The supply of infants relinquished for adoption: did access to abortion make a difference?
One largely undocumented fertility outcome is the option to relinquish an infant for adoption. The dramatic drop in the supply of infants relinquished during the 1980s is puzzling in that it parallels a trend toward greater state regulation of abortion, increases in nonmarital birth rates, and high demand for healthy, unrelated infants,(1) Motivated by a theory of desired fertility, this study uses aggregated state level data from three years to examine the impact of abortion access on the supply of infants relinquished. The highly controversial status of adoption and abortion in the United States today makes this a particularly compelling issue, especially given the move toward further restrictions on abortion at both the state and federal levels.(2)
The premise of the model is that a woman with an unintended pregnancy faces three options: to abort the pregnancy, to relinquish the infant for adoption, or to keep the infant. Although the 1973 Supreme Court decision Roe v. Wade guaranteed a woman's right to terminate a pregnancy, the Court also specified that the ruling must be considered against state interests. The resulting interstate variation in abortion availability provides a natural means of estimating the effect of abortion regulation on the supply of infants relinquished for adoption. Economic theory of desired fertility hypothesizes that if the relative cost of keeping an infant is very high, then, all else equal, increasing the relative cost of abortion should increase the supply of infants relinquished.
The empirical equation is motivated by this individual choice model. I estimate a model with the ratio of infants relinquished to total pregnancies and the ratio of infants relinquished to nonmarital pregnancies as a function of controls for observed heterogeneity across states over time and abortion access variables including parental involvement laws and restrictions on public funding. Subsequent models attempt to account for unobserved heterogeneity in two ways. First, I add an indicator in the empirical model for states in which restrictive abortion laws are enjoined or overturned. Laws that are enacted but not enforced should only have an indirect effect on the supply of infants relinquished through anti-abortion sentiment. Second, I add variables hypothesized to be highly correlated with unobserved state characteristics. I also extend the empirical analysis to (1) examine the potential endogeneity of abortion access by using a two-stage least squares (2SLS) procedure and (2) model the multinomial choice of pregnancy resolution: aborting the pregnancy, relinquishing the infant, or keeping the infant.
This study attempts to fill the gap in previous empirical work on pregnancy resolution by taking advantage of available state-level data on the number of infants relinquished for adoption to examine the effect of abortion access over time. The current noneconomic literature examining a woman's choice to relinquish an infant for adoption largely focuses on a selective sample, for example, adolescents residing in or affiliated with a maternity home, and most of the economics literature on unintended pregnancy has ignored the option to relinquish an infant, because of data availability. The one previous attempt at estimating the supply of infants relinquished, by Medoff , only includes a cross-section of data in 1980. This analysis is not able to capture the effect of variations in abortion availability and abortion law throughout the 1980s and does not consider the potential endogeneity of some of the variables.
In contrast to Medoff , I show evidence that abortion access affected the supply of infants relinquished during the 1980s in two different ways. First, consistent with a theory of desired fertility, the availability of abortion providers has the expected effect of reducing the supply of infants relinquished, particularly relative to the demand for abortion. Second, abortion law has an unexpected negative effect. This suggests that as abortion laws become more restrictive the total number of unwanted births may decrease.
My results further suggest that omitted variable bias confounds the effect of laws that restrict public funding of abortions, which is consistent with empirical findings by Currie, Nixon, and Cole , and that abortion access is not endogenous. Of the explanatory variables, education plays an important role. A greater percentage of women with higher opportunity costs (e.g., women expecting to complete or having completed college) increases abortion demand relative to the supply of infants relinquished. To the extent that the increase in occurrences of nonmarital births and single parenthood during the 1980s can be explained by variation in Aid to Families with Dependent Children (AFDC) payments, I do not find evidence that the increased occurrence of single parenthood affected the supply of infants relinquished.
The next sections review previous empirical work and present a model of desired fertility that motivates the estimating equation. This is followed by the estimation strategy and a description of the data. The paper concludes with a discussion of the results and policy implications.
II. PREGNANCY RESOLUTION AND PREVIOUS EMPIRICAL WORK
Beside keeping the infant (and getting married), the most frequent legal method of resolving an unintended pregnancy during the 1950s and 1960s was to relinquish the infant for adoption. Previous to the decision of Roe v. Wade in 1973, the direct and indirect costs of an illegal abortion generally were prohibitively high and the stigma associated with premarital motherhood was severe. Indeed, relative to the 1970s, the early 1980s showed an increase in abortion rates, nonmarital birth rates, and rates of single motherhood. By 1987, over half of unintended - mistimed and unwanted - pregnancies to all women, and 75% to never married women, ended in abortion, according to Brown and Eisenberg . Over roughly the same time period, the ratio of infant relinquishment to nonmarital births fell. based on national survey data, the percentage of infants relinquished from premarital births fell from 8.7% before 1973 to 2.0% between 1982 and 1988. By race, the change was most dramatic for whites: 19% of infants from premarital births were relinquished before 1973 compared with 8% in 1972 through 1981 and 3% in 1982 through 1988. In comparison, for blacks the percent of infants from premarital births that were relinquished remained at approximately 1% from before 1973 to 1988. (For statistics, see Bachrach, Stolley, and London .)
Most of the economics literature examining pregnancy resolution focuses on teenage mothers and the impact of public assistance on their decision to keep an infant.(3) Lundberg and Plotnick [1990, 1995] consider the impact of state contraceptive and abortion policy, in addition to public assistance, in a sequential decision model, which includes the option to abort. They find that young white women are more likely to abort than carry the pregnancy to term if they reside in states that offer abortion funding assistance or in states with greater availability of abortion. Because of sample size restrictions, however, this literature ignores the option of placing an infant for adoption.
Other empirical studies utilize variations in both abortion access and the abortion rate by state and over time to answer the question of whether states with the lowest abortion utilization rates are also the states that enforce abortion restrictions.(4) For the most part, these studies suggest that Medicaid funding restrictions and parental involvement laws have a negative impact on demand for abortion, even controlling for unobserved state heterogeneity.
Kane and Staiger  and Levine, Trainor, and Zimmerman  extend the analysis of examining the effect of abortion law to include other aspects of fertility besides abortion. If pregnancy rates remain constant, then as access to abortion becomes more restrictive birth rates should increase. This empirical work finds that recent restrictions on abortion access are associated with declines or no effect on total birth rates. The authors argue that this result is consistent if abortion restrictions alter fertility behavior such that the total number of pregnancies are reduced.
Only Medoff  explores an analogous question: are states with the highest infant relinquishment rates also the states with the greatest number of enforced abortion restrictions or the states with the least access to abortion? If abortion were the primary cause of the decrease in the supply of infants relinquished during the 1980s, then states with relatively easier access to abortion should have lower rates of infant relinquishment. Medoff  addresses this issue by estimating the supply of infants relinquished using a single cross-section of data from 1980. He finds that the coefficients on public funding of abortion and abortion price are not statistically significant and concludes that abortion and relinquishing an infant for adoption are not substitutes. Since Medoff uses one year of data from 1980, his analysis cannot take advantage of the variations in abortion access over time.(5) Most anti- and proabortion voter initiatives began in the 1980s. For example, the majority of parental consent and notification laws were instituted during the 1980s.
III. THEORY OF DESIRED FERTILITY AND FAMILY SIZE
In their analysis of pregnancy resolution and birth weight production functions, Grossman and Joyce  outline a model that incorporates the decision of a pregnant women into a model of fertility control. A woman maximizes her lifetime utility, which implies an optimal child production program. This lifetime program yields an optimal stock of children in time t, [Mathematical Expression Omitted]. If [C.sub.it] represents the actual stock of children a woman will have in time t, she will become pregnant and have a birth provided [Mathematical Expression Omitted]. If a woman becomes pregnant and [[Pi].sub.it] [less than or equal to] 0, then a discrepancy has occurred between the desired and actual number of children and the pregnancy is unintended. A woman with an unintended pregnancy faces three options: to abort the pregnancy, carry the pregnancy to term and relinquish the infant, or carry the pregnancy to term and keep the infant.
Assuming each of these pregnancy resolution options are mutually exclusive, the optimal outcome will be determined via pairwise utility comparisons; a woman will choose to abort if [U.sub.abort] [greater than] [U.sub.relinquish] and [U.sub.abort] [greater than] [U.sub.keep]. These utility comparisons imply the three probabilities of pregnancy resolution conditional on the probability of having an unintended pregnancy for the ith woman at time t, which can be estimated by a two-equation system. The first equation indicates whether a woman has had an unintended pregnancy, and the second equation indicates her choice to either abort the pregnancy, relinquish the infant for adoption, or keep the infant. The joint distribution of the error terms generates the probabilities of the events that form the basis of an estimable likelihood function. Unfortunately, in large, nationally representative data sets the incidence of relinquishing an infant for adoption is rare.(6) The small sample size prohibits estimation of a model using individual level data. This theory does, however, generate an empirical estimation of the supply of infants relinquished using aggregate data. At an aggregate level (that is, by county or state) the individual decisions may be reflected in a dependent variable where the denominator represents the first-stage fertility decision and the numerator represents the second-stage pregnancy resolution decision.
The explanatory variables in this system of equations include proxies for the determinants of the optimal number of children and the optimal spacing of births, such as the costs of contraception and the direct and indirect costs of abortion, namely, abortion access. The question of interest is "has the relatively higher cost of abortion during the 1980s altered the outcome of an unintended pregnancy?" Or, as implemented here, have abortion laws or direct changes in abortion access affected the outcome probabilities? More specifically, if the higher direct and indirect costs of abortion have increased the probability of relinquishing the infant then the coefficient on abortion law should be positive and the coefficient on abortion access should be negative. While the predictions for the impacts on the probability of aborting are straightforward (negative on abortion law and positive on abortion access), the impacts on the probability of keeping an infant are ambiguous.
While this theory offers clear predictions, the effect of abortion laws and measures of access on the probability of relinquishing an infant for adoption may be an empirical question. As found in recent empirical work, abortion law may decrease the actual number of children by affecting fertility behavior such that the probability of an unintended pregnancy also decreases. In this case, the coefficient on abortion law may be negative (as pregnancies decrease, total unwanted births may decrease as may the potential pool of infants to be relinquished) and the coefficient on measures of abortion access may be positive.
IV. EMPIRICAL ESTIMATION AND DATA
Weighted Least Squares (WLS) Model
If [n.sub.i] is the total number of unintended pregnancies in state i and [m.sub.i] is the number of infants relinquished in state i, then the empirical probability is [Mathematical Expression Omitted]. The minimum chi-squared method as summarized by Maddala  estimates this linear probability function with weights determined by the variance of the error term. The empirical estimation for the supply of infants relinquished in state i and year t is:
(1) [Mathematical Expression Omitted]
for i = 1, ..., 51 and t = 1982, 1986, and 1989.
Equation (1) is estimated in log-linear form using the minimum chi-squared method (weighted least squares) with pooled time-series cross-section data for three years (1982, 1986, and 1989) and 50 states plus Washington, D.C. Data on the number of infants relinquished in each state are available from the National Council for Adoption (NCFA). Although the federal government collects some adoption data, the NCFA publishes the most current information available on adoption. The NCFA state-level data include details on the type of adoption, limited characteristics of the child, and adoption regulations based on surveys sent to state departments of public welfare, vital statistics offices, state courts, and adoption agencies.
The supply of infants relinquished is modeled in a number of ways to extract as much information as possible from aggregate data and to examine the robustness of the results. The primary dependent variables follow from the structural model. As previously discussed the individual fertility decisions may be reflected at an aggregate level with the following denominators: (1) the sum of births and abortions and (2) the sum of births to and abortions by single women. Note that these two measures of the number of pregnancies excludes fetal loss due to miscarriage, since these data are not available at an aggregate level. The variable nonmarital pregnancies is pregnancies to divorced, never married, separated, or widowed women and attempts to represent the population most likely to have an unintended pregnancy and the population for which unintended pregnancy is the most costly. In Brown and Eisenberg  calculations using 1987 data show that the percentage of unintended pregnancies to never married women (88.2%) and formerly married women (68.5%) is much higher than the percentage of unintended pregnancies to currently married women (40.1%). One cost of using the denominator nonmarital pregnancies is that it confounds the effects of endogenous changes in marital status. The first measure, total pregnancies, is included to examine the effects of this endogeneity.
I repeat the empirical analysis using alternative measures of the dependent variable to replicate analyses by Medoff . Infants relinquished per total or nonmarital births excludes the option to abort the pregnancy and considers how timing may change an individual's options to resolve a pregnancy. More specifically, the decision to relinquish an infant for adoption may occur after the safe period for an abortion. In addition, I consider infants relinquished per 1000 single females as a measure of the actual supply of infants relinquished relative to the potential pool of suppliers. The results for these alternative specifications of the dependent variable are noted in the text and are available upon request from the author.
In equation (1), [X.sub.it] is a vector of state-level variables: percentage of women with high school degrees, percentage of women with college degrees, the female unemployment rate, the real average hourly wage, real maximum AFDC benefits, the incarceration rate, the percentage of black females, and the percentage of individuals who attend church. [Year.sub.t] are year-specific fixed effects for 1986 and 1989, and ??? is a random error term. The state-level variables (education, unemployment, and real hourly wage) are intended to capture the opportunity costs of fertility decisions by measuring the value of a female's time. Because the real wage is not differentiated by gender, it also may serve as a proxy for the status of the pool of marriageable men. Similar to Clarke and Strauss , I also include the incarceration rate as a proxy for the pool of marriageable men. An income transfer, such as an AFDC payment, may relax a woman's budget constraint and consequently increase the likelihood of keeping the infant. If it is the case that the opportunity costs of fertility decisions are adequately captured via the education variables, then the real wage variable may capture an income effect of fertility decisions. This will occur only to the extent that the real wage reflects the male real wage as a proxy for female unearned income. The percentage of black females and the percentage attending church are proxies for tastes and preferences for pregnancy resolution. Case studies of women who relinquish their infants emphasize the importance of race as an explanatory factor.(7)
Parental Involvement Law and Restrictive Public Funding measure the impact of abortion laws, each taking a value of I if the restriction is enforced. Enforced parental involvement laws require either parental consent, parental notification, or a judicial bypass for females under the age of 18 to obtain an abortion. Restrictive Public Funding refers explicitly to the passage of the Hyde Amendment, in 1977 by the U.S. Congress, which prohibits federal expenditures for abortion via the Medicaid program except in cases where a woman's life is in danger. Consequently, while some states continue to provide public funding for abortions voluntarily or by court order, other states restrict funding according to the federal guidelines. Over 95% of public funding for abortions comes from state funds, according to Gold and Daley , Gold and Macias , Gold and Nestor . Abortion Access is the percentage of counties with providers who annually performed five or more abortions and acts as a proxy for the cost of abortion. Parental Involvement Law and Restrictive Public Funding in Border states controls for the effect of living close to states with restrictive access to abortion.
When possible, the dependent and explanatory variables are constrained to measure women ages 15-44 or, when applicable, single women ages 15-44. Appendix Table AI provides more detail regarding the variable definitions and sources, and Appendix Table All provides descriptive statistics. The data on infants relinquished and abortions best reflect occurrences by state of residence. Although NCFA does not separate adoptions by state of residence, interstate adoption is strictly regulated. NCFA  notes that states often require that at least one adoptive parent resides in the state for a minimum of one year or that the adoption takes place through a licensed agency. Compared to data by state of occurrence, which examines the effect of policy on in-state infant relinquishment rates, state-of-residence data provide a better measure of the effect of abortion policy on a woman's pregnancy resolution decision.(8)
The empirical equation attempts to control for observed state heterogeneity. But, as originally noted by Ellwood and Bane , omitted variables or unobserved taste factors that reflect the social and political structure of a state, such as antiabortion sentiment or stigma effects, may influence both the restrictiveness of abortion and the supply of infants relinquished for adoption. The common technique for controlling for these omitted variables is fixed effects estimation. Unfortunately, with only 51 observations per year, three years of data were not sufficient to estimate the fixed effects specification.(9) Given this limitation, I attempt to account for unobserved heterogeneity in two ways. These approaches are designed to capture anti- or proabortion sentiment.
First, I add variables such as the overall number of abortion regulations enacted (excluding parental involvement law and public funding of abortion) between 1973 through 1989, as compiled by Halva-Neubauer , and the percentage of the population that lives in a metropolitan statistical area (MSA). The number of abortion regulations enacted reflect a state's relative tolerance for the liberalization of abortion. Between 1973 and 1989 state legislatures reacted very differently toward the liberalization of abortion. For example, as documented by Halva-Neubauer , Kansas and New Hampshire each enacted one abortion restriction, while Illinois and Utah each enacted over 10.
Second, the empirical model includes indicators for states in which abortion regulations are enforced and indicators for states in which abortion regulations are enjoined or overturned, as implemented by Haas-Wilson  and Currie, Nixon, and Cole .(10) The courts may enjoin or overturn restrictive state abortion legislation only if it is deemed unconstitutional by federal law. If the decisions of state legislatures reflect the wishes of voters in the state, then the preferences of voters is likely similar in states that enforce abortion laws with states in which these laws are enjoined. Therefore, the indicator for an enjoined law may capture the cost of abortion via state-level unobserved characteristics, such as antiabortion sentiment. In contrast, the indicator for an enforced law may capture the cost of abortion via both unobserved characteristics, such as antiabortion sentiment, and observed characteristics, such as higher transaction costs associated with enforcement of the law. A comparison of the indicator for states in which restrictive laws are passed and enforced with the indicator for states in which restrictive laws are passed and enjoined may indicate any potential omitted variable bias on the parameter for abortion law.(11) For example, if the variable Restrictive Public Funding is truly picking up the effect of enforcement of the law then it should be significantly different from Restrictive Public Funding Enjoined. Using this technique, Currie, Nixon, and Cole  find evidence that the effect of restrictive public funding of abortions on child birth weight reflects omitted characteristics of the state rather than a true effect of the law. In contrast, Haas-Wilson  finds that the effect of parental involvement law on abortion demand is not confounded by omitted state-level characteristics.
Another potential problem with the empirical specification that is considered in the empirical work of Blank, George, and London  and Currie, Nixon, and Cole  is the endogeneity of the Abortion Access variable. The availability of abortion providers is partly determined by the demand for abortion, which is hypothesized to affect the supply of infants relinquished for adoption. To ensure that the coefficient on Abortion Access is not biased, a first-stage equation is estimated using the number of obstetricians and gynecologists and the total number of hospitals as instruments. These instruments are related to the overall demand for medical services in a state but are likely independent from abortion demand. The estimated result from the first stage, the predicted value of Abortion Access, is then used in place of the actual variable in estimating equation (1) and standard errors are adjusted accordingly.
Multinomial Choice Framework: Log-Odds Model
As mentioned in the introduction, most unintended pregnancies end in abortion. As a percentage of pregnancies terminated, abortion is highest among unmarried women, women aged 40 and older, teenagers, and nonwhite women. The annual abortion rate per 1000 nonwhite women between 1981 and 1988 remained stable at approximately 56 versus approximately 23 for white women. A growing number of unintended pregnancies also are resolved through single parenthood. According to the U.S. Bureau of the Census , between 1980 and 1990, the incidence of births to single women increased from 18.4% in 1980 to 29.2% of all births in 1991.
The following log-odds ratio model fully explores the relative relationships between the demand for abortion, the supply of infants relinquished for adoption and the proportion of infants kept. The probability of aborting, relinquishing the infant, or keeping the infant, conditional on the probability of woman i having an unintended pregnancy in period t is depicted by:
(2) [Mathematical Expression Omitted]
for j [not equal to] k; 1 = 1, ..., 51; and t = 1982, 1986, and 1989; where j = aborting, relinquishing the infant for adoption, or keeping the infant and [X.sub.it] is the vector of state-level observed characteristics outlined previously. The model is analogous to multinomial logit models estimated with individual data as discussed by Greene .
WLS Regression of Supply of Infants Relinquished
The first column of Table I for each dependent variable presents the results from estimation of equation (1), the second column extends the model in column 1 to include an additional abortion regulation variable and the Percentage Residing in MSAs, and the third column includes the abortion law in border states variable.(13) Overall, it is interesting to note that the model with nonmarital pregnancies in the denominator has better predictive power.
Abortion Access is negative and statistically significant at the 1% level in every specification, as well as those specifications using total births, nonmarital births, and single females in the denominator of the dependent variable.(14) A 10% increase in the percentage of counties with abortion providers will decrease the ratio of infants relinquished to total pregnancies by 0.5 points (a 9% decrease from the mean) and the ratio of infants relinquished to nonmarital pregnancies by 2.2 points (a 13% decrease from the mean). Controlling for the total number of abortion restrictions enacted and percentage of the population living in MSAs, states with greater access to abortion providers (a lower price of abortion) also have lower ratios of infant relinquishment. Also, abortion law in border states, for example, parental involvement laws, does have an effect on the supply of infants relinquished. This is one indication that restrictive abortion laws in surrounding states decrease abortion demand.
The opposite holds for the within-state abortion law variables. Although Parental Involvement Law and Restrictive Public Funding are individually and jointly significant, the coefficients have a counterintuitive sign. This holds true for all specifications, including those with births or single women in the denominator of the dependent variable. All else equal, states with more restrictive abortion laws also have lower ratios and rates of infant relinquishment by 1 to 5 points (a 20% to 30% decrease from the mean). As mentioned previously, this result is consistent if abortion law affects fertility behavior. When birth rates fall due to abortion restrictions, the total number of unwanted births may fall, which then may decrease the potential supply of infants relinquished. Kane and Staiger  and Levine, Trainor, and Zimmerman  similarly find that abortion law may affect fertility behavior, and Kane and Staiger  explain this effect via a theoretical model in which pregnancy is endogenous. Note that the size of the abortion law coefficients are nearly 3-5 times smaller than the size of the coefficient on Abortion Access.(15)
The last column for each dependent variable of Table I presents the 2SLS results. The first stage results are presented in Appendix Table AIII.(16) With the exception of Percentage Completed Four Year College and the Incarceration Rate, the signs and significance of the explanatory variables do not change. Hausman tests for these specifications indicate that the coefficients in the weighted OLS and 2SLS are not significantly different. Thus, the effect of Abortion Access on the supply of infants relinquished for adoption is not likely a result of endogeneity bias.
The control variables have similar effects across model specifications including those with births or single women in the denominator of the dependent variable. Consistent with Medoff , the positive coefficient on high school education reflects higher opportunity costs of raising a child for women with completed high school education or that high school dropouts are more likely to grow up in disadvantaged homes (e.g., single parent, female-headed households) and therefore are less likely to be exposed to and consider adoption. Controlling for completion of high school the negative coefficient on Percentage Completed College and the Real Hourly Wage [TABULAR DATA FOR TABLE I OMITTED] may be an indication that states with a greater percentage of women who expect to complete (or have completed) college or have a higher value of time are likely to be states with lower rates of infant relinquishment.(17) Conditional on the number of pregnancies, these states may exhibit either higher abortion demand or higher own-demand for children. It appears that Percentage Completed College is not only negatively correlated with the supply of infants relinquished but possibly positively correlated with abortion demand; the absolute value of the coefficient decreases in the specifications with births in the denominator. The correlation with abortion demand for both the Percentage Completed College and the Real Hourly Wage is further discussed in the log-odds ratio results.
In contrast to Medoff's results, the coefficient on Real AFDC Maximum Benefits is not statistically different from zero.(18) While the evidence is mixed on whether or not public assistance explains any of the recent increase in nonmarital birth rates or the incidence of single parenthood, according to Moffitt , my results suggest that variation in the availability of public assistance did not significantly affect the supply of infants relinquished. The negative coefficient on Percentage Black and Percentage Residing in MSAs is consistent with descriptive evidence. As noted by Sobol and Daly , of the unmarried women who carry their pregnancies to term, women who relinquish their infants are more likely to be white and from rural areas. The incarceration rate has a positive and significant effect on the infant relinquishment rate. The smaller pool of marriageable men associated with a higher incarceration rate may have a positive impact on the infant relinquishment rate by decreasing the probability of marriage.
Recall that states with enforced and enjoined laws should have the same unobserved characteristics. If Restrictive Public Funding is truly the effect of the law and not just the effect of unobserved antiabortion sentiment, then its coefficient should be significantly different from Restrictive Public Funding Enjoined. The results in Table II provide some evidence that the abortion law variables may not be depicting the true effect of abortion law. The coefficients on Restrictive Public Funding and Restrictive Public Funding Enjoined are not significantly different from each other. In contrast the parental consent and notification enforced and enjoined law variables are jointly significant and significantly different from zero. Note, however, that the estimates of this variable may be confounded, since a parental involvement law affects only females under the age of 18.
Table II also compares the coefficients on abortion laws and abortion laws enjoined with and without controls for Abortion Access. Since it is likely that states with restrictive abortion policies are also states with few abortion providers, Abortion Access may act as one proxy for the unobserved sentiment effects of abortion law. Thus, the coefficient on abortion law should be less negative when Abortion Access is excluded. When excluding the Abortion Access variable, the coefficient on Restrictive Public Funding is less negative, as expected, but the coefficient on Parental Involvement Law changes only slightly. In the specification with the enjoined laws excluding Abortion Access, Parental Involvement Law Enjoined is positive and becomes larger. While the coefficient on Restrictive Public Funding loses significance, the enjoined variable surprisingly remains negative and gains significance.
If in states with enjoined public funding laws women have greater access to abortions, then Restrictive Public Funding Enjoined may be a proxy for Abortion Access. In fact while 18% of the counties in an average state with enforced restrictive funding have abortion providers of five or more abortions, in an average state with enjoined public funding this percentage jumps to 72%. The comparable difference is very small for parental involvement laws: 20% versus 18%, respectively. Thus, there are mixed results on the effects of abortion law on the supply of infants relinquished for adoption. First, states with abortion [TABULAR DATA FOR TABLE II OMITTED] laws may only appear to have lower incidences of infant relinquishment because abortion laws act as a proxy for characteristics of states that are associated with both the passage of specific laws and lower incidences of infant relinquishment. In this case, restrictive public funding law is a proxy for Abortion Access when Abortion Access is excluded. Or, since this result is not consistent for Parental Involvement Laws, it is more likely that abortion law may affect fertility behavior such that the total number of unwanted births decreases.
Log-Odds Ratio Regressions
Tables III and IV present the results from the log-odds ratio models. The first model, which is the particular specification of interest, uses the probability of aborting a pregnancy as the reference choice, while the latter two models use the probability of keeping the infant as the reference choice. All three models are consistent under the assumption that the probability of event j is independent of event k. A positive coefficient is associated with the increased probability of the numerator occurring (or a decreased probability of the denominator occurring); a negative coefficient is associated with the increased probability of the denominator occurring.
In column 1, relative to the demand for abortion, the supply of infants relinquished is associated with a higher percentage of women with a high school education, the Incarceration Rate, and the percentage of individuals who attend church. These results are consistent with the WLS estimates: Abortion Access, Percentage Completed College, the Real Hourly Wage, and Percentage Black are negatively associated with the supply of infants relinquished. This estimation now allows the interpretation that these significant variables are associated with the demand for abortion. The hourly wage may be interpreted as the female value of time and the higher opportunity cost of carrying a pregnancy to term relative to aborting. Percentage Completed College may also partially reflect higher opportunity costs.
Results similar to the WLS estimates also hold when closely examining the effect of abortion access and abortion law. In the log-odds ratio estimation, relative to abortion, abortion access affects the supply of infants relinquished. This effect is relatively small: a 10 percentage increase in the percentage of counties with abortion providers will decrease the ratio of infants relinquished to pregnancies aborted by less than 0.01 points. In comparison column 5 shows that a 10 percentage increase in the percentage of counties with abortion providers will increase the ratio of pregnancies aborted to live births not relinquished by a little over .05 points. Table IV presents the results of the log-odds ratio with the addition of the enjoined law variables. One interpretation of the Restrictive Public Funding Enjoined variable is that it may act as a proxy for Abortion Access as discussed previously.
In column 3, Table III, relative to keeping the infant, the supply of infants relinquished is associated with the Percentage Completed High School, the Incarceration Rate, and Percentage Attend Church. Not only is Percentage Completed College and Percentage Black associated with abortion demand (column l) relative to relinquishing an infant but also with the demand to keep an infant. Furthermore, the coefficients on the abortion law variables show the first indication that constraining access to abortion may increase the probability, in aggregate, of keeping the infant (rather than relinquishing the infant for adoption). Parental Involvement Law Enjoined in Table IV is positively associated with the supply of infants relinquished, and since it is significantly different from the enforced law, it likely depicts the separate cost effect of antiabortion sentiment.
The education and wage coefficients reveal a consistent pattern. Looking across all three log-odds ratios, Percentage Completed High School is positively associated with relinquishing an infant. And, though not significant, Percentage Completed High School is associated with keeping an infant relative to aborting. Relative to relinquishing an infant, Percentage Completed College is associated with aborting the pregnancy, as is the Real Hourly Wage. While not significant, these coefficients are also associated with aborting relative to keeping an infant.(19) Finally, relative [TABULAR DATA FOR TABLE III OMITTED] [TABULAR DATA FOR TABLE IV OMITTED] to aborting, the female unemployment rate is associated with keeping the infant. These results in the log-odds specification confirm the interpretation of the WLS regression; a higher percentage of women with higher opportunity costs, for example, expecting to complete college, having completed college, higher wages, or lower unemployment rates, is correlated with abortion demand.
The second column in each of these respective log-odds ratio models presents the 2SLS results. In the models Log (Relinquishing Infant / Keeping the Infant) and Log (Aborting / Keeping the Infant) a Hausman test indicates that the OLS and 2SLS coefficients are significantly different. This is the only indication that endogeneity bias may be affecting the coefficient on Abortion Access.
VI. DISCUSSION AND CONCLUSION
This paper examines the effect of abortion regulation and access on the supply of infants relinquished throughout the 1980s. With a cross-section of data from 1980, Medoff  finds that abortion and relinquishing an infant for adoption are not substitutes, and that AFDC payments negatively affected the supply of infants relinquished.
In contrast, I find that abortion access is negatively correlated with the supply of infants relinquished during the 1980s in two different ways. First, consistent with a theory of desired fertility, the availability of abortion providers has the expected effect of reducing the supply of infants relinquished. Second, abortion law has an unexpected negative effect. This suggests that as abortion laws become more restrictive the total number of unwanted births may decrease. The empirical results also suggest that omitted variable bias may confound the effect of some types of abortion law. Finally, to the extent that AFDC payments affect the incidence of single parenthood, I further find that the incidence of single parenthood is not correlated with the supply of infants relinquished.
These results should be interpreted carefully in the context of the limitations of the methodology employed. While I find that access to abortion during the 1980s is correlated with the decline in the supply of infants relinquished, I cannot claim that abortion access was the main cause of the decline in the supply of infants relinquished or that further restrictions on abortion would increase the supply of infants relinquished. The changing opportunity costs of women over this same time period may provide a better explanation than the direct effects of public policy. In both the WLS and the log-odds specifications, I find that a greater percentage of women with high opportunity costs, for example, expecting to complete college or having completed college, is correlated with abortion demand. These women who have an unintended pregnancy would find it more costly to continue with a pregnancy, give birth and relinquish, or keep the infant.
APPENDIX TABLE AI
Data Definitions and Sources
The number of infants relinquished is the total number of domestic adoptions of infants under two years of age adopted by persons not related to the infant by blood or marriage. Data from 1982 and 1986 from the National Committee for Adoption, Adoption Factbook 1985 and 1989. The number of infants relinquished in 1989 is calculated using the proportion of total adoptions that are domestic adoptions of infants from 1986 and applying this to the total number of domestic adoptions in 1989. For Alabama, California and Georgia, data are from 1990 [Flango and Flango, 1994].
Total pregnancies are the sum of total births per 1000 females aged 15-44 and abortions per 1000 females. Nonmarital pregnancies are the sum of nonmarital births per 1000 females aged 15-44 and abortions per 1000 single females. Abortions are by state of residence based on published data from the Alan Guttmacher Institute and personal correspondence with Stanley Henshaw, The Alan Guttmacher Institute. The percentage of abortions obtained by divorced, widowed or never married women are published by the Centers for Disease Control, CDC Surveillance Summaries, Morbidity and Mortality Weekly Report.
The number of total births and nonmarital births per 1000 females aged 15-44 is from Vital Statistics of the United States, various years, which is published by the U.S. Department of Health, Education, and Welfare. Nonmarital births are births to divorced, widowed or never married females.
Proportions of single females aged 15-44 are calculated from the March 1982 and 1986 Current Population Survey using weighted means, and 1990 Census of Population, Social and Economic Characteristics. These proportions are then multiplied by the total number of residents for each state in each respective year from Statistical Abstract of the United States, various years. Single is defined as separated, divorced, widowed, or never married.
Parental Involvement Law equals one if each state in each respective year had an enforced parental notification or consent law. Parental Involvement Law Enjoined equals one if for each state in each respective year in which the notification or consent law was enjoined [Greenberger and Connor, 1991; Haas-Wilson, 1996; and the NARAL Foundation]. Parental Involvement Law in Border States is the unweighted average in all border states of the variable parental involvement law.
Restrictive Public Funding equals one if each state in each respective year provided public funding of abortions only for low income women who are survivors of rape or incest, or if her life is in danger. Restrictive Public Funding Enjoined equals one if each state in each respective year provided unrestrictive public funding by court order only. Data reflect available information closest to the years in the analysis as reported by Gold and Daley , Gold and Macias , and Gold and Nestor . Restrictive Public Funding in Border States is the unweighted average in all border states of the variable restrictive public funding.
Number of Abortion Restrictions enacted between 1973 and 1989 is a measure of state policies designed to regulate the market for abortion services since Roe v. Wade. These policies include: conscience clauses (35 states), fetal experimentation (23 states), postviability requirements (29 states), postviability standards of care (29 states), memorials to Congress (25 states), calls for a constitutional convention (19 states), feticide laws (9 states), fetal disposal laws (11 states), informed consent laws (17 states), spousal notification (7 states), second-trimester hospitalization requirements (17 states), and insurance restrictions (10 states). Halva-Neubauer  and Haas-Wilson .
Abortion Access is the percentage of counties with providers who performed five or more abortions in each relevant year. Data reflect available information closest to the years in the analysis as reported by Henshaw and Van Vort, of the Alan Guttmacher Institute .
Percentage Completed High School and Percentage Completed Four Year College is calculated for women aged 25 through 34 from the March 1982 and 1986 Current Population Survey using weighted means, and 1990 Census of Population, Social and Economic Characteristics.
Female Unemployment Rate from Bureau of Labor Statistics Bulletin 2340, Handbook of Labor Statistics, 1989. Data for 1989 from Statistical Abstract of the United States.
Real Hourly Wage is the hourly wage of production and nonsupervisory workers in manufacturing from Bureau of Labor Statistics Bulletin 2340, Handbook of Labor Statistics, 1989, and Bulletin 2411, Hours and Earnings, States and Areas, 19871992. The Hourly Wage is deflated to 1982 dollars using the CPI Index.
Real AFDC maximum benefit is the AFDC payment to a family with three children and one adult. AFDC payment is deflated to 1982 dollars using the CPI index. 1982 data from Quarterly Public Assistance Statistics, U.S. Department of Health and Human Services, Table 25, April-June 1982. 1986 and 1989 from unpublished data, U.S. Department of Health and Human Services.
Incarceration Rate is the number of sentenced prisoners in the state per 100,000 resident population from United States Department of Justice, Bureau of Justice Statistics, Sourcebook of Criminal Justice Statistics, 1991.
Percentage Black Females is the proportion of females ages 15-44 whose race is identified as black, and when applicable, are separated, divorced, widowed or never married, in each state for each year, calculated from March 1982 and 1986 Current Population Survey, using weighted means, and 1990 Census of Population, General Population Characteristics.
Percentage Attend Church is the proportion of the total population in each state that adhere to a religious denomination (attend church regularly). Data are from Churches and Church Membership in the United States, 1980, Table 3. [Quinn, B, H. Anderson, M. Bradley, P. Goetting and P. Shriver, 1980]
Percentage of population living in a metropolitan statistical area from Statistical Abstract of the United States, various years. Data for 1982 is actual data from 1983, and data for 1989 is actual data as of April 1990.
Number of OBGYNs is the total number of obstetricians and gynecologists in total patient care for each state in 1982, 1986 and 1989 from the American Medical Association, Physician Characteristics and Distribution in the United States, 1983, 1987 and 1992, Table 9. Number of Hospitals is the total number of hospitals in each state that are American Hospital Association members from the American Hospital Association, Hospital Statistics, 1983, 1987 and 1990, Table 10B.
[TABULAR DATA FOR APPENDIX TABLE AII OMITTED]
[TABULAR DATA FOR APPENDIX TABLE AIII OMITTED]
This paper was written while the author was at Cornell University. A previous version of this paper was presented at the 1996 Population Association of America annual meetings, New Orleans. For helpful comments, I thank Deborah J. Anderson, Susan Averett, Carlena Cochi Ficano, Robert Hutchens, H. Elizabeth Peters, Ramsey D. Shehadeh, and an anonymous referee. For the invaluable service of data provision, I thank Deborah J. Anderson, Stanley Henshaw of the Alan Guttmacher Institute, and Nicole McLaughlin of the National Abortion Rights Action League Foundation. Finally, I thank Nicholas Keifer, whose course requirements inspired the first version of this paper. The author is solely responsible for all views expressed in this manuscript and any remaining errors or omissions.
Gennetian: Research Associate, Manpower Demonstration Research Corporation, New York, New York, Phone 1-212-532-3200, Fax 1-212-684-0832 E-mail firstname.lastname@example.org
1. Mosher  estimates that of the 2.7 million American couples with impaired fertility, approximately 1 million are interested in adoption.
2. The New York Times  reports that opponents of a recent Tennessee law, which allows adoptees to get information about their birth parents, argue that opening adoption records will invade a birth mother's privacy and encourage more women to choose abortion. With regard to restricting abortions, the recent Supreme Court decision, Planned Parenthood of Southeastern Pennsylvania v. Robert Casey, in 1992, ruled that state restrictions, such as parental consent requirements, would be upheld unless they placed an undue burden on a woman obtaining an abortion.
3. For examples of this literature see An, Haveman, and Wolfe , Clarke and Strauss , Moffitt , Murray , and Ratcliffe .
4. The abortion rate per 1000 women aged 15-44 increased from 16.3 in 1973 to 29.3 in 1980, reached a plateau during the 1980s at approximately 28 and declined slightly through the late 1980s and early 1990s to approximately 27. The trend among unmarried women is similar. For examples of this literature see Singh , Haas-Wilson , Blank, George, and London , Levine, Trainor, and Zimmerman .
5. At the time of Medoff's study, only one year of state-level adoption data was available.
6. For example, in any given year in the National Longitudinal Survey of Youth (NLSY) the approximate number of infants relinquished is less than 10.
7. Most studies find that white women are much more likely to relinquish their infants for adoption than are black women. According to Sobol and Daly , some of the reasons include the small number of potential black adoptive parents, the greater presence of the biological mother's extended family, and cultural norms.
8. See Blank, George, and London , for further discussion comparing use of state of residence versus state of occurrence abortion data.
9. While I could not formally test between the fixed effects and random effects specification, I was able to estimate a random effects model that does account for correlation of the error terms across states over years. The empirical results are robust to the random effects specification.
10. State policy may also vary by region or stem from the political composition of the state. I consider the following measures: regional indicators for the Midwest, South, and West (with Northeast as the default category); an indicator for a Republican governor; the percentage of the state legislature; and percentage of the state senate that is Republican in each state for each year. In all model specifications these variables were not significantly different from zero and thus were dropped from the analysis.
11. If the latent structure of the model is [[Beta].sub.2a] Restrictive Public Funding = [[Alpha].sub.1][Z.sub.1t] + [[Alpha].sub.2][Z.sub.2t], where [Z.sub.1t] is Restrictive Public Funding and [Z.sub.2t] is antiabortion sentiment (and unobserved). The parameter of interest is [[Alpha].sub.1]. In general if covariance ([Z.sub.1t], [Z.sub.2t]) [greater than] 0, then [Mathematical Expression Omitted] when [Z.sub.2t] is excluded.
12. Throughout this section, the marginal effects for continuous variables are calculated as: [Mathematical Expression Omitted]. The marginal effects for discrete variables are calculated as: [Mathematical Expression Omitted]. In both cases [Mathematical Expression Omitted].
13. For the variables of interest, results are robust to alternative functional forms (for example, linear) of the dependent variable and are available from the author.
14. In one specification, I replace Abortion Access with the Number of Abortion Providers per 1000 single females. The coefficient on Number of Abortion Providers is not statistically significant from zero and the signs and sizes of the other coefficients remain the same. Note that unlike the number of Abortion Providers, Abortion Access captures the location and distribution of abortion availability.
15. In a theoretical model in which pregnancy is an endogenous decision, Kane and Staiger  find that a small increase in the cost of abortion works on the pregnancy margin while more dramatic increases in the cost of abortion work on the abortion margin. If changes in abortion laws represent more costly changes in the cost of abortion relative to changes in Abortion Access, then the empirical results are consistent with this theory.
16. In the first stage results, Number of OBGYNS is positive and significant and the Number of Hospitals is negative and significant. Since the majority of abortions are performed in clinics and clinics often locate in counties without hospitals, the negative coefficient on Number of Hospitals is not surprising.
17. The large positive and significant coefficient on percentage completed high school remains when excluding percentage completed college. The large negative and significant coefficient on percentage completed college remains when excluding percentage completed high school in the specification with nonmarital pregnancies.
18. In the specification with the denominator single women in the dependent variable, real AFDC maximum benefits are positive and significant. In a state aggregate analysis using single women as the denominator, Clarke and Strauss  find that AFDC benefits may be endogenous. The sign and significance of this variable may in part reflect this endogeneity.
19. Note that the association of Percentage Completed Four Year College with keeping an infant relative to relinquishing an infant may suggest the effects of timing on pregnancy resolution. Highly educated women who surpass the safe period of time for an abortion may be more likely to keep the infant rather than relinquish the infant for adoption.
An, Chong-Bum, Robert Haveman, and Barbara Wolfe. "Teen Out-of-Wedlock Births and Welfare Receipt: The Role of Childhood Events and Circumstances." Review of Economics and Statistics, May 1995, 195208.
Bachrach, Christine, Kathy S. Stolley, and Kathryn A. London. "Relinquishment of Premarital Births: Evidence from National Survey Data." Family Planning Perspectives, January/February 1992, 27-32.
Blank, Rebecca, Christine C. George, and Rebecca London. "State Abortion Rates: The Impact of Policies, Providers, Politics, Demographics, and Economic Environment." Journal of Health Economics, October 1996, 513-53.
Brown, Sarah S., and Leon Eisenberg, eds. The Best Intentions: Unintended Pregnancy and the Well-Being of Children and Families. Washington, D.C.: National Academy Press, 1995.
Clarke, G. R. G., and Robert P. Strauss. "Children as Income Producing Assets: The Case of Teen Illegitimacy and Government Transfers." Mimeo, Department of Economics, University of Rochester, 1995.
Centers for Disease Control and Prevention. "Abortion Surveillance: United States." Morbidity and Mortality Weekly Report. Washington, D.C.: U.S. Government Printing Office, various years.
Currie, Janet, Lucia Nixon, and Nancy Cole. "Restrictions on Medicaid Funding of Abortion: Effects on Birth Weight and Pregnancy Resolutions." Journal of Human Resources, Winter 1996, 157-88.
Cushman, Linda F., Debra Kalmuss, and Pearila B. Namerow. "Placing an Infant for Adoption: The Experiences of Young Birthmothers." Social Work, May 1993, 264-72.
Ellwood, David, and Mary Jo Bane. "Slipping Into and Out of Poverty: The Dynamics of Spells." Journal of Human Resources, Winter 1986, 1-23.
Flango, Victor E. "Agency and Private Adoptions by State." Child Welfare, May/June 1990, 263-75.
Flango, Victor E, and Carol R. Flango. The Flow of Adoption Information from the States. Williamsburg, Va.: National Center for State Courts, U.S. Department of Health and Human Services, 1994.
Gold, Rachel, and Daniel Daley. "Public Funding of Contraceptive, Sterilization and Abortion Services, Fiscal Year, 1990." Family Planning Perspectives, September/October 1991, 204-11.
Gold, Rachel, and J. Macias. "Public Funding of Contraception, Sterilization and Abortion Services, 1985." Family Planning Perspectives, November/December 1986, 259-64.
Gold, Rachel, and Barry Nestor. "Public Funding of Contraceptive, Sterilization, and Abortion Services, 1982." Family Planning Perspectives, May/June 1984, 128-33.
Greenberger, Marcia D., and Katherine Connor. "Parental Notice and Consent for Abortion: Out of Step with Family Law Principles and Policies." Family Planning Perspectives, January/February 1991, 31-5.
Greene, William H. Econometric Analysis. New York: MacMillan, 1990.
Grossman, Michael, and Theodore J. Joyce. "Unobservables, Pregnancy Resolutions, and Birth Weight Production Functions in New York City." Journal of Political Economy, October 1990, 983-1007.
Haas-Wilson, Deborah. "The Impact of State Abortion Restrictions on Minor's Demand for Abortions." Journal of Human Resources, Winter 1996, 140-58.
Halva-Neubauer, Glen. "Abortion Policy in the Post-Webster Age." Publius, Summer 1990, 327-44.
Henshaw, Stanley K., and Jennifer Van Volt. Abortion Factbook 1992 Edition: Readings, Trends, and State and Local Data to 1988. New York: Alan Guttmacher Institute, 1992.
Kalmuss, Debra, Pearila B. Namerow, and Linda F. Cushman. "Adoption versus Parenting among Young Pregnant Women." Family Planning Perspectives, January/February 1991, 17-21.
Kane, Thomas J., and Douglas Staiger. "Teen Motherhood and Abortion Access." Quarterly Journal of Economics, May 1996, 467-506.
Leibowitz, Arleen, Marvin Eisen, and Winston K. Chow. "An Economic Model of Teenage Pregnancy Decision-Making." Demography, February 1986, 67-77.
Levine, Phillip B., Amy B. Trainor, and David J. Zimmerman. "The Effect of Medicaid Abortion Funding Restrictions on Abortion, Pregnancy and Births." Journal of Health Economics, October 1996, 555-78.
Lundberg, Shelly, and Robert Plotnick. "Adolescent Premarital Childbearing: Do Economic Incentives Matter?" Journal of Labor Economics, April 1995, 177-200.
-----. "Effects of State Welfare, Abortion, and Family Planning Policies on Premarital Childbearing among White Adolescents." Family Planning Perspectives, November/December 1990, 246-75.
Maddala, G. S. Limited Dependent and Qualitative Variables in Econometrics. Cambridge: Cambridge University Press, 1983.
Maza, Penelope. "Adoption Trends: 1944-1975." Child Welfare Research Notes 9, 1994.
Medoff, Michael. "An Empirical Analysis of Adoption." Economic Inquiry, January 1993, 59-70.
Moffitt, Robert. "Incentive Effects of the U.S. Welfare System: A Review." Journal of Economic Literature, March 1992, 1-61.
Mosher, William D. "Infertility: Why Business Is Booming." American Demographics, 1987, 42-3.
Murray, Charles. "Welfare and the Family: The U.S. Experience." Journal of Labor Economics, January 1993, s224-62.
NARAL Foundation. Who Decides? A State-by-State Review of Abortion Rights. Washington, D.C.: NARAL Foundation, various years.
National Committee for Adoption (NCFA). 1989 Adoption Factbook: United States Data, Issues, Regulations and Resources. Washington, D.C.: National Committee for Adoption, 1989.
-----. Adoption Factbook: United States Data, Issues, Regulations and Resources. Washington, D.C.: National Committee for Adoption, 1985.
New York Times. "Tennessee is Focus of Debate on Adoptees' Birth Records." March 18, A1, 1996.
Ratcliffe, Caroline. "Intergenerational Transmission of Welfare Participation: How Large Is the Causal Link?" Mimeo, Department of Economics, Cornell University, 1995.
Singh, Susheela. "Adolescent Pregnancy in the United States: An Interstate Analysis." Family Planning Perspectives, September/October 1986, 210-20.
Sobol, Michael, and Kerry J. Daly. "The Adoption Alternative for Pregnant Adolescents: Decision Making, Consequences, and Policy Implications." Journal of Social Issues 48(3), 1992, 143-61.
U.S. Bureau of the Census. Statistical Abstract of the United States. Washington, D.C.: Reference Press, 1994.
U.S. Department of Health, Education and Welfare. "Adoptions." Publication no. (SRS) 72-03259. Center for Social Statistics Report E-10, various years.
|Printer friendly Cite/link Email Feedback|
|Author:||Gennetian, Lisa A.|
|Date:||Jul 1, 1999|
|Previous Article:||The demand for illicit drugs.|
|Next Article:||An analysis of women's return-to-work decisions following first birth.|