DoesParental Consent for Birth Control affect Underage Pregnancy Rates? The case of Texas
November 2011
Preliminary version: comments are most welcome, but please do not cite without permission.
Sourafel Girma
School of Economics
Sir Clive Granger Building
Nottingham University
University Park
Nottingham
NG7 2RD
United Kingdom
Tel: + 00 44 115 951 5482
Email:
and
David Paton*
Nottingham University Business School
Jubilee Campus
Wollaton Road
Nottingham
NG8 1BB
United Kingdom
Tel: + 00 44 115 846 6601
Email:
*Corresponding author
Acknowledgements:
We are very grateful to Janice Jackson at the Texas Department of State Health Services for her assistance with the birth and abortion data. We thank also participants at a staff seminar at the University of Surrey for useful comments.
DoesParental Consent for Birth Control affect Underage Pregnancy Rates? The case of Texas
Abstract
In 2003, Texas became the second state to mandate parental consent for the provision of state-funded contraception to minors. Previous work based on conjectural responses of minors, predicted that the ruling would lead to a large increase in the Texas underage pregnancy rate. In this paper we examine the actual impact of the parental consent mandate on the under-18 pregnancy rate using both state- and county-level data and using 18-year olds as a control group. The county-level data allows us to use a statistical matching estimator to compare the effect of parental consent in counties with and without family planning clinics that are subject to the ruling. We control for both observable and unobservable characteristics that may be systematically correlated with the presence of affected clinics by combiningdifference-in-differences and difference-in-difference-in-differences strategies with propensity-score weighted regressions. The estimates provide little evidence that requiring parental consent for state-funded contraception led to an increase in the underage pregnancy rate.
Keywords:propensity score weights; parental consent; family planning; teenage pregnancy; abortion.
JEL Classifications: C21, I18, J13.
DoesParental Consent for Birth Control affect Underage Pregnancy Rates? The case of Texas
1. Introduction
In this paper we use treatment effects evaluation estimators to test whether mandatory parental consent for the provision of state-funded birth control to minors leads to an increase in underage pregnancies. Policy discussions on the appropriate level of parental involvement in sexual health decisions of minors generate a huge amount of controversy. The underlying ethical debate rests on the balance between the rights of minors to make autonomous healthcare decisions and the rights of parents and carers to be involved in decisions affecting those for whom they have legal responsibility. In practice, policy arguments often hinge on the likely impact of mandatory parental involvement on the sexual health of minors and, in particular, rates of underage pregnancy.
Legislation to mandate parental involvement for birth control in the U.S. has been considered recently both at the federal level and by numerous state legislatures. Other countries such as England, the Republic of Ireland and the Philippines are in the process of debating proposals either to abolish existing parental consent provisions for contraception or to put new ones in place. Despite the level of activity in the policy arena, the academic literature has paid very little attention indeed to this issue. This is in stark contrast to the considerable body of evidence, much of it conducted by economists, relating to parental involvement for abortion.
In January 2003, Texas became the second U.S. jurisdiction to mandate parental consent for the provision of state-funded contraception to teenagers below the age of 18.[1] Franzini et al (2004) use survey data on conjectural responses of teenagers to changes in confidentiality along with data on failure rates of contraception to estimate the likely effect of the Texas regulations. They projected that the law would lead to more than 5,000 additional births and 1,600 additional abortions per year amongst minors in the State. This implies an increase in the underage pregnancy (defined as abortions plus births) rate of about 20%, with an estimated direct medical cost of about $44 million.
Whether or not these projections translate into actual additional pregnancies and costs depends on the actual, rather than behavioral, response of minors. To date no research has been undertaken to estimate the actual effect of the Texas parental consent law on pregnancy rates amongst minors and it is this gap which we are trying to fill in this paper. The results are likely to be of interest not only to policymakers in Texas but also to those in other states and countries which may be considering the introduction of similar laws. Further, the results should provide insights into whether adolescents respond to new incentives induced by regulatory change in ways predicted by standard neo-classical models.
The empirical approach taken in the paper is first to undertake some preliminary analysis to explore whether the parental consent law for family planning actually affected the take-up of contraception amongst adolescents. We then test for a state-wide effect of parental consent on pregnancy by comparing changes in rates amongst minors before and after the rulingrelative to changes amongst older teenagers.
Establishing a causal effect at the statelevel is potentially complicated by state-wide changes to minor pregnancy rates caused by other contemporaneous factors. For example, in September 2005, Texas strengthened its parental involvement law for abortions on minors, mandating parental consent rather than just notification. To control for such effects, we exploit the fact that parental consent for contraception is only required for state-funded, rather than federally-funded, contraception. That means that family planning clinics receiving funding under the (federal) Title X scheme, should be relatively unaffected by the ruling, whilst non-Title X funded clinics will be affected. By identifying those counties with such affected clinics, we are able to use statistical matching methods to help isolate any causal effect of the parental consent mandate. Weare careful to control for both observable and unobservable characteristics that may be systematically correlated with the presence of affected (i.e. non-Title X) clinics. We do so by combining, difference-in-differences and difference-in-difference-in-differences strategies with propensity-score weighted regressions.
2. Parental Consent for contraception – the evidence to date
Following the seminal work of Becker (1963), the economic literature has offered several theoretical models in which teenagers make decisions regarding sexual activity, contraceptive use and pregnancy resolution based on their subjective evaluation of expected costs and benefits of uncertain outcomes. Akerlof, Yellen and Katz (1996) and Paton (2002) argue that easier access to contraception may lower the perceived costs of underage or extra-marital sexual activity and, as a result, can have an ambiguous effect on unwanted pregnancy or abortion rates. Similar models have been presented in the context of sex education (Oettinger, 1999) and abortion (Levine, 2003, 2004). More recently, Arcidiacono, Khwaja and Ouyang (2011) develop an inter-temporal model in which “habit persistence” is a feature of teen sexual behavior. Once a relationship progresses to sexual intercourse, the authors argue that it becomes harder for the couple to switch back to abstinence. The prediction, confirmed in their empirical analysis, is that more costly access to contraception leads to an increase in teen pregnancies in the short run, but to a decrease in the long run.
In the context of sucheconomic models, the predicted effects of a parental consent requirement for contraceptives on underage pregnancies are unclear. Such policies might discourage some young women from having sexual intercourse, which should lead to a reduction in pregnancies among minors. However, pregnancies could increase if some young women who would have used prescription contraceptives absent the parental consent policy have intercourse anyway but use less effective methods of birth control or no contraception at all. Pregnancies would not be affected if minors, who would have used prescription birth control absent the policy,either obtain parental consent after the policy change or, instead,manage to access supplies from other providers.
A number of studies have used surveys to obtain conjectural responses from teenagers on their likely response to hypothetical parental involvement requirements. These studies suggest that parental consent laws might indeed affect both teens’ sexual activity and contraceptive use. In a survey conducted at Planned Parenthood clinics in Wisconsin, 59% of minors said they would discontinue use of contraceptive services if parental involvement were required (Reddy, Fleming and Swain, 2002). In surveys conducted at publicly funded family planning clinics across the U.S., 40% of young women said they would not go to publicly funded clinics that required parental consent for prescription birth control (Jones et al., 2005). Some minors said they would switch to condoms or other non-prescription forms of birth control. About 6% said they would continue to have intercourse but not use any contraceptives. About 7% said they would stop having intercourse, although only 1% indicated this as their only likely response. However, 60% of minors in one study (Jones et al., 2005) and 45% of teens in another (Harper et al., 2004) said a parent was already aware they obtained sexual health services at a clinic, suggesting that many minors might not change their behavior in response to a parental consent policy.
The results of these conjectural responses form the basis for the predictions by Franzini et al (2004) that the parental control mandate in Texas would lead to a large increase in underage pregnancies. However, it is difficult to assess the actual effect of a parental involvement requirement from surveys about hypothetical policy changes. Indeed, previous studies of the impact of more general changes to availability of family planning to minors are at best inconclusive (Girma and Paton, 2006, 2011; Kearney and Levine, 2009; Raymond et al., 2007; Evans, Oates and Schwab, 1992).
In the first place, it may be that teen’s actual responses will be very different to the hypothetical responses declared to researchers. Second, the samples for such studies tend to be drawn from teens attending birth control clinics and who are likely already to be sexually active. The work by Arcidiacono et al (2011) suggests that the response may be very different amongst young people who are not yet sexually active but who will consider such a transition in the future.
Very few studies have examined the actual impact of limitations on the confidentiality for contraception on pregnancy rates (rather than just on the uptake of services). Paton (2002) examines the impact of the Fraser ruling which meant that for most of 1985, family planning could not be provided to underage girls without parental involvement in England and Wales. Take-up at family planning clinics amongst this age group dropped by about 30% in 1985, yet the underage conception rate in England decreased slightly relative to the rate amongst older teenagers. Similarly, the rate also did not increase relative to the underage conception rate in Scotland where the Fraser ruling did not apply.
A previous study of the imposition of parental consent for family planning in a single county (McHenry) in Illinois found that the requirement led to an increase in the percentage of births to women under age 19 living in McHenry County relative to three other nearby counties but not to an increase in the percentage of abortions or pregnancies occurring among teens (Zavodny, 2004; Zavodny, 2005). To date no studies at all have examined the impact of the parental involvement restrictions for family planning at a state-wide level.
There does exist a much larger literature on the related issue of parental involvement for abortion. Methodological difficulties, such as controlling adequately for abortions carried out in neighboringstates without such a law, make it hard to draw firm inferences from the data, but the majority of studies to date find that parental involvement laws have significant effects on teenage fertility and sexual behavior. For example, relative decreases in underage abortion rates are observed by New (2011), Joyce, Kaestner and Colman (2006) and Levine (2003) whilst the latter also concludes that these laws lead to reduction in total pregnancy rates amongst minors. Klick and Stratmann (2008) use rates of sexually transmitted infections (STIs) as a proxy for risky sexual activity and find that states implementing parental involvement laws experienced relative reductions in some STIs amongst minors. There is some evidence that parental involvement laws have differential effects for older and younger teenagers. For example, Colman, Joyce and Kaestner (2008) provide evidence that the 2000 law mandating parental notification before abortions on minors in Texas decreased both abortions and births amongst girls aged 17 at the time of the birth or abortion. However, amongst a slightly older cohort (girls aged 17 at the time of conception) abortions decreased by a lower amount whilst births increased (albeit by a statistically insignificant amount).
3. Methodology
3.1 State-level Estimates
Based on aggregate state level data, we employ logit regression for grouped data to determine the impacts of the parental consent mandate on underage pregnancy rates. We measure the impacts as changes from 2002 (the year before the mandate) to the three years after the mandate for two treatment groups: under-18s (U18) and under-16s (U18), relative to the 18 years olds (the control group).[2] The Texas parental consent mandate applies to those aged under-18. There are several reasons for selecting the second treatment group (under-16s). In the first place, previous research (Colman et al, 2008) suggests that the impact of parental consent laws may differ between older and younger teens. One reason for this is that parental consent may be more strongly binding to younger teens who are less likelyto be able to travel independently to access alternative sources. Further, policy makers tend to be relatively more concerned about pregnancies amongst younger teens than amongst those approaching the age of majority. Finally, some older minors (for example those that are married) are considered ‘emancipated’ in Texas state law and are exempt from the parental consent ruling.
Since our estimation methodology involves “before” and “after” periods for the control and treatment groups, the relevant estimates can be interpreted as difference-in-differences estimates. The estimating equation in this analysis can be expressed as:
(1)
where p is the number of pregnancies (abortions plus births) divided by the population for the relevant age group at the estimated time of conception, i represents the three age groups; t=2002,….,2005, and Du16 and Du18 denote group dummies for under-16s and under-18s respectively. We repeat the analysis for the whole sample and for the sample of Non-Hispanic Whites, Non-Hispanic Blacks and Hispanics separately.
3.2 County Level Estimates
The availability of county-level data allows us to estimate the average treatment effect of the binary treatment, parental consent for family planning, on the scalar outcome variable, county-level pregnancy rate. Our identification strategy exploits the fact that parental consent for contraception is only required for state-funded, rather than federally-funded, contraception. That means that family planning clinics receiving funding under the Title X scheme, should be relatively unaffected by the ruling compared to non-Title X funded clinics. In our empirical model we are careful to control for two potential sources of bias: (i) unobservable characteristics that are correlated with both the treatment and outcome; and (ii) the possibility that assignment to the treatment is systematically correlated with observable pre-treatment characteristics, or the self-selection problem.
Following the seminal contribution of Rosenbaum and Rubin (1983), one can remove estimation bias by using propensity score matching, where the propensity score is defined as the probability of assignment to the treatment conditional on pre-treatment characteristics, say Z. More recent work has shown that greater efficiency relative to propensity score matching can be obtained if one transforms the propensity score estimates into weights in the relevant regressions, the so-called the propensity score reweighted regression (Hirano et al, 2003 and Busso et al, 2009). In this approach, treatment observations receive a weight of , whereas control observations receive a weight of .
It is worth noting that controlling for self-selection by adjusting for the propensity score would not remove all biases as long as there are unobservable characteristics affecting the potential outcomes and the decision to select into treatment. Fortunately, the influence of time-invariant unobservable characteristics can be controlled by using country-level fixed effects in the regressions, which is equivalent to using the difference-in-differences (DD) estimator in a two-period (pre- and post-treatment) panel data model. Hence our baseline econometric model consists of combining the DD approach with propensity score reweighted regression. The steps involved in this approach can be summarized as follows:
(i)Estimate the propensity score via a probit model of the probability treatment in 2003 as a function of pre-treatment (2002) characteristics:. The characteristics used to obtain the propensity score are as follows:the percentage of the population living in poverty (poverty); the two-year lagged pregnancy rate (lag pregnancies); an indicator variable for whether or not the county is urban in nature (urban); an indicator variable for whether or not the county borders either Mexico or another U.S. state (border); the proportion of the relevant population that are Hispanic (Hispanic); the proportion of the population that are Black (Black); the number of Title X (i.e. unaffected) clinics present in the county as of 2001 (Title X (2001)). Based on the estimated propensity scores, we only keep observations on the common support by deleting control observations whose propensity score is smaller (higher) than the minimum (maximum) propensity score of the treatment group.