Towards a Local-Level Resilience Composite Index: Introducing Different Degrees of Indicator Quantification

When it comes to the operationalization of indicators, there is a wide range between indistinctly defined indicators and fully quantified indicators, a fact that is frequently left unaddressed in the research literature. To address this gap, a classification of different indicator operationalization stages is proposed. To illuminate those different stages or degrees of indicator operationalization, access to credit is introduced as one dimension of drought resilience in rural areas of east India (Jülich 2013, 2015). This setting serves as an example to illustrate seven stages of indicator quantification.

Household access to credit is an important resilience factor in eastern India. For many households borrowing money is a major coping strategy during and after times of drought. In addition to what kind of security can be presented by a household, the determining factor of what kind of credit conditions a household has access to is its social networks and reputation within the village community. In the sphere of formal institutions, there is credit raising from the State Bank of India. For subsistence farmers who own land and have proof of this in the form of a land title, the interest rate per annum (pa) is 12%, if the land functions as security for the credit. Households have potential access to this form of credit, if they own enough land for security. The land quality and fertility has to be sufficient to be hypothecated and a formal land property title is necessary. If formal credit raising fails, households are dependent on borrowing money based on informal institutions. Some households are in the position to borrow money from a relative outside the village, usually without interest to be paid. The rest of the informal sphere is dominated by local traders, whose interest rates are much higher than those of formal moneylenders. If the household can present a certain amount of gold as security, mostly in the form of jewelry worn by women, the interest rate is 3% per month (pm); resulting in 36% pa, if the interest is paid regularly each month. The situation for the household is worse if only agricultural tools or animals can be presented as security. In this case, an interest rate of 5% pm/60% pa has to be paid. However, particularly during droughts some households cannot present any security like gold, tools, or animals because they have already been sold to buy food. If no security is presented, the credit relation is based on trust and the borrower has to be known to the trader. The interest rate in that case is 10% pm/120% pa. To be dependent on such credits has to be seen as a sign of low social resilience. In the short term, these credits might help to cope with drought, but research indicates that, in the long term, they very often lead into overindebtedness (Jülich 2015). Such excessive indebtedness depletes future options for development and, as a consequence, the access to credit determines the risk of overindebtedness that comes with certain credit conditions and affects resilience. Credits by the State Bank of India, by traders with 3% pm interest, and credit from relatives bear a relatively low risk of overindebting, at least compared to the other alternatives present in the region.

A qualitative investigation of the causal connections between drought resilience and access to credit is the basis for the development of an indicator that captures credit as one dimension of drought resilience. Drawing on this example, Table 1 shows seven stages of indicator operationalization, with an increasing degree of quantification from stage 1 through stage 7.

The quantitative indicator development in the emBRACE case study was guided by the general hypothesis: Resilience against natural hazards varies at the local level and can be characterized by measurable characteristics that indicate the degree of disaster resilience. From this hypothesis the central research question was derived: Are there measurable differences in resilience at the local level? In order to answer this main research question, the following secondary questions arose: Which socioeconomic or demographic characteristics can be employed to measure the disaster resilience of populations at the local level? How can these characteristics be utilized to give an indication of disaster resilience? Since disaster resilience is a complex phenomenon with various dimensions, it cannot be captured by a single partial indicator. Several indicators are needed to reflect the multidimensional nature of disaster resilience.

To investigate these dimensions of resilience, expert interviews with various stakeholders from the field of natural hazard prevention, disaster response, and information platforms were conducted in the canton of Grisons, Switzerland. For this, a matrix was developed that shows on one axis all the natural harzard-induced disasters that can occur in the study region (intense rainfall and snowfall, snow avalanches, storms, wind, hail, floods, debris flows, rockslides, rockfalls, landslides, earthquakes, droughts), and shows on the other axis the following guiding questions: Who in particular was affected during past disasters and who was not affected? What measures helped against the disaster? Who is very well informed, aware, and prepared for the disaster and who is not? Who would recover best from a disaster and who would severely struggle to recover? Who would even have positive externalities from a disaster? Who has more human, social, or financial capital than others? Who is resilient and who is not? The aim of these guiding questions was to identify measurable characteristics that can be employed as a measure for disaster resilience differences.

With this matrix, the stakeholders were questioned on each combination of possible disaster-generating hazards in the region and asked the guiding questions listed above. The guiding questions were used as interview openers to identify thematic indicator complexes (for example, local disaster prevention knowledge). Once such a thematic complex was identified, it was investigated in depth for all relevant aspects connected to it (for example, households’ residence time). This was the qualitative basis for the quantitative indicator development. Three thematic indicator complexes—residence time, past disasters, and warning systems—resulted from the matrix interviews and were taken as exemplary outputs for the development of quantitative indicators. These complexes are discussed in the next three sections. All three presented indicators are completely quantified, but partially lower quantification stages are offered to discuss possibilities and constraints of different stages of quantification.

If several completely quantified partial indicators are developed, it is crucial to transform the input parameters to the same numerical dimension reflecting the level of resilience. Otherwise the partial indicators cannot be combined in the form of a composite index. In this study, values between 0.0 and 1.0 were chosen as the indicator value range, with consistently 0 indicating lowest/no resilience and 1 representing highest resilience. Most composite indices, like the Human Development Index (HDI) developed by the United Nations Development Programme, operate between 0 and 1 (UNDP 2007). The main advantage of this numerical range is that all mathematical operations have the same effect within the whole range, contrary to numerical ranges like 0–10, or 0–100. This is because mathematically numbers between 0 and 1 act differently than numbers above 1 in response to operations like squaring: for instance, figures between 0 and 1 decrease if squared, and figures greater than 1 increase. It would distort results in some cases, if input indicator values above and below 1 were used with the same equations. That is why for all developed partial indicators a transformation of input variables into numbers between 0.0 and 1.0 was chosen.

The three developed indicators aim to quantify the subcomponent “Risk/loss perception” within the domain “Learning” in the emBRACE community disaster resilience framework (Fig. 1). Without further indicators quantifying the other domains of resilience, the developed indicators are not able to capture the complex phenomena of resilience in their entirety.

This partial resilience indicator aims at capturing the risk/loss perception subcomponent of resilience. All questioned disaster experts confirmed a positive relation between the residence time of households and natural hazard awareness. In this form the indicator is already operationalized at stage 4 or 5 (Table 1), depending on the determination of the unit of analysis.

For further quantification, unpublished empirical data collected for a study by Buchecker et al. (2016) were employed. The study explored factors that can positively influence local populations’ attitudes towards integrated risk management. They conducted a household survey in two Swiss Alpine valleys, in which a disastrous flood event had taken place 2 years before (Demeritt et al. 2013). A total of 2100 standardized questionnaires were sent to all households in the Lötschen valley and to a random sample of the households in the larger Kander valley. The response rate was 30%. Table 2 shows the results for two questions on the residents’ disaster prevention knowledge, itemized according to how long the respondents’ already live in the valley.

In general the residence time of all respondents in both valleys is relatively long. In the Kander valley there is a clear association between increasing residence time in the valley and how well the respondent assessed her/his information level on disaster prevention. In the Lötschen valley there is a clear association between residence time in the valley and knowledge about places at risk in the village. The data suggest that prevention knowledge increases for up to 40 years for someone living in the same valley. The interviewed disaster experts confirmed a steep learning curve within the first 10 years of residence time at one place. In terms of quantification this led to a minimum goalpost of 0 years, a maximum goalpost of 40 years and a logarithmic run of the curve. Formula 1 captures all three characteristics. Unit of analysis is a household and the only input parameter is the time of residence of the household within the village.

Formula 1 creates just values between (and including) 0 for lowest resilience and (including) 1 for highest resilience. Values above 1.0 are not allowed by this minimizing function. Hence, the concept of goalposts (UNDP 2007, p. 356) or minimum/maximum (OECD 2008, p. 85) is employed. Higher values than a residence time of 40 years have no further effect and would also result in an indicator value of 1.0. According to the classification of indicator operationalization stages proposed in Table 1, the indicator formula 1 is fully quantified at stage 7. Figure 2 visualizes the run of the resulting curve of formula 1. On the horizontal axis the input parameter is shown, and on the ordinate axis the resulting level of resilience according to the resilience indicator formula 1 is shown.

Residence time is only suitable for capturing the risk/loss perception of disasters that occur relatively frequent. When it comes to more infrequent disasters, like high magnitude earthquakes or tsunamis with a probability of occurrence exceeding one or two generations, residence time is not an appropriate measure for risk/loss perception.

The interviewed disaster experts pointed out that disaster impacts are predominantly negative at the time of occurrence. But once direct impacts of a disaster are endured, the disaster starts to act positively in terms of awareness building. Hence, past disasters have positive effects on the awareness of people and increase their resilience. The manifestation of hazard in the form of a disaster increases in the aftermath with the willingness to invest in mitigation measures. Research even indicates that disasters have acted as catalysts for the construction of preventive measures, which had been previously planned, but for which no political support was available through which to mobilize funding.

This leads to a stage 3 operationalized indicator: The occurrence of disasters in the region in the past increases resilience. By making this disaster specific the indicator becomes stage 4 operationalized: The occurrence of a specific type of disaster (flood, debris flow, avalanche, and so on) in the region increases the resilience against that type of disaster.

To further refine this stage 4 indicator to stage 5 the unit of analysis can be defined, as canton or municipality for instance. Data collection for this indicator through empirical research is relatively easy if operationalized between stage 3 and stage 5. However, the dimension of time, the intensity of past disasters, and their spatial dimensions remain indefinite. For a stage 6 or 7 indicator quantification, these three subfactors have to be defined.

The indicators “residence time” and “past disaster awareness” produce continuous resilience values between 0 and 1. The thematic complex of warning systems has been picked to demonstrate how in terms of stage 7 quantification binary indicators can be transferred into this numerical dimension.

Research indicates that persons and households who are subscribed to one of the natural hazard warning services present in the study region (for example, MetoSwiss and respective public cantonal building insurance systems) are more resilient than others. A warning message received in time and interpreted properly can effect individuals’ efficacy in getting themselves or their belongings to safety—for example by proceeding to safe zones, bringing valuables upstairs in the case of flooding, parking the car in the garage in the case of hail or on higher ground in the case of a river flood, being on the right side if the only road of a closed off valley is likely to be blocked by avalanche or debris flows, and so on.

This indicator is constructed as an all or nothing indicator. As a result, the indicator allows only two values: 0.0 if the analyzed household or person is not subscribed to a warning service, or 1.0 if the household or person is subscribed to at least one natural hazard warning service. Therefore, the value of formula 7 is defined by an indicator function.The most appropriate unit of analysis for this partial resilience indicator is single individuals or households. If subparts of municipalities are supposed to be compared in terms of disaster resilience, all households in an area can be surveyed or a random sample can be taken and mean values can be calculated.