The value of a statistical life in Mexico

ABSTRACT

Mexico has been making an increased use of cost–benefit analyses to inform its public policy decisions, in line with other OECD countries. The monetary values used in these analyses have generally been based on benefit transfer methods. There are several reasons to be cautious about the use of benefit transfer values, including that the resulting figures from the procedure can vary widely depending on assumptions. This is particularly pressing where large allocation decisions are at stake, as tends to be the case with air pollution policy and climate change policy. This study collected primary data to produce a value of statistical life for Mexico, using a questionnaire previously employed in several countries, including in the USA (where benefit transfer values tend to be sourced from by Mexican authorities). The analysis produces a value of statistical life of USD 210,880. The evidence suggests that benefit transfer values currently being used by Mexican authorities are being driven by the source values themselves, rather than by inappropriate assumptions about income elasticities or by being affected unduly by other factors. The conclusions offer some recommendations for future use of value of statistical life figures in Mexico.

KEYWORDS:

• Air pollution policy
• benefit transfer
• contingent valuation
• cost–benefit analysis
• Mexico
• monetisation of health benefits
• non-market valuation
• value of statistical life

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1 The MVMA includes the Federal District of Mexico City and part of the surrounding States.

2 Data collection was funded and conducted by Mexico’s Instituto Nacional de Ecología y Cambio Climático (INECC).

3 The application of the VSL to air pollution can, for example, be extended to climate change policy: most of the benefits stemming from climate change policy in the short-term result from ancillary reductions to air pollution (from reduced human mortality rates).

4 There are also differences between the US and other countries in preferred methodological approach when setting ‘administratively approved’ VSL figures. The OECD (Biausque et al. 2012) notes that there is a

reliance on revealed preference methods in terms of wage risk studies in the United States (where most such studies have been conducted), while Europe, Canada and Australia rely more on stated preference methods, eliciting people’s willingness to pay (WTP) for changes in mortality risks.

There are frequently significant differences between revealed preferences and stated preferences VSLs, with the former generally producing higher money value estimates (i.e. U.S.A. administrations tend to use a higher VSLs than other administrations due to the sourcing of VSL from revealed preference methods). For the purposes of environmental policy there is a general trend though away from wage-risk studies and into contingent valuation as well as, more recently, experimental methods. This is due to primarily to: (1) differences in the nature of the underlying risk being measured between wage risk and environmental risk studies: wage risk is associated with accidental/traumatic death (with no latency or a negligible latency period), whereas environmental risk tends to be associated with chronic exposure (and with death occurring after a period of illness); (2) differences in the relevant populations and in their associated preferences (wage risk studies focus on working age populations in risky jobs, whereas environmental risks often affect the population at large or even primarily children or the elderly); and (3) potential effect in exposure to occupational risk being generally voluntary, whereas environmental risks are less so (Krupnick 2007; Cropper, Hammitt and Robinson 2011; Biausque et al. 2012).

5 A 1 in 10,000 over 10 years mortality risk reduction option was also considered but rejected as in previous applications of the survey (in other countries) this size of risk was found to generate a very large number of zero WTP answers (being perceived as too small to be distinguishable from the status quo), and to be difficult to understand in focus groups. In focus group tests for Mexico it was found that opinions were roughly equally divided on whether a 1 in 10,000 mortality risk reduction was ‘too small to matter’.

6 In the following, total WTP is reported rather than annual WTP for the risk reductions.

7 Life expectancy in Mexico at the time of the survey was about 76 years old, according to World Health Organisation data.

8 The full text of the questionnaire and several screenshots of its online version for illustration can be found in de Lima (2016).

9 Whereas other studies found no starting point bias (Thayer 1981). The presence of a starting point bias is thus an empirical rather than systematic issue, but there is sufficient evidence for this to be of concern for researchers.

10 Another group that is also particularly at risk are children. However, specific survey designs are required to measure the WTP (of parents) for reductions in fatality risks for children, which fall outside the scope of the current study. There is some evidence of a ‘child-premium’, i.e. that parents would be WTP a higher amount to reduce the risk of dying prematurely of their children than the amount they would be WTP for the same risk reduction for themselves (Alberini and Loomes 2011).

11 For socio-economic groups using the standardised ‘niveles socioeconómicos’, established by the Mexican Government, https://nse.amai.org/niveles-socio-economicos/.

12 The pilot was used to test whether there were significant problems in the implementation of the survey, from a qualitative perspective. The performance was assessed by checking descriptive statistics and seeing the percentage of the logical test questions that were satisfactorily passed, and whether the remaining data seemed within reasonable bounds. This information helped determine the need for a relatively large sample size.

13 The online version of the survey was used to test the effect of survey sponsorship on WTP (publication forthcoming).

14 The unused FLAGs were redundant or had little impact on the resulting subsamples.

15 See the ‘construct validity’ subsection below for a different form of the external scope test, done within a regression framework.

16 As well as, in some instances, in the online version of the questionnaire referred to before.

17 Scope insensitivity problems were also found in the other applications of the survey (see table 14).

18 OECD (2018), Education at a Glance (http://www.oecd.org/education/education-at-a-glance/).

19 INEGI (2019), Población, http://www.beta.inegi.org.mx/temas/estructura/.

20 From the perspective of a policy-maker if a proposed policy passes a CBA analysis even when using the conservative estimate of WTP then the case for implementing the policy is particularly strong. There is however a risk that real WTP is higher than the conservative estimate and that cost-effective policies are discarded. From the perspective of researchers, there may be reasons to be concerned that respondent’s stated WTP may overestimate real WTP. For example ‘prestige effects’ may occur, whereby the respondent seeks to impress or please the surveyor by stating a large WTP (Getzner 2000). Conversely, there may also be reasons to suspect that the respondent’s stated WTP is lower than real WTP, for example due to ‘consumer-collaboration effects’, whereby respondents seek to put downward pressure on a product’s market prices (Hanna and Dodge 1995). The latter may be relevant in the context of the Krupnick et al. suite of studies, in which the present study is included, as the study design is such that the product being described to respondents has the characteristics of a private good rather than a public good (but see also the results of the online-based survey sponsor effects analysis, forthcoming). The questionnaire used has previously produced results that have been found to be on the lower end of stated values for WTP for mortality risk reductions in meta-analyses of the relevant literature conducted in the USA (i.e. the evidence available suggests that it is intrinsically conservative).

21 The religion question asks respondents: ‘how would you describe yourself?’. The answer options are: ‘very religious; somewhat religious; neither religious, nor non-religious; somewhat non-religious; non-religious’.

22 The income questions ask for respondent’s household and personal income, after taxes.

23 The term ‘value of a statistical life’ has in some cases been mistakenly taken to mean the value of an individual human life. It is in fact the value of reducing the likelihood of death in a human population by an amount equivalent to one individual in that population. Not monetising explicitly reductions in the probability of dying that would result from public policy is, de facto, monetising them implicitly (thus less transparently so; see Viscusi 1993). To avoid the confusion that in some cases the term VSL seems to generate other terms that are deemed clearer have been proposed, such as the ‘value of a prevented fatality’ or ‘value of a mortality risk’.

24 The stricter subsample choice procedure from which sample C* results is also inherently more conservative than the less strict subsample choice procedures: its associated WTP is significantly lower than for the other subsamples, except sample D*. In the case of sample D*, however, sample size is small and significantly less representative of the population.

25 The VSL resulting from a 5 in 10,000 risk reduction is greater than the VSL resulting from a 10 in 10,000 risk reduction. This is due to the ratio of WTP values associated with the two risk reduction measures not being proportional to the ratio between the two risk reduction values. The VSL based on the 10 in 10,000 mortality risk reduction is MEX 894,267, or USD 111,783.

26 The VSL is the yearly payment divided by the risk reduction over 10 years.

Additional information

Funding

This work was supported by Fundação para a Ciência e a Tecnologia [grant number SFRH / BD / 77070 / 2011]; London School of Economics and Political Science.