Figure 1-20. Percentage of women in labor force (total and agricultural). Source: World Bank, 2004b; ILO, 2007.

of values and valuation, dealing with uncertainty, dealing with different knowledge systems, as well as modeling issues and developing scenarios.

1.4.2 Working with indicators

What are indicators for? Indicators are used both for specialist purposes and in everyday life. In specialist applications the purposes are defined within the domain of expertise. In everyday life, they form part of the repertoire of heuristics-simple rules for making decisions when time is pressing, information limited or partial, and deep reflection a luxury (Gigerenzer et al., 1999). Indicators become part of what we observe in the world around us as we attempt to detect patterns and extract information relevant to effective action. In this everyday sense, they can be accurate and powerful (Gigerenzer et al., 1999) but also, if wrongly observed or interpreted, contribute to systemic failures (Dörner, 1996).

Referents and contexts. All indicators require a referent measurement situation. To allow meaningful interpretation of indicators and utilization that will appropriately inform policy processes, there is also a need for awareness of the context of use. Strictly speaking, indicators require application in a controlled environment (with/without, before/after). Rarely, however, is such a design possible in reality, for obvious practical and ethical reasons. Thus the present assessment has to accept that information is not perfect. One approach to handle uncertainty is through scenarios that are built on available indicators and assumptions.

State variables and trend indicators. The IAASTD uses two kinds of indicators, describing either state or trends. State variables, of high precision and generality, tend to be favored in science, as they represent the current state of an object or process and are thus measurable. In everyday life,

there is a strong preference for accurate trend indicators. Especially at policy level, information is required on whether situations are improving or worsening, and whether policy objectives are getting closer to their goals or farther away. Trend indicators tend to focus more on identifying thresholds that might indicate an imminent change of state, and less on constant values-the more favored emphasis of many sciences. In many usages trend indicators are also used as learning devices, leading to reestimation of achievement and redefinition of goals as trend data move through time.

Precision, accuracy, and generality. There is agreement in the philosophy of logic and statistics that precision, accuracy, and generality cannot be simultaneously optimized. Any pair of the three may be. The construction and choice of indicator thus has to take into consideration which combination is the most pertinent to the problem or situation for which the indicator might be used. There is a need to identify appropriate indicators and the relationships of these when used at various spatial and temporal hierarchical levels. This is partly a matter of scale and structure of systems hierarchies, and partly a matter of whether it is the state variables or dynamics that the user considers important to observe and monitor.

The dilemmas of interpretation and meaning

An indicator does not exist independently of the observer: as mentioned above, a range of pre-analytic choices are made before an indicator is constructed or brought into use. These choices are inevitably value-laden, and enriched with meaning that the indicator itself does not possess. Take, for example, poverty indicators: one can construct incomebased, nutrition-based, gender-based (etc.) indicators. Each type of indicator both reveals what is important for the user's purpose but also conceals what is not considered preanalytically to be of importance.