Abstract
This paper examines the relationship between averting expenditures / choices and perceived health risks. Models in the literature often employ risk perceptions as explanatory variables without addressing the potential endogeneity of the perceived risk. We examine the implications of ignoring endogeneity in this context, using an application to both drinking water choices and expenditures and perceived health risks. Our data are from an Internet-based cross-Canada survey that employs a novel interactive risk ladder to elicit mortality risk perceptions relating to water. We employ two fundamentally different methods to assess the impact of risk perceptions on behavior: an analysis of expenditures on alternate water sources and a model of proportional choice of water sources. Results suggest the presence of averting behavior with respect to perceived mortality risks and that the estimated effect of water risks is greater than 3 times higher when using approaches that correct for endogeneity compared to models that do not.
Similar content being viewed by others
Notes
This event took place in the year 2000 in Walkerton, Ontario where E. coli contamination in local drinking water supplies led to total costs of nearly $65 million (Livernois 2001). Other notable water contamination events took place in 2001 in North Battleford, Saskatchewan where the presence of cryptosporidium, a parasitic organism, led to an estimated 4–7 thousand illnesses in the region (Stirling et al. 2001), and in the aboriginal community of Kashechewan, Ontario where E. coli resulted in the evacuation of the community and a total cost of over $16 million (CBC 2006). Between 1993 and 2008, at least 48 water-borne disease events were reported by public health officials in Canada (Wilson et al. 2009).
An illustrative video of the risk ladder in use is provided in the online supplementary material.
A much more detailed discussion of this literature is presented in Schram (2009).
The survey was developed using the aid of 7 focus groups, and a pretest with follow-up calls. The pretest was implemented by Ipsos-Reid, and resulted in 128 completed surveys. Particular consideration was given to the design of the risk ladder for gathering risk perception information. The goal for the survey was 1000 respondents. In order to achieve this, 5556 invites were sent out to the Ipsos-Reid online panel. 1304 individuals completed the survey, which would indicate a response rate of 23.5%. The 4252 non-responders include those who quit the survey partway through, as well as those that did not choose to activate their survey link.
The reason the census median age is lower than our sample is that it includes all Canadians whereas our sample includes only adults.
When compared, for Quebec the difference in regional population between data sources is 1.45% and is overrepresented in the survey sample.
Following Abdalla et al. (1992) we used 10 years or 120 months for tap attachment filters. For container style filters, which are likely to see much more wear and tear, 5 years or equivalently 60 months was considered the useful life of the product.
A double-bounded binary choice scenario was presented to respondents where they chose between a receiving $100 in 1 month from now or $116 in 7 months. A follow-up question with a higher ($128) or lower ($105) monetary amount received in seven months was presented to respondents depending how they answered the first question.
These rates are slightly high, however they are consistent with responses in the survey, and on average only a small decrease (less than $1.00) was noted when the same calculations were done using 10% rates for all respondents.
Although there may be an implicit cost associated with this feature of a refrigerator, the cost of the appliance was not gathered in the survey. A total of 82 individuals reported themselves to be refrigerator water filter users.
In some cases, individuals indicated a positive consumption amount, but did not know the costs they incurred for that consumption. These cases mostly arose in filtering expenditures, as there are many components to expenditures on filtration for which “don’t know” was a possible answer (e.g. replacement cost, replacement frequency, system cost). In the case where an individual indicated positive consumption, but did not know a specific expenditure, the average cost specific to each water alternative was used.
For example, if an individual reported spending approximately $1.00 for 1% of their monthly consumption, 100% consumption would cost them approximately $100.00.
We removed two observations with very high monthly bottled water costs of $5000 and $10,000. Including these two outliers, the mean and standard deviation is $119.38 and $347.85 and we used 10 standard deviations of the mean (\(\sim \)$3600) as the outlier cut-off threshold. The 1302 remaining observations have bottled water costs within 10 standard deviations of the mean.
That is, each exponential decrease (ex. \(10^{-5}\) to \(10^{-6}\)) in the level of risk was given its own linear section in the risk ladder, in which the appropriate decreases (ex. 0.00045–0.00040%, a decrease of 0.00005%) were represented in a linear fashion. The “semi-logarithmic” property of the risk ladder describes the appearance of the change between each exponential section.
The 90 deaths per year figure is confirmed cases based on extrapolated data from the United States and may underestimate actual numbers due to under-reporting (Edge et al. 2001).
The 192 deaths are derived from a mortality rate of 20 bladder cancer deaths per 100,000 people over a 35-year period reported in the survey used by Adamowicz et al. (2011). This rate was multiplied by the 33.6 million people in Canada in 2009. The mortality rate information is based on studies by Wigle (1998) and Canadian Cancer Society’s Advisory Committee on Cancer Statistics (2015).
The estimated age-standardized mortality rate for skin cancer in Canada is 2.3 per 100,000 (Canadian Cancer Society’s Advisory Committee on Cancer Statistics 2015). Applied to the 2009 Canadian population of 33.6 million, this mortality rate implies approximately 773 deaths per year from skin cancer in Canada.
The ratio of the objective risk levels between skin cancer and drinking water risks is calculated using the 773 annual deaths from skin cancer and 282 estimated deaths from drinking water (192 from bladder cancer and 90 from microbial infections).
The empirical model does not specify a particular value of \(\rho \), but rather the proportions represent the average drinking water consumption shares over the month. Without the reproduction of such occasions in an experimental fashion, knowledge of the number of choice occasions for drinking water that one faces in a month is difficult to obtain in a survey format. Most individuals are not likely to know how many times they drink water in each month.
For example, reducing risk from 1 in 100,000,000 chance of death in a year to zero has the same effect on behavior as reducing risk from 101 in 100,000,000 to 100 in 100,000,000.
Preliminary analysis using the continuous risk probability variable yields no statistically significant relationship between risks and water source choices. This can be interpreted in two ways. Either, there is no relationship or risk reductions do not have linear effects on behaviour (but may have nonlinear effects depending on the base level of risk). The advantage of using the dummy variable approach is that the nonlinearity of the effect of risk reductions can be modelled quite flexibly. One disadvantage of the dummy variable approach is choosing the appropriate threshold level between low and high levels of risk. In order to illustrate this we employ four different specifications in the paper.
We chose gender, age, and language as they are the most plausible exogenous socio-demographic variables available, but cannot be assured these variables are completely endogenous. To check the robustness of the results to different socio-demographic variables, Table 9 also summarizes models using alternative socio-demographic variables. Results are similar across the different specifications.
Lewbel et al. (2013) highlight that the control function approach is less robust than IV methods when the endogenous variable is not continuous and the model is nonlinear. Wooldridge (2014) provides a more in-depth theoretical explanation of the control function approach and binary endogenous variables and argues that the control function approach can be applied.
The generalized residual can be computed as the derivative of the log likelihood with respect to the constant term and is equal to the inverse mills ratio for a probit model.
One potential issue with the control function approach is accounting for the new distribution of errors that is induced by the residuals included in the second stage equation. One solution is to apply the log-odds transformation and estimate the model in linear form as opposed to as a logit (Blass et al. 2010). The dependent variable is now ln[\(\alpha _{ij}/(1-\alpha _{i\mathrm{j}})\)] where \(\alpha _{ij}\) is the share of consumption for individual i of alternative j. The benefit of this procedure is that the distribution of the errors around the mean doesn’t affect consistency as long as the errors have zero conditional mean. However, this approach does not work if the shares for each alternative are not strictly positive. This is, in fact, the case with our data since almost 65% of respondents do not consume any filtered water and 18% drinking tap water only.
There is some potential for the price coefficient to be also endogenous. While consumers are price takers for each bottle of water or package of bottled water, they may be able to somewhat control the price-per-litre by varying the amount of bulk purchases of water which may carry a lower price per litre.
Models with additional classes were considered but not justified based on the AIC/BIC criteria or had difficulties converging.
Note that similar to Models 2 and 4, Model 6 uses constructed variables in the selection and expenditure equations and the standard errors of the parameter estimates will not be valid (Wooldridge 2010). Therefore, we use nonparametric bootstrap replication to calculate estimates of empirical standard errors.
For the expenditure models, the estimated coefficient for Drisk1e represent the value individuals place on a reduction in tap water mortality risk level of 0.0002814%. For the choice models, we can divide the Drisk1c coefficient by the cost coefficient to derive the implied value for a reduction in general water mortality risk level of 0.0002119%.
We also ran the models using a more limited sample which excluded the 5% of individuals with the highest tap water risk perceptions. Welfare measures estimated using models without these extreme individuals were generally higher by 20–40% across the different risk levels. For example, using the Drisk0 risk level and the limited sample, the VSL estimate for the multinomial logit model (Model 2) is estimated to be $10.7 million (32% higher than the full sample result) while the estimate for the latent class model (Model 4) is estimated to be $4.8 million (41% higher).
References
Abdalla CW, Roach BA, Epp DJ (1992) Valuing environmental quality changes using averting expenditures: an application to groundwater contamination. Land Econ 68(2):163–169
Abrahams NA, Hubbell BJ, Jordan JL (2000) Joint production and averting expenditure measures of willingness to pay: do water expenditures really measure avoidance costs? Am J Agric Econ 82(2):427–437
Adamowicz W, Dupont D, Krupnick AJ (2004) The value of good quality drinking water to Canadians and the role of risk perceptions: a preliminary analysis. J Toxicol Environ Health-Part A-Curr issues 67(20–22):1825–1844
Adamowicz W, Dupont D, Krupnick A, Zhang J (2011) Valuation of cancer and microbial disease risk reductions in municipal drinking water: an analysis of risk context using multiple valuation methods. J Environ Econ Manag 61(2):213–226
Adamowicz W, Dickie M, Gerking S, Veronesi M, Zinner D (2012) Collective rationality and environmental risks to children’s health. Department of economics working paper, University of Central Florida
Alberini A, Cropper M, Krupnick A, Simon NB (2004) Does the value of a statistical life vary with age and health status? evidence from the US and Canada. J Environ Econ Manag 48(1):769–792
Ancker JS, Senathirajah Y, Kukafka R, Starren JB (2006) Design features of graphs in health risk communication: a systematic review. J Am Med Inform Assoc 13(6):608–618
Bartik TJ (1988) Evaluating the benefits of non-marginal reductions in pollution using information on defensive expenditures. J Environ Econ Manag 15:111–127
Ben-Akiva ME, Lerman SR (1985) Discrete choice analysis: theory and application to travel demand. MIT Press, Cambridge
Bhattacharya S, Alberini A, Cropper ML (2007) The value of mortality risk reductions in Delhi India. J Risk Uncertain 34(1):21–47
Blass AA, Lach S, Manski CF (2010) Using elicited choice probabilities to estimate random utility models: preferences for electricity reliability. Int Econ Rev 51:421–440
Canadian Cancer Society’s Advisory Committee on Cancer Statistics (2015) Canadian cancer statistics 2015. Canadian Cancer Society, Toronto
Carlsson F, Johansson-Stenman O, Martinsson P (2004) Is transport safety more valuable in the air? J Risk Uncertain 28(2):147–163
CBC (2006) Kashechewan: water crisis in northern Ontario. http://www.cbc.ca/news/background/aboriginals/kashechewan.html. Accessed 29 June 2009
Chestnut LG, Mills D, Rowe RD, Civita PD, Stieb D (1999) Air quality valuation model version 3.0 (AQVM 3.0), report 2: methodology. Stratus Consulting, Colorado
Connelly NA, Knuth BA (1998) Evaluating risk communication: examining target audience perceptions about four presentation formats for fish consumption health advisory information. Risk Anal 18(5):649–659
Corso PS, Hammitt JK, Graham JD (2001) Valuing mortality-risk reduction: using visual aids to improve the validity of contingent valuation. J Risk Uncertain 23(2):165–184
Courant PN, Porter RC (1981) Averting expenditure and the cost of pollution. J Environ Econ Manag 8(4):321–329
Dickie M, Gerking S (1996) Formation of risk beliefs, joint production and willingness to pay to avoid skin cancer. Rev Econ Stat 78(3):451–463
Dupont DP, Jahan N (2012) Defensive spending on tap water substitutes: the value of reduced perceived health risks. J Water Health 10(1):56–68
Edge T, Byrne JM, Johnson R, Robertson W, Stevenson R (2001) Waterborne pathogens. In: Canada Environment (ed) Threats to sources of drinking water and aquatic ecosystem health in Canada. National Water Research Institute, Burlington, pp 1–5
Freeman AMI (1993) The measurement of environmental and resource values: theory and methods. Resources for the Future Press, Washington
Guimaraes P, Lindrooth RC (2007) Controlling for overdispersion in grouped conditional logit models: a computationally simple application of Dirichlet-multinomial regression. Econom J 10(2):439–452
Hagihara K, Asahi C, Hagihara Y (2004) Marginal willingness to pay for public investment under urban environmental risk: the case of municipal water use. Environ Plan C: Gov Policy 22(3):349–362
Hammitt JK, Graham JD (1999) Willingness to pay for health protection: inadequate sensitivity to probability? J Risk Uncertain 18(1):33–62
Itaoka K, Saito A, Krupnick A, Adamowicz W, Taniguchi T (2006) The effect of risk characteristics on the willingness to pay for mortality risk reductions from electric power generation. Environ Resour Econ 33(3):371–398
Jakus PM, Shaw WD, Nguyen TN, Walker M (2009) Risk perceptions of arsenic in tap water and consumption of bottled water. Water Resour Res 45:W05405
Johannesson M, Jonsson B, Borgquist L (1991) Willingness to pay for antihypertensive therapy: results of a swedish pilot-study. J Health Econ 10(4):461–474
Johannesson M, Johansson P-O, O’Connor RM (1996) The value of private safety versus the value of public safety. J Risk Uncertain 13(3):263–275
Jones-Lee MW (1974) The value of changes in the probability of death or injury. J Polit Econ 82(4):835–849
Jones-Lee MW, Hammerton M, Philips PR (1985) The value of safety: results of a national sample survey. Econ J 95(377):49–72
Jordan JL, Elnagheeb AH (1993) Willingness to pay for improvements in drinking-water quality. Water Resour Res 29(2):237–245
Konishi Y, Adachi K (2011) A framework for estimating willingness-to-pay to avoid endogenous environmental risks. Res Energy Econ 33:130–154
Krupnick A, Alberini A, Cropper M, Simon N, O’Brien B, Goeree R, Heintzelman M (2002) Age, health and the willingness to pay for mortality risk reductions: a contingent valuation survey of Ontario residents. J Risk Uncertain 24(2):161–186
Larson BA, Gnedenko ED (1999) Avoiding health risks from drinking water in Moscow: an empirical analysis. Environ Dev Econ 4(4):565–581
Lee CK, Kwak SJ (2007) Valuing drinking water quality improvement using a Bayesian analysis of the multinomial probit model. Appl Econ Lett 14(4):255–259
Lee SJ, Liljas B, Neumann PJ, Weinstein MC, Johannesson M (1998) The impact of risk information on patients’ willingness to pay for autologous blood donation. Med Care 36(8):1162–1173
Lewbel A, Dong Y, Yang TT (2013) Comparing features of convenient estimators for binary choice models with endogenous regressors. Canad J Econ 45(3):809–829
Lindhjem H, Navrud S, Braathen NA, Biausque V (2011) Valuing mortality risk reductions from environmental, transport, and health policies: a global meta-analysis of stated preference studies. Risk Anal 31(9):381–407
Livernois J (2001) The economic costs of the walkerton water crisis. Walkerton inquiry commissioned paper No. 14
Loomis JB, duVair PH (1993) Evaluating the effect of alternative risk communication devices on willingness to pay: results from a dichotomous choice contingent valuation experiment. Land Econ 69(3):287–298
McConnell KE, Rosado MA (2000) Valuing discrete improvements in drinking water quality through revealed preferences. Water Resour Res 36(6):1575–1582
Nauges C, van den Berg C (2009) Perception of health risk and averting behavior: an analysis of household water consumption in Southwest Sri Lanka. Toulouse school of economics working paper 09-139
Orgill J, Shaheed A, Brown J, Jeuland M (2013) Water quality perceptions and willingness to pay for clean water in peri-Urban Cambodian communities. J Water Health 11(3):489–506
Petrin A, Train K (2010) A control function approach to endogeneity in consumer choice models. J Mark Res 47(1):3–13
Ridell M (2012) Comparing risk preferences over financial and environmental lotteries. J Risk Uncertain 45(2):135–157
Rosado MA, Cunha-E-Sa MA, Ducla-Soares MM, Nunes LC (2006) Combining averting behavior and contingent valuation data: an application to drinking water treatment in Brazil. Environ Dev Econ 11(6):729–746
Schram C (2009) Health risk perceptions, averting behavior, and drinking water choices in Canada. Master’s Thesis, Department of Rural Economy, University Of Alberta
Schwiebert J (2012) Revisiting the composition of the female workforce: a Heckman selection model with endogeneity. Leibniz Universität Hannover discussion paper series
Shaw WD, Woodward RT (2008) Why environmental and resource economists should care about non-expected utility models. Resour Energy Econ 30(1):6–89
Stirling R, Aramini J, Ellis A, Lim G, Meyers R, Fleury M, Werker D (2001) Waterborne cryptosporidiosis outbreak, north Battleford, Saskatchewan. Health Canada. http://www.health.gov.sk.ca/nb-waterborne-cryptodporidiosis-outbreak. Accessed 15 July 2014
Swait J (2006) Chapter 9 advanced choice models. In: Kanninen BJ (ed) Valuing environmental amenities using stated choice studies: a common sense approach to theory and practice. springer, Dordrecht
Treasury Board of Canada Secretariat (2007) Canadian cost-benefit analysis guide: regulatory proposals. Treasury Board of Canada Secretariat
Tsuge T, Kishimoto A, Takeuchi K (2005) A choice experiment approach to the valuation of mortality. J Risk Uncertain 31(1):73–95
Um M-J, Kwak S-J, Kim T-Y (2002) Estimating willingness to pay for improved drinking water quality using averting behavior method with perception measure. Environ Resour Econ 21(3):287–302
Viscusi WK, Aldy JE (2003) The value of a statistical life: a critical review of market estimates throughout the world. J Risk Uncertain 27(1):5–76
Whitehead JC (2006) Improving willingness to pay estimates for quality improvements through joint estimation with quality perceptions. South Econ J 73(1):100–111
Wigle DT (1998) Safe drinking water: a public health challenge. Chronic Dis Canada 19(3):103–107
Wilson J, Aramini J, Clarke S, Novotny M, Quist M, Keegan V (2009) Retrospective surveillance for drinking water-related illnesses in Canada, 1993–2008: final report. Report prepared by Novometrix Research Inc. for the Public Health Agency of Canada. http://www.ncceh.ca/sites/default/files/DW_Illnesses_Surveillance_Aug_2009.pdf. Accessed 1 Sept 2015
Wooldridge JM (2010) Econometric analysis of cross section and panel data, 2nd edn. MIT Press, Cambridge
Wooldridge JM (2014) Quasi-maximum likelihood estimation and testing for nonlinear models with endogenous explanatory variables. J Econom 182(1):226–234
Wu PI, Huang CL (2001) Actual averting expenditure versus stated willingness to pay. Appl Econ 33(2):277–283
Zerah MH (2000) Household strategies for coping with unreliable water supplies: the case of Delhi. Habitat Int 24(3):295–307
Acknowledgements
Thanks to Shelby Gerking for comments on an earlier version of this work. We would like to thank the Canadian Water Network for financial support of the survey.
Author information
Authors and Affiliations
Corresponding author
Electronic supplementary material
Below is the link to the electronic supplementary material.
Rights and permissions
About this article
Cite this article
Lloyd-Smith, P., Schram, C., Adamowicz, W. et al. Endogeneity of Risk Perceptions in Averting Behavior Models. Environ Resource Econ 69, 217–246 (2018). https://doi.org/10.1007/s10640-016-0075-6
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10640-016-0075-6
Keywords
- Averting behavior
- Risk perceptions
- Water quality
- Human health
- Latent class models
- Expenditure model
- Endogeneity