Benefits of Preventing a Death Associated With Child Maltreatment: Evidence From Willingness-to-Pay Survey Data

Abstract

Although assessing the costs of an intervention to prevent child maltreatment is straightforward, placing a monetary value on benefits is challenging. Respondents participating in a statewide random-digit-dialed survey were asked how much they would be willing to pay to prevent a death caused by child maltreatment. Our results suggested that society may value preventing a death from child maltreatment at $15 million. If a child maltreatment intervention is effective enough to save even 1 life, then in many cases, its benefits will outweigh its costs.

Child maltreatment is a serious public health problem affecting at least 1 of 5 children during their lifetimes.¹ In 2008, approximately 1740 children in the United States died from maltreatment.² Evidence-based interventions designed to prevent maltreatment and reduce its consequences are emerging³; however, to move policy development forward, information is needed to estimate the returns that investments in these interventions will provide to society.

Benefit-cost analysis, which compares an intervention's costs with that intervention's benefits, both measured in monetary units, is an approach to estimate returns on investment.⁴ Although assessing the costs to implement an intervention is relatively straightforward,⁵ the challenge is estimating a monetary value for potential benefits of these interventions, particularly when the benefits include saving lives. One approach is the human capital method, which relies on future wage rates to assess the value of life as lost productivity to society.^6–8 Corso et al.⁹ used this method and recently estimated that per-person productivity losses and medical costs for violent deaths experienced by children aged 0 to 4 years was approximately $1 million (in 2000 US$).

Although the human capital method captures most of the tangible benefits associated with death resulting from child maltreatment, it does not capture any intangible benefits, such as the pain, suffering, and grief experienced by a community when a child dies, which may be at least as important to taxpayers.¹⁰ Contingent valuation analysis provides another method for valuing societal benefits of preventing a death, of which one approach is asking members of society how much they would be willing to pay to reduce a fatality risk.^11–13 We estimated an alternative measure of the benefits of preventing a death associated with child maltreatment by asking survey respondents their willingness to pay for small reductions in the risk of child maltreatment mortality to estimate the value of a life saved from child maltreatment.

METHODS

In fall 2008, a random-digit-dialed telephone survey of adult residents in Georgia was conducted by the Survey Research Center at the University of Georgia. The purpose of the survey was to learn respondents' opinions about several areas of health and other domestic topics. As part of this survey, we included a series of questions about respondents' willingness to pay to reduce child maltreatment mortality in their community, exposure to abuse and neglect during childhood, and perceptions of child maltreatment prevalence in their community. The literature suggests that the former (willingness to pay for a nonmarket good such as prevention of a child maltreatment death) is influenced by exposure or knowledge of the good under consideration and the perception of the risk in one's own community. Ludwig and Cook,¹⁴ for example, confirmed that the amount people are willing to pay to reduce the risk of a gunshot injury depends on how it affects them, their families, and their communities.

We asked respondents their willingness to pay for a 50% annual reduction in the risk of a child being killed by a parent or caregiver from 2 per 100 000 to 1 per 100 000. We first provided respondents with an initial willingness-to-pay value, to which they responded “yes” or “no,” followed by a second willingness-to-pay value that was either higher (yes) or lower (no) than the first willingness-to-pay value, depending on the answer to the first question. To correct for biases associated with the initial willingness-to-pay value, initial values of $25 to $250 were randomly selected.

To test for biases in how one would pay for the risk reduction,^15,16 we asked respondents in a split sample about their willingness to pay by either increased taxes or charitable donations. Wiser,¹⁷ for example, found that respondents typically stated a higher willingness to pay when confronted with a collective payment mechanism than with a voluntary payment mechanism; others^18,19 reported that taxes are viewed as more credible than donations, particularly if respondents want to ensure that everyone pays for the good under consideration.

Other data elicited from respondents that have been shown to influence willingness to pay in other studies included age, gender, race/ethnicity, income, and political affiliation. For example, strong evidence indicated that income, which is correlated with age, gender, and race/ethnicity, significantly and positively affected willingness to pay.²⁰ We added both age and age squared into the regression to account for the possible nonlinear relation between age and willingness to pay.^21–23 Political affiliation was included because previous literature has found that politically conservative individuals expressed a lower willingness to pay for public goods compared to their political counterparts^24,25

To test for significance of the effect of the covariates on the dependent variable (willingness to pay), we estimated the maximum likelihood function with the interval regression command in Stata version 9.2 (Stata Corp, College Station, TX). Regression results were then used to estimate individual willingness to pay. We used bootstrap standard errors (300 replications) to calculate the 95% confidence intervals (CIs) on the mean and median willingness to pay.

RESULTS

Of the 1132 eligible respondents contacted in the random-digit-dialed telephone survey, 44% yielded complete interviews. Willingness to pay to prevent child maltreatment mortality was obtained randomly from approximately half of the sample (n = 199). Table 1 provides a summary of the sociodemographic characteristics of the final sample included in this study. Most of the respondents were female and White, owned their own home, were younger than 50 years, and did not report a history of child maltreatment. Of the respondents 78.4% reported household income: 33.7% reported an income of $50 000 or less, and 44.7% reported an income of greater than $50 000. Approximately 77% of the participants responded that they thought the risk of child maltreatment in their community was equal to or less than the average risk presented (2 per 100 000).

TABLE 1.

Respondent Demographics: Willingness to Pay to Prevent Child Maltreatment Survey, Georgia, 2008

Variable	Mean (SD) or %
Age, y	48.3 (17.2)
Female	70.4
White	62.8
Household income, $
< 25 000	12.1
25 000–50 000	21.6
50 001–75 000	16.1
> 75 000	28.6
Missing	21.6
Own home	75.9
Child maltreatment history	28.6
Neighborhood child maltreatment riska
< Average	44.7
About average	32.2
> Average	17.6
Missing	5.5
Willingness to pay by increased taxes	53.3
Party affiliation
Democrat	39.6
Republican	32.7
Independent	17.6
Other	10.1

^aThe following scale was used for the question asking respondents' perception of child maltreatment risk in their neighborhood: 1 = much greater than average, 2 = somewhat greater than average, 3 = about average, 4 = somewhat less than average, 5 = much less than average. The responses for this question were recoded into a variable with 3 categories (> average = codes 1–2; about average = code 3; < average = codes 4–5).

As shown in Table 2, the mean estimated willingness to pay for a 50% reduction in the risk of death associated with child maltreatment was $148 (95% CI = $121, $176), whereas the median willingness to pay was $151 (95% CI = $120, $186). In the regression results provided in Table 3, the only significant predictor of willingness to pay to prevent a death associated with child maltreatment was payment type; respondents asked to pay taxes were willing to pay twice as much as those asked to make charitable donations. These findings are similar to what others have observed.^18,19 Because either type of payment likely would be used to support child maltreatment prevention programs, however, we did not account for this variation or any of the insignificant covariates in estimating the overall average willingness to pay.

TABLE 2.

Mean and Median Estimated Willingness to Pay for a 1 in 100 000 Reduction in the Mortality Risk Associated With Child Maltreatment: Willingness to Pay to Prevent Child Maltreatment Survey, Georgia, 2008

Willingness to Pay	Estimated Value, $ (95% Confidence Interval)
Mean	148 (121, 176)
Median	151 (120, 186)
Value of statistical life based on mean	14.8 million (12.1, 17.6 million)
Value of statistical life based on median	15.1 million (12.0, 18.6 million)

TABLE 3.

Interval Regression Results of Willingness to Pay for 50% Reduction in the Risk of Death Associated With Child Maltreatment: Willingness to Pay to Prevent Child Maltreatment Survey, Georgia, 2008 (n = 199)

	Coefficient	P
Age	2.20	.685
Age squared	−0.01	.802
Non-White (Ref)	1.00
White	−3.03	.926
Male (Ref)	1.00
Female	−32.53	.3
Democrat (Ref)	1.00
Republican	−40.69	.237
Independent	−46.19	.262
Other political affiliations	−128.74	.017
Neighborhood child maltreatment risk < average (Ref)a	1.00
Neighborhood child maltreatment risk > average	24.78	.544
Neighborhood child maltreatment risk about average	−16.09	.633
Household income < $25 000 (Ref)	1.00
Household income $25 000–50 000	99.57	.06
Household income $50 001–75 000	78.57	.186
Household income > $75 000	38.83	.502
No history of child maltreatment (Ref)	1.00
History of child maltreatment	−28.91	.384
Pay through donation (Ref)	1.00
Pay through taxes	120.66	<.001
Constant	59.80	.634

^aThe following scale was used for the question asking respondents' perception of child maltreatment risk in their neighborhood: 1 = much greater than average, 2 = somewhat greater than average, 3 = about average, 4 = somewhat less than average, 5 = much less than average. The responses for this question were recoded into a variable with 3 categories (> average = codes 1–2; about average = code 3; < average = codes 4–5).

DISCUSSION

In contrast with the human capital approach for valuing deaths prevented by child maltreatment, which would include only medical costs and productivity losses,⁹ our results suggested that the benefits of preventing a death associated with child maltreatment should be valued much higher. If each person is willing to pay approximately $150 to reduce the number of deaths resulting from child maltreatment by 1 per 100 000 per year, then the value of a life saved from child maltreatment (or the value of a statistical life) is $15 million (value of a statistical life = mortality risk reduction rate × average willingness to pay). Although value of a statistical life has no application to an identifiable child or to very large reductions in mortality risks, its purpose is to help describe better the likely benefits of a policy or an intervention that saves a child from death caused by maltreatment. In comparison, a substantial majority of value of statistical life estimates for adults range from approximately $1 million to $10 million.^26,27

Thus, the value of saving a child from maltreatment implied in our study has considerable policy implications. First, our higher value of a statistical life estimate may reflect the intangible benefits of preventing a death, such as the subjective value of safety, concern about others' welfare, and prevention of pain and suffering. Second, our estimates may reflect a value for preventing the death of a child that is different from the value society may place on preventing the death of an adult. Third, if evidence-based child maltreatment prevention interventions, such as home visiting programs,²⁸ can save even 1 life and have a cost of less than $15 million, then the benefits of the intervention likely will outweigh the costs.

Several study limitations must be considered. First, this study represented pilot data collected within a narrowly defined and small population, thus potentially limiting the statistical significance of some of the covariates and the generalizability of our findings. For example, unlike other studies,¹⁴ we found no effect of exposure to child maltreatment or perception of greater community risk of child maltreatment on willingness to pay to prevent child maltreatment mortality. We also did not find that any of the sociodemographic characteristics were significant in the model. This finding could be a result of the small sample size or because valuation of death may not be influenced as readily by factors that may affect the valuation of nonfatal, nonmarket goods. Second, the risk reduction level (1 in 100 000) may have been cognitively challenging to the survey respondents, thus biasing results. In a full-scale replication of this study, one could test for this bias with several techniques such as providing visual aids to convey risk or eliciting willingness to pay for different risk reduction levels in separate samples to test for sensitivity to magnitude.²⁹ Third, although not unique to our study, willingness-to-pay estimates sometimes are affected by the hypothetical nature of the good being valued such that stated and real willingness to pay differ substantially.³⁰ We attempted to test for this by asking a follow-up question to respondents as to their confidence in their stated willingness to pay, which was found not to be significantly associated with willingness to pay (results not shown; P = .66).

Despite these limitations, the real strength of this research was the estimation, for the first time, of the monetary benefits of preventing a death associated with child maltreatment. These data can prove useful to policymakers and advocates who want to effectively lobby for more resources to prevent child maltreatment. Finally, once these results are validated with national data and expanded to address nonfatal child maltreatment, the resulting willingness-to-pay estimates can be used to determine the cost-benefit ratios of policies and programs designed to prevent child maltreatment. These results will help advance the field of child maltreatment prevention by helping to identify more clearly the best return on public health investment in child maltreatment prevention.