Journal of Family and Economic Issues

, Volume 28, Issue 4, pp 583–599

Price, Restrictions and Abortion Demand

Authors

    • Department of EconomicsCalifornia State University, Long Beach
Original Paper

DOI: 10.1007/s10834-007-9080-9

Cite this article as:
Medoff, M.H. J Fam Econ Iss (2007) 28: 583. doi:10.1007/s10834-007-9080-9

Abstract

This study uses pooled time-series data to estimate the effects of various restrictive abortion laws on the demand for abortion. This study differs from prior pooled time-series cross-section research in that it explicitly includes the price of an abortion in the abortion demand equation. State Medicaid funding is found to increase the abortion demand of women of childbearing age; while the price of an abortion, parental involvement, parental consent, and parental notification laws all have a negative effect on the demand for abortions. State mandatory waiting periods have no statistically significant impact on abortion demand. The empirical results remain robust for the abortion demand of teen minors.

Keywords

Abortion demandRestrictive abortion laws

Introduction

On January 22, 1973, the U.S. Supreme Court in Roe vs Wade ruled that there existed a constitutionally protected right to privacy which allowed women to obtain an abortion. The Roe vs Wade decision established a woman’s right to have an abortion, but the ruling did not mandate access to an abortion. The decision permitted substantial interstate differences in access to an abortion. What followed in the ensuing years after the Roe vs Wade decision was a wide variety of laws enacted by states that placed restrictions on the ability of women to obtain an abortion. Because the Roe vs Wade decision was vague about the precise parameters of permissible state restrictions, every state law was immediately challenged in federal court as being an infringement of a woman’s constitutional right to obtain an abortion.

In 1992 the Supreme Court ruled, in Planned Parenthood vs Casey, that a state was allowed to impose restrictions on a woman’s access to an abortion provided that the restrictions did not impose an undue burden on a woman’s right to obtain an abortion. The Casey decision had two effects. First, it gave states considerable discretion and latitude to restrict a woman’s access to an abortion. An abortion restriction may not be a prima facie undue burden, but in practice may substantially restrict or curtail a woman’s access to an abortion. Second, prior to the Casey decision a state could only exercise discretion in restricting public funding of abortions or restricting minors’ access to an abortion. After the Casey decision, however, states could and did enact a variety of new abortion restrictions. The Casey decision enhanced the need for empirical research on the effect of these restrictions to determine if any of them constitute an undue burden on women seeking to obtain an abortion.

This study examines two issues of public policy interest. (1) What effect do various restrictive state abortion laws (i.e., Medicaid funding, parental involvement, and waiting periods) have on the demand for abortion? (2) How responsive is abortion demand to changes in abortion price and income? This study differs from previous studies in a number of significant ways. This study uses state data pooled over the Census years 1982, 1992, and 2000. The advantage of this is that the price of an abortion within each state is available on a consistent basis for all 3 years. The elapsed time between the three sample years is sufficiently long enough to detect changes in the demand for abortion due to state abortion restrictions. The three time periods are also far enough apart so that there is ample variation in the independent control variables to detect their impact on abortion demand. Using Census years also provides the opportunity to utilize as controls socioeconomic variables that only measure women of childbearing age.

Literature Review

Price and Income Elasticity of Demand

An elasticity of abortion demand is a numerical measure of the responsiveness in the number of abortions per 1,000 pregnancies resulting from a 1% change in either the price of an abortion or income. There have been four studies, all of which used a model based on the economic theory of fertility control, to estimate the demand for abortion. These studies provided a variety of estimates of the price and income elasticity of abortion demand because of differences in specifications, estimation techniques, variable definitions, and time periods.

Table 1 provides a summary of the elasticity figures from the four studies. All four studies used either individual state-level or pooled time-series cross-section state data and found that the fundamental law of demand was applicable to abortion and the price elasticity of demand for abortion tended to be inelastic (relatively unresponsive to price increases—in absolute value a number less than one). A 1% increase in the price of an abortion will reduce the demand for abortions by less than 1%. Abortion was also found to be a normal good—increases in income increase the purchases of abortion services. The estimated income elasticity of abortion demand ranges from .27 to 1.33 depending upon how income is measured.
Table 1

Estimates of the price and income elasticity of abortion demand in previous studies

Author

State level data, year(s)

Estimation method

Price elasticity

Income elasticity

Income measure

Medoff (1988)

1982

2SLS

−.81

.79

Average income women, 15–44

Garbacz (1990)

1982

OLS

−.68, −1.04

.84–1.3

Disposable income per capita

Gohmann and Ohsfeldt (1993)

1985

2SLS

−.75, −1.3

.62–1.0

Per capita income

Pooled: 1982–1987

2SLS

−1.1

.35

Per capita income

Medoff (1997)

Pooled: 1982 & 1992

2SLS

−.70, −.99

.27–.35

Average income women, 15–44

State Medicaid Funding

In 1976, Congress approved the Hyde amendment which prohibited federal Medicaid funding of abortions for indigent women, except when the life of the woman is endangered. Abortion funding for low-income women was left to the discretion of each individual state. A number of states continue to use public funds to pay for Medicaid abortions for low-income women. State Medicaid funding of abortions reduces the out-of-pocket cost of abortion services to zero for low-income women and as a consequence one would expect the demand for abortions in those states to be greater, other things equal.

Table 2 provides a summary of studies that used multivariate analysis on multi-year panel data of states to estimate the impact of state Medicaid abortion funding. All these studies estimated the demand for abortion using state, year, and state trend fixed effects, and none of the studies included the price of an abortion. In general, the empirical results were somewhat supportive of the prediction that state Medicaid funding of abortions increases the demand for abortions. State Medicaid funding was found to increase the abortion rate of women of childbearing ages (15–44) by between 3% and 6% depending on the time period examined and the number of fixed effects controls. The numerical impact of state Medicaid funding of abortions tended to be smaller (in some cases statistically insignificant) the greater the number of fixed effects controls included in the estimation.
Table 2

Estimates of the numerical impact of state Medicaid funding on abortion demand in previous studies

Author

Pooled state level years

Estimation method

Numerical impact

Fixed effects variables

Blank et al. (1996)

1974–1988

OLS, 2SLS

n.s.

State and year

Women (15–44)

Haas-Wilson (1996)

1976–1990

Weighted OLS

+9%, +15%

State and year

Women (15–19)

+14%, +16%

State, year, and state trend

Levine et al. (1996)

1977–1988

Weighted OLS

+5.5%

State and year

Women (15–44)

+3.0%

State, year, and state trend

Matthews et al. (1997)

1978–1988

Weighted OLS

+2.9%

State and year

Women (15–44)

n.s.

State, year, and state trend

Bitler and Zavodny (2001)

1974–1997

Weighted OLS

n.s.

State and year

Women (15–44)

n.s.

State, year, and state trend

Note: n.s. = not statistically significant

Parental Involvement Laws

Since the 1973 Roe vs Wade decision, the Supreme Court has permitted a state to enact parental involvement laws. Parental involvement laws require unmarried minors (less than 18 years of age) to notify or get permission from one or both parents before obtaining an abortion. Parental involvement laws increase the indirect costs of an abortion (higher emotional costs as well as out-of-pocket money expenses) and would be expected to decrease the demand for abortion. The impact of parental involvement laws on abortion demand depends upon a number of factors. (1) The laws may be merely symbolic and not binding if minors routinely involve their parents in the abortion decision. (2) The laws vary widely in their degree of restrictiveness. Some states allow grandparents, other adult relatives, or other specified health professionals to be involved in the abortion decision. (3) The impact of a parental involvement law may be diminished depending on a minor’s proximity to abortion providers in bordering states without such laws.

Case studies of individual states consistently found that a parental involvement law has either a very small or no impact on teen’s demand for abortion. Joyce and Kaestner (1996) found that parental involvement laws had a small impact on teen minors’ demand for abortion in South Carolina and Tennessee. Rogers et al. (1991) found that teen minor abortion rates decreased slightly after Minnesota enacted a parental involvement law. Cartoff and Klerman (1986) and Henshaw (1995) found parental involvement laws had no impact on teen abortion rates in Massachusetts and Mississippi, respectively.

Table 3 provides a summary of the six studies that used pooled time-series cross-section state data to examine the impact of parental involvement laws on state abortion rates. Five of these studies used a fixed effects model, and none of the five studies included the abortion price in their estimation of the abortion demand. The empirical results of these five studies (which are the last five listed in Table 3) were ambiguous.
Table 3

Estimates of the numerical impact of state parental involvement law on abortion demand in previous studies

Author

Pooled state level years

Estimation method

Numerical impact

Fixed effects variables

Ohsfeldt and Gohmann (1994)

1984–1987

OLS

−18%, −22%

Year

Women (15–17)

Blank et al. (1996)

1974–1988

OLS, 2SLS

n.s.

State and year

Women (15–44)

Haas-Wilson (1996)

1978–1990

Weighted OLS

−13%, −22%

State and year

Women (15–19)

Matthews et al. (1997)

1978–1988

Weighted OLS

−3.2%

State and year

Women (15–44)

n.s.

State, year, and state trend

Bitler and Zavodny (2001)

1974–1997

Weighted OLS

−5.5%

State and year

Women (15–44)

n.s.

State, year, and state trend

Levine (2003)

1985–1996

Weighted OLS

−21%

State and year

Women (15–17)

−15%

State, year, and state trend

Note: n.s. = not statistically significant

Blank et al. (1996) found that parental involvement laws had no statistically significant impact on the abortion rate of women of childbearing age. Matthews et al. (1997) and Bitler and Zavodny (2001) reported that parental involvement laws reduced the abortion demand of women of childbearing age by 3.2% and 5.5%, respectively, when a state and year fixed effects model was used. But both studies found that the effect of parental involvement laws became statistically insignificant when a state trend fixed effect variable was added to their model. Haas-Wilson (1996) and Levine (2003) found that parental involvement laws reduced the abortion demand of teenagers (ages 15–17) by between 13% and 22%.

The only study listed in Table 3 that included the abortion price (but did not include state fixed effects) in their model was Ohsfeldt and Gohmann (1994). Ohsfeldt and Gohmann reported that parental involvement laws reduced the ratio of abortion rates of minors (ages 15–17) to older teens (ages 18–19) by 22% for the 3 years 1984, 1985, and 1986. The problem with their empirical results is that the abortion price in their model had an improbable positive, though statistically insignificant, impact on abortion demand (Ohsfeldt and Gohmann’s study was not included in Table 1 because they did not estimate a price elasticity of demand value).

Mandatory Waiting Periods

Many states, after the 1992 Casey decision, enacted mandatory waiting period laws. These laws mandate that all women wait a specified time period, typically 24 h, before obtaining an abortion. During the waiting period most states also require that these women receive or be offered abortion-related material. The information normally includes a list of available adoption agencies, material on fetal viability and development, the health and psychological risks associated with an abortion procedure, and information about the financial legal liability of the biological father for child support. Parenthetically, no state requires that a woman certify that she actually read or received the materials. In most cases all that is required is that the woman be given or is informed of the availability of the materials.

A state’s rationale for having a mandatory waiting period is to provide women the time and information necessary to make an informed choice. The implicit intent of waiting period laws is to increase the emotional costs to women and dissuade women from terminating an unplanned pregnancy. Mandatory waiting period laws also increase other indirect costs of an abortion to women. Waiting period laws usually require women to make two visits to an abortion provider. This increases the indirect cost of an abortion to women in terms of travel time, lost work time, possible overnight accommodations and additional childcare expenses, as well as emotional costs. The predicted effect of mandatory waiting periods is to decrease the demand for abortions.

Because mandatory waiting period laws are a relatively new type of restriction, studies have focused on data that compare the impact on abortion trends in an individual state before and after the law was adopted. The adoption of a mandatory waiting period law in Mississippi in 1992 was associated with a fall in the number of abortions performed in Mississippi, an increase in the number of out-of-state abortions performed on Mississippi residents, and a rise in the number of abortions performed during the second trimester of pregnancy to Mississippi residents (Althaus and Henshaw 1994; Joyce and Kaestner 2000, 2001).

There remain several unexplored questions and weaknesses in the literature. An important shortcoming of many of the aforementioned studies is the omission of the price of an abortion in their estimation of the demand for abortion (notable exceptions are the studies listed in Table 1). The reason for this omission is because a state’s abortion price is not available on a consistent year-to-year basis. Virtually all the studies which omit the abortion price included state fixed effects variables (dummy variables for the states) in their models. Moffitt (1994) was among the first to argue that the justification for the inclusion of fixed effects variables was to control for unmeasured or otherwise difficult to quantify explanatory variables that vary across states and that might influence the abortion demand. It might be argued that the exclusion of the abortion price in an estimation of the abortion demand is somewhat mitigated since some of the variation in the abortion price across states may be subsumed within each of the state fixed effects variables. However, the variation in the abortion price across states is controlled for with state fixed effects only if the difference across states does not vary over time.

Moreover, one of the ways that restrictive abortion laws affect abortion demand is through their impact on the abortion price. A restrictive abortion law increases the financial costs (e.g., lost work time, extra travel time, more required visits to a provider) incurred by a woman in terminating a pregnancy as a result of the law. But restrictive abortion laws also impose additional operating expenses on abortion providers in order to comply with the restrictive abortion laws (Althaus and Henshaw 1994). All these additional operating expenses increase a provider’s cost of providing abortion services resulting in a higher abortion price. A consequence of excluding a state’s abortion price in the estimation of abortion demand is that the resulting coefficient of a restrictive abortion law shows the total effect (including the effect of the abortion law that operates through the omitted abortion price) of a change in a restrictive abortion law on abortion demand. The inclusion and independent impact of the abortion price in an estimation of the demand for abortion is of interest to researchers concerned with the issue of how much of the impact of restrictive abortion laws operates outside of the abortion price effect.

Many of the prior studies also reflect the authors’ ambiguity about when various abortion restrictions were in effect in a state since many of these restrictions were in flux due to legal challenges. As a consequence, many of these studies are inconsistent with one another in their modeling of the time periods during which the restrictions are enforced. In addition, it is difficult to detect small changes in state abortion rates using year-to-year data, particularly over a short time span. This means that estimates of the effect of abortion restrictions may not be reliable. In order to estimate the effect of an abortion restriction on abortion demand, an extended period of time needs to be examined.

Some of the prior studies use abortion data based on the state of occurrence of the abortion procedure. But for most studies the relevant data is abortions by state of residence. The reason is that abortions by state of occurrence does not capture the impact of a restriction within a state if women are able to travel to a bordering state to obtain an abortion. The actual impact of the restriction may be just to change the location of the abortion procedure rather than reduce the demand for abortion. The consequence is that the empirical results on the effect of an abortion restriction on abortion demand using abortions by state occurrence will be spurious.

Data

This study uses state data pooled over three Census years to estimate the effects of the price of an abortion and various restrictive abortion laws on the demand for abortion. A complete listing of the dependent and independent variables used in this paper, their means and standard deviations, and their data sources is provided in Table 4. All the economic variables by state used in this paper are for the Census years 1980, 1990, and 2000, whereas the abortion demand dependent variable and the abortion price by state are only available from the Alan Guttmacher Institute (2004) for the years 1982, 1992, and 2000. There is no bias in using economic data from 1980 and 1990 rather than 1982 and 1992 economic data since the economic variables are state averages which change relatively slowly over a 2-year period (Schmidt 2005, p. 244).
Table 4

Means and standard deviations

Variable

Mean (standard deviation)

Abortions/1,000 pregnancies women (ages 15–44)

224.87 (86.79)

Abortions/1,000 pregnancies minors (ages 15–17)

378.09 (137.6)

Female income (in 1982 dollars)

12741.29 (2744.35)

Labor force participation women (ages 16–44)

73.51 (9.77)

% Single women (ages 16–44)

34.81 (4.10)

% Education 12+  years women (ages 25–44)

76.78 (6.49)

% Evangelical Christians

14.55 (13.67)

Abortion price (in 1982 dollars)

219.38 (41.71)

State Medicaid funding

.29 (.45)

Parental involvement law

.40 (.49)

Parental border involvement

38.95 (33.76)

Parental consent law

.22 (.42)

Parental notification law

.18 (.38)

Note: The number of abortions/1,000 pregnancies of women and minors and the abortion price for the years 1982, 1992, and 2000 are from the Alan Guttmacher Institute (2004). All economic variables are for the years 1980, 1990, and 2000 and are from the U.S. Bureau of the Census, U.S. Census of Population, State Reports 1980, 1990, and 2000 (1983, 1993,2003). Each state’s abortion laws for the years 1982 and 1992 are from Merz et al. (1995) and from the Alan Guttmacher Institute (2004) for the year 2000

Methodology

The theoretical model of fertility control argues that the determinants of abortion demand are opportunity cost factors, revealed tastes for children, and the direct and indirect costs of an abortion (Michael 1973; Willis 1973). The abortion demand equation to be estimated is:
$$ \begin{aligned}{} & {\text{A}}_{{st}} \, = \,b_{0} + b_{1} {\text { Abortion Price}}_{{st}} + b_{2} {\text {Female Income}}_{{st}} + b_{3} {\text {X}}_{{st}} + b_{4} \text {Re} {\text{strictive Abortion Laws}}_{{st}} \\ & + {\text{ }}b_{5} {\text {Time}}_{{st}} + {\text{ u}}_{{st}} ,\quad s = {\text{ }}1,{\text{ }}2, \ldots {\text{ }},50;{\text{ }}t = 1982,{\text{ }}1992,{\text{ }}2000 \\ \end{aligned} $$
(1)

The dependent variable is the abortion rate, the number of abortions (by state of residence) per 1,000 pregnancies of women of childbearing ages (15–44 years) in state s during time period t.

The Abortion Price is the average cost (in 1982 dollars) of an abortion performed during the first 10 weeks in a nonhospital facility in each state, s, during time period, t. Ninety percent of abortions performed in the United States each year are during the first trimester and over 93% are performed in a nonhospital facility. The Alan Guttmacher Institute (2004) calculates the average state-level abortion price by weighing the abortion price charged by each provider within a state by the number of abortions performed by that provider. Aggregation basis would be a problem if there were a large variation in the price charged by providers within a state. However, large abortion clinics (400 or more abortions per year) perform more than 94% of all the abortions in the United States. Virtually all these large abortion clinics (+99%) are located in a large metropolitan area within a state (Finer and Henshaw 2003). These large abortion clinics, as noted by Perloff (2006, p. 256), operate in a highly competitive market and very close to their break-even point. The consequence is that, within a state, there is very little variation in the abortion price since, in order to be competitive, each large abortion clinic must keep its abortion price close to its competitors’ abortion price.

The variable Female Income is the average income (in 1982 dollars) of women who worked full-time year-round during time period, t. It has been argued that the full-time income of women is an excellent measure of the opportunity cost of a woman’s time and of potential income foregone (Mincer 1963).

The vector X includes other opportunity cost and revealed tastes measures that have been found in the literature to be determinants of abortion demand. They include Labor Force Participation rate of women ages 16–44 years, percentage of women ages 16–44 who are Single, percentage of women ages 25 to 44 who have completed 12 years or more of Education, and percentage of the population who are affiliated with an Evangelical Christian denomination that believes that the Bible dictates that abortion is murder.

The vector Restrictive Abortion Laws includes measures of a state’s abortion restrictions. The variable State Medicaid Funding equals 1 if state, s, funded Medicaid abortions for low-income women during time period, t. One would expect the demand for abortion to be greater in these states since out-of-pocket cost is no longer a consideration in the utilization of abortion services. The variable Parental Involvement Law equals 1 if state, s, had a parental involvement law in effect during time period, t. Parental involvement laws raise the indirect costs of an abortion and would be expected to reduce the demand for abortion. However, the impact of a parental involvement law may be attenuated if minors can obtain an abortion from a provider in a neighboring border state without such a law. The variable Parental Border Involvement is a weighted average of the percentage of all states that border to state, s, which enforce a parental involvement law during time period, t. The weights are the inverse distance between the capital of state, s, and the capital of each border state (the reported empirical results are unchanged if the weights are the inverse distance between the most populated metropolitan area in state, s, and the most populated metropolitan area of each border state or if an unweighted average is used). The value of the Parental Border Involvement variable ranges from 0 (none of the border states enforce a parental involvement law) to 100% (all of the border states enforce a parental involvement law).

Two obvious concerns are (a) collinearity between the restrictive abortion laws and the other variables in the abortion demand equation and (b) that a state’s restrictive abortion laws are not exogenous, but are a reaction by state legislators to the level of abortion demand in their state. None of the zero-order correlations between either of the restrictive abortion laws and the other regressors in Eq. (1) was greater than .25. In addition, both Cohen and Barrilleaux (1993) and Medoff (2002) found that a state’s passage of restrictive abortion laws was not a reaction to a state’s abortion rates. Each of these studies found that a state’s restrictive abortion policy was a function of the political strength of well-organized and highly mobilized interest groups opposed to abortion.

The last variables are two time trend variables. Time 2000 equals 1 for all states in the year 2000 and Time 1992 equals 1 for all states in the year 1992 and zero otherwise. These time trend variables control for time-varying factors such as greater awareness of the possible deleterious health effects from unprotected sexual activity, more awareness and availability of alternative contraceptive methods (including abstinence), and greater promotion and practice of safe sex that are common to the populace in all states over the time period 1982–2000 (Kaiser Family Foundation 1999; Levine 2001).

Empirical Results

In the abortion demand Eq. (1), the price of an abortion is endogenous since as noted by Levine (2004), “...prices are determined by the behavior of those who supply abortion services and those who demand them” (p. 114). Using a Hausman test, the null hypothesis that the abortion price is exogenous is rejected at the .01 level of significance; the coefficient of the residual variable was −.576, and the t-statistic was −5.19). The econometric solution to this problem (Gujarati 1995, p. 604) is to find instruments for the abortion price that are correlated with the abortion price, but do not directly affect the demand for abortion and then estimate the abortion demand equation using two-stage least-squares. Following Blank et al. (1996), the instruments selected for the abortion price are (a) the number of non-OBGYN physicians per 100,000 females ages 15–44; (b) number of hospitals; and (c) the average weekly wage of employees in clinics and offices of physicians. These three variables are related to the overall level of availability and accessibility of health care providers of general medical services for women in a state, but should not be significantly affected by the demand for abortion within a state (the first-stage instrumental variable results are available upon request).

The two-stage least-squares estimation results of the abortion demand equation appear in Table 5, column 1. The empirical results provide substantial support for the a priori expectations of the fertility control model and prior research. Women of childbearing age in the labor force have a significantly (p < .01) greater demand for abortion, and educated women and evangelicals have a significantly (p < .01) lower demand for abortion. Income is positive (p < .01). Increases in the income of women increases their opportunity cost of childbearing (potential income foregone) which increases their demand for abortions.
Table 5

2SLS estimates of abortion demand of women (ages 15–44)

Independent variables

Dependent variable

Number of abortions/1,000 pregnancies, women (ages 15–44)

(1)

(2)

(3)

(4)

Constant

114.784 (.99)

141.510 (1.15)

168.817 (1.31)

183.022 (1.35)

Abortion price

−1.101 (2.65)***

−1.294 (2.66)***

−1.306 (2.75)***

−1.411 (2.67)***

Female income

.022 (5.48)***

.023 (5.11)***

.023 (5.29)***

.024 (4.91)***

Labor force participation

5.656 (3.88)***

5.499 (3.55)***

5.102 (3.19)***

5.004 (3.01)***

Single

.963 (.48)

.216 (.09)

.377 (.17)

−.119 (.05)

Education

−3.163 (2.06)**

−2.656 (1.61)*

−2.646 (1.61)*

−2.333 (1.68)*

Evangelical Christians

−1.194 (2.99)***

−1.163 (2.70)***

−1.239 (2.88)***

−1.211 (2.67)***

State Medicaid funding

33.495 (2.52)**

36.535 (2.48)**

35.665 (2.52)**

37.665 (2.44)**

Parental involvement law

−28.355 (2.61)**

−24.211 (2.04)**

Parental border involvement

.282 (1.40)

.322 (1.45)

.373 (1.67)*

.403 (1.68)*

Parental consent law

−25.944 (1.84)*

−20.401 (1.63)*

Parental notification law

−31.745 (2.07)**

−28.461 (1.79)*

Waiting period law

1.036 (.03)

1.697 (.05)

Waiting period border states

−.476 (.99)

−.510 (1.01)

F-statistic

26.15***

19.98***

18.69***

15.42***

Note: Number of observations = 150. All regressions include year fixed effects

Absolute value of t-statistics in parentheses:

* Estimated coefficient statistically significant at .10 level

** Estimated coefficient statistically significant at .05 level

*** Estimated coefficient statistically significant at .01 level

McCloskey and Ziliak (1996) argue that it is more useful to describe the impact of a regression coefficient of an economic variable in terms of its numerical impact on the dependent variable (i.e., its elasticity of demand). An elasticity of demand is a numerical measure of the responsiveness of the dependent variable in Eq. (1)—the number of abortions per 1,000 pregnancies of women of childbearing ages—resulting from a 1% change in the value of an independent economic variable (e.g., price or income). An elasticity is calculated at the mean of the dependent and independent variable. For a linear relationship, such as the abortion demand Eq. (1), the elasticity (Ei ) of an independent variable Xi equals (Gujarati 1995, p. 168): Ei = bi • (mean of the independent variable Xi/mean of the demand for abortion ), where bi is the estimated regression coefficient of the independent variable Xi that appears in Table 5, column 1. For the independent variable female income the income elasticity of demand is equal to 1.24. This figure is consistent with the range of prior income elasticity of demand values that appears in Table 1.

The price of an abortion is negative (p < .01). The fundamental law of demand is applicable to abortion. The price elasticity of demand is −1.07. This value indicates that the demand for abortion is more responsive to price over an extended period of time than the figures reported in previous cross-section studies listed in Table 1. The larger figure for the price elasticity of demand over time weakly supports the hypothesis that the cost of an abortion may alter women’s decisions regarding ex ante contraceptive usage that affects the likelihood of becoming pregnant (Kane and Staiger 1996; Levine 2004). This hypothesis argues that an increase in the cost of an abortion raises the cost of engaging in noncontracepted sexual activity relative to contracepted sexual activity and induces women, over time, to adopt alternative birth control methods (other than abortion) resulting in a decrease in both the number of pregnancies and the subsequent number of abortions. Thus if abortion costs alter women’s pregnancy avoidance behavior, then the impact of an increase in the price of an abortion should have a numerically greater impact on abortion demand over time (i.e., a larger price elasticity of demand value).

The State Medicaid Funding variable is positive and significantly (p < .05) different from zero. State Medicaid funding increases the number of abortions by 33.5 per 1,000 pregnancies of women of childbearing age (or equivalently an increase of 3.3 percentage points in the average state’s abortion rate from nearly 22.5 to 25.8% of all pregnancies ending in abortion).

The parental involvement law is negative and significantly (p < .05) different from zero. The numerical impact of a state parental involvement law is to reduce the abortion rate by 28 abortions per 1,000 pregnancies of women of childbearing age. There are two types of parental involvement laws. Parental notification laws require that a parent be notified of a minor’s intent to have an abortion. However, a parent may not prevent the minor from obtaining an abortion. Parental consent laws require informed consent of at least one parent usually 24–48 h before the abortion procedure. Parental consent laws give parents the right to deny a minor an abortion and are in theory more restrictive than parental notification laws. This suggests that there may be empirical differences between the two laws with parental consent laws expected to have a stronger negative impact on abortion demand.

In order to test for the possibility that there are empirical differences between the two types of parental involvement laws, the Parental Involvement Law variable was disaggregated into two separate variables: Parental Consent Law = 1 if state, s, had a parental consent law in effect during time period, t, and Parental Notification Law = 1 if state, s, had a parental notification law in effect during time period, t. The empirical results appear in Table 5, column 2. Both the parental consent law and the parental notification law are negative and significantly different from zero. The null hypothesis of equality of coefficients between the parental consent law and parental notification law cannot be rejected. Even though parental consent laws are more restrictive than parental notification laws, empirically there are no statistically significant differences on their negative impact on abortion demand.

Waiting period laws are intended to dissuade women from having an abortion by increasing the time, emotional, or other indirect costs of obtaining an abortion. To ascertain the effect of waiting period laws on abortion demand, the abortion demand equation was reestimated with the addition of the variable Waiting Period Law equal to 1 if state, s, had a waiting period law in effect during time period, t. However, women can circumvent complying with a state’s waiting period law by traveling to a nearby border state without a waiting period to obtain an abortion. To account for travel across state lines the variable Waiting Period Border States, which is a weighted average of the percentage of states that border to state, s, that also have a waiting period law, is added to the abortion demand equation. The value of the Waiting Period Border States variable ranges from 0 (no border states have a waiting period law) to 100% (all border states have a waiting period law).

The empirical results appear in Table 5, column 3 when the Parental Involvement Law variable is included in the abortion demand equation and column 4 when the disaggregated Parental Consent Law and Parental Notification Law variables are included. In both specifications neither the Waiting Period Law variable nor the Waiting Period Border States variable is significantly different from zero. Also, the coefficients of the other variables that appear in Table 5, column 3 and in column 4 are virtually identical to the corresponding estimates in column 1 and in column 2, respectively (the null hypothesis of equality of coefficients between the estimates appearing in column 1 and column 3 and between the estimates in column 2 and column 4 cannot be rejected). The empirical results suggest that state waiting period laws do not have a statistically significant effect on the number of abortions performed or the location of the procedure. Columns 3 and 4 of Table 5 also find that the parental border involvement variable has a positive, but numerically small impact on abortion demand. In-state abortion rates increase slightly when parental involvement laws are enacted by border states. This result suggests that one effect of border states enforcing parental involvement laws may be to induce minors to involve a parent in their decision to have an abortion (Henshaw and Kost 1992).

A common contention in the literature is that when estimating the abortion demand it is necessary to use state fixed effects variables in order to control for unobservable time-invariant factors across states (Blank et al. 1996). Specifically, the unobservable factor is thought to be differences in antiabortion sentiment across states (the determinants of antiabortion sentiment are discussed in Deitch 1983; Jenks 1985; Leahy et al. 1983). This raises the question: Do differences in antiabortion sentiment across states affect the estimates reported in Table 5 for the price and restrictive abortion law variables? In order to answer this question a direct measure of antiabortion sentiment within a state is needed.

The Council of State Governments (1990) compiled an ordinal index which lists how many of 24 specific types of antiabortion laws, regulations, or policies such as reporting and record-keeping requirements, ban on all abortions, conscience clauses, fetal protection laws, spousal notification laws, insurance restrictions, gender selection laws, and viability test laws were enacted by a state’s legislature since 1973. The passage of these 24 antiabortion laws within each state did not all occur during one particular legislative session. Their passage was a function of the time-varying relative political strength and influence of groups who were abortion rights supporters and opponents (Cohen and Barrilleaux 1993). The ordinal State Antiabortion Sentiment variable is the cumulative total number of the 24 antiabortion laws or regulations that were enacted by each state’s legislature through time period, t. The State Antiabortion Sentiment variable is calculated for each state separately for the years 1982, 1992, and 2000. The value of the State Antiabortion Sentiment variable ranges from 0 (no antiabortion laws or regulations passed by a state’s legislature) to 24 (all 24 of the possible antiabortion laws or regulations were passed by a state’s legislature). The greater the value, the more antiabortion sentiment within a state (the correlation between the State Antiabortion Sentiment variable and each of the restrictive abortion laws: State Medicaid Funding, Parental Involvement Law, Waiting Period Law, Parental Consent Law, Parental Notification Law is −.38, .34, .12, .31, .09, respectively).

The empirical results when the State Antiabortion Sentiment variable is included in the estimation of the abortion demand equation appear in Table 6, column 1 (with Parental Involvement Law), column 2 (with Parental Consent Law and Parental Notification Law), column 3 (with Parental Involvement Law and Waiting Period), and column 4 (with Parental Consent Law, Parental Notification Law, and Waiting Period). The empirical results show that for all four specifications that appear in Table 6 the State Antiabortion Sentiment variable is not significantly different from zero, and the estimated coefficients of all the other variables in the abortion demand equation are virtually identical to their respective counterparts that appear in Table 5 (null hypothesis of equality of coefficients cannot be rejected).
Table 6

2SLS estimates of abortion demand of women (ages 15–44) with antiabortion sentiment

Independent variables

Dependent variable

Number of abortions/1,000 pregnancies, women (ages 15–44)

(1)

(2)

(3)

(4)

Constant

127.905 (1.08)

136.082 (1.11)

180.272 (1.36)

192.779 (1.39)

Abortion price

−1.188 (2.72)***

−1.254 (2.62)***

−1.374 (2.88)***

−1.471 (2.76)***

Female income

.023 (5.37)***

.023 (5.03)***

.023 (5.16)***

.024 (4.82)***

Labor force participation

5.589 (3.75)***

5.543 (3.62)***

5.460 (3.10)***

4.960 (2.93)***

Single

.657 (.31)

.337 (.14)

.115 (.05)

−.382 (.14)

Education

−2.964 (1.93)*

−2.771 (1.66)*

−2.503 (1.61)*

−2.203 (1.60)*

Evangelical Christians

−1.194 (2.87)***

−1.176 (2.72)***

−1.249 (2.77)***

−1.218 (2.59)**

State Medicaid funding

34.874 (2.39)**

36.218 (2.33)**

36.998 (2.39)**

39.043 (2.34)**

Parental involvement law

−28.605 (2.52)**

−24.448 (1.97)*

Parental border involvement

.298 (1.43)

.315 (1.42)

.388 (1.69)*

.418 (1.69)*

Parental consent law

−26.505 (1.91)*

−20.189 (1.61)*

Parental notification law

−31.324 (2.04)**

−29.317 (1.77)*

Waiting period law

1.826 (.05)

2.464 (.07)

Waiting period border states

−.502 (1.01)

−.533 (1.03)

State antiabortion sentiment

.077 (.06)

.107 (.08)

.160 (.11)

.184 (.13)

F-statistic

22.12***

19.06***

16.03***

13.32***

Note: Number of observations = 150. All regressions include year fixed effects

Absolute value of t-statistics in parentheses:

* Estimated coefficient statistically significant at .10 level

** Estimated coefficient statistically significant at .05 level

*** Estimated coefficient statistically significant at .01 level

Robustness of the Abortion Cost Results

Teenagers have a higher incidence of unplanned pregnancies than any other age group. In 1998, approximately 10% of teenagers became pregnant. An estimated 35% of those pregnancies resulted in an abortion (Henshaw and Feivelson 2000). One way to test whether the previous estimated effects of restrictive abortion laws on the demand for abortions are spurious is to examine their effects on the abortion rates of minors (ages 15–17). If the previous empirical results are measuring the actual causal effects of abortion costs then the impact of the restrictive abortion laws on minors’ abortion demand should be greater than their impact on all women of childbearing age.

The dependent variable in the abortion demand equation is the abortion rate of minors, the number of abortions per 1,000 pregnancies of teen minors (ages 15–17), in state, s, during time period, t. Tomal (1999) argued that minors used the opportunity costs of older women in their assessment of the probable consequences of their pregnancy resolution decision. This holds whether the pregnant teen makes her own decision, has the decision forced upon her by her parents, or makes the decision in consultation with other adults. Accordingly, the income, labor market, and marital variables used for women of childbearing age that appear in Eq. (1) are used as measures of the opportunity cost of a pregnancy to a minor. The empirical results of the abortion demand for minors appear in Table 7, column 1 (when Parental Involvement Law is included), column 2 (when Parental Consent and Parental Notification Law are included), column 3 (when Parental Involvement and Waiting Period Law are included), and column 4 (when Parental Consent, Parental Notification, and Waiting Period Law are included).
Table 7

2SLS estimates of abortion demand of minors (ages 15–17)

Independent variables

Dependent variable

Number of abortions/1,000 pregnancies, minors (ages 15–17)

(1)

(2)

(3)

(4)

Constant

−430.905 (2.87)***

−357.651 (2.17)**

−366.586 (2.24)**

−337.741 (1.96)**

Abortion price

−.251 (1.61)*

−.809 (1.71)*

−.521 (1.65)*

−.756 (1.62)*

Female income

.015 (2.76)***

.018 (2.95)***

.015 (2.77)***

.017 (2.78)***

Labor force participation

7.947 (4.20)***

7.513 (3.61)***

7.283 (3.59)***

7.077 (3.33)***

Single

4.664 (1.52)

2.307 (.62)

3.887 (1.45)

2.629 (.83)

Education

2.394 (1.20)

3.935 (1.69)*

3.123 (1.76)*

3.881 (1.67)*

Evangelical Christians

−2.396 (4.61)***

−2.271 (3.91)***

−2.408 (4.41)***

−2.318 (4.01)***

State Medicaid funding

61.069 (3.52)***

70.622 (3.55)***

63.825 (3.55)***

68.874 (3.50)***

Parental involvement law

−52.603 (3.74)***

−46.287 (3.08)***

Parental border involvement

.221 (.84)

.347 (1.16)

.339 (1.19)

.414 (1.34)

Parental consent law

−40.887 (2.17)**

−34.288 (1.75)*

Parental notification law

−67.838 (3.29)***

−59.798 (2.95)***

Waiting period law

−15.822 (.42)

−14.429 (.36)

Waiting period border states

−.401 (.65)

−.474 (.74)

F-statistic

44.97***

32.21***

35.05***

29.06***

Note: Number of observations = 150. All regressions include year fixed effects

Absolute value of t-statistics in parentheses:

* Estimated coefficient statistically significant at .10 level

** Estimated coefficient statistically significant at .05 level

*** Estimated coefficient statistically significant at .01 level

State Medicaid funding of abortions significantly (p < .05) increases the demand for abortions by minors in a state by between 61 and 70 per 1,000 pregnancies of minors (or equivalently an increase of between 6 and 7 percentage points in a state’s abortion rate of minors as compared to states which do not fund Medicaid abortions). The null hypothesis that the State Medicaid Funding coefficient for minors is equal to or less than the coefficient for all women of childbearing age is rejected at the .05 level of significance. State Medicaid funding of abortions has a significantly greater positive impact on the abortion demand for minors than on all women of childbearing age.

The presence of a parental involvement law reduces the abortion demand of minors by between 46 and 52 abortions per 1,000 pregnancies (or equivalently a decrease of between 4.6 and 5.2 percentage points in a state’s abortion rate of minors compared to states without a parental involvement law). The numerical impact of a parental involvement law on the abortion demand for minors is also found to be significantly (p < .05) greater than the impact of the law on the abortion demand for all women of childbearing age. When the parental involvement law is disaggregated, both the parental consent law and the parental notification law are significantly related to lower abortion demand by minors, and both types of parental involvement laws have a significantly (p < .05) greater impact on reducing minors’ abortion demand than these two types of parental involvement laws do for all women of childbearing age. Regardless of the specification, a waiting period law has no statistically significant impact on the demand for abortions by minors.

The empirical results in Table 7 are consistent with our ex ante expectations about the relative size of the effects of indirect abortion costs on the demand for abortions by minors and by all women of childbearing age. Compared to the demand for abortion by all women of childbearing age, state Medicaid funding, parental involvement laws, and parental consent and parental notification laws are found to have a significantly larger numerical impact on the demand for abortion by teen minors. The numerical estimates of the impact of a parental involvement law and Medicaid abortion funding on minor’s demand for abortion are lower than prior studies (see Tables 2, 3) which may be attributed to the inclusion of the statistically and numerically significant negative impact of a state’s abortion price on minor’s demand for abortion.

The abortion demand Eq. (1) was also estimated for non-minor teens (ages 18–19). Since non-minor teens are not subject to parental involvement laws, they serve as a control group to determine if the estimated impact of parental involvement laws reported in Table 7 is causal. The empirical results (which are available upon request) show that parental involvement laws have no statistically significant impact on the abortion demand of non-minor teens. When the parental involvement law variable is disaggregated the empirical results find that a parental consent law has no statistically significant impact on the abortion demand of non-minor teens, but a parental notification law is negatively related (p < .05) to the abortion demand of non-minor teens. This latter result is consistent with Tomal’s (1999) finding that parental notification laws may have caused a change in the sexual behavior of minor teens which is perpetuated as they age.

Conclusion

This study estimates the abortion demand of all women of childbearing age using pooled time-series cross-section state data from three Census years. This study differs from much of the previous research in that it includes the price of an abortion in each state for each of the three sample years. The empirical results find that state Medicaid funding of abortion increases the number of abortions by between 33 and 37 per 1,000 pregnancies of women of childbearing ages compared to states that do not fund Medicaid abortions. These figures are consistent with prior research (see Table 2) by Levine et al. (1996) and Matthews et al. (1997) who found that states which fund Medicaid abortions had a 3–5% higher abortion rate than those states which restrict the funding of Medicaid abortions. The presence of a state parental involvement law decreases the number of abortions by between 24 and 28 per 1,000 pregnancies of women of childbearing age compared to states without a parental involvement law. These figures are less than the estimates by Matthews et al. (1997) and Bitler and Zavodny (2001) who found that states with a parental involvement law had a 3–6% lower abortion demand by women of childbearing age than states which do not have a parental involvement law (see Table 3). Since neither study included the price of an abortion in their estimation, their figures for the estimated impact of a state parental involvement law may include the effects of the omitted abortion price. The empirical results also show that there are no significant numerical differences between the impact of parental consent and parental notification laws on abortion demand. Waiting period laws are found to have no statistically significant impact on abortion demand.

Teenagers have a higher rate of unplanned pregnancies and abortions than any other age group. One confirmation of the causal effects reported in this paper would be the a priori expectation that restrictive laws should have a greater impact on teen minors’ abortion rates than on the abortion rates of all women of childbearing age. The empirical results support this prediction. State Medicaid funding, parental involvement laws, parental consent and parental notification laws are all found to have a significantly larger numerical impact on teen minors’ demand for abortion than on demand of all women of childbearing age.

Copyright information

© Springer Science+Business Media, LLC 2007