al. attempted to capture three aspects of environmental vulnerability, namely:

The level of risks (or pressures) which act on the environment forming the risk exposure sub-index (REI).
Intrinsic resilience of the environment to risks, forming the intrinsic resilience sub-index (IRI) which refers to characteristics of a country which would tend to make it less/more able to cope with natural and anthropogenic hazards.
Extrinsic vulnerability or resilience as a result of external forces acting on the environment, forming the environmental degradation sub-index (EDI) which describes the ecological integrity or level of degradation of ecosystems. (This included 39 indicators of risk (REI), five indicators of resilience (IRI) and 13 indicators of environmental integrity or degradation (EDI).)

To date there has not been a serious attempt to create a super-composite index which combines environmental and economic vulnerability.

Benefits of the index
There are at least two benefits that can be derived from the construction of a composite vulnerability index:

The index can draw attention to the issue of economic and environmental vulnerability of SIDS (depending on the aspect which the index is supposed to measure).
The index presents a single-value measure of vulnerability based on meaningful criteria which can be considered by donor countries and organizations when taking decisions regarding the allocation of financial and technical assistance or for assigning special status to SIDS.

The indicators share a number of weaknesses, principally associated with the subjectivity in their computation, in particular with regard to the choice of variables, the method of measurement and the averaging procedure.

Subjective choice of variables
The question of subjective choice of variables is difficult to resolve. This is, however, not a problem peculiar to the vulnerability indices but to most empirical work, especially that which purports to quantify data which is essentially qualitative.

Problems of measurement
The measurement problems arise in part because of an absence of data for certain variables or for certain countries; different methods of statistical compilation across countries; and errors in measurements of the variables.

Weighting
Composite indices are averages of different sub-indices, and the single value which they produce may conceal divergencies between the individual components or sub-indices, possibly hiding useful information. Furthermore, a composite index implies some form of trade-off between the sub-indices of the composite index and averaging would conceal, for example, situations where the effect of one variable cancels out the effect of another. In addition there is the problem of whether to take a simple average or a weighted average and, in the latter case, which weights are to be assigned to the different variables. In general, the weighting problem remains in the realm of subjectivity, with the simple average having a favourable edge on grounds of simplicity.

Some of these problems will probably never be resolved, and their acceptability and operationality will in the end depend on some form of consensus.