Seccion III: Estimación de impactos de la minería en los hogares urbanos y rurales de
3.2. Resultados de estimaciones de impacto usando emparejamiento de hogares
3.2.1. Impactos de la minería metálica en los hogares de la sierra en general
A 3-step process
Economic evaluations of ex-ante accountability studies of proposed inter- ventions of air pollution control consist of three core steps. The latter have been systematised by some softwares, such as Benefits Mapping and Analysis Program (BenMAP) from Abt Associates, which is used by the US EPA for the assessment of air pollution control policies (US EPA, 2006, 2011, 2012). In the UK, the Department for Environment Food and Rural Affairs (DEFRA) refers to this 3-step approach as the “full-impact pathway” (DEFRA, 2013).
Step 1:
The first step involves the modelling of the change in air pollution con- centrations to which the target population is exposed, based on the expected change in emissions associated with policy scenarios. This typically requires sophisticated GIS-based dispersion modelling tools, e.g. CMAQ, RAINS etc. An extensive description of such tools and of the algorithms underpinning them may be found in Yerramilli et al. (2011).
Step 2:
The second step consists of a quantitative health impact assessment, us- ing the method to impact computation described above. Typically effects are computed per year, by applying epidemiological risk estimates to annual background rates of endpoint incidence. The most comprehensive regulatory impact assessments of proposed interventions of particulate air pollution con- trol consider the following health endpoints: (1) premature deaths (or life ex- pectancy impacts see section 2.3.4), (2) chronic bronchitis, (3) hospital admis- sions for cardio-pulmonary causes, (4) upper and lower respiratory symptoms, (5) asthma exacerbations and (6) restricted activity days. For Europe, the
World Health Organisation (WHO) has recently compiled a report of recom- mended coefficients of concentration-response function for each of these health endpoints, as part of the HRAPIE 2 project (WHO, 2013).
Most regulatory analyses place particular emphasis on mortality effects, as recommended by WHO (2013). The latter are estimated based on risk esti- mates from cohort studies of long-term exposure, which capture the effects from both short-term peaks and long-term background exposure to pollution (K¨unzli et al., 2001). As mentioned in section 2.1.3, cohort studies of air pollu- tion have investigated the mortality effects of long-term air pollution exposure on both all natural causes of death and specific causes of death. However, in order to avoid under-estimating the overall mortality burden (see Chapter 3 for further details), WHO recommends to use all-causes of death risk estimates in evaluations of air pollution control interventions (WHO, 2013).
In contrast to mortality impacts, with the exception of chronic bronchitis, morbid endpoints are typically considered only for acute exposure, i.e short- term peak in pollution above daily recommendation measures. It follows that, although the reduction in life expectancy associated with the development of chronic conditions associated with long-term exposure is expected to be captured in overall the mortality effect (K¨unzli et al., 2001), long-term quality of life impacts are completely ignored.
Step 3:
In a third step, in accordance with welfare theory roots of CBA which is the preferred decision tool of regulatory impact assessments (see section 2.3.1), the attributable change in each health endpoint is monetized using WTP (VSL) values for relevant health risk reduction. When WTP values do not exist, for instance for hospital admissions, cost of illness estimates are used. However as mentioned in section 2.2.4, since cost of illness estimates do not account for quality of life impacts from restricted lifestyle, pain and psychological suffering, their use is equivalent to setting quality of life impacts to zero.
It is worth highlighting that in most studies, mortality benefits drive the overall benefits (WHO, 2013). This is not surprising given that: (i) mortality
impacts are the primary focus of such analyses; (ii) morbidity effects are com- monly monetized using cost of illness estimates which as mentioned in section 2.2.4, do not account for quality of life impacts from restricted lifestyle, pain and psychological suffering.
As mentioned in section 2.2.4, WTP values for health risk reduction, be it revealed or stated, vary greatly. This results in substantial differences in gov- ernments’ recommended values for monetizing health impacts. For instance, Scapecchi (2008)’s comparison of recommended values (mean estimates, ex- pressed in 2006 $) for monetizing mortality impacts from PM exposure showed that the VSL estimates for the US and Canada were much higher (respectively $ 7,4 and 6 million) than those recommended at EU level ($ 1,8 million). In the UK, the DEFRA-commissioned study of Chilton et al. (2004) suggested a VSL of about $ 6 million.
Example: DEFRA’s damage costs
Based on this 3-step impact pathway approach, in the UK, DEFRA pro- duced standardised damage costs estimates per tonne of pollutant emitted (DEFRA, 2011, 2015), in order to support the evaluation of small-scale pro- posals (below £50 million).
The health endpoints and the magnitude of effects considered in damage costs computations are in line with recommendations from the UK Commit- tee on the Medical Effects of Air Pollutants (COMEAP), which independently advises the government on matters concerning the health effects of air pollu- tants. For particulate matter, the health endpoints included are: (i) mortality effects associated with chronic exposure and (ii) hospital admissions for cardio- respiratory causes following acute exposure, i.e. following a short-term peak in pollutants concentrations over a few days (DEFRA, 2013).
Health impacts were computed for different densities of population e.g. “Central London”, “Urban medium” or “Rural” and monetized based on WTP values or cost of illness estimates. For instance, according to DEFRA’s damage costs calculator, an annual reduction of one tonne of PM emission in central
London is expected to yield an annual gain of £2.4 million (in 2015 prices). By contrast, the same reduction in a small rural area would be expected to provide an annual monetized benefit of £375,000 (in 2015 prices).
Dealing with uncertainty
In addition to considering uncertainties inherent to the modelling of air pol- lution concentrations, many large scale ex-ante studies consider uncertainty in key parameters, typically concentration-response coefficients and VSL/WTP values, via probabilistic sensitivity analysis (PSA). This consists in fitting a probability distribution to each uncertain input parameter, where uncertainty is indicated by the 95% confidence interval, and to propagate joint parameter uncertainty in total aggregated benefits via Monte Carlo simulations. The lat- ter are readily integrated in software packages such as previously mentioned BenMAP or @RISK (Palissade Corporation).
If one takes into account the nature of the data used for estimating param- eters as well as parameters’ logical bounds, only a few distributions remain that are appropriate candidates for a given type of parameter (Briggs et al., 2006). Typically, the beta and Dirichlet distributions will be appropriate to model transition probabilities derived from respectively binomial and multi- nomial data, whereas the gamma distribution will be appropriate for costs and dis-utilities and the log-normal distribution will be adequate for relative risks (Briggs et al., 2006). By contrast the use of the triangular distribution, which simply requires a maximum, a minimum and a mode, has been largely discouraged since it is not statistically related to the estimation process of the data and thus, very difficult to parametrize correctly (Briggs et al., 2006).
Interestingly, probabilistic sensitivity analyses in past evaluations of air pollution control have departed from the above recommendations, which may bring their quality into question. For instance, the US EPA fitted triangular distributions to model utility parameters and WTP values in past regulatory analyses of air pollution control policies (US EPA, 2006, 2012). For the CBAs
of Clean Air for Europe (Holland et al., 2005b) and Revisions of the E.U. Gothenburg Protocol (Holland et al., 2011), triangular distributions were fitted to baseline incidence rates of health endpoints.
It should be underlined that recently published European guidelines for uncertainty analysis in health impact assessment and cost benefit analysis of air pollution control policies (Holland, 2014) do not suggest to rule out the triangular distribution. On the contrary, they recommend to use the latter for background mortality and morbid endpoint incidence rates as well as for risk estimates when “there is great confidence in central estimates but adoption of a normal distribution would imply that ranges were based on more data than is the case” (pp 41). Alternatively, guidelines suggest to chose among two other distributions for risk estimates. The normal distribution is recommended in the case of a higher probability of values towards the range midpoint than towards its extremes, whereas the uniform distribution is recommended when all values in the range are thought to be of equal probability. Uncertainty pertaining to WTP values for mortality risk reduction are recommended to be dealt with in univariate sensitivity analysis.