Printer Friendly

State abortion restrictions and child fatal-injury: an exploratory study.

1. Introduction

In 1973, the U.S. Supreme Court decision in Roe v. Wade established that all women in the U.S. had the constitutional right to terminate a pregnancy via abortion. This decision made it illegal for states to implement laws prohibiting women from obtaining abortions. However, subsequent court decisions have given states the authority to restrict abortion access in specific ways. A number of states have responded by limiting public funding of abortion services, instituting mandatory waiting periods, and requiring that at least one parent give consent or be notified before a minor can obtain an abortion. A new bill, approved by the United States Senate, imposes substantial penalties on non-parent adults who help minors cross state borders to obtain an abortion without parental involvement. It will, in all probability, be signed into law by the President before this article appears in print. There is also some speculation as to whether the latest changes in the nation's Supreme Court will eventually lead to an overturning of Roe v. Wade and leave the decision of permitting legal abortions to individual states. South Dakota has passed a law making almost all abortions illegal, which seems to be designed as a challenge to Roe v. Wade, and a similar bill is currently under consideration in Mississippi.

Numerous studies have considered the effects of restrictive state abortion policies on the incidence of abortion (Oshfeldt and Gohmann 1994; Blank, George, and London 1996; HaasWilson 1996; Joyce and Kaestner 1996; Levine, Trainor, and Zimmerman 1996; Matthews, Ribar, and Wilhelm 1997; Cook et al. 1999; Levine 2003; Joyce, Kaestner, and Colman 2006), and most find that the policies are associated with declines in abortion. Other studies, detailed later in the paper, have considered the effects of the policies on teen births, female-headed household formation, teen sexual behavior and contraception use, and state-level prevalence of sexually transmitted diseases (STDs). Only recently have researchers turned to investigating another aspect--whether such policies impact outcomes for children. The premise is that abortion restrictions may result in more unplanned and unwanted pregnancies being carried to term, and also in disproportionately more children being born to women of low socioeconomic status (SES), minor women, single women, and women with inadequate parenting skills and resources. This provides grounds for hypothesizing that children born in the presence of abortion restrictions will experience poorer outcomes compared to children born in absence of such restrictions.

This work is an exploratory study that considers the effects of restrictive state abortion policies on three negative outcomes for young children that are measured at the state level--fatal-injury rates due to violence (hereafter homicide), fatal-injury rates due to accidents (hereafter unintentional causes), and fatal-injury rates due to accidents other than motor vehicle crashes where the child was in the car (hereafter non-motor unintentional). Homicide-resultant fatal injuries and unintentional fatal injuries are among the five leading causes of death for young children (Anderson 2002), thus, understanding what factors influence their prevalence is important from a public health perspective. The analysis utilizes state-level cross-sectional time-series data and two-way fixed effects empirical models. The results find some associations between abortion restrictions and increases in child fatal-injury rates--specifically, between parental consent laws and homicide-resultant fatal-injury rates for white children; mandatory delays and non-motor unintentional fatal-injury rates for white children as well as homicide-resultant fatal-injury rates for black children; and no public funding and unintentional fatal-injury rates for black children.

2. Background

A report based on the 1995 National Survey of Family Growth (Henshaw 1998) informs that in 1994, about 44.7% of all pregnancies were unintended, and 71% of pregnancies among 15 to 19-year-old women were unintended; 54% of all unintended pregnancies ended in abortion, whereas 45.3% of unintended pregnancies among 15 to 19-year-olds ended in abortion. In short, abortion services continue to be widely utilized in the United States, and state-level restrictions to abortion access can potentially affect a wide segment of the female population.

Since 1973, when the U.S. Supreme Court legalized abortion nationwide, most states have adopted at least some strategies designed to reduce abortion access. Three of the most widely adopted policies are restrictions on publicly funded abortions, mandatory waiting periods, and parental involvement laws. Restrictions on publicly funded abortions began in 1976 with the Hyde Amendment, which eliminated federal Medicaid funding for most abortions. This legislation left Medicaid funding of abortions at the discretion of states and most states responded by adopting restrictions. At present, 16 states fund all abortions sought by Medicaid recipients, 32 states only fund abortions resulting from rape or incest or life-threatening pregnancies, and two states only fund abortions in cases of life-threatening pregnancies. (1) Mandatory waiting periods were first introduced in Mississippi, and started becoming more widely adopted after the 1992 Supreme Court decision in the case of Planned Parenthood of Southeastern Pennsylvania v. Casey, which upheld a Pennsylvania law mandating a 24-hour waiting period between when a woman sought an abortion at a clinic and when the abortion could actually be performed. By 1995, 11 states had implemented similar mandatory waiting periods, which were as long as 72 hours, and in all cases the woman was required to receive, in person from the clinic, state-mandated information regarding abortion-related complications, fetal development, and alternatives to abortion in that interim period. (2) While mandatory delays do not appear to be a major restriction at first glance, when coupled with the fact that many states have only a few abortion clinics and those sometimes offer abortion services only on selected days during the week, mandatory delays can impose time and travel costs that are prohibitive for women with limited resources. (3) States began introducing parental involvement laws soon after abortion was legalized in 1973. (4) In several states, these laws were initially enjoined by court order or not enforced, but by 1996, 26 states had binding parental-involvement laws in place. The design of these laws varies across states--some states require the consent of one or both parents for a minor seeking an abortion, while other states only required that the abortion provider notify one or both parents before the abortion is performed. (5)

As stated earlier, numerous studies have found that the stricter policies on abortion were associated with reductions in the incidence of abortion. However, the extent to which this implies changes in birth outcomes is affected by whether the restrictions prompt women to take precautions against unwanted pregnancies, like abstinence or better contraceptive use. Readers are referred to the influential paper by Kane and Staiger (1996), which presents a theoretical model for why restrictions may even reduce births. (6) In their empirical analysis, Kane and Staiger find that presence of parental involvement laws appears to reduce birthrates among minor white teens, but they also appear to do so among non-minor teens and young adults--a counterintuitive result, since these age groups are outside the scope of such laws. Thus, they conclude that no definitive conclusions can be drawn about the effects of these abortion restrictions on birthrates. Recently published work by Joyce, Kaestner, and Colman (2006) finds an increase in birthrates among teens within the scope of parental involvement laws compared to 18-year-olds (who are outside the scope of the laws). Evans et al. (1993); Currie, Nixon, and Cole (1996); and Cook et al. (1999) find that no public funding for abortion results in increases in birthrates, whereas Levine, Trainor, and Zimmerman (1996) and Matthews, Ribar, and Wilhelm (1997) do not find any increases.

Among studies that directly explore the effects of these laws on sexual behavior and contraception use, Levine (2001); Argys, Averett, and Rees (2002); Levine (2003); and Sen (2006) find no significant reductions in sexual activity, and, at best, weakly significant and small increases in contraception use. Studies using state-level STD prevalence rates as a proxy for prevalence of risky sexual behavior also find no significant associations between STD rates and no public funding (Sen 2003a, b) or parental involvement laws (Dee and Sen 2005).

Thus, on balance, it appears that while the presence of abortion restrictions has reduced the incidence of abortion, it has not led to substantial changes in sexual behavior or precautions against (unwanted) pregnancies. This makes it likely that the presence of the restrictions has led to some live births that would not have occurred in their absence. There is good reason to speculate that this, in turn, might affect child outcomes. Levine et al. (1999) show that the legalization of abortion was associated with relative declines in births to teen women, single women, and non-white women. Gruber, Levine, and Staiger (1999) find that cohorts born after abortion legalization were less likely to be poor, welfare-dependent, and in single-parent households. Bitler and Zavodny (2002a) find that abortion legalization was associated with fewer children being given up for adoption--which they interpret as fewer 'unwanted' children being born, and also find that (Bitler and Zavodny 2002b) abortion legalization was associated with reductions in reports of child maltreatment. Finally, Grossman and Jacobowitz (1981) and Gruber, Levine, and Staiger (1999) find that abortion legalization was associated with reductions in neonatal mortality rates and infant mortality rates. Analogously, it is probable that making abortion inaccessible or less accessible will be associated with relative increases in births to teen women and single women, proportionately more children in poverty, more "unwanted" births, and more child maltreatment. These, in turn, may increase the risk of situations that, at the margin, lead to increases in child fatalities from violence/homicide and from accidents/unintentional causes. One such risk factor is child physical abuse and neglect. For instance, Zuravin (1987) finds that experiencing unplanned pregnancies is strongly correlated to whether a mother or another caregiver was physically abusive to her children, and whether the children were neglected--with the dimensions of neglect including lack of physical safety at home, lack of supervision of activities, and being left alone without substitute childcare arrangements. Another risk factor is young maternal age. Numerous studies document the correlation between low maternal age at childbirth and increased risk of child physical abuse (for example, Zuravin 1988; Connelly and Strauss 1992), and a recent comprehensive study on infant homicide in the U.S. from 1983-91 (Overpeck et al. 2006) finds that the strongest predictors of infant homicide deaths included maternal age at birth less than 17 years. Note that these findings do not imply that it is necessarily the mothers who are the perpetrators of violence against children. While some mothers may be more punitive or abusive towards an unwanted child, it is also very possible that mothers who are young and of low socioeconomic status simply lack the ability and resources to protect their children from other adults--for example, an abusive boyfriend. In a study on child deaths from maltreatment in the state of Missouri, Schnitzer and Egelman (2005) find that the strongest predictor of such a death was residing in a household with an unrelated male, and in more than 80% of cases this male was the mother's boyfriend. Stiffman et al. (2002) identify the presence of a mother's boyfriend in the household as a major risk factor in child maltreatment deaths, but, additionally, find high risks associated with living in step, foster, or adoptive households. Low maternal socioeconomic status and poverty are also strong predictors of overall childhood fatal injury (Nersesian et al. 1985; Emerick, Foster, and Campbell 1986). This may be because poverty-enhanced stress increases the risk of child physical abuse and neglect by the mother, because poverty correlates with increased risk of the mother forming relationships with potentially abusive men, or because poverty correlates with factors like parental substance abuse, mental illness, and cognitive disorders which, in turn, correlates with increased risk of child abuse and neglect (Reid, Macchetto, and Foster 1999; Anda, Whitfield, and Felitti 2002). Furthermore, poverty can correlate with unsafe neighborhoods and hazardous housing conditions that increase the risk of unintentional injuries--for example, numerous reports state that poor children are more likely to live in old and decrepit housing that poses a high fire risk. Finally, poor and young mothers may have to rely on substandard and unsafe childcare arrangements if they are working or at school. For all these reasons, it may be hypothesized that policies that potentially increase the relative number of unplanned pregnancies and the relative proportions of children born to young, poor, and single mothers will contribute to an increase in child fatal-injury rates from homicide and from unintentional causes.

To my knowledge, only two extant studies (Bitler and Zavodny 2002b, 2004) look at the effects of state abortion restrictions on child outcomes. The first of these considers the effects of restrictions on reports of child abuse and finds no clear results. The second finds that public funding restrictions were associated with increases in abuse reports and increases in murder by parents and relatives, but no apparent effects of the other restrictions. It is possible that in both these studies, effects that manifest themselves for specific age groups (e.g., very young children) or just for one race are masked. This is because both studies constrain the effects of abortion restrictions (at time of birth) to be the same for all children in the age group 0-17 years, and analyze data aggregated by race.

This study focuses on the effects of abortion restrictions on child fatal injuries among the age group 0-4 years. Evidence suggests that this age group is the most vulnerable to fatalities associated with abuse or neglect. The Child Welfare Information Gateway, a service provided by the U.S. Department of Health and Human Services, reports that this age group accounted for more than 75% of all child fatalities either directly attributed to abuse/neglect or where abuse/neglect were reported as being a contributing factor. The report states, "This population of children is the most vulnerable for many reasons, including their dependency, small size and inability to defend themselves." In cases where abuse or neglect does not play a role, children of this age group are still likely to be at greater risk of fatal injury than their older counterparts. For example, a child of one year can drown more easily in a bathtub in a genuine accident than a child of eight years. Hence, I focus on this age group rather than combining them with older age groups, which may potentially obfuscate the effects. (7)

The analysis is done separately for white and black children. Rates of fatal injuries from homicide and from accidents are considerably higher for black than white children, indicating that the underlying causes for such fatal injuries may also differ by race. Previously mentioned studies like Gruber, Levine, and Staiger (1999) found that effects of abortion legalization varied across white and black children, which raises the possibility that effects of abortion restrictions might also vary across white and black children. Black women are more likely to be in poverty than their white counterparts, hence, some restrictions--like no public funding for abortions--may have greater relevance for them. In case of parental involvement, Joyce and Kaestner (1996) found that parental consent laws affected abortion among white teens but not black teens, while Lichter, McLaughlin, and Ribar (1998) found that parental consent laws were associated with the formation of more white single-mother households, whereas no public funding was associated with formation of more black single-mother households. All of this supports doing the analysis separately by race. (8)

3. Data

This study uses state-level longitudinal data on fatal-injury rates over the period 1981-2002 for all states and the District of Columbia. The data is obtained from the Web-based Injury Statistics Query and Reporting System (WISQAR), made publicly available by Centers for Disease Control and Prevention. It is based on the number of fatal injuries by cause of injury available from the National Center for Health Statistics (NCHS), and corresponding population estimates from the Census Bureau. The WISQAR data has two limitations. The first is that it is only available aggregated by age group (0-4, 5-9, 10-14, etc.) for all years of the study. While it is available in single age-years after 1990, there is much less within-state variation in abortion policies after 1990, which poses difficulties when using a two-way fixed effects empirical technique that includes state fixed effects. The second limitation is that the WISQAR data provides the state of death of the subject, but not the state of birth, or as is more relevant in this case, the state of residence of the mother during course of the pregnancy. To the extent that this is different from the state where the child's death occurred, there can be some measurement error in the relevant abortion restriction variables. Finally, since the data is aggregated at the state level, there is no information either on age of the mother (namely, whether she was a minor and within scope of parental involvement laws) or on her socioeconomic status (namely, whether she was Medicaid eligible). Thus, since I cannot identify the children for whom the restrictions would have personally applied when in utero, I frame the hypothesis of interest in more general terms--namely, that there will be a higher incidence of child fatal injuries due to homicide and due to unintentional causes among child cohorts born in the presence of state policies that restrict abortion access.

Three categories of child fatal injuries are considered--fatal injuries resulting from homicide, fatal injuries due to all unintentional causes, and fatal injuries due to unintentional causes other than motor vehicle crashes where the child was a passenger in the car (hereafter referred to as 'non-motor unintentional causes'). The rationale for the last category is that while risk of most other unintentional injuries can be exacerbated by parental negligence or other parental characteristics, motor vehicle crashes (where the child is a passenger) are often caused by third parties, and also endanger the driver of the car. Thus, there is arguably less likelihood of a statistical correlation between such crashes and child neglect or poor parental skills. (9)

The abortion restriction policies considered are enforced parental consent laws, enforced parental notification laws, no public funding for abortion, and mandatory delay laws. The data on abortion restrictions up to the year 1990 corresponds closely to that used by Blank, George, and London (1996), which was compared against the data from two other studies, by Haas-Wilson (1996) and by Levine (2002, Table 1). In case of irreconcilable discrepancies (which occurred for Utah and Arizona), I directly contacted the Alan Guttmacher Institute (Washington D.C. branch), which provided detailed information on the evolvement of the parental involvement laws, which was then used to code the laws for these two states. Information for abortion restrictions in successive years is obtained using reports prepared by Sollom (1995; 1997), Levine (2002, Table 1), and from the Alan Guttmacher Institute. An additional variable, the number of neighboring states with no parental involvement laws for each state-year cell, is constructed using the information on parental consent and notification laws. Finally, the abortion providers in the state per 100,000 women aged 10-50 years is included. This is calculated using data on the number of abortion providers for each state-year cell from the Alan Guttmacher Institute and population estimates from the Census Bureau. The data on number of abortion providers is available for all years up to 1982. Thereafter, it is available for the years 1984, 1985, 1987, 1988, 1991, 1992, 1995, 1996, 1999, and 2000. In the missing years, the number is computed as the weighted average of the last preceding and first succeeding year of available data, rounded off to the nearest integer. For 2001 and 2002, the rate from 2000 is used. Table A in the Appendix presents details of the abortion policies in each state.

The basic empirical approach involves identifying the effects of abortion restrictions on child fatal injuries by exploiting variations in the timing of these laws within states over time. I also control for additional regressors representing other time-varying, state-level characteristics and policies that may have additionally influenced child fatality rates. These include the percentage of state population in poverty, the state unemployment rate, the percentage of state population living in rural areas, the apparent per capita consumption of alcohol in the state, the maximum AFDC (Aid to Families with Dependent Children) benefit level for a family of three (in 1982-4 dollars), and a binary indicator for a 'family cap' law, which prohibits increases in AFDC or TANF (Temporary Assistance for Needy Families) payments for a child conceived when the mother is already a recipient in those programs. It would have been useful to have also controlled for state spending on child welfare agencies. However, reliable and consistent data on this variable was not available for this study period. Hence, I use the number of police officers per 100,000 people in the state as a proxy variable.

4. Empirical Methods and Results

The choice of the empirical model is driven by the underlying count nature of the fatal injury data. A substantial number of the state-year cells have very limited numbers of child fatal injuries and a non-trivial proportion have zero child fatal injuries for at least one race (details of the distribution of the data are available from the author upon request), therefore, employing a conventional fatality rate regression model could lead to a poor fit and weak statistical power. Thus, I use a negative binomial model, which assumes that the mean of the fatal injury count [y.sub.st] from a particular cause and for a given race is:

E([y.sub.st]) = [[lambda].sub.st] = exp ([A.sub.st][beta] + [X.sub.st][alpha] + [u.sub.s] + [v.sub.t] + [LnP.sub.st]). (1)

[A.sub.st] is a vector of relevant abortion restriction policies, [X.sub.st] is a vector of the other time-variant state characteristics and policies that were previously listed, and [u.sub.s] and [v.sub.t] respectively represent vectors of binary state and year effects. The state effects minimize bias from state-level time-invariant unobservables and the year effects minimize bias from temporal factors that apply to all states in a given year. [LnP.sub.st] is the natural log of the population of children aged 0-4 of that race in the state-year cell, whose coefficient is constrained to be 1. The exponential functional form along with inclusion of [LnP.sub.st] with its coefficient constrained to be 1 permit the estimates to be interpreted as proportionate changes in the rate of that particular type of fatal injury as well as the proportionate change in the fatal injury count. Standard errors are clustered within state to circumvent the problem of overstated precision of estimates, which may occur if serial correlation exists within state (Bertrand, Duflo, and Mullainathan 2004).

I turn next to describing how the abortion policy variables are created. Among the policies, enforced parental consent, enforced parental notification, no public funding, and mandatory delays are represented by binary variables that are 1 if the policy is in place, 0 otherwise. However, recall that WISQAR data only makes fatality counts and rates available by age group over the relevant study period, and this paper uses the data on 04 years of age. Since the relevant abortion policies are those that were in place when the children were in utero, this requires some manipulations of the abortion policy variables to obtain the 'pertinent' ones. This is done in the following method. First, it is assumed that births are evenly distributed through the year and that all pregnancies last nine months. Therefore, among children in the age cohort of 0-< 1 in year t, for 25% (i.e., those born in the last quarter of the year t) the pertinent abortion policies when in utero are the policies in place in year t, and for the remaining 75% (those born in the first three quarters of year t) it is those in place in year t - 1. Parallel assumptions are made for children in age cohorts 1, 2, 3, and 4 years. Thus, in general, for children in age cohort j in year t and state s, the relevant abortion policies when they were in utero is denoted as

[A.sup.j.sub.st] = 0.75 [a.sub.s,t-j-1] + 0.25 [a.sub.s,t-j], (2)

where [a.sub.s,t-k] is 1 if the relevant abortion restriction was in place in state s in the chronological year t-k, 0 otherwise. Because the overall population of 0- to 4-year-olds is approximately uniformly distributed over the five age cohorts of 0, 1, 2, 3, and 4-year-olds, the relevant abortion policies for the 0-4 age group are, therefore, defined as the mean of [A.sup.j.sub.st],

[A.sub.st] = [4.summation over (j = 0)] [A.sup.ju.sub.st]/5. (3)

By construction, the above variable takes values between 0 and 1 for enforced parental consent, enforced parental notification, no public funding, and mandatory delays, with 0 denoting the case that none of the age cohorts had that particular abortion restriction in place when in utero, 1 denoting the case that all five age cohorts had that restriction in place. Thus, the estimated effect associated with any of the abortion restriction variables may be interpreted as the difference in child fatality when all the age cohorts in the 0-4 year age range have that specific abortion restriction present in utero versus when none of them have it present. In the pooled state-year sample, enforced parental consent, enforced parental notification, no public funding, and mandatory delay restrictions take the value 1 for 13%, 8.8%, 52.6%, and 3.9% of the state-year cells, respectively; the value 0 for 79%, 79.6%, 27.5%, and 88.9% of the state-cells respectively; and values between 0 and 1 for the remaining state-year cells. (10) Detailed distributions for each restriction are available upon request. The number of border states with no parental involvement laws and the number of abortion providers per 100,000 women are constructed following the same methods, but by using the actual values, and are in the form of continuous variables.

Table 1 presents sample means for fatal injury counts, fatal-injury rates, presence of abortion restrictions, the average number of border states with no parental involvement laws, the average abortion provider rates, and the state-level control variables.

Table 2 presents estimation results from negative binomial models for fatal injuries for 0-to 4-year-olds, by race. For each race, the first set of results pertains to homicide-resultant fatal injuries, the second set to unintentional fatal injuries, and the third set to non-motor unintentional fatal injuries. For brevity, only the results pertaining to abortion restrictions, border states, and abortion provider rates are presented. Estimated results are presented in the form of Exp([beta])-1, and may be interpreted as semi-elasticities.

For white children, the presence of parental consent laws is found to be significantly associated with a 20% increase in homicide-resultant fatal injuries (p < 0.06). That is, state-years when all the age cohorts in the 0-4 years age groups were subject to enforced parental consent laws in utero have, on average, 20% more homicide-resultant fatal injuries than state-years where none of age cohorts were subject to such laws. Congruently, more border states with no parental involvement laws are associated with a decrease in homicide-resultant fatal injuries. On average, the presence of one more border state with no parental involvement laws reduces such fatal injuries by approximately 6% (p < 0.01). Mandatory delays also appear to increase homicide-resultant fatal injuries for white children by 23.5% with weak significance (p < 0.09). No public funding and enforced parental notification have positive but insignificant effects.

In the case of unintentional fatal injuries for white children, only mandatory delays are found to have a statistically significant result--they are associated with a 9% increase in such fatal injuries (p < 0.01). However, when we turn to non-motor unintentional fatal injuries, no public funding appears to also be associated with about a 6% increase in such injuries, with weak statistical significance (p < 0.09). Notably, more border states with no parental involvement laws continues to be associated with a reduction in unintentional fatal injuries even though presence of parental consent laws and parental notification laws have no significant effects. The presence of abortion providers has a negative relationship with all three categories of fatal injuries for white children, but in no case do they approach conventional levels of statistical significance.

Turning next to the results for black children and homicide-resultant fatal injuries, we see that the presence of mandatory delay laws is associated with statistically significant increases in such fatal injuries by 30% (p < 0.01). The estimated effects of the other restrictions are positive and that of more border states with no parental involvement laws is negative, however, none of these is statistically significant. In the case of unintentional fatal injuries for black children, on the other hand, it is the lack of public funding for abortion that has significant effects, increasing overall unintentional fatal-injury rates by about 15% (p < 0.02) and non-motor unintentional fatal injuries by 19% (p < 0.01). One noteworthy finding is that parental consent laws and parental notification laws appear to have a negative relationship with unintentional fatal-injury rates among black children, and while the results are statistically imprecise for parental consent laws, they are significant at the 10% level or better for parental notification laws. This suggests that, among blacks, parental involvement laws might have an effect similar to that proposed in the aforementioned paper by Kane and Staiger (1996); namely, these restrictions may lead to women taking precautions against pregnancies, and prevent some births where the (born) infants might otherwise have been at high risk of death from unintentional causes.

To explore whether the relationships between abortion restrictions and child fatal-injury rates are biased due to state-level time-variant unobservables that correlate to both, I also estimated a set of counterfactual equations where the fatal injury among adults aged 25-65 years from homicides and from unintentional causes were regressed on the same abortion restrictions and other control variables. While arguably abortion restrictions may enhance stress levels and thereby increase the risk of violence and of accidents among adults too, it is unlikely that the magnitude of effects from such sources will be large enough to be statistically discernible over the age group of 25-65 years. Hence, using data for the adults seems to be a reasonable technique for detecting spurious correlations between abortion restrictions and fatal injuries at the state-level.

Detailed results for the counterfactual models are available upon request. For the large part, no statistically discernible associations were found between abortion restrictions and the adult fatal-injury rates, but there were some exceptions. Particularly, mandatory delay laws appeared to be significantly associated with increases in homicide-resultant fatal injuries for adults of both races and unintentional fatal injuries for black adults, while parental consent and notification laws both appeared to be associated with decreases in unintentional fatal injuries for black adults and no public funding appeared to be associated with decreases in homicide-resultant fatal injuries for white adults. This suggests that there exist at least some state-level unobservables that correlate to certain abortion policies and to fatal injuries among the general population. The problem is deciphering whether the previously reported significant associations between abortion restrictions and child fatal injuries are entirely an artifact of these unobservables. In an attempt to decipher this, I estimate a second set of regression equations for child fatal injuries where, in addition to the existing covariates, I also include the corresponding fatal-injury rates for 25- to 65-year-old adults as a proxy for these confounding unobservables. These results are presented in Table 3. The notable differences between these results and the ones in Table 2 are as follows: The relationship between mandatory delays and homicide-resultant fatal injuries among white children is still positive, but ceases to be statistically significant; and the relationships between the parental involvement laws and unintentional fatal injuries for black children are still negative, but also cease to be statistically significant. All other results, including the association between mandatory delays and homicide-resultant fatal injuries for black children, remain very similar to those in Table 2.

Among the other control variables whose results were not shown here, having a greater percentage of the population living in rural areas was associated with increases in child fatal-injury rates. Higher apparent per capita alcohol consumption was strongly associated with increases in unintentional fatal-injury rates for white children. Unemployment rates were negatively and sometimes significantly associated with fatal-injury rates for children of both races. (11) The remaining variables had mixed results. Full results are available upon request.

There is the possibility that the abortion restrictions in utero for the different age cohorts in the 0-4 age group will not have the same impact on the aggregate fatal injuries in the age group. For example, data from the Bureau of Justice Statistics, available online, show that in the last three decades, the large majority of homicide victims in the 0-4 age group were 1 year or 1-<2 years of age--hence abortion restrictions relevant for the younger age cohorts may have a different impact on homicide-resultant fatal-injury rates than restrictions relevant for the older age cohorts. To explore this, five additional model specifications were estimated where the abortion policies were

[A.sub.st] = [A.sup.j.sub.st], j = 0 ... 4, (4)

where [A.sup.j.sub.st] is defined as in Equation 2.

The results revealed some variations in effects across age cohorts. For example, in the case of homicide-resultant deaths among white children, parental consent laws pertaining to age cohorts 0, 1, and 2 years had large (17%, 21%, and 19%, respectively) and statistically significant effects, whereas for age cohorts 4 and 5, the size of the effects were smaller (11% and 12%) and fell just short of significance at the 10% level. In case of unintentional fatal injuries among black children, presence of no public funding for age cohorts of 0, 1, 2, and 3 years were statistically significant, but not for age cohort 5. These findings tentatively suggest that any negative effects of abortion restrictions on child outcomes manifest themselves more strongly when the children are young, and the effects weaken somewhat as the children become older. Again, the full results are available upon request.

One concern with the results presented in Tables 2 and 3 is the potential collinearity between enforced parental consent or notification, no public funding, and mandatory delays. To see the extent to which the estimated effects of the policies are sensitive to the inclusion of each other, sets of equations are estimated where each of the policies are omitted in turn. In all cases, border states with no parental involvement laws, abortion provider rates, the other state-level variables, and the corresponding fatal-injury rates for adults aged 25-65 years are controlled for. Table 4 presents these results. For white children, the relationships between parental consent laws and homicide-resultant fatal injuries and between mandatory delays and unintentional fatal injuries remain qualitatively similar after the exclusion of other restrictions, but there is noticeable gain in statistical precision in most cases with the exception of mandatory delays and non-motor unintentional fatal injuries. More interestingly, the estimated relationship between no public funding and unintentional fatal injuries among white children gains substantially in magnitude and becomes statistically significant, particularly when the parental involvement laws are omitted. For black children, the relationship between no public funding and unintentional fatal injuries are robust to the exclusion of the other restrictions. However, parental consent laws and no public funding each become positively and significantly associated with homicide-resultant fatal injuries when the other is omitted. Notably, also, for black children, the effect of parental notification laws on unintentional non-motor fatal injuries is negative and statistically significant when mandatory delay laws are omitted. Thus, the issue of collinearity cannot be dismissed, and the models estimated in Tables 2 and 3 may fail to detect some of the relationships between abortion restrictions and child fatal injury. Therefore, one final set of models are estimated and presented in Table 4, where the individual abortion policies are replaced by a single additive 'index of restrictiveness' variable that is simply created by summing the restriction variables that were created using Equation 2. By construction, this variable ranges between zero and three points, where three points defines the cases where all of the age cohorts in the 0-4 age group were subject to no public funding, mandatory delays, and either parental consent or parental notification when in utero, and zero points defines the cases where none of the age cohorts in that age group were subject to any of the restrictions when in utero.

In this final case, increases in the index of restrictiveness are found to be associated with increases in all fatal injuries among white children and homicide-resultant fatal injuries among black children. Specifically, with a one-point increase in this index, homicide resultant fatal injuries among white and black children increase on average by 11% and 12%, respectively (p < 0.01 in both cases), and unintentional and non-motor unintentional fatal injuries increase among white children by 3% and 4%, respectively (p < 0.05 in both cases). Unintentional fatal injuries among black children are also positively associated with increases in the index of restrictiveness, but the results fall well short of statistical significance at the 10% level.

Other sensitivity analyses were also conducted. Specifically, the abortion restriction variables from Equation 2 were redefined using each of the following two methods:

[A.sup.j.sub.st] = [a.sub.s,t-j-1] (5)

and

[A.sup.j.sub.st] = 0.5 [a.sub.s.t-1] + 0.5 [a.sub.s,t-j]. (6)

These were then used to recalculate the final variables as in Equation 3, and the negative binomial models from Table 3 were re-estimated using these newly defined variables by turn. The index of restrictiveness was also recreated using these newly defined variables, and those equations were re-estimated. The empirical estimations were also repeated using Poisson rather than Negative Binomial models. The results in all cases remained very similar to the ones reported. (12)

5. Discussion and Conclusion

The results from this study indicate that associations exist between certain restrictive state abortion policies and increases in homicide-resultant child fatal-injury rates for white and black children, and increases in unintentional fatal-injury rates for white children. Unintentional fatal injuries among black children are positively associated with no public funding, but in certain model specifications they appear to be negatively associated with parental notification laws. Overall increases in the index of abortion restrictiveness have positive but statistically imprecise associations with such fatal injuries.

On the whole, the evidence suggests that abortion restrictions tend to lead to detrimental outcomes for children in terms of fatal-injury rates. While I do not dismiss the possibility that these results might be an artifact of measurement errors in the data or of unaccounted confounding factors, the results nonetheless display certain logical patterns that argue against them being purely spurious. For example, it is parental consent laws rather than parental notification laws that are correlated with increments in homicide-resultant fatal-injury rates for whites. Of the two types of parental involvement laws, clearly parental consent is the more restrictive, since it allows parents to compel a minor to carry a pregnancy to term against her own wishes if they themselves are anti-abortion. Parental notification laws do not give parents such abilities. Thus, parental consent laws are likely to increase the probability of 'unplanned and unwanted' births more so than parental notification laws and, therefore, should be more strongly correlated with outcomes that are potentially associated with unwanted children. The finding that unavailability of public funding has a stronger association with fatal-injury rates of black children than white children is also viable. Since blacks are at higher risk of poverty and dependency on public funding than whites, it is logically congruent that the unavailability of public funding will have a greater impact on black women's choice between aborting and carrying the pregnancy to term and, therefore, have stronger effects on outcomes for black children than their white counterparts.

It should also be noted that the finding that parental consent laws have stronger effects on fatal-injury rates of white children than black, whereas no public funding affects black children rather than white children, is broadly congruent with the findings of Lichter, McLaughlin, and Ribar (1998) that parental consent laws increase single-parent households among white women while unavailability of public funding increases that status among black women.

Some questions remain about the results. It is perplexing as to why no public funding for abortion only relates to unintentional fatal injuries among black children whereas mandatory delay laws only relate to homicide-resultant fatal injuries among black children, and why mandatory delays relate to unintentional fatal injuries among white children but not to homicide-resultant fatal injuries among white children. At best it can be speculated that women who are prevented from having abortions due to lack of public funding are different at the margin than women prevented from doing so due to mandatory delay laws. However, the nature of this data does not allow any further deciphering of this issue. It is also puzzling as to why, in some models, more border states with no parental involvement laws appear to reduce fatal injuries when neither parental consent nor parental notification laws within that state have any statistically discernible effects. One reason might be that this occurs because the bordering states with no parental involvement laws have other unobserved characteristics--for example, fewer hostile protestors at abortion clinics--that makes getting an abortion in those states easier in general. This requires further research for verification.

The data limitations of this study have been previously discussed. An added limitation of aggregated data is that we cannot decipher the precise paths through which the restrictions might affect child fatal-injury rates. For example, we cannot tell whether a fatal injury occurs because the mother (or father, if he is involved) is more punitive or negligent towards an 'unwanted' child, whether the mother is unable to protect the child from abusive relatives or boyfriends, or whether bad neighborhoods or poor housing quality play a role. Thus, there is a need for further studies with more detailed data before definitive conclusions can be reached regarding the mechanism underlying the relationship between state abortion restrictions and child fatal injuries.

There is also scope for further research on effects of abortion restrictions using data specific to infant mortality. The Compressed Mortality Files from the CDC provides detailed state-level data on neonatal and infant mortality rates from various causes including homicide-resultant injuries. One planned direction of future research is exploring how abortion restrictions associate with neonatal and infant mortality rates from a broad range of causes that include but are confined to deaths from homicide or unintentional causes.

What are the policy implications of this study? Abortion rights and restrictions are among the most controversy-fraught issues of these times, thus policy implications are likely to be subject to the stance that the general public takes in the debate. Two extreme positions are possible. One of these is that abortions are 'costless' events from the moral and social viewpoint, and since the post-natal death of a child from fatal injury carries social and moral costs, any restrictions linked with increases in child fatal injuries are undesirable policies. The other extreme position is that the aborting of a fetus and the post-natal death of a child from fatal injury have equal cost from a moral and social viewpoint. Hence, since only a fraction of prevented abortions end up as cases of child fatal injury, the restrictions still 'save lives,' and are therefore desirable. (13) Between these extremes lie an infinitum of possible opinions and positions regarding what is an acceptable tradeoff between prevented abortions and a child's death from fatal injury. Ultimately, the desirability of the abortion restriction policies in light of their potentially having incremental effects on child fatal injuries becomes a normative issue. What seems clear is preventing abortions by restricting abortion access may not be sufficient to 'protect life' if the restrictions lead to future adverse consequences like fatal injuries for those same lives. Thus, the morally unambiguous policy implication is that abortion restrictions should be accompanied by more societal resources being devoted toward family counseling, financial support, and educational services for those women who are most likely to carry pregnancies to term because of the presence of the restrictions. At the same time, more resources should be devoted to state and local services responsible for detecting and preventing cases of child abuse and neglect. The other policy implication is the prevention of unplanned pregnancies. In that case, it becomes important to weigh the effectiveness of abstinence-only versus other types of sexual educational programs, and also to consider whether low-income adult women should be offered more help with family planning and access to contraception so that they can better avoid unintentional pregnancies.
Appendix
State Laws Pertaining to Abortion

States Parental Consent Parental Notification

Alabama Effected 1987 No law
Alaska No law No law
Arizona No law Effected 1982; revoked/
 enjoined 1987
Arkansas No law Effected 1989
California No law No law
Colorado No law No law
Connecticut No law Effected 1990; revoked/
 enjoined 1998
Delaware No law Effected 1995
District of Columbia No law No law
Florida No law No law
Georgia No law Effected 1991
Hawaii No law No law
Idaho Effected 2000 Effected 1997; revoked/
 enjoined 2000
Illinois No law No law
Indiana Effected 1984 Effected 1982; revoked/
 enjoined 1984
Iowa No law Effected 1997
Kansas No law Effected 1992
Kentucky Effected 1994 No law
Louisiana Effected 1991 Effected 1992
Maine Effected 1989; No law
 revoked/enjoined
 1997
Maryland No law Effected 1992
Massachusetts Effected 1979 No law
Michigan Effected 1991 No law
Minnesota No law Effected 1980, 1990;
 revoked/enjoined 1987
Mississippi Effected 1993 No law
Missouri Effected 1979 No law
Montana No law No law
Nebraska No law Effected 1991
Nevada No law No law
New Hampshire No law No law
New Jersey No law No law
New Mexico No law No law
New York No law No law
North Carolina Effected 1995 No law
North Dakota Effected 1981 No law
Ohio No law Effected 1986

Oklahoma No law No law
Oregon No law No law
Pennsylvania Effected 1994 No law
Rhode Island Effected 1982 No law
South Carolina Effected 1990 No law
South Dakota No law Effected 1997
Tennessee Effected 1998 Effected 1990; revoked/
 enjoined 1998
Texas No law Effected 1999
Utah No law Effected 1976
Vermont No law No law
Virginia No law Effected 1998; revoked/
 enjoined 2002
Washington No law No law
West Virginia No law Effected 1984
Wisconsin Effected 1998 No law
Wyoming Effected 1989 No law

States No Public Funding

Alabama Effected 1978, 1981; revoked/enjoined 1980
Alaska Effected 1997
Arizona Effected 1976, 1978, 1982;
 revoked/enjoined 1977, 1980
Arkansas Effected 1978, 1981; revoked/enjoined 1980
California No law
Colorado Effected 1986
Connecticut Effected 1978; revoked/enjoined 1980
Delaware Effected 1978, 1981; revoked/enjoined 1980
District of Columbia Effected 1989
Florida Effected 1978, 1981; revoked/enjoined 1980
Georgia Effected 1978, 1982; revoked/enjoined 1980
Hawaii No law
Idaho Effected 1978, 1981; revoked/enjoined 1980,
 1994
Illinois Effected 1981; revoked/enjoined 1994
Indiana Effected 1978, 1981; revoked/enjoined 1980
Iowa Effected 1978, 1981; revoked/enjoined 1980
Kansas Effected 1978, 1981; revoked/enjoined 1980
Kentucky Effected 1978
Louisiana Effected 1980
Maine Effected 1978, 1981; revoked/enjoined 1980
Maryland No law
Massachusetts No law
Michigan Effected 1989
Minnesota Effected 1978, 1982; revoked/enjoined 1980,
 1994
Mississippi Effected 1978, 1981; revoked/enjoined 1980
Missouri Effected 1978, 1981; revoked/enjoined 1980
Montana Effected 1978, 1981; revoked/enjoined 1980,
 1994
Nebraska Effected 1978, 1981; revoked/enjoined 1980
Nevada Effected 1978, 1981; revoked/enjoined 1980
New Hampshire Effected 1978, 1981; revoked/enjoined 1980
New Jersey No law
New Mexico Effected 1978, 1981; revoked/enjoined 1980,
 1994
New York No law
North Carolina Effected 1997
North Dakota Effected 1976
Ohio Effected 1976, 1978, 1981;
 revoked/enjoined 1977, 1979
Oklahoma Effected 1978, 1981; revoked/enjoined 1980
Oregon No law
Pennsylvania Effected 1986
Rhode Island Effected 1978
South Carolina Effected 1978, 1981; revoked/enjoined 1980
South Dakota Effected 1978
Tennessee Effected 1978, 1981; revoked/enjoined 1980
Texas Effected 1977, 1981; revoked/enjoined 1979
Utah Effected 1978
Vermont Effected 1978; revoked/enjoined 1983
Virginia Effected 1978
Washington No law
West Virginia No law
Wisconsin Effected 1979, 1981; revoked/enjoined 1980
Wyoming Effected 1978

 Abortion Abortion
 Provider Provider
States Mandatory Delays Rate (1980) Rate (2000)

Alabama Effected 2002 3.274 1.086
Alaska No law 7.616 3.600
Arizona No law 5.111 1.461
Arkansas Effected 2001 1.999 0.934
California No law 7.501 4.002
Colorado No law 8.391 3.103
Connecticut No law 4.942 5.158
Delaware No law 3.813 3.932
District of Columbia No law 7.231 8.366
Florida No law 5.744 2.561
Georgia No law 4.522 1.045
Hawaii No law 16.508 15.022
Idaho Effected 1995 4.367 1.868
Illinois No law 1.704 1.020
Indiana Effected 1997 1.709 0.852
Iowa No law 2.888 0.977
Kansas Effected 1992 3.707 0.910
Kentucky Effected 1998 1.387 0.256
Louisiana Effected 1995 1.242 0.971
Maine No law 6.694 4.119
Maryland No law 3.905 2.642
Massachusetts No law 4.852 2.568
Michigan Effected 1998 2.847 1.730
Minnesota Effected 2001 1.477 0.830
Mississippi Effected 1992 1.211 0.476
Missouri No law 1.893 0.373
Montana Effected 1996; 9.083 3.131
 revoked/
 enjoined
 1999
Nebraska Effected 1993 1.553 1.021
Nevada No law 6.845 2.319
New Hampshire No law 6.443 3.845
New Jersey No law 3.875 3.520
New Mexico No law 6.504 2.073
New York No law 5.418 4.320
North Carolina No law 6.254 2.387
North Dakota Effected 1994 1.081 1.102
Ohio Effected 1994 1.616 1.070
Oklahoma No law 1.825 0.613
Oregon No law 7.032 3.584
Pennsylvania Effected 1993 3.284 2.121
Rhode Island No law 1.807 1.964
South Carolina Effected 1995 (a); 1.561 0.851
 revoked/
 enjoined
 1997
South Dakota Effected 1994 1.036 0.938
Tennessee No law 3.688 1.028
Texas No law 2.882 1.088
Utah Effected 1994 1.155 0.583
Vermont No law 10.901 6.177
Virginia Effected 2002 3.484 2.284
Washington No law 7.084 3.173
West Virginia No law 1.434 0.598
Wisconsin Effected 1998 2.234 0.645
Wyoming No law 4.969 2.093

(a) Only a 1-hour waiting period was imposed.


I am grateful to Sara Markowitz, Frank Sloan, Chris Ruhm, Meredith Kilgore, Michael Morrisey, and David Bishai, participants of the SHESG 2005 conference and participants at Lister Hill Center (UAB) November 2005 seminar for very helpful comments. Janeen Burlinson provided valuable research assistance and Christine Campbell provided valuable assistance with manuscript preparation. The responsibility for all errors and opinions is mine.

Received April 2006; accepted July 2006.

References

Anda, R. F., C. L. Whitfield, and V. J. Felitti. 2002. Adverse childhood experiences, alcoholic parents, and later risk of alcoholism and depression. Psychiatric Services 53:1001-9.

Anderson, R. N. 2002. Deaths: Leading causes for 2000. National Vital Statistics Report 50(16). Hyattsville, MD. National Center for Health Statistics. Last accessed 28 February 2006. Available http://www.cdc.gov/nchs/data/ nvsr/nvsr50/nvsr50_16.pdf.

Argys, L. M., S. L. Averett, and D. I. Rees. 2002. The impact of government policies and neighborhood characteristics on teenage sexual activity and contraceptive use. American Journal of Public Health 92:1173-8.

Bertrand, M., E. Duflo, and S. Mullainathan. 2004. How much should we trust differences in differences estimates? Quarterly Journal of Economics 119:249-76.

Bitler, M. P., and M. Zavodny. 2002a. Child abuse and abortion availability. AEA Papers and Proceedings 92(2):363-7.

Bitler, M. P., and M. Zavodny. 2002b. Did abortion legalization reduce the number of unwanted children? Evidence from adoptions. Perspectives on Sexual and Reproductive Health 34:25-33.

Bitler, M. P., and M. Zavodny. 2004. Child maltreatment, abortion availability, and economic conditions. Review of Economics of the Household 2:119-41.

Blank, R. M., C. C. George, and R. A. London. 1996. State abortion rates: The impact of policies, providers, and economic environment. Journal of Health Economics 15:513-53.

Bureau of Justice Statistics. The number of homicides of children under age 5 increased through the mid 1990s, but declined recently. Accessed 15 July 2006. Available http://www.ojp.usdoj.gov/bjs/homicide/children.htm.

Child Welfare Information Gateway, U.S. Department of Health and Human Services. Child Abuse and Neglect Fatalities: Statistics and Interventions. Accessed 15 July 2006. Available http://www.childwelfare.gov/pubs/ factsheets/fatality.cfm.

Connelly, C. D., and M. A. Strauss. 1992. Mother's age and physical child abuse. Child Abuse and Neglect 16:709-18.

Cook, P. J., A. M. Parnell, M. J. Moore, and D. Pagnini. 1999. The effects of short-term variation in abortion funding on pregnancy outcomes. Journal of Health Economics 18:241-57.

Currie, J., L. Nixon, and N. Cole. 1996. Restrictions on Medicaid funding of abortion: Effects on birth weight and pregnancy resolutions. Journal of Human Resources 31:159-88.

Dee, T., and B. Sen. 2005. Do abortion restrictions promote safe-sex among teenagers? Unpublished paper, University of Alabama at Birmingham.

Emerick, S. J., L. R. Foster, and D. T. Campbell. 1986. Risk factors for traumatic infant death in Oregon, 1973 to 1982. Pediatrics 77:518-22.

Evans, M. I., E. Gleicher, E. Feingold, M. Johnson, and R. J. Sokol. 1993. The fiscal impact of the Medicaid abortion funding ban in Michigan. Obstetrics and Gynecology 82:555-60.

Grossman, M., and S. Jacobowitz. 1981. Variations in infant mortality rates among counties of the U.S.: The role of public policies and programs. Demography 18:695-713.

Gruber, J., P. Levine, and D. Staiger. 1999. Abortion legalization and child living circumstances: Who is the "marginal child"? Quarterly Journal of Economics 114:263-91.

Haas-Wilson, D. 1996. The impact of state abortion restrictions on minors' demand for abortions. Journal of Human Resources 31:140-58.

Henshaw, S. K. 1998. Unintended pregnancies in the United States. Family Planning Perspectives 30:24-9, 46.

Joyce, T., and R. Kaestner. 1996. State reproductive policies and adolescent pregnancy resolution: The case of parental involvement laws. Journal of Health Economics 15:579-607.

Joyce, T., R. Kaestner, and S. Colman. 2006. Changes in abortions and births and the Texas parental notification law. New England Journal of Medicine 354:1031-8.

Kane, T. J., and D. Staiger. 1996. Teen motherhood and abortion access. Quarterly Journal of Economics 111:467-506.

Levine, P. 2001. The sexual activity and birth control use of American teenagers. In Risky behavior among American youth: An economic analysis, edited by J. Gruber. University of Chicago Press, pp. 167-218.

Levine, P. 2002. The impact of social policy and economic activity throughout the fertility decision tree. NBER Working Paper No. 9021.

Levine, P. 2003. Parental involvement laws and fertility behavior. Journal of Health Economics 22:861-78.

Levine, P., D. Staiger, T. J. Kane, and D. J. Zimmerman. 1999. Roe v. Wade and American fertility. American Journal of Public Health 89:199-203.

Levine, P., A. Trainor, and D. J. Zimmerman. 1996. The effect of Medicaid abortion funding restrictions on abortions, pregnancies and births. Journal of Health Economics 15:555-78.

Lichter, D. T., D. K. McLaughlin, and D. Ribar. 1998. State abortion policy, geographic access to abortion providers and changing family formation. Family Planning Perspectives 30:281-7.

Matthews, S., D. Ribar, and M. Wilhelm. 1997. The effects of economic conditions and access to reproductive health services on state abortion rates and birthrates. Family Planning Perspectives 29:52-60.

Nersesian, W. S., M. R. Petit, R. Shaper, D. Lemieux, and E. Naor. 1985. Childhood death and poverty: A study of all childhood deaths in Maine, 1976 to 1980. Pediatrics 75:41-50.

Oshfeldt, R., and S. Gohmann. 1994. Do parental involvement laws reduce adolescent abortion rates? Contemporary Economic Policy 12:65-76.

Overpeck, M. D., R. A. Brenner, A. C. Trumble, L. B. Trifiletti, and H. W. Berendes. Risk factors for infant homicide in the United States. New England Journal of Medicine 339:1211-6.

Reid, J., P. Macchetto, and S. Foster. 1999. No safe haven: Children of substance-abusing parents. National Center on Addiction and Substance Abuse at Columbia University (CASA).

Resnick, M. D., P. S. Bearman, and R. W. Blum, et al. 1997. Protecting adolescents from harm: Findings from the National Longitudinal Study on Adolescent Health. Journal of the American Medical Association 278:823-32.

Ruhm, C. J. 2000. Are recessions good for your health? Quarterly Journal of Economics 115(2):617-50.

Schnitzer, P. G., and B. G. Egelman. 2005. Child deaths resulting from inflicted injuries: Household risk factors and perpetrator characteristics. Pediatrics 116:687-93.

Sen, B. 2003a. A preliminary investigation of the effects of restrictions on Medicaid funding for abortions on female STD rates. Health Economics 12:453-64.

Sen, B. 2003b. An indirect test for whether restricting Medicaid funding for abortion increases pregnancy-avoidance behavior. Economics Letters 81:155-63.

Sen, B. 2006. Frequency of sexual activity among unmarried adolescent girls: Do state policies pertaining to abortion access matter? Eastern Economic Journal. In press.

Sollom, T. 1995. State actions on reproductive health issues in 1994. Family Planning Perspectives 27:83-7.

Sollom, T. 1997. State actions on reproductive health issues in 1996. Family Planning Perspectives 29:35-40. \ Stiffman, M. N., P. G. Schnitzer, P. Adam, R. L. Kruse, and B. G. Egelman. 2002. Household composition and risk of fatal child maltreatment. Pediatrics 109:615-21.

Web-based Injury Statistics Query and Reporting System, Centers for Disease Control and Prevention, Accessed 30 September 2006. Available http://www.cdc.gov/ncipc/wisqars'.

Zuravin, S. J. 1987. Unplanned pregnancies, family planning problems, and child maltreatment. Family Relations 36:135-9.

Zuravin, S. J. 1988. Fertility patterns: Their relation to child physical abuse and chi

(1) A special case is North Carolina, which financed abortion among poor women from a separate state fund, and ceased such financing in the years when that fund was prematurely exhausted. Cook et al. (1999) exploit this variation to examine the effects of abortion funding on abortion and birth rates. In 1995, that fund was reduced to $50,000 per year, and restricted to financing abortions due to rape and incest only.

(2) South Carolina imposed a mandatory waiting period of the fairly trivial period of one hour in 1995, and later revoked it.

(3) See, for example, http://www.aclu.org/ReproductiveRights/ReproductiveRights.cfm?ID=9045&c=143. (4) In 1992, the U.S. Supreme Court upheld the general constitutionality of parental-involvement laws, which suggests that they will be an important abortion policy in the foreseeable future.

(5) Certain states waive the involvement requirement in cases of proven parental abuse or neglect, medical emergencies, and pregnancies resulting from rape or incest. All states except Utah also provide for a judicial bypass mechanism.

(6) Briefly, these authors posit a model where, in a world of easily accessed abortion services, some women would get pregnant and then make the decision to abort versus carry-to-term based on further information (for example, whether the impregnator is willing to marry her). Depending on the new information, some of these pregnancies would result in live births and some would be terminated. In a world where abortion services are difficult to access, these women would avoid pregnancy altogether. Thus, abortion restrictions would not only prevent some pregnancies that would have otherwise been terminated, but they would also prevent some pregnancies that would otherwise have resulted in live births.

(7) Combining age groups leads to further scope for obfuscation since among older children, particularly teens, peer violence and own behaviors are major contributors to fatal injuries (Resnick et al. 1997).

(8) Statistical tests conducted also rejected pooling of the black and white samples in this study.

(9) On the other hand, child passenger deaths may occur if the child is not in a safety seat or not otherwise appropriately secured in the car. They may also occur due to alcohol or drug use by the driver of the car. Insofar as these factors are influenced by parental concern for the child's well-being, parental age and parental SES status, there may be a correlation between abortion restrictions and child fatal injuries in motor vehicle crashes.

(10) It should be noted that parental consent laws and parental notification laws are mutually exclusive. The above numbers can be used to calculate the percentage of cases where either one of parental consent or notification laws were enforced.

(11) This is consistent with earlier findings by Ruhm (2000) on the relationship between unemployment, total mortality rates, and infant and neonatal mortality rates.

(12) I also repeated the analysis using a conventional OLS model where the dependent variable is the natural log of the fatal-injury rate (with a rate of '0' replaced by '0.00001' for purposes of converting to log). The results remained qualitatively similar, but the magnitudes of the effects were often improbably large. For example, presence of parental consent laws appeared to increase white homicide resultant injury rates by 55%. This is probably indicative of the poor fit of the OLS model, given the distribution of the fatal injury counts and the non-negligible number of state-year cells with 0 child fatal injuries for at least one race.

(13) It should be noted, however, that fatal injuries are the most extreme manifestation of abuse, neglect, or an otherwise unsafe environment. There may also be negative outcomes for children in the form of non-fatal injuries or mental trauma that are associated with abortion restrictions.

Bisakha Sen, Department of Healthcare Organization and Policy, University of Alabama at Birmingham, RPHB 330, 1665 University Boulevard, Birmingham, AL 35294, USA; E-mail bsen@uab.edu.
Table 1. Pooled Sample Means for Fatal-Injury Counts,
Fatal-Injury Rates, Abortion Restrictions, and Other
Variables Used in Empirical Analysis

 Standard
Variable Mean Deviation

Homicide resultant fatal injuries,
 white children, 0-11 years 7.53 10.14
Homicide resultant fatal injuries,
 black children, 0-4 years 5.50 7.13
Unintentional fatal injuries,
 white children, 0-4 years 46.72 56.94
Unintentional fatal injuries,
 black children, 0-4 years 15.99 19.58
Non-motor unintentional fatal injuries,
 white children, 0-4 years 39.47 48.29
Non-motor unintentional fatal injuries,
 black children, 0-4 years 14.17 17.46
State population, white children, 0-4 years 291,733.0 339,540.0
State population, black children, 0-4 years 57,864.00 68,833.38
Homicide resultant fatal-injury rate,
 per 100,000 white children, 0-4 years 2.55 1.88
Homicide resultant fatal-injury rate,
 per 100,000 black children, 0-4 years 9.67 26.06
Unintentional fatal-injury rate,
 per 100,000 white children, 0-4 years 16.54 7.16
Unintentional fatal-injury rate,
 per 100,000 black children, 0-4 years 26.20 35.82
Unintentional non-motor fatal-injury rate,
 per 100,000 white children, 0-4 years 13.87 6.10
Unintentional non-motor fatal-injury rate,
 per 100,000 black children, 0-4 years 23.00 33.64
Enforced parental consent laws 0.167 0.35
Enforced parental notification laws 0.142 0.31
No public funding 0.648 0.44
Mandatory delays 0.068 0.23
Border states with no parental involvement
 laws 2.96 1.67
Abortion provider rates per 100,000
 women aged 10-50 years 3.54 2.85
Percent of population in rural areas 30.21 15.10
State poverty rate 13.23 4.15
Maximum monthly AFDC payments 304.37 131.74
Family cap law in place 0.13 0.34
Apparent per capita alcohol use 2.46 0.65
State unemployment rate 6.10 2.27
Police per 100,000 population in state 27.45 8.65
Homicide resultant fatal-injury rate,
 per 100,000 white adults, 25-65 years 5.48 3.06
Homicide resultant fatal-injury rate,
 per 100,000 black adults, 25-65 years 33.38 10.18
Unintentional fatal-injury rate,
 per 100,000 white adults, 25-65 years 31.56 21.79
Unintentional fatal-injury rate,
 per 100,000 black adults, 25-65 years 39.60 23.00

N = 1122. Means are based on pooled sample of all states
and the District of Columbia for 1981-2002.

Table 2. Estimates for Relationships between State Abortion
Restrictions and Child Fatal Injury from Homicide and from
Unintentional Causes

 0- to 4-Year-Old White Children

 Homicide Non-Motor
 Resultant Unintentional Unintentional
 Fatal Injury Fatal Injury Fatal Injury

Parental
 consent 0.20 * (1.89) 0.03 (0.97) 0.04 (0.88)
Parental
 notification 0.10 (1.17) 0.03 (1.27) 0.05 (1.54)
No public
 funding 0.03 (0.41) 0.02 (0.71) 0.06 * (1.69)
Mandatory
 delay 0.23 * (1.70) 0.09 *** (2.60) 0.07 * (1.72)
Border states
 with no PI
 laws -0.06 *** (-2.93) -0.02 ** (-2.47) -0.03 ** (-2.32)
Abortion
 provider rate -0.03 (-0.91) -0.003 (-0.24) -0.008 (-0.51)

 0- to 4-Year-Old Black Children

 Homicide Non-Motor
 Resultant Unintentional Unintentional
 Fatal Injury Fatal Injury Fatal Injury

Parental
 consent 0.07 (0.95) -0.09 (-1.17) -0.11 (-1.36)
Parental
 notification 0.10 (0.75) -0.09 * (-1.64) -0.11 * (-1.83)
No public
 funding 0.10 (1.22) 0.16 ** (2.51) 0.19 *** (2.73)
Mandatory
 delay 0.30 *** (2.58) 0.12 (1.11) 0.13 (1.05)
Border states
 with no PI
 laws -0.04 (-1.62) 0.01 (0.44) -0.003 (0.15)
Abortion
 provider rate 0.04 (1.20) -0.002 (-0.05) -0.006 (-0.16)

Negative Binomial Models are used. Estimated coefficients are
presented as Exp([beta])-1 with t-statistics in parentheses. Standard
errors and t-statistics adjusted for clustering upon state. All models
include controls for % of state population in poverty, state
unemployment rate, % of state population living in rural areas,
apparent per capita consumption of alcohol in the state, the maximum
AFDC benefit level, "family cap" laws, police officers per 100,000
state population, the log of the population of children of that race
aged 0-4 years with its coefficient constrained to be 1, and binary
vectors for state and year fixed effects. Standard errors and
t-statistics adjusted for clustering upon state. PI, parental
involvement.

* Significant at <0.10.

** Significant at <0.05.

*** Significant at <0.01.

Table 3. Estimates for Relationship between State Abortion
Restrictions and Child Fatal Injury from Homicide and from
Unintentional Causes, with Controls for Adult Fatal Injury Rates

 0- to 4-Year-Old White Children

 Homicide
 Resultant Unintentional
 Fatal Injury Fatal Injury

Parental consent 0.18 * (1.77) 0.03 (1.06)
Parental notification 0.11 (1.20) 0.03 (0.98)
No public funding 0.01 (0.13) 0.02 (0.79)
Mandatory delay 0.19 (1.35) 0.09 ** (2.06)
Border states with no PI
 laws -0.05 ** (-2.06) -0.02 ** (-2.35)
Abortion provider rate -0.03 (-0.77) -0.004 (-0.33)
Adult (25-65 years) rate 0.02 * (1.88) 0.004 * (1.88)

 0- to 4-Year-Old 0- to 4-Year-Old
 White Children Black Children

 Non-Motor Homicide
 Unintentional Resultant
 Fatal Injury Fatal Injury

Parental consent 0.04 (0.92) 0.07 (0.93)
Parental notification 0.04 (1.38) 0.09 (0.73)
No public funding 0.06 * (1.71) 0.11 (1.29)
Mandatory delay 0.07 * (1.71) 0.28 ** (2.30)
Border states with no PI
 laws -0.02 ** (-2.29) -0.04 (-1.43)
Abortion provider rate -0.007 (-0.46) 0.03 (1.09)
Adult (25-65 years) rate 0.002 (1.06) 0.003 (1.46)

 0- to 4-Year-Old Black Children

 Non-Motor
 Unintentional Unintentional
 Fatal Injury Fatal Injury

Parental consent -0.06 (-0.70) -0.08 (-0.94)
Parental notification -0.07 (-1.20) -0.09 (-1.43)
No public funding 0.15 ** (2.40) 0.17 *** (2.64)
Mandatory delay 0.08 (0.74) 0.09 (0.74)
Border states with no PI
 laws 0.02 (0.73) 0.001 (0.05)
Abortion provider rate -0.01 (-0.40) -0.02 (-0.46)
Adult (25-65 years) rate 0.006 *** (3.66) 0.006 *** (3.19)

Negative Binomial Models state are used. Estimated coefficients
are presented as Exp([beta])-1; t-statistics in parentheses. Standard
errors and t-statistics adjusted for clustering upon state. Apart
from adult fatal-injury rates by race from homicides and unintentional
causes, all models also include controls for percentage of state
population in poverty, state unemployment rate, percentage of state
population living in rural areas, apparent per capita consumption of
alcohol in the state, the maximum AFDC benefit level, "family cap"
laws, police officers per 100,000 state population, the log of the
population of children of that race aged 0-4 years with its
coefficient constrained to be 1, and binary vectors for state and
year fixed effects. Standard errors and t-statistics adjusted for
clustering upon state. PI, parental involvement.

* Significant at <0.10.

** Significant at <0.05.

*** Significant at <0.01.

Table 4. Estimates from Alternate Model Specifications

 0- to 4-Year-Old White Children

 Homicide
 Resultant Unintentional

Model 1
 Parental consent 0.19 ** (1.97) 0.06 (1.40)
 Parental notification 0.11 (1.20) 0.04 (1.26)
 Mandatory delay 0.19 (1.36) 0.05 * (1.78)

Model 2
 No public funding 0.06 (0.81) 0.05 ** (2.29)
 Mandatory delay 0.24 * (1.81) 0.07 * (1.75)

Model 3
 Parental consent 0.23 ** (2.29) 0.03 (0.87)
 Parental notification 0.12 (1.42) 0.03 (0.94)
 No public funding 0.01 (0.19) 0.05 * (1.81)

Model 4
 Abortion
 restrictiveness index 0.11 *** (3.16) 0.03 ** (2.01)

 0- to 4-Year-Old 0- to 4-Year-Old
 White Children Black Children

 Unintentional Homicide
 Non-Motor Resultant

Model 1
 Parental consent 0.03 (1.25) 0.13 * (1.85)
 Parental notification -0.03 (-1.03) 0.11 (0.87)
 Mandatory delay 0.07 (1.62) 0.27 ** (2.20)

Model 2
 No public funding 0.09 *** (3.00) 0.14 * (1.93)
 Mandatory delay 0.05 (1.55) 0.29 ** (2.32)

Model 3
 Parental consent 0.03 (0.87) 0.14 (1.62)
 Parental notification 0.04 (1.25) 0.09 (0.70)
 No public funding 0.09 ** (2.50) 0.10 (1.17)

Model 4
 Abortion
 restrictiveness index 0.04 ** (2.30) 0.12 *** (3.92)

 0- to 4-Year-Old Black Children

 Unintentional
 Unintentional Non-Motor

Model 1
 Parental consent -0.02 (-0.29) 0.04 (0.46)
 Parental notification -0.07 (-1.21) -0.08 (-1.40)
 Mandatory delay 0.11 (1.05) 0.13 (0.99)

Model 2
 No public funding 0.12 ** (2.21) 0.13 ** (2.48)
 Mandatory delay 0.10 (0.88) 0.11 (1.05)

Model 3
 Parental consent -0.07 (-0.94) -0.10 (-1.15)
 Parental notification -0.11 * (-1.64) -0.13 * (-1.79)
 No public funding 0.17 *** (2.62) 0.19 *** (2.83)

Model 4
 Abortion
 restrictiveness index 0.03 (0.87) 0.03 (1.00)

Negative binomial models are used. Estimated coefficients
are presented as Exp([beta])-1; t-statistics in parentheses.
Standard errors and t-statistics adjusted for clustering upon state.
Model 1 excludes no public funding, Model 2 excludes the parental
involvement laws, Model 3 excludes mandatory delays, Model 4 replaces
individual restrictions with a single "index of restrictiveness."
All models also control for border states with no parental involvement
laws, abortion provider rates, the other state-level controls, adult
fatal-injury rates, log of the population of children of that race
aged 0-4 years with coefficient constrained to be 1, and binary
vectors for state and year fixed effects. Standard errors and
t-statistics adjusted for clustering upon state.

* Significant at <0.10.

** Significant at <0.05.

*** Significant at <0.01.
COPYRIGHT 2007 Southern Economic Association
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2007, Gale Group. All rights reserved. Gale Group is a Thomson Corporation Company.

Article Details
Printer friendly Cite/link Email Feedback
Comment:State abortion restrictions and child fatal-injury: an exploratory study.
Author:Sen, Bisakha
Publication:Southern Economic Journal
Article Type:Report
Geographic Code:1USA
Date:Jan 1, 2007
Words:12000
Previous Article:Introduction.
Next Article:The effect of reimbursement on the intensity of hospital services.
Topics:


Related Articles
Psychological aftereffects of abortion: the rest of the story.
Abortion law reform ruled out for now.
A woman's right is in peril.
Reasons U.S. women have abortions: quantitative and qualitative perspectives.
The economics of abortion access in the US: restrictions on government funding for abortion is the post-Roe battleground.
The abortion-breast cancer link: a medical-legal nightmare on the horizon.
Will Canada escape the new abortion debate?
Time for a new abortion law.
Abortion and Human Rights: legal developments in Europe may open the door to wider availability of abortion in Ireland.

Terms of use | Privacy policy | Copyright © 2020 Farlex, Inc. | Feedback | For webmasters