View in EDS HTML Full Text PDF Full Text

Three Strategies to Prevent Unintended Pregnancy

Saved in:
Bibliographic Details
Title: Three Strategies to Prevent Unintended Pregnancy
Language: English
Authors: Thomas, Adam
Source: Journal of Policy Analysis and Management. Spr 2012 31(2):280-311.
Availability: Wiley-Blackwell. 350 Main Street, Malden, MA 02148. Tel: 800-835-6770; Tel: 781-388-8598; Fax: 781-388-8232; e-mail: cs-journals@wiley.com; Web site: http://www.wiley.com/WileyCDA
Peer Reviewed: Y
Physical Description: PDF
Page Count: 32
Publication Date: 2012
Document Type: Journal Articles
Reports - Research
Education Level: Secondary Education
Descriptors: Medical Services, Family Planning, Prevention, Sexually Transmitted Diseases, Pregnancy, Change Strategies, At Risk Students, Simulation, Public Policy, Program Descriptions, Cost Effectiveness, Contraception, Sex Education, Youth Opportunities
DOI: 10.1002/pam.21614
ISSN: 0276-8739
Abstract: This paper presents results from fiscal impact simulations of three national-level policies designed to prevent unintended pregnancy: A media campaign encouraging condom use, a pregnancy prevention program for at-risk youth, and an expansion in Medicaid family planning services. These simulations were performed using FamilyScape, a recently developed agent-based simulation model of family formation. In some simulation specifications, policies' benefits are monetized by accounting for projected reductions in government expenditures on medical care for pregnant women and infants. In a majority of these specifications, policies' fiscal benefit-cost ratios are less than 1. However, in specifications that account additionally for projected savings to programs that provide a broader range of benefits and services to young children, all three policies have benefit-cost ratios that are comfortably greater than 1. The results from my preferred specifications suggest that the simulated policies would produce returns to taxpayers on each dollar spent of between $2 to $6. On the whole, the results of these simulations imply that all three policies are sound public investments. (Contains 50 footnotes, 5 tables, and 1 figure.)
Abstractor: As Provided
Number of References: 69
Entry Date: 2012
Accession Number: EJ969569
Database: ERIC
Full text is not displayed to guests.
FullText Links:
  – Type: pdflink
    Url: https://content.ebscohost.com/cds/retrieve?content=AQICAHj0k_4E0hTGH8RJwT4gCJyBsGNe_WN95AvKlDbXJGqwxwEnO9lezFbUBd3Apo1wmdJsAAAA5zCB5AYJKoZIhvcNAQcGoIHWMIHTAgEAMIHNBgkqhkiG9w0BBwEwHgYJYIZIAWUDBAEuMBEEDCoDT2vV6OczcuLe5wIBEICBn3TCrg1XtSfgccbg3f9eeRxY_Q7VMnFLccyo2t1cADhBY0SfsB57SFlWVrK3mzfcH0yIAq8vkqPZQA4NB0nSCbNqwVvsHQTISzhRRymS7933gV5WJK-eQgFaYX2ZJxvkGLdyqsA7RFhuMVY7X4EeCZtkr2cu1DXUW9cN-wci1uWng7uMBHuzoIA7NpGt6mtJ-DR1VXyPO0gdcJvWQjm4nw==
Text:
  Availability: 1
  Value: <anid>AN0076170450;jpa01apr.12;2019Jun03.10:32;v2.2.500</anid> <title id="AN0076170450-1">Three Strategies to Prevent Unintended Pregnancy. </title> <p>This paper presents results from fiscal impact simulations of three national‐level policies designed to prevent unintended pregnancy: A media campaign encouraging condom use, a pregnancy prevention program for at‐risk youth, and an expansion in Medicaid family planning services. These simulations were performed using FamilyScape, a recently developed agent‐based simulation model of family formation. In some simulation specifications, policies' benefits are monetized by accounting for projected reductions in government expenditures on medical care for pregnant women and infants. In a majority of these specifications, policies' fiscal benefit‐cost ratios are less than 1. However, in specifications that account additionally for projected savings to programs that provide a broader range of benefits and services to young children, all three policies have benefit‐cost ratios that are comfortably greater than 1. The results from my preferred specifications suggest that the simulated policies would produce returns to taxpayers on each dollar spent of between $2 to $6. On the whole, the results of these simulations imply that all three policies are sound public investments.</p> <p>Unintended pregnancy has garnered considerable interest among social scientists and policymakers in recent years. Roughly half of all pregnancies in the United States are unintended, and such pregnancies have been found to be associated with a host of negative outcomes for women and children.[<reflink idref="bib1" id="ref1">1</reflink>] For example, women who experience unintended pregnancies are less likely than women who experience intended pregnancies to be married, are more likely to endure physical abuse, have a higher incidence of mental health problems, and are more likely to delay the initiation of prenatal care. Moreover, children whose births resulted from unintended pregnancies have been found to be less likely than other children to be breastfed, to be at greater risk of experiencing negative physical and mental health outcomes, and to be more likely to engage in delinquent behavior during their teenage years.[<reflink idref="bib2" id="ref2">2</reflink>]</p> <p>Much of the evidence on these relationships is correlational. However, some studies have exploited plausibly exogenous variation over time in state laws governing access to birth control (which lowers women's risk of experiencing an unintended pregnancy) and abortion (which lowers women's risk of experiencing an unintended birth) in an attempt to identify the causal relationship between pregnancy and childbearing intentions and maternal and child outcomes. For instance, Ananat and Hungerman ([<reflink idref="bib2" id="ref3">2</reflink>]), Goldin and Katz ([<reflink idref="bib18" id="ref4">18</reflink>]), and Bailey ([<reflink idref="bib5" id="ref5">5</reflink>]) use variation across states and over time in laws governing access to oral contraception and in female education and labor market outcomes in order to estimate the effects of the former on the latter. These studies' findings suggest that expansions in access to the pill have raised women's educational attainment (in the case of the first two studies) and increased their rates of labor force participation (in the case of the latter study).</p> <p>Other research focuses on child outcomes. For example, in a paper in which she instruments for mothers' age at first birth using information on biological fertility shocks such as the occurrence of miscarriages, Miller ([<reflink idref="bib42" id="ref6">42</reflink>]) finds that delayed childbearing leads to higher standardized test scores for firstborn children. In addition, Ananat et al. ([<reflink idref="bib3" id="ref7">3</reflink>]); Donohue and Levitt ([<reflink idref="bib11" id="ref8">11</reflink>]); and Gruber, Levine, and Staiger ([<reflink idref="bib21" id="ref9">21</reflink>]) all attempt to identify the effect of access to legal abortion on the characteristics of a given birth cohort by exploiting variation in state abortion laws over time, and they find that expansions in the availability of abortion caused the members of the affected birth cohort(s) to be more likely to graduate from college (Ananat et al., [<reflink idref="bib3" id="ref10">3</reflink>]); to commit less crime (Donohue & Levitt); to be less likely to live in poverty or to die as an infant (Gruber, Levine, & Staiger, [<reflink idref="bib21" id="ref11">21</reflink>]); and to be less likely to claim welfare assistance (Ananat et al., [<reflink idref="bib3" id="ref12">3</reflink>]; Gruber, Levine, & Staiger, [<reflink idref="bib21" id="ref13">21</reflink>]).</p> <p>Taken as a whole, this literature suggests that unintended pregnancy and childbearing have important implications for mothers, for their children, for the government programs that serve these individuals, and for society at large. It is therefore unsurprising that policymakers have turned their attention to the task of devising strategies to address this issue. For instance, the recently enacted Patient Protection and Affordable Care Act. H.R. 3590, 11th Cong (2010) contains provisions that expand access to publicly subsidized family planning services, provide funding for interventions to reduce teen pregnancy, and restore funding for abstinence‐only programs (National Campaign to Prevent Teen and Unplanned Pregnancy, [<reflink idref="bib45" id="ref14">45</reflink>]). Social scientists have produced a substantial body of credible research documenting the near‐term behavioral effects of policy initiatives of this sort. However, this research is kaleidoscopic—it contains varying estimates from several different literatures of the likely effects of a number of policies on a range of outcomes. In light of the pressing nature of this topic and the diffuse state of the relevant evidence, this paper synthesizes that evidence and uses a recently developed agent‐based simulation model to analyze it in a systematic fashion. More specifically, I conduct fiscal impact simulations of a mass media campaign encouraging condom use, an evidence‐based pregnancy prevention program targeted on at‐risk teens, and an expansion in eligibility for family planning services provided by state‐operated Medicaid programs.</p> <p>The ultimate goal of this paper is to provide new insights to policymakers and researchers seeking to determine how best to allocate scarce public funds in order to reduce the prevalence of unintended pregnancy. Given the prevailing fiscal and political winds—public officials at all levels of government are considering or have already implemented cuts to a range of benefits and services—policymakers who are interested in reducing unintended pregnancy might be reluctant to pursue more than one strategy at the same time. As such, I present simulation results that afford "apples‐to‐apples" comparisons of these policies' projected effects. However, a more sensible approach would be to implement any policy that is capable of generating taxpayer savings in excess of the amount required to finance it. My findings suggest that, depending upon how broadly one measures their estimated fiscal effects, the savings produced by each of these policies would likely exceed their costs. The results of my preferred specifications indicate that the three policies would produce returns to taxpayers on each dollar spent of between $2 and $6. Thus, although there is inevitable uncertainty associated with many of the simulations' parameters, these results provide suggestive evidence that all three policies represent sensible public investments.</p> <p>The rest of this paper is organized as follows. I begin by discussing the analytical framework for the policy simulations. I then describe the simulation model and the data used to parameterize it, summarize the relevant evidence as to the behavioral effects and costs of the policies simulated here, review a variety of key assumptions underpinning the simulations, and present simulation results. I conclude by discussing the policy implications of my findings. The assumptions used to conduct these simulations are also documented in more detail in three technical papers that are available online.[<reflink idref="bib3" id="ref15">3</reflink>]</p> <hd id="AN0076170450-2">ANALYTICAL FRAMEWORK</hd> <p>In light of the evidence suggesting that unintended pregnancy and childbearing have negative effects on mothers, on their children, on taxpayers, and on society at large, one would ideally prefer that an analysis such as this account for all such considerations. However, it would be difficult to pinpoint with any precision the effects of preventing an unintended pregnancy (pregnancies) on many of the outcomes enumerated above (e.g., levels of maternal educational attainment or crime rates) or to monetize any such effects that could be estimated. I therefore take the relatively conservative approach of evaluating policies only from the government's perspective. More specifically, I measure policies' costs in terms of the number of taxpayer dollars that would be required to implement them, and I measure their benefits in terms of the amount of public sector cost savings that they would generate via reduced government transfers to pregnant women and to children from birth through their fifth birthdays.[<reflink idref="bib4" id="ref16">4</reflink>] I focus on these benefits in particular because there are readily available data that allow them to be measured in a straightforward fashion; because they have direct impacts on government budgets in the near term and may therefore be more likely to factor into public policy decisions; and because the results of the simulations are relatively insensitive to the choice of discount rate used to calculate policies' benefits over such a short period of time.</p> <p>I follow standard practice for analyses of this sort by monetizing the benefits and costs for each policy such that they are expressed in constant dollars corresponding to a common base year (Levin & McEwan, [<reflink idref="bib37" id="ref17">37</reflink>]). Specifically, I measure both in 2008 real dollars and on an annualized basis. Because I focus only on the government's perspective, this exercise is more appropriately classified as a fiscal impact analysis than as a benefit‐cost analysis. However, when I report ratios of the savings that these policies would generate to their costs, I refer to them as "benefit‐cost ratios" for ease of exposition. The reader should nonetheless bear in mind that the policies studied here may generate a variety of societal benefits for which I do not account. If sufficient data were available to allow me to consider the full gamut of effects produced by the prevention of unintended pregnancies, my estimates of these policies' monetized benefits would likely be even higher.</p> <p>In order to estimate the cost savings generated by a given policy, I simulate its effects on key individual‐level behaviors such as contraceptive use and sexual frequency. These changes in behavior produce lower simulated rates of pregnancy and childbearing, which in turn reduces government spending on pregnant women and young children to the extent that these impacts are concentrated among women who—had they become pregnant—would have claimed publicly subsidized benefits for themselves and their newborn children. Critical to these simulations, then, are the questions of how these interventions affect individual behavior and of whose behavior is affected. In a subsequent section, I review the literature on each policy's behavioral effects. Before doing so, however, I describe the model that is used to conduct these simulations.</p> <hd id="AN0076170450-3">DESCRIPTION OF THE SIMULATION MODEL</hd> <p>The policy analyses are performed using FamilyScape, which is an agent‐based simulation tool that allows the user to model the impacts of policy changes on family‐formation outcomes. There is no single feature(s) that definitively distinguishes agent‐based models from related approaches such as dynamic microsimulation models (e.g., the Congressional Budget Office's CBOLT model) or static microsimulation models (e.g., the Urban Institute's TRIM model).[<reflink idref="bib5" id="ref18">5</reflink>] In general, however, agent‐based models are more likely than the latter classes of models to account for spatial dynamics, to simulate local interactions between members of the simulation population, and to allow for extensive heterogeneity in terms of agents' attributes and the ways in which they behave when confronted with a given set of circumstances.[<reflink idref="bib6" id="ref19">6</reflink>]</p> <p>FamilyScape possesses all of these characteristics to varying degrees. It simulates the key antecedents of pregnancy (e.g., sexual activity, contraceptive use, and female fecundity) and many of its most important outcomes (e.g., pregnancy and childbearing within and outside of marriage and among teenaged and non‐teenaged women, and abortion). Behaviors and outcomes of interest are simulated at the individual level and are then summed across the simulation population in order to produce aggregate estimates of phenomena of interest. As is true with most agent‐based models, the individuals within FamilyScape are heterogeneous: Each of them is assigned a set of demographic and behavioral attributes that govern the decisions that they make over the course of the simulation. The simulation population's gender, age, race, education, socioeconomic status (SES), and marital‐status characteristics are consistent with the demographic profiles of the members of a nationally representative data set.[<reflink idref="bib7" id="ref20">7</reflink>]</p> <p>Figure diagrams FamilyScape's overall structure and delineates the various stages of the simulation. During the first stage, the model is populated with a group of 10,000 agents (i.e., individuals) aged 15 to 44 whose demographic characteristics are nationally representative.[<reflink idref="bib8" id="ref21">8</reflink>] In the second stage, opposite‐sex relationships of varying duration are formed among some individuals.[<reflink idref="bib9" id="ref22">9</reflink>] In the third stage, sexual activity (or lack thereof) is simulated among married and unmarried couples, and contraceptive use (or lack thereof) is simulated among couples who have sex. In the fourth stage, some sexually active couples become pregnant, and each pregnancy eventually results in a birth, an abortion, or a fetal loss. The model's fifth and final stage accounts for the fact that each birth is either to a married couple or to a single mother. As a function of the structure of the family into which each child is born and of his or her mother's demographic characteristics, a poverty status is also assigned to each newborn child during the model's final stage. The model has a daily periodicity, and most of these dynamics are simulated on an ongoing basis. Thus, at any given point in analysis time—i.e., on any given day within the simulation—relationships are being formed and ended, pregnancies are occurring, children are being born, and so forth.</p> <p>Graph: Summary Diagram of FamilyScape Simulation Model.</p> <p>All of the model's input modules are aligned to real‐world benchmarks that were produced via analysis of several different external data sources. For instance, data from the General Social Survey and the National Survey of Family Growth (NSFG) were used to develop FamilyScape's benchmarks for the number of single individuals in relationships; data from the NSFG were used to calibrate the model's module on coital frequency and to set benchmarks for the annual number of sexual partners among sexually active individuals; data from the Guttmacher Institute, the National Vital Statistics System, and the NSFG were used to develop the model's parameters that impute an outcome (live birth, abortion, or fetal loss) for each pregnancy; data from the Current Population Survey were used to parameterize the module that assigns a poverty status to each newborn child; and the results of a number of different clinical studies were used to develop the model's parameters governing the gestation periods for various pregnancy outcomes and the probability that a woman will become pregnant from a given act of intercourse as a function of her contraceptive regime, her age, and the day in her menstrual cycle. FamilyScape is designed to produce demographic variation in these dynamics that is similar to the equivalent variation that is observed in the real world.</p> <p>After about a year's worth of analysis time has elapsed under FamilyScape's baseline parameterization, the model reaches a quasi steady‐state in which simulated data from any given slice of 365 periods should be roughly reflective of the conditions observed in the real‐world data used to parameterize the model.[<reflink idref="bib10" id="ref23">10</reflink>] Most of FamilyScape's dynamics are probabilistic: Random processes govern the pool of potential partners with whom a given individual might, for example, enter into a new relationship at a given point in analysis time, the probability that he or she will in fact enter into a relationship, the likelihood that a given couple will have sex on a given day, the probability that a couple will use contraception if and when they have sex, and so forth. Thus, no two runs of the model are exactly alike. As such, the results reported below are annualized averages calculated using simulated data that are produced over the course of 50 different ten‐year‐steady‐state simulation runs.[<reflink idref="bib11" id="ref24">11</reflink>] Generally speaking, the annualized data produced by the model for its key output variables are roughly normally distributed, and these data tend to be clustered reasonably closely around their means (Thomas & Monea, [<reflink idref="bib63" id="ref25">63</reflink>]).</p> <p>None of FamilyScape's input components was manipulated to ensure that it produces realistic outcomes. Rather, the model is validated to the extent that its inputs are able to work in concert to grow aggregate‐level outputs (e.g., rates of pregnancy and childbearing within and outside of marriage, the frequency of abortion, and so forth) that are consistent with their real‐world equivalents. The model generally performs well in this regard, especially for the unmarried population (Thomas & Monea, [<reflink idref="bib63" id="ref26">63</reflink>]). For instance, the real‐world pregnancy rate among unmarried women is only about 1 percent higher than the simulated rate of pregnancy for the same group. For married couples, the real‐world pregnancy rate is about a third higher than the equivalent simulated rate. I have conducted sensitivity analyses in which the effects of simulated policies were reestimated using alternative (but arguably plausible) behavioral assumptions that produce baseline rates of simulated pregnancy among married couples that are more closely aligned to their real‐world equivalents. The results from these alternative policy simulations were qualitatively similar to the ones that were produced under the model's original specification. The model's original parameter set therefore serves as the baseline specification for the simulation results reported below.</p> <p>FamilyScape lends itself readily to policy simulations, since its parameters can be changed relatively easily under the assumption that a given intervention has an effect on individual behavior. For instance, if one believes that a policy has a particular effect on contraceptive use or on sexual frequency, that effect can be simulated at the individual level by altering the model's baseline behavioral parameters, and the policy's impacts can then be estimated on (say) the number of teenage pregnancies, the frequency of out‐of‐wedlock childbearing, and the incidence of abortion. This approach capitalizes on the strengths of the existing research literature, since more is generally known about pregnancy prevention interventions' behavioral effects than about their implications for outcomes such as rates of teen pregnancy or unwed childbearing. The simulations also help to fill an important gap in the literature by using the best available estimates of policies' behavioral impacts to project their effects on outcomes of policy interest.</p> <p>In addition, the appendix at the end of this paper shows that the FamilyScape simulations produce different and more credible estimates of these policies' effects than does a simpler, back‐of‐the‐envelope approach.[<reflink idref="bib12" id="ref27">12</reflink>] In proportional terms, the differences between the simulation results and a set of back‐of‐the‐envelope estimates range from a little more than 10 percent (for the Medicaid expansion) to almost 90 percent (for the media campaign). These differences are attributable to the fact that FamilyScape models a variety of dynamics—e.g., the diversity of contraceptive methods used during intercourse, the covariance between coital frequency and contraceptive use, and the covariance between fecundity and contraceptive use—that would be difficult if not impossible to incorporate into a series of simple back‐of‐the‐envelope calculations. Thus, FamilyScape estimates these policies' impacts with a greater degree of precision than is afforded by a more naïve approach.</p> <hd id="AN0076170450-4">DESCRIPTIONS OF SIMULATED POLICIES</hd> <p>Sawhill, Thomas, and Monea ([<reflink idref="bib54" id="ref28">54</reflink>]) identify three broad reasons for the prevalence of unintended and out‐of‐wedlock pregnancy and childbearing: First, some individuals lack the motivation to avoid risky sexual activity; second, some individuals who are properly motivated lack information about how best to realize their intentions; and third, individuals armed both with accurate information and with good intentions are sometimes hindered by limited access to effective contraception. I simulate policies that are designed to address each of these considerations and for which there is credible evidence of their behavioral impacts. Specifically, I model a mass media campaign that seeks to motivate individuals to use condoms, an evidence‐based teen pregnancy prevention program that is designed to provide at‐risk youth with information about how to avoid becoming pregnant, and an expansion in access to Medicaid‐subsidized family planning services.</p> <p>I estimate these policies' impacts by comparing the results of simulations that were conducted under FamilyScape's baseline assumptions to the results of simulations that were conducted under assumptions that reflect their presumed effects on contraceptive use and coital frequency. In Thomas ([<reflink idref="bib62" id="ref29">62</reflink>]), I analyze in detail the literatures on all three policies, and I describe the way in which I synthesized the relevant evidence in order to develop various assumptions about these policies' behavioral effects and their costs. I summarize these assumptions in the top panel of Table , and I provide further detail on the reasoning behind them in the subsections that follow.</p> <p>Summary of key assumptions for simulations of policies to prevent unintended pregnancy</p> <p> <ephtml> <table><tbody><tr><td><bold>Mass Media Campaign</bold></td></tr><tr><td>• <italic>Behavioral effects</italic>: Three percent more adult male members of target audience use condoms as a result of the campaign (assumption developed based on findings and analysis in Noar, <xref ref-type="bibr" rid="bibr47">47</xref>, and in Snyder et al., <xref ref-type="bibr" rid="bibr57">57</xref>). Smaller effect (1.5 percent) assumed for teens, since they have a higher baseline condom‐usage rate.</td></tr><tr><td>• <italic>Target population</italic>: Unmarried males aged 15 to 44.</td></tr><tr><td>• <italic>Total annual cost of national campaign for initial specification</italic>: $100 million (based on cost estimates for the <italic>Truth</italic> and <italic>VERB</italic> campaigns).</td></tr><tr><td>• <italic>Total annual cost of national campaign for alternative specification</italic>: $250 million (based on cost estimates for the Lexington condom campaign and the National Youth Anti‐Drug Media Campaign).</td></tr><tr><td><bold>Evidence‐based teen pregnancy prevention program</bold></td></tr><tr><td>• <italic>Behavioral effects for initial specification</italic>: Within target population, a 7.5 percent increase in number of teens who are sexually inactive in any given three‐month period and a 12.5 percent increase in number of contraceptive users.</td></tr><tr><td>• <italic>Behavioral effects for alternative specification</italic>: Same change in contraceptive use as under initial specification; no effect on sexual activity (based on findings reported in Scher, Maynard, & Stagner, <xref ref-type="bibr" rid="bibr55">55</xref>).</td></tr><tr><td>• <italic>Target population</italic>: Unmarried, low‐SES youth.</td></tr><tr><td>• <italic>Program intensity</italic>: Assume that members of the target population will participate in the program once every 2 years.</td></tr><tr><td>• <italic>Annual program cost per member of the target population</italic>: $50.</td></tr><tr><td>• <italic>Total annual cost</italic>: $145 million</td></tr><tr><td>• <italic>Program behavioral effects and costs</italic>: Estimated using data detailed in Table.</td></tr><tr><td><bold>Expanded access to Medicaid Family Planning</bold></td></tr><tr><td>• <italic>Behavioral effects for initial specification</italic>: Five percent fewer sexually active women fail to use contraception after policy is implemented (Kearney & Levine, <xref ref-type="bibr" rid="bibr32">32</xref>).</td></tr><tr><td>• <italic>Behavioral effects for alternative specification</italic>: Half as large as for initial specification.</td></tr><tr><td>• <italic>Target population</italic>:</td></tr><tr><td> • Behavioral effects concentrated among females under 200 percent of poverty.</td></tr><tr><td> • Implemented for a share of the simulation population that is consistent with the portion of the U.S. population living in non‐waiver states.</td></tr><tr><td> • Marriage rate among women who take up family planning services under the policy is similar to the rate of marriage among women who currently use subsidized contraception.</td></tr><tr><td>• <italic>Annual cost per woman served</italic>: $188 (Kearney & Levine, <xref ref-type="bibr" rid="bibr32">32</xref>).</td></tr><tr><td>• <italic>Take‐up</italic>: 5.4 percent of women of childbearing age newly take up services under expansion (Kearney & Levine, <xref ref-type="bibr" rid="bibr32">32</xref>).</td></tr><tr><td>• <italic>Total Annual Cost</italic>: $235 million.</td></tr><tr><td><bold>Quantifying policies' benefits</bold></td></tr><tr><td>• Estimates of average public cost savings produced by the prevention of an unintended birth:</td></tr><tr><td> • Spending on publicly subsidized prenatal care, delivery, and postpartum care equals $1,700 for teen births, $2,000 for non‐teen births.</td></tr><tr><td> • All of above plus 1 year's worth of publicly subsidized infant medical care equals $4,000 for teen births, $5,000 for non‐teens births.</td></tr><tr><td> • All of above plus 5 years' worth of publicly subsidized children's benefits equals $19,000 for teen births, $24,000 for non‐teen births.</td></tr><tr><td>• Estimate of average public cost savings produced by the prevention of a fetal loss equals $750.</td></tr><tr><td>• Public cost savings from prevention of abortions are quite small, therefore not considered in simulation.</td></tr><tr><td>• Savings amounts are expressed as averages across all members of the population assumed to be eligible to participate in means‐tested programs.</td></tr><tr><td>• Estimated savings from preventing an unintended birth are adjusted to account for the fact that some pregnancies are reported to be mistimed and may therefore simply occur at a later point in time. Adjustments are age‐specific.</td></tr><tr><td>• Eligibility for publicly subsidized pregnancy care, infant medical care, and children's benefits are imputed based on whether a pregnant woman or young child is estimated to fall below 200 percent of the federal poverty line.</td></tr><tr><td>• Effects of each policy on the incidence of pregnancy and childbearing are estimated using the FamilyScape simulation model.</td></tr><tr><td><bold>Alternative pregnancy‐outcome assumptions:</bold></td></tr><tr><td>• Alternate versions of each simulation incorporate additional information on unintended pregnancy outcomes from special data tabulations from the Guttmacher Institute.</td></tr></tbody></table> </ephtml> </p> <hd id="AN0076170450-5">Mass Media Campaign: Behavioral Effects</hd> <p>In recent decades, numerous media campaigns have been mounted to encourage individuals to engage in healthy behavior. These campaigns have focused on outcomes ranging from seat belt use and tooth decay to eating habits and drug abuse (Snyder et al., [<reflink idref="bib57" id="ref30">57</reflink>]). Many of these campaigns have been evaluated for their effects on the behavior of members of their target audiences. It is usually not possible, however, to evaluate their effects using random assignment. The best existing studies tend to rely instead on a difference‐in‐differences approach in which behavioral changes over time in the geographic area(s) in which the campaign was implemented are compared to equivalent changes in a demographically similar area(s) in which it was not implemented.</p> <p>In their meta‐analysis of the evaluations of these campaigns, Snyder et al. ([<reflink idref="bib57" id="ref31">57</reflink>]) find that, on average, they changed the behavior of about 8 percent of the members of their target populations.[<reflink idref="bib13" id="ref32">13</reflink>] Some of the media campaigns included in the authors' analysis attempted to encourage contraceptive use. Among campaigns falling into this category, all of them have focused specifically on the promotion of condom use as a means of preventing the transmission of sexually transmitted infections (STIs). Snyder and her co‐authors conclude that, on average, these campaigns caused 6 percent of the members of their target populations to increase their use of condoms. However, the true average effect of such campaigns is likely smaller than is suggested by the authors' analysis because, first, they only considered the results of published studies, which are presumably more likely to have reported evidence of a positive effect; and, second, they rely on identification strategies that are imperfect at best.[<reflink idref="bib14" id="ref33">14</reflink>] I therefore adopt Noar's ([<reflink idref="bib47" id="ref34">47</reflink>]) assumption that these campaigns' true average effects were probably about half as large as is suggested by Snyder et al.'s conclusions. In other words, I assume that these campaigns actually affected the behavior of about 3 percent of the members of their target populations.</p> <p>Given that the campaigns described above all encouraged condom use in particular as a means of avoiding STI transmission—and because their target populations were comprised of groups of men who were either entirely or predominantly unmarried—I assume that the simulated campaign would only affect condom use and would only have an effect among unmarried males. In light of the fact that a much smaller proportion of teenaged males than of adult males fail to use condoms during intercourse, I assume that the simulated campaign's percentage‐point effect on the share of teens who use condoms would be about half as large as the equivalent effect among adult men.[<reflink idref="bib15" id="ref35">15</reflink>] And because there is evidence that previous campaigns' effects have faded after they were completed, I assume that the simulated campaign must be ongoing continuously in order for it to produce the effects described above, which were typically measured in the midst of the campaigns in question or shortly after their completion.[<reflink idref="bib16" id="ref36">16</reflink>]</p> <hd id="AN0076170450-6">Mass Media Campaign: Cost</hd> <p>I assume that the campaign would be implemented on a national scale. I estimate its cost using data on the itemized expenses of four health‐related campaigns that have been implemented in recent years. Specifically, I use cost data for the <emph>Truth</emph>, <emph>VERB</emph>, and <emph>National Youth Anti‐Drug Media</emph> (NYADMC) campaigns, and for a campaign encouraging condom use in Lexington, KY. The first three of these campaigns were implemented nationally. Because the fourth campaign was implemented locally, I use estimates of the relative sizes of the Lexington and national media markets in order to calculate what such a campaign might cost if it were implemented nationwide. I estimate that the annual costs of the <emph>Truth</emph> and <emph>VERB</emph> campaigns were $100 million or less, and I estimate the average cost of the NYADMC campaign and a national version of the Lexington campaign to be a little more than $250 million per year. Thus, I implement two different specifications for this simulation: In the initial specification, I assume that the campaign would cost $100 million annually, and in the alternative specification, I assume that its annual cost would be $250 million. One of the least effective of these campaigns (NYADMC) was also one of the most expensive.[<reflink idref="bib17" id="ref37">17</reflink>] I therefore assume that the simulated campaign's effectiveness does not vary with its cost. Given that the findings reviewed here suggest that a well‐designed campaign can affect behavior at a comparatively lower cost, I take the initial specification of this simulation (which assumes a smaller cost) to be its preferred specification.</p> <hd id="AN0076170450-7">Evidence‐Based Teen Pregnancy Prevention Program: Behavioral Effects</hd> <p>Sex education programs can assume a variety of different forms. The most common type of intervention is based on a written curriculum designed to discourage intercourse, to encourage contraceptive use, or both. Some other types of programs focus both on parents and on their teenaged children; are implemented on a community‐wide basis; or attempt to encourage youth development in a more holistic sense, rather than focus solely on encouraging healthy sexual behavior (Kirby, [<reflink idref="bib33" id="ref38">33</reflink>]). Many of these interventions have been evaluated using random assignment. These evaluations suggest that some programs have had substantial effects on such outcomes as sexual activity, contraceptive use, and/or childbearing. Other interventions appear to have had little if any effect on such outcomes. The most effective of these interventions tend to combine an emphasis on sexual abstinence with instruction on proper contraceptive use.</p> <p>I synthesize the evaluation results of a group of effective programs to develop a set of assumptions about what the impacts might be of a well‐designed pregnancy prevention program that is implemented on a national scale. To be clear, the interventions whose evaluation results are used to parameterize this simulation are not typical of all teen pregnancy prevention programs. Rather, they are among the most effective such programs that have been rigorously evaluated. An assumption underlying this simulation is that the national‐level program being modeled would draw on the best practices of interventions that have been found effective on a smaller scale.[<reflink idref="bib18" id="ref39">18</reflink>] This assumption is consistent with recently developed federal guidelines for funding evidence‐based teen pregnancy programs: The Office of Adolescent Health has made $75 million available to fund the replication of programs for which there is a strong evidentiary basis demonstrating their effectiveness (National Campaign to Prevent Teen and Unplanned Pregnancy, [<reflink idref="bib46" id="ref40">46</reflink>]). This simulation can be thought of as modeling the effects of a similar policy.</p> <p>Interventions were included in the synthesis undertaken to parameterize this simulation if they primarily served high school students, if they were evaluated using random assignment, and if they were found to have had statistically significant effects on both contraceptive use and some measure of sexual frequency. Interventions were excluded from the synthesis if their effects were found to have faded over a relatively short period of time, if attempts to replicate them have been generally unsuccessful, or if their evaluations showed that they ultimately had little or no effect on pregnancy or childbearing. I have identified five programs that meet these criteria: <emph>Becoming a Responsible Teen</emph>, <emph>HIV Prevention for Adolescents in Low‐Income Housing Developments</emph>, <emph>Safer Choices</emph>, and two programs that were developed using a core curriculum designed by John and Loretta Jemmott: <emph>Be Proud! Be Responsible!</emph> and <emph>¡Cuídate!</emph>. See Table for an overview of each program's evaluation design, estimated effects, and costs per participant. I conclude that these programs collectively increased the proportion of participants who used contraception at last sex by an average of about 25 percent and reduced the number of individuals who were sexually active in a typical three‐month period by an average of about 15 percent. These assumptions are roughly consistent with Kirby's ([<reflink idref="bib33" id="ref41">33</reflink>]) qualitative conclusion that curriculum‐based sex education programs reduce risky sexual behavior by around one third.</p> <p>Behavioral impacts and costs of selected teen‐pregnancy prevention programs found to have affected both sexual activity and contraceptive use. *</p> <p> <ephtml> <table><thead><tr><th align="center">Among Interventions that have been Evaluated using Random Assignment</th></tr><tr><th>Name of intervention</th><th align="center">Details of evaluation design for initial study</th><th align="center">Estimated program effects on sexual abstinence/initiation of sex</th><th align="center">Estimated program effects on frequency of intercourse</th><th align="center">Estimated program effects on male contraceptive use</th><th align="center">Estimated program effects on female contraceptive use</th><th align="center">Estimated program cost per participant (in 2008 dollars)</th><th align="center">Replication information</th></tr></thead><tbody><tr><td><italic>Becoming a Responsible Teen</italic></td><td><sup>*</sup><italic>Randomized controlled experiment</italic> serving African American youth. Participants were recruited from a low‐income community in Jackson, MS.<sup>*</sup><italic>Treatment group</italic> participated in eight sessions in a community‐based setting, each one lasting 90 to 120 minutes. Curriculum designed specifically to prevent HIV infection among African American adolescents.<sup>*</sup><italic>Control group</italic> received one‐time, two‐hour HIV‐prevention session.<sup>*</sup><italic>N</italic> = 246 at baseline; 225 at follow‐up 1 year after completion of the intervention.</td><td>One year after the end of the intervention, treatment group members were about <bold>65%</bold> as likely as control group members to report having had sex during the previous 2 months.</td><td>No results reported for sexual frequency in evaluations of this program.</td><td><italic>Two months after the end of the intervention</italic>: About <bold>57%</bold> more sexual occasions from the previous 2 months were reported to have involved the use of a condom among males in the treatment group than among males in the control group.<italic>One year after the end of the intervention</italic>: No significant difference between treatment‐group and control‐group males in the proportion of sexual occasions protected by a condom. However, combined‐sex analyses showed a significant difference at 1 year: Almost <bold>30</bold>% more sexual occasions from the previous 2 months were reported to have involved the use of a condom among males and females in the treatment group than among males and females in the control group.</td><td><italic>Two months after the end of the intervention</italic>: About <bold>16%</bold> more sexual occasions from the previous 2 months were reported to have involved the use of a condom among females in the treatment group than among females in the control group.<italic>One year after the end of the intervention</italic>: About <bold>44%</bold> more sexual occasions from the previous 2 months were reported to have involved the use of a condom among females in the treatment group than among females in the control group.</td><td align="center">≈ <bold>$70</bold></td><td><italic>One successful replication</italic>: Curriculum fully implemented in drug rehabilitation facility; increased abstinence and condom use.<italic>One unsuccessful replication</italic>: Curriculum shortened by more than half and implemented in a state juvenile reformatory; no significant program effects on sex or contraceptive use.</td></tr><tr><td><italic>HIV prevention for Adolescents in Low‐Income Housing Developments</italic></td><td><sup>*</sup><italic>Randomized controlled experiment</italic> serving adolescents aged 12 to 17. Participants were recruited from 15 low‐income housing communities.<sup>*</sup><italic>Primary treatment group</italic> was comprised of residents of the housing developments that were randomly assigned to receive community treatment. Treatment consisted of distribution of free condoms and brochures, two 3‐hour workshops on HIV prevention, and a community‐wide program with various neighborhood initiatives and workshops for parents.<sup>*</sup><italic>Control group</italic> was comprised of residents of control developments that received free condoms and brochures, watched a videotape about HIV prevention, and discussed the video after viewing.<sup>*</sup><italic>N</italic> = 1,172 at baseline; 763 at follow‐up 2 months after completion of the intervention.</td><td>Among participants who were sexually inexperienced at baseline, treatment‐group members were about <bold>88%</bold> as likely as control‐group members to report having initiated sex within 2 months of the end of the intervention.</td><td>No results reported for sexual frequency in evaluations of this program.</td><td>Self‐reports indicate that, as of the follow‐up 2 months after the completion of the intervention, a condom was used at last sex about <bold>24%</bold> more often among treatment group members than among control group members.</td><td>Cost information not available from team that designed, implemented, and evaluated the intervention.</td><td>No published evaluations of any attempts to replicate program.</td></tr><tr><td><italic>Safer Choices</italic></td><td><sup>*</sup><italic>Randomized controlled experiment</italic> implemented for freshmen and sophomores in 20 high schools in California and Texas.<sup>*</sup><italic>Treatment group</italic> was comprised of students in the schools that were randomly assigned to receive treatment. Intervention was implemented for all students in each treatment school and consisted of 20 sessions focusing on improving students' knowledge about condom use and sexually transmitted infections (STIs) and on changing their perception of abstinence in order to make it a more appealing option. In addition, clubs and councils were created and speaker series and parenting‐education initiatives were implemented in order to change the cultures of the treatment schools.<sup>*</sup><italic>Control group</italic> was comprised of students at control schools that received standard, five‐session sexual education curriculum and a few other school‐wide activities that varied from school to school.<sup>*</sup><italic>N</italic> = 3,869 at baseline; 3,058 at follow‐up about 1 year after completion of the intervention.</td><td><italic>Among all members of the analysis sample</italic>: No statistically significant difference 1 year after completion of the intervention (or at earlier follow‐ups) in the self‐reported odds of having initiated sex between treatment and control group members who were sexually inexperienced at baseline.<italic>Among Latino members of the analysis sample</italic>: About 1 year after completion of the intervention, sexually inexperienced treatment group members were significantly less likely than control group members to report that they had initiated sex (odds ratio = <bold>.57</bold>).</td><td>About one year after completion of the intervention, no significant differences between treatment‐ and control‐group members in the self‐reported frequency of sexual intercourse over the previous 3 months (nor were such differences observed at earlier follow‐ups).</td><td>About 1 year after completion of the intervention, males in the treatment group were significantly more likely to report having used contraception at last sex (odds ratio = <bold>1.64</bold>).</td><td>About one year after completion of the intervention, no statistically significant difference between females in the treatment and control groups in the self‐reported use of contraception at last sex (results for female contraceptive use not reported for earlier follow‐ups, but evaluators found a significant difference in the self‐reported use of contraception at last intercourse for the combined male and female samples while the intervention was ongoing; odds ratio = <bold>1.76</bold>).</td><td align="center">≈ <bold>$110</bold></td><td>No published evaluations of any attempts to replicate program.</td></tr><tr><td><italic>Be Proud! Be Responsible!</italic></td><td><sup>*</sup><italic>Randomized controlled experiment</italic> serving urban, African American males aged 13 to 18 in the Philadelphia, PA metropolitan area. Participants were recruited from a local medical clinic, a neighborhood high school, and a local YMCA.<sup>*</sup><italic>Treatment group</italic> participated in 5‐hour intervention designed to prevent HIV infection. Intervention techniques included small group discussions, videos, and role‐playing.<sup>*</sup><italic>Control group</italic> participated in career planning intervention of similar length.<sup>*</sup><italic>N</italic> = 157 at baseline; 150 at follow‐up 3 months after the intervention.</td><td>No statistically significant difference observed 3 months after completion of the intervention between treatment‐ and control‐group members in the share of participants who reported having had sex over the previous 3 months (among boys only).</td><td>Three months after the intervention, treatment‐group members reported having engaged in about <bold>40%</bold> as much sex as control group members over the previous 3 months (among boys only).</td><td>Three months after the intervention, a significant difference was observed between average self‐reported treatment‐ and control group scores (<bold>4.4</bold> vs. <bold>3.5</bold>, respectively) on condom‐use scale where 1 = "never" and 5 = "always."</td><td>Intervention was for boys only.</td><td align="center">≈ <bold>$120</bold></td><td><italic>One successful replication</italic>: Implemented in different communities from original for boys and girls, rather than just for boys; and was evaluated over 6 months, rather than over just 3 months. Found to have reduced the incidence of unprotected sex over the evaluation period.<italic>One unsuccessful replication</italic>:Implemented in high school classrooms during school day. Not found to have any effect on sexual behavior, perhaps because it was mandatory (original version of the program was optional).</td></tr><tr><td>Modified Version of "Be Proud!": <italic>¡Cuídate!</italic></td><td><sup>*</sup><italic>Randomized controlled experiment</italic> serving Latino youth aged 13 to 18 in the Philadelphia, PA metropolitan area. Participants were recruited from three local high schools and various community organizations.<sup>*</sup><italic>Treatment and control groups</italic> received interventions similar to the ones described above for "Be Proud," although <italic>¡Cuídate!</italic> was tailored specifically for Latinos and Latinas rather than for African Americans.<sup>*</sup><italic>N</italic> = 656 at baseline; 553 at follow‐up 1 year after the intervention.</td><td>Using data from follow‐ups conducted 3 months, 6 months, and 1 year after the intervention, evaluators concluded that treatment‐group members were significantly less likely than control group members to report having had sexual intercourse in the previous 3 months. At each of the 3 follow‐ups, treatment‐group members were about <bold>85%</bold> as likely as control‐group members to report having had sex over the previous 3 months.</td><td>No results reported for sexual frequency in evaluations of this program.</td><td>Using data from follow‐ups conducted 3 months, 6 months, and 1 year after the intervention, evaluators concluded that treatment group members were significantly more likely to report using condoms consistently. Across the three follow‐ups, treatment group members were between about <bold>50%</bold> and <bold>65%</bold> more likely than control group members to report having used condoms consistently over the previous 3 months. However, no statistically significant difference observed using data from the three follow‐ups between treatment and control group members in the share of participants who reported having used condoms at last sex.</td><td /><td>No published evaluations of any attempts to replicate program directly. However, <italic>Making Proud Choices!</italic> (MPC), like <italic>¡Cuídate!</italic>, was based on the <italic>Be Proud!</italic> curriculum. MPC was implemented for black boys and girls aged 11 to 13; found to have reduced self‐reported sexual frequency and increased self‐reported contraceptive use. See above for information on successful and unsuccessful <italic>Be Proud!</italic> implementations.</td></tr></tbody></table> </ephtml> </p> <ulist> <item>2 Notes</item> <item>3 Unless otherwise indicated, all findings listed here are statistically significant at or beyond the 0.05 level. The summaries for all five programs are based in large part on information taken from two overviews: Kirby ([<reflink idref="bib33" id="ref42">33</reflink>]) and Suellentrop ([<reflink idref="bib61" id="ref43">61</reflink>]). Additional information on <emph>Becoming a Responsible Teen</emph> is taken from Child Trends ([<reflink idref="bib8" id="ref44">8</reflink>]), Manlove et al. ([<reflink idref="bib41" id="ref45">41</reflink>]), St. Lawrence et al. ([<reflink idref="bib59" id="ref46">59</reflink>]), and St. Lawrence et al. ([<reflink idref="bib60" id="ref47">60</reflink>]); additional information on <emph>HIV Prevention for Adolescents in Low‐Income Housing Developments</emph> is taken from Sikkema et al. ([<reflink idref="bib56" id="ref48">56</reflink>]); additional information on <emph>Safer Choices</emph> is taken from Coyle et al. ([<reflink idref="bib9" id="ref49">9</reflink>]), Kirby et al. ([<reflink idref="bib34" id="ref50">34</reflink>]), Olaiya ([<reflink idref="bib48" id="ref51">48</reflink>]), and Wang et al. ([<reflink idref="bib68" id="ref52">68</reflink>]); and additional information on <emph>Be Proud! Be Responsible!</emph>, on <emph>Making Proud Choices!</emph>, and on <emph>¡Cuídate!</emph> is taken from Jemmott, Jemmott, and Fong ([<reflink idref="bib27" id="ref53">27</reflink>]), Manlove et al. ([<reflink idref="bib41" id="ref54">41</reflink>]), United Way of Rochester ([<reflink idref="bib66" id="ref55">66</reflink>]), and Villaruel, Jemmott, and Jemmott ([<reflink idref="bib67" id="ref56">67</reflink>]). Bolded values represent specific estimates used to develop assumptions for policy simulations.</item> </ulist> <p>All of these programs were implemented on a small scale (the sample sizes for their evaluations ranged from 157 to 3,869, and only about half of these subjects were members of their studies' treatment groups). If programs of this sort were taken to a larger scale, their effects would almost certainly be diluted, due in large part to the difficulty of replicating the commitment and enthusiasm of the teams that originally implemented them. Thus, I assume that, if such an intervention were implemented nationally, its effects would be about half as large as the impacts of the small‐scale programs upon which it is based. For the purposes of the initial simulation of a teen pregnancy prevention program, I therefore assume that the simulated program increases the proportion of teens who use contraception by about 12.5 percent and reduces the number of teens who have sex during an average three‐month period by about 7.5 percent. Because almost all of the programs described above were created for what might be considered to be at‐risk teens (e.g., minority adolescents in urban areas or teens living in public housing developments), and since very few teenagers are married, I implement this simulation only for unmarried and teenaged members of the simulation population who are tagged as low‐SES.</p> <p>I also develop an alternative specification for this simulation based on the findings of Scher, Maynard, and Stagner ([<reflink idref="bib55" id="ref57">55</reflink>]), who re‐estimate the results of numerous evaluations of teen pregnancy prevention programs in order to correct for methodological flaws in the original studies.[<reflink idref="bib19" id="ref58">19</reflink>] Evaluations for two of the five programs incorporated into my synthesis—<emph>Becoming a Responsible Teen</emph> and <emph>HIV Prevention for Adolescents in Low‐Income Housing Developments</emph>—were also included in Scher, Maynard, and Stagner's reanalysis. Although the original evaluations of both interventions found that they affected the probability of initiating sexual activity, the authors' reanalysis suggests that neither program actually had an effect on this margin of behavior. The authors did, however, find that both programs affected the probability of having had unprotected sex. The latter outcome is essentially a composite measure of sexual initiation, contraceptive use, and coital frequency among individuals who have already initiated sex. It is possible that the reduction in unprotected sex observed by the authors is the result of changes in the latter two behavioral margins. However, the authors' finding that these programs did not affect sexual initiation also suggests the possibility that the observed reduction in unprotected sex is driven solely by changes in contraceptive use. The authors' analytical approach does not allow for these hypotheses to be disentangled. The alternative specification for this simulation is premised on an assumption that the Scher, Maynard, and Stagner findings reflect programmatic impacts only on contraceptive use. I therefore assume that the simulated program affects contraceptive use in the manner described above, but that it has no effect on sexual activity. However, given that Scher, Maynard, and Stagner's results cannot rule out the assumptions underlying the initial parameterization for this simulation, I consider that specification to be the preferred one.</p> <hd id="AN0076170450-8">Evidence‐Based Teen Pregnancy Prevention Program: Cost</hd> <p>Using data summarized in Table on the itemized expenses of the interventions described above, I calculate that the average, per‐participant cost of these interventions is about $100.[<reflink idref="bib20" id="ref59">20</reflink>] After having compared the time periods over which each of these interventions was implemented and the timing of their most recent evaluations, I conclude that the measured effects incorporated into my synthesis can best be thought of as representing these programs' average impacts over a two‐year period. I therefore assume that, in order for the effects described above to be achieved, teens must participate in the simulated program once every 2 years. Thus, I assume the annual, per‐participant cost of the program to be $50. I multiply the per‐participant cost cited above by my estimate of the national number of unmarried, low‐SES teens to calculate that the annual cost of the program would be about $145 million.</p> <hd id="AN0076170450-9">Expanded Access to Medicaid Family Planning: Behavioral Effects</hd> <p>In fiscal year 2006, public expenditures on family planning services totaled about $1.85 billion, and spending by state‐run Medicaid programs accounted for more than two‐thirds of that total (Sonfield, Alrich, & Gold, [<reflink idref="bib58" id="ref60">58</reflink>]). Eligibility for Medicaid family planning services has historically been limited to pregnant women and mothers whose incomes place them below a very low threshold—typically, about 90 percent of the federal poverty line.[<reflink idref="bib21" id="ref61">21</reflink>] Starting in 1995, however, the federal government began granting waivers to some states allowing them to serve all income‐eligible women—regardless of whether they were pregnant or already had children—and, in most cases, to raise their income‐eligibility thresholds as well. In total, about half of states were granted Medicaid family planning waivers over a period of about a decade and a half. Most of these states raised their income‐eligibility thresholds to somewhere between 185 percent and 200 percent of poverty. More recently, the newly enacted health care reform legislation grants states the option to set their income thresholds for Medicaid family planning services at a level that is less than or equal to the thresholds that they use to determine eligibility for Medicaid pregnancy‐related care (National Campaign to Prevent Teen and Unplanned Pregnancy, [<reflink idref="bib45" id="ref62">45</reflink>]; Patient Protection and Affordable Care Act. H.R. 3590, 11th Cong, [<reflink idref="bib51" id="ref63">51</reflink>]).</p> <p>This provision is unlikely to have a substantial effect on the group of states that have already expanded their income‐eligibility criteria (hereafter, "waiver states"), but it could have a considerable effect on states that have not yet done so (hereafter, "non‐waiver states").[<reflink idref="bib22" id="ref64">22</reflink>] I simulate the effects of an expansion in access to Medicaid family planning services in non‐waiver states under the assumption that all of these states would set their income‐eligibility thresholds for these services equal to their thresholds for Medicaid pregnancy care. The average income‐eligibility threshold for Medicaid family planning services in waiver states is about 190 percent of poverty, and the average eligibility threshold for Medicaid pregnancy care in non‐waiver states is about 195 percent of poverty.[<reflink idref="bib23" id="ref65">23</reflink>] Given the similarity between these thresholds, I assume that take‐up of the new family planning option by non‐waiver states would affect contraceptive use within those states in much the same way that the implementation of income‐based waivers affected the same margin of behavior in waiver states.</p> <p>I develop parameters for this simulation using results reported by Kearney and Levine ([<reflink idref="bib32" id="ref66">32</reflink>]), who use a difference‐in‐differences strategy to estimate the effect of the implementation of income‐based Medicaid family planning waivers on contraceptive use among women between the ages of 15 and 44 in waiver states. Kearney and Levine's results suggest that the implementation of income‐based waivers caused 5 percent fewer non‐teenaged, sexually active women in waiver states to fail to use contraception at last sex. Their result for the same outcome for teenaged girls is not statistically significant. However, the standard error for the relevant estimate is quite large, and the authors present other results (discussed further below) suggesting that waiver implementation did affect the number of births to teen mothers. As such, I make the simplifying assumption that expansions in access to family planning services in waiver states had the same impact on contraceptive use among teenagers as among non‐teenagers. For the initial specification of this simulation, I therefore assume that, if non‐waiver states were to take up the new family planning option, about 5 percent of sexually active teenaged and non‐teenaged females in those states would newly use contraception at a given act of intercourse as a result.</p> <p>Kearney and Levine also find that the implementation of income‐based waivers reduced the birth rate in waiver states by about 2 percent among non‐teenaged women and by about 4 percent among teenaged girls. By contrast, when I model an increase in contraceptive use in the manner described above, the corresponding reductions in teen and non‐teen birth rates within the simulation are higher than is suggested by the authors' findings.[<reflink idref="bib24" id="ref67">24</reflink>] There are, however, large confidence intervals around the authors' point estimates of the waivers' impacts on contraceptive use and on the incidence of birth, and there are a range of estimates in the former interval that produce reductions in the simulated incidence of birth that are consistent with a range of values contained in the latter interval. For example, a simulation with an increase in contraceptive use that is half as large as is suggested by the authors' point estimate (but is well within the relevant confidence interval) produces reductions in simulated teen and non‐teen birth rates that fall within the confidence intervals for the authors' estimates of both effects. Thus, I implement an alternative specification for this simulation in which I assume that an expansion in access to subsidized contraception would cause 2.5 percent fewer sexually active females to fail to use contraception at a given act of intercourse. This specification's results in terms of the policy's effect on contraceptive use, childbearing, and the number of abortions are all contained within the corresponding confidence intervals reported by Kearney and Levine. Because this specification generates results that are closer to the estimates produced by Kearney and Levine, I consider it to be the preferred specification.</p> <p>For both specifications, I model increases in contraceptive use in such a way as to ensure that the marriage rate among new contraceptors is equivalent to the rate of marriage among current users of publicly subsidized contraception.[<reflink idref="bib25" id="ref68">25</reflink>] And because I assume that the average eligibility threshold for family planning services in states that take up the new option would be about 200 percent of poverty, I concentrate all of this increase in contraceptive use among women who are estimated to fall below this threshold.</p> <hd id="AN0076170450-10">Expanded Access to Medicaid Family Planning: Cost</hd> <p>Kearney and Levine conclude that the implementation of income waivers caused about 5.4 percent of women of childbearing age in waiver states to take up Medicaid family planning services. They also estimate that the average cost of the program per woman served is $188. I combine these figures with my tabulation of the number of women of childbearing age in non‐waiver states to estimate that the expansion would cost about $235 million annually.[<reflink idref="bib26" id="ref69">26</reflink>]</p> <hd id="AN0076170450-11">QUANTIFYING POLICIES' BENEFITS</hd> <p>I measure the benefits produced by a given policy in terms of the estimated amount of taxpayer savings that it would generate. The middle panel of Table summarizes my savings estimates, which account for government expenditures on unintended pregnancies in the form of publicly subsidized prenatal care, deliveries, medical treatment for fetal losses, postpartum care, and benefits provided to children after they are born. Regarding the last of these cost categories, the choice of time period over which to measure government spending on children is, at least in part, an arbitrary one. I make the relatively conservative choice to measure such costs only through the first 5 years of childhood, in part because this approach is compatible with readily available data. Note, then, that my savings estimates do not account for taxpayer spending on public education. If one were to extend the time period over which to measure the cost of children's benefits in order to capture spending on public schooling, the benefit‐cost ratios for the policies analyzed here would be higher. On the other hand, many of the children in question will become productive members of the labor force as they age into adulthood, after which their economic contributions may be sufficient to offset and eventually exceed the financial costs that they impose over the course of their lifetimes. For purposes of practicality, I ignore these longer‐term considerations here.</p> <p>For a given simulation, I produce three different benefit‐cost ratios, each one corresponding to a different measure of the savings produced by the policy in question. The first measure accounts only for cost savings associated with government expenditures on prenatal care, deliveries, postpartum care, and medical treatment for fetal losses. The second measure accounts for all of the aforementioned plus the cost savings associated with the provision of 1 year of publicly subsidized medical care to qualifying infants. And the third measure accounts for all of the aforementioned plus the cost savings associated with the provision of a wide range of means‐tested benefits and services to qualifying children through their fifth birthdays. All three measures were developed using findings reported in other studies that estimate spending on pregnant women and children.[<reflink idref="bib27" id="ref70">27</reflink>] I ignore public spending on abortions because it is severely curtailed by legal restrictions (such spending is dwarfed by public expenditures on births and on young children), and because it would be difficult for me to realistically model eligibility for abortion subsidies within the simulations.[<reflink idref="bib28" id="ref71">28</reflink>] Most public spending on pregnancies and on benefits for young children is means‐tested, and the substantial majority of these means‐tested benefits are provided to pregnant women and children who are below 200 percent of the federal poverty line. I therefore measure savings only to means‐tested programs serving women and children who are below this threshold.[<reflink idref="bib29" id="ref72">29</reflink>]</p> <p>My cost‐savings estimates account for the fact that some prevented pregnancies are merely delayed, while others are averted altogether. The public savings produced by delaying a pregnancy are considerably smaller than the savings produced by preventing it from occurring at all. I make the simplifying assumption that the prevention of pregnancies that are reported to be mistimed will simply delay the point in time at which they occur and that the prevention of pregnancies that are reported to be unwanted will avert them altogether.[<reflink idref="bib30" id="ref73">30</reflink>] I assume further that delayed teen and non‐teen births will on average occur 4.5 years and 2 years in the future, respectively. These assumptions are based on Chandra et al.'s ([<reflink idref="bib7" id="ref74">7</reflink>]) analysis of data on pregnancies that are reported to be mistimed in the 2002 NSFG. I calculate the savings generated by the prevention of an unintended pregnancy as the weighted average of the savings produced by delaying a mistimed pregnancy and by averting an unwanted pregnancy altogether, in which the weighting accounts for the share of unintended pregnancies that are reported to be mistimed and unwanted. Unintended births to teens are more likely reported to be mistimed (71 percent) than are unintended births to adults (54 percent).[<reflink idref="bib31" id="ref75">31</reflink>] I therefore specify separate cost‐savings amounts for pregnancies that are prevented to teenaged and non‐teenaged women.</p> <p>Researchers have found that the distinction between mistimed and unwanted pregnancies is murkier in practice than in theory.[<reflink idref="bib32" id="ref76">32</reflink>] For instance, Santelli et al. ([<reflink idref="bib53" id="ref77">53</reflink>]) show that pregnancies originally reported to be unwanted are often described as having simply been mistimed (or even intended) in subsequent retrospective reports. As a result, the approach described above may lead to an overestimation or to an underestimation of the cost savings associated with the prevention of unintended pregnancies. On one hand, consider the assumption that a policy that delays a reportedly mistimed birth will simply induce the mother in question to put off that birth until such time as it is no longer unintended. The implication of this assumption is that delaying a mistimed pregnancy will have no effect on a woman's total lifetime fertility. This presumption is, on the margins, almost certainly incorrect, given that the occurrence of at least some pregnancies that are reported as being mistimed probably increases women's total lifetime fertility levels. In other words, it is almost certainly not true that all mistimed pregnancies prevented by a policy are merely delayed. It is also likely that, if truly mistimed (and publicly subsidized) births were delayed for long enough, some of the mothers in question would be in more favorable financial positions and would therefore no longer require means‐tested benefits for themselves or for their children. If I were to account for either of these considerations, my cost‐savings estimates would be higher.</p> <p>On the other hand, it is also probable that the prevention of some pregnancies that are reported to be unwanted will in fact simply cause them to occur at a later point in time. Accounting for this possibility would lower my estimates of the cost savings associated with the prevention of unintended pregnancies. On net, the exclusion from my analysis of the considerations described here probably exerts a downward bias on the estimated cost savings produced by pregnancy prevention policies. However, given the likely difficulty of quantifying with any precision the magnitudes of the causal relationships between the prevention of an unintended pregnancy and a woman's total lifetime fertility or between the delaying of a pregnancy and the likelihood that the mother will claim means‐tested benefits for herself or her child (ren) in the future, I do not account for these factors. The reader should therefore bear in mind that—for this and other reasons enumerated elsewhere in this paper—the benefit‐cost ratios for the policies analyzed here can be considered to be relatively conservative.</p> <p>I use a real discount rate of 3 percent to calculate the present value of 5 years' worth of publicly provided children's benefits and to calculate the present value of the public savings associated with delaying mistimed births.[<reflink idref="bib33" id="ref78">33</reflink>] My calculations suggest that, according to the measure of savings that accounts only for government spending on prenatal care, deliveries, and postpartum care, the prevention of a birth to a teenager who is under 200 percent of poverty produces an average of $1,700 in public cost savings. The equivalent estimate of cost savings for the prevention of a non‐teen birth is $2,000. The corresponding estimates for teen and non‐teen births that account additionally for public spending on infant medical care are $4,000 and $5,000, respectively, and the corresponding estimates for teen and non‐teen births that account additionally for spending on other benefits provided to young children are $19,000 and $24,000, respectively. The prevention of a fetal loss is assumed to produce an average of $750 of public savings.[<reflink idref="bib34" id="ref79">34</reflink>]</p> <hd id="AN0076170450-12">ALTERNATIVE PREGNANCY‐OUTCOME ASSUMPTIONS</hd> <p>FamilyScape does not explicitly account for the pregnancy intentions of members of the simulation population (although it does implicitly account for this dynamic to the extent that it is correlated with age, marital status, and the other demographic characteristics that are incorporated into the simulation). Unintended pregnancies are substantially more likely than intended pregnancies to result in abortions, and they are much less likely to result in live births (Finer & Henshaw, [<reflink idref="bib14" id="ref80">14</reflink>]). Under the assumption that the pregnancies prevented by a given policy would have been unintended, had they occurred, one therefore would expect them to be disproportionately likely to have resulted in abortions. Although the results of the original versions of the policy simulations are in fact consistent with this expectation, the share of prevented pregnancies that would have resulted in abortions is still lower than the share of unintended pregnancies that result in the same outcome.</p> <p>I therefore implement a second version of each policy simulation in which I assume that the number of pregnancies prevented by a given policy is the same as in the corresponding original version, but I use the results of back‐of‐the‐envelope calculations to re‐estimate the distribution of outcomes that these pregnancies would have produced, had they occurred.[<reflink idref="bib35" id="ref81">35</reflink>] Relative to the original versions of the simulations, policies are therefore assumed to prevent a smaller number of births and a larger number of abortions. Because it seems reasonable to assume that the distribution of pregnancy outcomes prevented by a given policy would be similar to the distribution of outcomes for real‐world unintended pregnancies, I consider the results of these alternate versions of the simulations to be somewhat preferable to the results originally produced by FamilyScape.</p> <hd id="AN0076170450-13">RESULTS</hd> <p>Table presents findings from the policy simulations. Shaded columns contain the results of my preferred specifications. The top portion of the table reports my estimates of these policies' effects on the incidence of pregnancy, abortion, and childbearing, and on the number of children who are born into poverty. The reader should bear in mind that a program might appear to be relatively more or less efficacious depending simply upon which reference group one chooses. For example, the teen pregnancy prevention intervention has a much smaller effect on overall pregnancy and birth rates than on rates of teenage pregnancy and childbearing. Similarly, the media campaign—which only affects the behavior of unmarried males—has a larger effect on out‐of‐wedlock pregnancy and childbearing rates than on pregnancy and childbearing rates overall.</p> <p>Estimated effects and costs of various pregnancy prevention interventions</p> <p> <ephtml> <table><thead><tr><th /><th align="center">Mass Media Campaign</th><th align="center">Evidence‐Based Teen Pregnancy Prevention Program</th><th align="center">Expanded Access to Medicaid Family Planning</th></tr><tr><th /><th align="center">Initial specification</th><th align="center">Alternative specification</th><th align="center">Initial specification</th><th align="center">Alternative specification</th><th align="center">Initial specification</th><th align="center">Alternative specification</th></tr><tr><th /><th><italic>Original pregnancy outcome assumptions</italic></th><th><italic>Unintended pregnancy outcome assumptions (preferred specification)</italic></th><th><italic>Original pregnancy outcome assumptions</italic></th><th><italic>Unintended pregnancy outcome assumptions</italic></th><th><italic>Original pregnancy outcome assumptions</italic></th><th><italic>Unintended pregnancy outcome assumptions (preferred specification)</italic></th><th><italic>Original pregnancy outcome assumptions</italic></th><th><italic>Unintended pregnancy outcome assumptions</italic></th><th><italic>Original pregnancy outcome assumptions</italic></th><th><italic>Unintended pregnancy outcome assumptions</italic></th><th><italic>Original pregnancy outcome assumptions</italic></th><th><italic>Unintended pregnancy outcome assumptions (preferred specification)</italic></th></tr></thead><tbody><tr><td>Percent reduction in pregnancies</td><td /><td /><td /><td /><td /><td /><td /><td /><td /><td /><td /></tr><tr><td> <italic>Overall</italic></td><td>1.7</td><td>1.7</td><td>1.7</td><td>1.7</td><td>0.8</td><td>0.8</td><td>0.8</td><td>0.8</td><td>3.8</td><td>3.8</td><td>1.9</td><td>1.9</td></tr><tr><td> <italic>Among unmarried females</italic></td><td>3.4</td><td>3.4</td><td>3.4</td><td>3.4</td><td>1.7</td><td>1.7</td><td>1.5</td><td>1.5</td><td>4.5</td><td>4.5</td><td>2.2</td><td>2.2</td></tr><tr><td> <italic>Among teenagers</italic></td><td>4.0</td><td>4.0</td><td>4.0</td><td>4.0</td><td>7.5</td><td>7.5</td><td>6.7</td><td>6.7</td><td>3.0</td><td>3.0</td><td>1.4</td><td>1.4</td></tr><tr><td>Percent reduction in abortions</td><td>3.1</td><td>3.9</td><td>3.1</td><td>3.9</td><td>0.9</td><td>1.4</td><td>0.9</td><td>1.2</td><td>4.5</td><td>7.1</td><td>2.3</td><td>3.5</td></tr><tr><td>Percent reduction in births</td><td /><td /><td /><td /><td /><td /><td /><td /><td /><td /><td /></tr><tr><td> <italic>Overall</italic></td><td>1.6</td><td>1.0</td><td>1.6</td><td>1.0</td><td>0.8</td><td>0.6</td><td>0.7</td><td>0.6</td><td>3.7</td><td>2.9</td><td>1.8</td><td>1.4</td></tr><tr><td> <italic>Among unmarried females</italic></td><td>4.0</td><td>2.5</td><td>4.0</td><td>2.5</td><td>2.1</td><td>1.6</td><td>1.7</td><td>1.4</td><td>4.6</td><td>3.1</td><td>2.3</td><td>1.5</td></tr><tr><td> <italic>Among teenagers</italic></td><td>4.6</td><td>3.4</td><td>4.6</td><td>3.4</td><td>8.1</td><td>6.2</td><td>6.8</td><td>5.6</td><td>3.2</td><td>2.7</td><td>1.4</td><td>1.2</td></tr><tr><td>Percent reduction in the number of children born into poverty</td><td>3.6</td><td>2.2</td><td>3.6</td><td>2.2</td><td>1.8</td><td>1.4</td><td>1.7</td><td>1.4</td><td>5.1</td><td>4.0</td><td>2.3</td><td>1.8</td></tr><tr><td>Program cost</td><td>$100,000,000</td><td>$100,000,000</td><td>$250,000,000</td><td>$250,000,000</td><td>$145,000,000</td><td>$145,000,000</td><td>$145,000,000</td><td>$145,000,000</td><td>$235,000,000</td><td>$235,000,000</td><td>$235,000,000</td><td>$235,000,000</td></tr><tr><td>Fiscal impact analysis</td><td /><td /><td /><td /><td /><td /><td /><td /><td /><td /><td /></tr><tr><td><italic>Public savings: Based on pregnancy care alone</italic></td><td /><td /><td /><td /><td /><td /><td /><td /><td /><td /><td /></tr><tr><td> <italic>Public cost savings from prevented pregnancies</italic></td><td>$57,889,415</td><td>$37,005,263</td><td>$57,889,415</td><td>$37,005,263</td><td>$45,808,023</td><td>$36,980,412</td><td>$53,299,067</td><td>$45,404,951</td><td>$311,287,747</td><td>$245,444,664</td><td>$153,708,083</td><td>$121,677,106</td></tr><tr><td> <italic>Benefit‐cost ratio</italic></td><td><italic>$0.58</italic></td><td><italic>$0.37</italic></td><td><italic>$0.23</italic></td><td><italic>$0.15</italic></td><td><italic>$0.32</italic></td><td><italic>$0.26</italic></td><td><italic>$0.37</italic></td><td><italic>$0.31</italic></td><td><italic>$1.32</italic></td><td><italic>$1.04</italic></td><td><italic>$0.65</italic></td><td><italic>$0.52</italic></td></tr><tr><td>Fiscal impact analysis</td><td /><td /><td /><td /><td /><td /><td /><td /><td /><td /><td /></tr><tr><td><italic>Public savings: Based on pregnancy care and infant medical care</italic></td><td /><td /><td /><td /><td /><td /><td /><td /><td /></tr><tr><td> <italic>Public cost savings from prevented pregnancies</italic></td><td>$141,904,217</td><td>$90,244,018</td><td>$141,904,217</td><td>$90,244,018</td><td>$101,267,942</td><td>$79,412,746</td><td>$117,238,181</td><td>$98,167,919</td><td>$734,629,793</td><td>$574,098,322</td><td>$362,866,102</td><td>$284,843,611</td></tr><tr><td> <italic>Benefit‐cost ratio</italic></td><td><italic>$1.42</italic></td><td><italic>$0.90</italic></td><td><italic>$0.57</italic></td><td><italic>$0.36</italic></td><td><italic>$0.70</italic></td><td><italic>$0.55</italic></td><td><italic>$0.81</italic></td><td><italic>$0.68</italic></td><td><italic>$3.13</italic></td><td><italic>$2.44</italic></td><td><italic>$1.54</italic></td><td><italic>$1.21</italic></td></tr><tr><td>Fiscal impact analysis</td><td /><td /><td /><td /><td /><td /><td /><td /><td /><td /><td /></tr><tr><td><italic>Public savings: Based on pregnancy care and children's benefits</italic></td><td /><td /><td /><td /><td /><td /><td /><td /><td /></tr><tr><td> <italic>Public cost savings from prevented pregnancies</italic></td><td>$678,767,789</td><td>$430,909,421</td><td>$678,767,789</td><td>$430,909,421</td><td>$462,963,071</td><td>$356,145,363</td><td>$534,232,402</td><td>$442,274,236</td><td>$3,421,483,267</td><td>$2,660,322,108</td><td>$1,689,948,584</td><td>$1,320,394,996</td></tr><tr><td> <italic>Benefit‐cost ratio</italic></td><td><italic>$6.79</italic></td><td><italic>$4.31</italic></td><td><italic>$2.72</italic></td><td><italic>$1.72</italic></td><td><italic>$3.19</italic></td><td><italic>$2.46</italic></td><td><italic>$3.68</italic></td><td><italic>$3.05</italic></td><td><italic>$14.56</italic></td><td><italic>$11.32</italic></td><td><italic>$7.19</italic></td><td><italic>$5.62</italic></td></tr></tbody></table> </ephtml> </p> <ulist> <item>5 Notes</item> <item>6 Estimates of program effects were generated using the FamilyScape simulation model. See Table for a summary of the assumptions underlying each policy analysis. For more information on FamilyScape, see Thomas and Monea ([<reflink idref="bib63" id="ref82">63</reflink>]).</item> </ulist> <p>The teen pregnancy program has by far the largest effects of the three simulated policies on rates of pregnancy and childbearing among teens, while the initial specification of the Medicaid simulation has the largest impact on the numbers of pregnancies and births overall. Indeed, the results for the alternative (and preferred) specification of the Medicaid expansion show that—even though the effects of the expansion are only about half as large as in the initial specification for this simulation—it is estimated to have a larger effect than the other two policies on these same outcomes. However, a comparison of the results from the preferred specifications of the simulated Medicaid expansion and mass media campaign suggests that the latter policy has a larger effect on out‐of‐wedlock childbearing and pregnancy and on the number of children born into poverty.</p> <p>It is also noteworthy that there is relatively little difference between the results for the initial and alternative specifications of the simulations of the teen pregnancy prevention program, even though the latter assumes no reduction in sexual activity. Most of the difference between the relative effects of the simulated increase in contraceptive use and the simulated reduction in sexual activity is a function of the fact that, at any given point in analysis time, a larger number of teens are affected by the former than by the latter. However, it is also the case that the effects of the two are not simply additive in the aggregate because the same individuals sometimes exhibit changes in both types of behavior. In other words, some of the simulated reduction in sexual activity is allocated to individuals who would already have had a substantially reduced probability of being involved in a pregnancy by virtue of the fact that they have begun to use contraception. In a separate analysis (results not shown here), I simulate the effects of a reduction in coital frequency without modeling any increase in contraceptive use. The results of this analysis suggested that the simulated diminution in sexual activity produces reductions in pregnancy and childbearing that are about a fifth as large as the effects produced by the simulated increase in contraceptive use.</p> <p>Benefit‐cost ratios for each simulated policy are reported in the bottom portion of the table. The simulated expansion of Medicaid family planning services is specifically targeted on women who are below 200 percent of poverty, which is to say that it serves women who, if they were to become pregnant, would meet the income‐eligibility criteria for a variety of means‐tested benefits for themselves and their newborn children. One might therefore expect this policy to produce the highest benefit‐cost ratios, and the results shown in Table confirm this expectation. If one uses a measure of public savings that is based on public expenditures on pregnancy‐related care, the only simulations that produce benefit‐cost ratios greater than 1 are the initial specifications of the Medicaid expansion. If one instead uses a public savings measure that accounts additionally for public spending on infant medical care, the benefit‐cost ratios for all four Medicaid‐expansion simulations are greater than 1, as is the ratio for one of the four simulations of the media campaign. And, if one uses a public savings measure that accounts for public spending on children until their fifth birthdays, all three policies have ratios that are greater than one in all specifications.</p> <p>The results of the preferred specifications for each simulation show that, even though it is easily the most expensive of the three policies, the Medicaid expansion has the highest benefit‐cost ratios. As is discussed above, this result partially reflects the fact that the Medicaid expansion disproportionately affects lower income women who are likely to qualify for the government benefits and services upon which my cost‐savings estimates are based. But it also reflects the fact that the Medicaid expansion has comparatively little "leakage": When money is spent on improving access to Medicaid‐funded contraceptive services, a relatively large share of those dollars are spent on the provision of contraception to women who are likely to use it. By contrast, the simulated sex education program serves large swaths of teens whose behaviors are unchanged as a result of the intervention. Similarly, while the media campaign reaches a large number of people relatively cheaply, it only changes the behavior of a small share of those individuals. Thus, the campaign's benefit‐cost ratios are generally higher than those of the teen pregnancy program but are generally lower than those of the Medicaid expansion.</p> <hd id="AN0076170450-14">Robustness of Results</hd> <p>As is made clear throughout this paper, I was required to make a number of simplifying assumptions in order to perform these simulations and, as a result, there is uncertainty as to the simulated policies' precise costs and effects. However, some of the qualitative conclusions implied by these findings are relatively insensitive to large changes in the assumptions underlying the analysis. For example, the results of the preferred specifications suggest that, even if the cost of the Medicaid expansion were twice as high as I assume it to be—or if the benefits of the teen pregnancy prevention program were twice what they are estimated to be—the benefit‐cost ratios for the former would still be at least as large as for the latter. Moreover, under any of the numerous specifications for these policy analyses, they almost always produce benefit‐cost ratios of less than 1 if the programs' benefits are monetized by accounting only for cost savings on pregnancy care. On the other hand, the results from the preferred specifications that account for spending on children through age 5 suggest that, even if these programs were half as effective (or twice as expensive) as I assume them to be, all of them would have benefit‐cost ratios of greater than 1.</p> <hd id="AN0076170450-15">DISCUSSION</hd> <p>This paper documents results from a series of fiscal impact policy simulations that were parameterized using estimates from the best available empirical work on the behavioral effects of three pregnancy prevention programs. I present results from simulations that adopt a variety of different assumptions regarding these policies' behavioral effects and the amount of cost savings that they might produce. Most of the variation in the estimated benefit‐cost ratios for a given policy is driven by the fact that different ratios are calculated using different cost‐savings assumptions. I would argue that the consideration of savings to programs that provide children's benefits produces a more complete assessment of these policies' fiscal implications. My preferred ratios are therefore the ones that account for spending on children up to their fifth birthdays. The benefit‐cost ratios from my preferred specifications for the simulations of a mass media campaign encouraging contraceptive use, an evidence‐based pregnancy prevention program for at‐risk teens, and an expansion of Medicaid family planning services are $4.31, $2.46, and $5.62, respectively.</p> <p>These estimates specifically reflect the projected return to taxpayers on each dollar spent. Thus, although many of the benefit‐cost ratios that do not account for public spending on children's benefits are less than 1, the results of my preferred specifications suggest that all three policies are likely to be sound public investments. Indeed, although there would probably be some overlap in their effects—for example, a mass media campaign might increase contraceptive use among some of the same at‐risk youth who fall into the target group for a teen pregnancy prevention campaign—the three policies' estimated benefit‐cost ratios are high enough that they could prove to be cost‐beneficial even if all of them were implemented simultaneously.</p> <p>I would also reiterate that I do not account for spending on children over the age of 5, for the private costs of unintended pregnancy (such as the prospect of diminished lifetime earnings on the part of the mother or of lower academic achievement on the part of the child), or for other potentially important societal costs (such as the effect that unintended childbearing might have on the contribution of a given birth cohort to the rate of crime or unemployment). Nor do I account for the likelihood that the programs studied here would slow the spread of STIs. This consideration is potentially important not only from a social perspective but also from the standpoint of public sector balance sheets—federal spending on HIV/AIDS treatment alone now totals nearly $12 billion—and it is especially relevant for interventions such as the simulated mass media campaign that focus specifically on encouraging condom use.[<reflink idref="bib36" id="ref83">36</reflink>] I do not account for these potential cost savings because of practical difficulties in measuring them. However, their inclusion in my analysis would increase the estimated benefit‐cost ratios for the simulated policies.[<reflink idref="bib37" id="ref84">37</reflink>]</p> <p>On the other hand, the reader should bear in mind that there are large confidence intervals around many of the policy simulations' most important parameters. For example, the teen pregnancy prevention program was parameterized using the results of evaluations of small‐scale interventions. Although I made an arguably conservative assumption about the likely effects that an equivalent program would have if it were implemented nationally, it is possible that such an intervention would be even more difficult to scale up effectively than is implied by my assumption. As is discussed in an earlier section of this paper, there are also reasonably large standard errors associated with the Kearney and Levine ([<reflink idref="bib32" id="ref85">32</reflink>]) estimates that are used to parameterize the simulation of the Medicaid expansion. In addition, the simulation of the national media campaign is parameterized using results from difference‐in‐differences analyses whose validity rests on an untestable and potentially incorrect identifying assumption: That, if not for the campaigns in question, changes in contraceptive use in the cities in which those campaigns were implemented would have been similar to the changes observed in "control cities" selected by these studies' authors. There are also varying levels of uncertainty in my estimates of these programs' costs and in many of the simulation model's baseline parameters. Given the uncertainty inherent in these simulations, the policy recommendations articulated here should be taken with an appropriately sized grain of salt.</p> <p>Even when viewed with a healthy dose of circumspection, however, the simulation results provide useful policy insights. My findings suggest that federal lawmakers should consider the possibility of increasing funding for evidence‐based programs designed to encourage safer sexual behavior and that statehouses in non‐waiver states should consider taking advantage of the option allowing them to expand access to Medicaid family planning services. In light of the prevalence of unintended pregnancies and the mounting personal and societal costs that they pose, policymakers would do well to give careful consideration to these measures in the near future.</p> <hd id="AN0076170450-16">ACKNOWLEDGMENTS</hd> <p>Many thanks to Emily Monea, who has been an invaluable partner in this enterprise and is the first author of one of the technical reports written in support of this paper. Thanks also to Alex Gold, Daniel Moskowitz, and Audrey Dufrechou for research assistance; to Miles Parker for his leadership in the conceptualization and development of the software and computer code that were used to create the FamilyScape model; to Katherine Suellentrop for providing helpful information on a number of different occasions; to Isabel Sawhill, Bobbie Wolfe, Julia Isaacs, Ron Haskins, Rebecca Maynard, Matt Stagner, James Trussell, Lauren Scher, Josh Epstein, Ross Hammond, Carol Graham, Sara McLanahan, Sarah Brown, Kelleen Kaye, Andrea Kane, Rachel Fey, Melissa Kearney, Larry Finer, Jennifer Frost, Kris Moore, Seth Noar, Douglas Kirby, Kathy Johnson, Janet St. Lawrence, Antonia Villaruel, Kathleen Sikkema, Susan Philliber, Jean Emans, Joshua Price, and Liz Ananat for their helpful comments and constructive insights; and to the William and Flora Hewlett Foundation for their support of this project.</p> <hd id="AN0076170450-17">Appendix</hd> <p></p> <hd id="AN0076170450-18">APPENDIX: DISCUSSION OF BACK‐OF‐THE‐ENVELOPE CALCULATIONS</hd> <p>I develop back‐of‐the‐envelope estimates of a given policy's effect on pregnancy rates by calculating annual pregnancy probabilities among women affected by the policy in question both before and after its implementation. Pregnancy probabilities are estimated as follows:</p> <p> <ephtml> <math display="block" xmlns="http://www.w3.org/1998/Math/MathML"><mrow><mi>P</mi><mi>r</mi><mi>e</mi><mi>g</mi><mi>P</mi><mi>r</mi><mi>o</mi><msub><mi>b</mi><mi>p</mi></msub><mo>=</mo><mrow><mo>(</mo><mn>1</mn><mo>−</mo><mrow><mo>(</mo><msub><mi>a</mi><mi>p</mi></msub><mo>×</mo><msup><mrow><mo>(</mo><mn>1</mn><mo>−</mo><msub><mi>f</mi><mi>p</mi></msub><mo>×</mo><mrow><mo>(</mo><mn>1</mn><mo>−</mo><mi>c</mi><mo>)</mo></mrow><mo>)</mo></mrow><msub><mi>x</mi><mi>p</mi></msub></msup><mo>+</mo><mrow><mo>(</mo><mn>1</mn><mo>−</mo><msub><mi>a</mi><mi>p</mi></msub><mo>)</mo></mrow><mo>×</mo><msup><mrow><mo>(</mo><mn>1</mn><mo>−</mo><msub><mi>f</mi><mi>p</mi></msub><mo>)</mo></mrow><msub><mi>x</mi><mi>p</mi></msub></msup><mo>)</mo></mrow><mo>)</mo></mrow><mo>,</mo></mrow></math> </ephtml> </p> <p>where <emph>PregProb<subs>p</subs></emph> is the average probability of becoming pregnant over the course of a year among the members of the target population <emph>p</emph>; <emph>a<subs>p</subs></emph> is the share of individuals in the target population who use contraception at a given act of intercourse; <emph>f<subs>p</subs></emph> is the average fecundity level of women in the target population (i.e., <emph>f<subs>p</subs></emph> is the average probability of conceiving from a single act of unprotected intercourse); <emph>c</emph> is a contraceptive efficacy rate (i.e., <emph>c</emph> is the proportional reduction in the probability of conceiving from a single act of intercourse that is brought about by the use of contraception); and <emph>x<subs>p</subs></emph> is the average number of times that women in the target population have intercourse over a period of 1 year.[<reflink idref="bib38" id="ref86">38</reflink>]</p> <p>The proportional effect of a given policy on the overall rate of pregnancy is then estimated as follows:</p> <p> <ephtml> <math display="block" xmlns="http://www.w3.org/1998/Math/MathML"><mrow><mrow><mi mathvariant="italic">Policy</mi><mspace width="0.28em" /><mi mathvariant="italic">Effect</mi></mrow><mo>=</mo><mfenced separators="" open="(" close=")"><mn>1</mn><mo>−</mo><mfenced separators="" open="(" close=")"><mi>P</mi><mi>r</mi><mi>e</mi><mi>g</mi><mi>P</mi><mi>r</mi><mi>o</mi><msubsup><mi>b</mi><mi>p</mi><mrow><mi>p</mi><mi>o</mi><mi>s</mi><mi>t</mi></mrow></msubsup><mo>/</mo><mi>P</mi><mi>r</mi><mi>e</mi><mi>g</mi><mi>P</mi><mi>r</mi><mi>o</mi><msubsup><mi>b</mi><mi>p</mi><mrow><mi>p</mi><mi>r</mi><mi>e</mi></mrow></msubsup></mfenced></mfenced><mspace width="0.16em" /><mo>×</mo><mspace width="0.16em" /><mi>P</mi><mi>r</mi><mi>e</mi><mi>g</mi><mi>S</mi><mi>h</mi><mi>a</mi><mi>r</mi><msub><mi>e</mi><mi>p</mi></msub><mo>,</mo></mrow></math> </ephtml> </p> <p>where <ephtml> <math display="inline" xmlns="http://www.w3.org/1998/Math/MathML"><mrow><mi>P</mi><mi>r</mi><mi>e</mi><mi>g</mi><mi>P</mi><mi>r</mi><mi>o</mi><msubsup><mi>b</mi><mi>p</mi><mrow><mi>p</mi><mi>o</mi><mi>s</mi><mi>t</mi></mrow></msubsup></mrow></math> </ephtml> is the annual pregnancy probability among members of the target group after the policy has been implemented, <ephtml> <math display="inline" xmlns="http://www.w3.org/1998/Math/MathML"><mrow><mi>P</mi><mi>r</mi><mi>e</mi><mi>g</mi><mi>P</mi><mi>r</mi><mi>o</mi><msubsup><mi>b</mi><mi>p</mi><mrow><mi>p</mi><mi>r</mi><mi>e</mi></mrow></msubsup></mrow></math> </ephtml> is the annual pregnancy probability of target group members before the policy was implemented, and <emph>PregShare<subs>p</subs></emph> is the share of all pregnancies that are accounted for by pregnancies within the target group.</p> <p>As is made clear from the discussion above, I use average characteristics of policies' target group members to perform these calculations. However, many of the relationships modeled in equation between the pregnancy rate and the antecedent factors that affect it are nonlinear. The reader should therefore note that the use of average quantities in this analysis—while necessary for the purpose of conducting back‐of‐the‐envelope calculations—is at least marginally inappropriate. As an illustration of this point, an early version of FamilyScape assigned to each member of the simulation population an identical "sexual‐proclivity" threshold. The value of this threshold was assigned so as to ensure that the average coital frequency among members of the simulation population was consistent with the comparable quantity as measured in survey data. We found that, under this approach, the model simulated about twice as many pregnancies as occur in the real world. We also discovered via a series of exploratory analyses that many more pregnancies are produced if a large number of people are assumed to have a small amount of sex than if a small number of people are assumed to have a large amount sex, even if the average amount of sex is the same in both scenarios.[<reflink idref="bib39" id="ref87">39</reflink>] Real‐world data suggest that, over a given time period, a small number of individuals have a substantial amount of sex, a large subset of individuals have little or no sex, and another large subset of individuals have only a moderate amount of sex. As such, FamilyScape was re‐parameterized in order to ensure that the model produced not only the correct average amount of intercourse but also the correct distribution of coital frequency. After this re‐parameterization, the outcomes produced by the model were dramatically more realistic.</p> <p>The discussion above demonstrates one of the values of a simulation model such as FamilyScape: It produces more realistic results by ensuring that the distributions of key input variables are correct. For the purposes of this exercise, however, I use only representative quantities (averages) to summarize those distributions. It is therefore unsurprising that the baseline pregnancy rates suggested by these back‐of‐the‐envelope calculations are about twice as high as they should be.[<reflink idref="bib40" id="ref88">40</reflink>] Nonetheless, it may be useful to examine the implied <emph>changes</emph> in pregnancy rates that are produced when one modifies the parameters for these calculations to reflect the presumed effects of various policy initiatives. This discussion will therefore focus not on the levels of the pregnancy rates produced by these calculations, but instead on proportional differences between those rates under various sets of assumptions.</p> <p>Summary of parameters used in back‐of‐the‐envelope calculations of simulated policies' effects.</p> <p> <ephtml> <table><tbody><tr><td><bold>Mass Media Campaign</bold></td></tr><tr><td /><td><italic>Percent of sexually active women who used contraception at last intercourse (policy specification)</italic></td><td>89.1</td></tr><tr><td /><td><italic>Average number of acts of intercourse per year (policy and baseline specifications)</italic></td><td>53</td></tr><tr><td /><td><italic>Conception probability for a single act of unprotected intercourse (policy and baseline specifications)</italic></td><td>0.045</td></tr><tr><td /><td><italic>Percent of sexually active women who used contraception at last intercourse (baseline specification)</italic></td><td>87.6</td></tr><tr><td /><td><italic>Pregnancies experienced by members of target population as a percentage of all pregnancies (policy and baseline specifications)</italic></td><td>43.9</td></tr><tr><td><bold>Teen pregnancy prevention program</bold></td></tr><tr><td /><td><italic>Percent of sexually active women who used contraception at last intercourse (policy specification)</italic></td><td>96.4</td></tr><tr><td /><td><italic>Average number of acts of intercourse per year (policy specification)</italic></td><td>31</td></tr><tr><td /><td><italic>Conception probability for a single act of unprotected intercourse (policy and baseline specifications)</italic></td><td>0.065</td></tr><tr><td /><td><italic>Percent of sexually active women who used contraception at last intercourse (baseline specification)</italic></td><td>83.1</td></tr><tr><td /><td><italic>Average number of acts of intercourse per year (baseline specification)</italic></td><td>34</td></tr><tr><td /><td><italic>Pregnancies experienced by members of target population as a percentage of all pregnancies (policy and baseline specifications)</italic></td><td>2.2</td></tr><tr><td><bold>Expanded Medicaid Family Planning services</bold></td></tr><tr><td> Married women</td><td><italic>Percent of sexually active women who used contraception at last intercourse (policy specification)</italic></td><td>84.8</td></tr><tr><td /><td><italic>Average number of acts of intercourse per year (policy and baseline specifications)</italic></td><td>88</td></tr><tr><td /><td><italic>Conception probability for a single act of unprotected intercourse (policy and baseline specifications)</italic></td><td>.045</td></tr><tr><td /><td><italic>Percent of sexually active women who used contraception at last intercourse (baseline specification)</italic></td><td>83.3</td></tr><tr><td /><td><italic>Pregnancies experienced by members of target population as a percentage of all pregnancies (policy and baseline specifications)</italic></td><td>9.3</td></tr><tr><td> Unmarried women</td><td><italic>Percent of sexually active women who used contraception at last intercourse (policy specification)</italic></td><td>88.9</td></tr><tr><td /><td><italic>Average number of acts of intercourse per year (policy and baseline specifications)</italic></td><td>50</td></tr><tr><td /><td><italic>Conception probability for a single act of unprotected intercourse (policy and baseline specifications)</italic></td><td>.045</td></tr><tr><td /><td><italic>Percent of sexually active women who used contraception at last intercourse (baseline specification)</italic></td><td>85.4</td></tr><tr><td /><td><italic>Pregnancies experienced by members of target population as a percentage of all pregnancies (policy and baseline specifications)</italic></td><td>7.8</td></tr><tr><td /><td><italic>Contraceptive efficacy rate (same value used for all calculations)</italic></td><td>.97</td></tr></tbody></table> </ephtml> </p> <p>7 Rates of contraceptive use and coital frequency are estimated via analysis of the 2002 National Survey of Family Growth (NSFG). The percent of all pregnancies that are experienced by members of policies' target populations are also estimated using the 2002 NSFG. Rates of contraceptive efficacy are estimated using data reported in Thomas and Monea ([<reflink idref="bib63" id="ref89">63</reflink>]). Conception probabilities are estimated using an equation and parameter estimates taken from Royston ([<reflink idref="bib52" id="ref90">52</reflink>]).</p> <p>Further explanation is required for some of the parameters contained in equation. The quantities <emph>a<subs>p</subs></emph> and <emph>x<subs>p</subs></emph> are estimated via tabulations of data from the 2002 NSFG, which was also used to develop many of FamilyScape's parameters. My estimate of <emph>a<subs>p</subs></emph> is based on the share of NSFG respondents who report that they used contraception at their most recent act of intercourse. A simplifying assumption is made for the purposes of this calculation that there is only one type of contraception. Thus, <emph>a<subs>p</subs></emph> reflects the share of individuals who reported having used any type of contraception at last intercourse, and the parameter <emph>c</emph> reflects a rough weighted average of the efficacy rates of the various types of contraception that are currently in use. I estimate that a roughly representative typical‐use efficacy rate as averaged across all contraceptive methods would be about.97.[<reflink idref="bib41" id="ref91">41</reflink>]</p> <p>The fecundity level <emph>f<subs>p</subs></emph> is estimated using the following functional form, which is taken from Royston ([<reflink idref="bib52" id="ref92">52</reflink>]):</p> <p> <ephtml> <math display="block" xmlns="http://www.w3.org/1998/Math/MathML"><mrow><mi>p</mi><msub><mrow><mo>(</mo><mi>c</mi><mi>o</mi><mi>n</mi><mi>c</mi><mi>e</mi><mi>p</mi><mi>t</mi><mi>i</mi><mi>o</mi><mi>n</mi><mo>)</mo></mrow><mrow><mi>i</mi><mo>,</mo><mi>t</mi></mrow></msub><mo>=</mo><mrow><mo>{</mo><msub><mi>κ</mi><mn>0</mn></msub><mo>−</mo><msub><mi>κ</mi><mn>1</mn></msub><mrow><mo>(</mo><msub><mi>A</mi><mi>i</mi></msub><mo>−</mo><mover><mi>A</mi><mo>¯</mo></mover><mo>)</mo></mrow><mo>}</mo></mrow><msub><mi>α</mi><mi>t</mi></msub><mo>,</mo></mrow></math> </ephtml> </p> <p>where <emph>p(conception)<subs>i,t</subs></emph> is the probability that individual <emph>i</emph> will conceive if she has unprotected sex on day <emph>t</emph>, <emph>A<subs>i</subs></emph> is the age of individual <emph>i</emph>, <ephtml> <math display="inline" xmlns="http://www.w3.org/1998/Math/MathML"><mover><mi>A</mi><mo>¯</mo></mover></math> </ephtml> is the mean age of all of the women in the sample, <emph>κ</emph><subs>0</subs> and <emph>κ</emph><subs>1</subs> are econometrically estimated parameters giving the age‐specific probability that a viable ovum is produced during a single menstrual cycle, and α<emph><subs>t</subs></emph> is a vector of generic probabilities of ovular fertilization indexed by the day in the menstrual cycle.[<reflink idref="bib42" id="ref93">42</reflink>] This is also the equation used to model conception probabilities within FamilyScape. Given that the probability of conception is allowed to vary by age and day in the menstrual cycle, I calculate an unweighted population‐wide average of <emph>p(conception)</emph> across all ages (individuals aged 15 to 44 are included in the simulation model) and all days in the typical menstrual cycle (1 through 28) for the purposes of this exercise.[<reflink idref="bib43" id="ref94">43</reflink>] This average is equal to about 4.5 percent.</p> <p>Because conception probabilities vary reasonably strongly by age, I calculate a separate teen‐specific fecundity level (about 6.5 percent) for the back‐of‐the envelope analysis corresponding to the teen pregnancy program. The quantities <emph>a<subs>p</subs></emph> and <emph>x<subs>p</subs></emph> are also estimated separately for each set of calculations, since (a) the simulated policies are each targeted on different demographic groups, and (b) there is important demographic variation in coital frequency and rates of contraceptive use. Given that the Medicaid simulation is presumed to affect both married and unmarried women, I perform separate calculations for those two groups. I then estimate an overall pregnancy rate as a weighted average of these two rates.</p> <p>Regarding the <emph>PregShare<subs>p</subs></emph> parameter in equation , the target population for the media campaign is comprised of unmarried individuals and the target population for the teen pregnancy prevention campaign is comprised of unmarried, low‐SES teens.[<reflink idref="bib44" id="ref95">44</reflink>] My analysis of the 2002 NSFG suggests that pregnancies within these two populations account for 43.9 percent and 2.2 percent of all pregnancies, respectively. The target population for the Medicaid expansion is comprised of both unmarried and married women who live in non‐waiver states and are below 200 percent of poverty. My analysis suggests that pregnancies among women in these two populations account for 7.8 percent and 9.3 percent of all pregnancies, respectively.</p> <p>Table A1 summarizes the parameters used to calculate back‐of‐the‐envelope estimates of each policy's effects. For the mass‐media campaign, recall that an additional 3 percent of adult unmarried males and an additional 1.5 percent of teenaged unmarried males are assumed to use contraception at a given act of intercourse as a result of the campaign. I assume that some of the men who newly use condoms will have sexual partners who are themselves already using contraception.[<reflink idref="bib45" id="ref96">45</reflink>] My tabulations of the 2002 NSFG indicate that, about 40 percent of the time when condoms are used by unmarried couples during intercourse, there are other contraceptive methods in use at the same time.[<reflink idref="bib46" id="ref97">46</reflink>] I therefore assume that 60 percent of new condom users will be using contraception during intercourse (as opposed to having unprotected sex) as a result of the campaign. According to my tabulations of the 2002 NSFG, 16.72 percent of unmarried men are teenaged. Thus, I estimate that the media campaign will cause ((.6 × (1−.1672) × 3%) + (.6 × (.1672) × 1.5%)) ≈ 1.5 percent more unmarried males to use contraception at a given sexual encounter.[<reflink idref="bib47" id="ref98">47</reflink>]</p> <p>The teen pregnancy prevention campaign is assumed to boost the number of contraceptive users by 12.5 percent among both males and females. However, the share of sexual encounters that involve the use of contraception can be expected to increase by somewhat more than this amount because new contraceptors of both genders may be paired up with partners who are not themselves using any form of protection. Using 2002 NSFG data on the share of teens whose sexual partners use contraception, I estimate that a 12.5 percent increase in contraceptive use among male and female teens would result in an increase of about 16 percent in the share of sexual encounters that involve the use of contraception. For this simulation's preferred specification, the campaign is also assumed to reduce the number of sexually active teens by 7.5 percent. Under the simplifying assumption that "newly‐celibate" teens are drawn equally from all parts of the coital‐frequency distribution, I model the effect of the intervention on this margin of behavior by assuming a 7.5 percent reduction in the average number of sexual encounters among members of the target population.[<reflink idref="bib48" id="ref99">48</reflink>]</p> <p>The preferred specification for the Medicaid expansion suggests that the policy would cause 2.5 percent more women to use contraception at a given act of intercourse. Based on the marriage rate of current users of publicly subsidized contraception, I assume that 30 percent of new contraceptive users will be married and that the remainder will be unmarried. According to my tabulations of the 2002 NSFG, roughly half of the members of this policy's target population are married. This implies that the policy would cause ((.3 × 2.5%)/.5) ≈ 1.5 percent more married women and ((.7 × 2.5%)/.5) ≈ 3.5 percent more unmarried women to use contraception at a given act of intercourse.[<reflink idref="bib49" id="ref100">49</reflink>]</p> <p>Table 5 summarizes the results of the back‐of‐the‐envelope calculations and compares them to the results of the preferred specifications for each policy simulation. For all three policies, the two sets of estimates are within 1.5 percentage points of each other in absolute terms, but the differences between them are quite a bit larger in relative terms. The back‐of‐the‐envelope estimates are about 88 percent larger and about 48 percent larger than the FamilyScape estimates for the media campaign and the teen pregnancy program, respectively. For the Medicaid simulation, the former is a little more than 11 percent smaller than the latter.</p> <p>Comparison of simulated policies' estimated effects on pregnancy rates based on FamilyScape simulations and back‐of‐the‐envelope calculations</p> <p> <ephtml> <table><thead><tr><th /><th align="center">Estimated Percent Reduction in Pregnancy Rates</th></tr><tr><th /><th>FamilyScape simulations</th><th>Back‐of‐the‐envelope calculations</th><th>Percent difference</th></tr></thead><tbody><tr><td>Mass media campaign</td><td>1.7</td><td>3.2</td><td>+88.1</td></tr><tr><td>Teen pregnancy prevention program</td><td>0.8</td><td>1.3</td><td>+47.7</td></tr><tr><td>Medicaid expansion</td><td>1.9</td><td>1.7</td><td>−11.1</td></tr></tbody></table> </ephtml> </p> <p>There are a number of explanations for the differences between these two sets of results. First and perhaps most importantly, I made a simplifying assumption for the back‐of‐the‐envelope calculations that there is only one type of contraception, and I used an estimated efficacy rate that is a rough weighted average of the efficacy rates of many different methods. In fact, condoms are somewhat less effective than is suggested by this weighted average, and oral contraception is somewhat more effective. FamilyScape models the use of various types of contraceptive methods and allows for variation in the efficacy of those methods. Thus, one might expect the back‐of‐the‐envelope estimate of the effect of the increase in condom use produced by the media campaign to be larger than the comparable estimate from the FamilyScape simulations, and one might expect the opposite to be true for the increase in pill use produced by the Medicaid expansion.</p> <p>Also important is the fact that the back‐of‐the‐envelope calculations ignore many of the subtleties captured by FamilyScape in the interrelationships between the various antecedents of pregnancy. For example, FamilyScape explicitly models the correlation between coital frequency and contraceptive use. Generally speaking, unmarried individuals who are in the top half of the coital‐frequency distribution are more likely than those in the bottom half to use female‐specific contraceptive methods and are less likely to use condoms. Thus, increases in condom use among the most sexually active unmarried individuals are disproportionately likely to involve men whose partners are already using another method. My back‐of‐the‐envelope calculations do not account for this fact and may therefore overstate the effects of interventions—such as the media campaign and the teen pregnancy program—that encourage condom use.</p> <p>A related consideration is the fact that, compared to unmarried older women, unmarried younger women tend to be more fecund but are also more likely to use contraception during intercourse. As a result, policies that increase contraceptive use among unmarried women may have a disproportionate effect on the behavior of those who are relatively less likely to become pregnant because of their lower fecundity levels.[<reflink idref="bib50" id="ref101">50</reflink>] FamilyScape accounts for this dynamic, but the back‐of‐the‐envelope calculations do not. One might therefore expect the back‐of‐the‐envelope calculations to show larger policy effects than the FamilyScape simulations, all things being equal.</p> <p>These are but representative examples of the many reasons why FamilyScape can be expected to produce different results from the back‐of‐the‐envelope calculations. Because it accounts in a more realistic way for the dynamics described here and others like them, FamilyScape can be assumed to have estimated these policies' impacts with a greater degree of precision than can the back‐of‐the‐envelope estimates.</p> <ref id="AN0076170450-19"> <title> Footnotes </title> <blist> <bibl id="bib1" idref="ref1" type="bt">1</bibl> <bibtext> On the share of pregnancies that are unintended, see Finer and Zolna [15].</bibtext> </blist> <blist> <bibl id="bib2" idref="ref2" type="bt">2</bibl> <bibtext> Information on the marriage rates of women who experience intended and unintended pregnancies is taken from unpublished tabulations of data gathered by the Guttmacher Institute. For a survey of the literature on the association between pregnancy intentions and the maternal and child outcomes described here, see Logan et al. ([38]). For a comparison, see Joyce, Kaestner, and Korenman ([28]).</bibtext> </blist> <blist> <bibl id="bib3" idref="ref7" type="bt">3</bibl> <bibtext> See Monea and Thomas ([43]) for a thorough treatment of the methods used to monetize policies' benefits; see Thomas ([62]) for more information on the way in which behavioral and cost parameters were developed for each policy simulation; and see Thomas and Monea ([63]) for a detailed description of the simulation model used to conduct these analyses.</bibtext> </blist> <blist> <bibl id="bib4" idref="ref16" type="bt">4</bibl> <bibtext> Some line items considered to be benefits from the government's perspective could be viewed as costs from another perspective. For example, consider a scenario in which a family would qualify for cash assistance after having experienced an unintended birth but would not qualify for this assistance if the birth does not occur. The prevention of that birth would produce a benefit for the government in terms of reduced public expenditures, but it might also be considered to have imposed a cost on the family in question, since the family "loses" financial resources that it would otherwise have had.</bibtext> </blist> <blist> <bibl id="bib5" idref="ref5" type="bt">5</bibl> <bibtext> For more information on the CBOLT model, see Harris, Sabelhaus, and Sevilla‐Sanz ([23]); for more information on the TRIM model, see Giannarelli, Morton, and Wheaton ([17]).</bibtext> </blist> <blist> <bibl id="bib6" idref="ref19" type="bt">6</bibl> <bibtext> This list of characteristics is a modified version of a similar list contained in Epstein's ([12]) description of agent‐based models.</bibtext> </blist> <blist> <bibl id="bib7" idref="ref20" type="bt">7</bibl> <bibtext> Agents are assigned to one of four age categories (15 to 19, 20 to 24, 25 to 29, and 30 to 44), to one of four race categories (white, black, Hispanic, and other), and to one of three educational‐attainment categories (less than high school, high‐school degree only, more than high school). Socioeconomic status is measured as a function of maternal educational attainment: Individuals are designated as low SES if their demographic profiles indicate that their mothers did not graduate from high school, and all other individuals are designated as high SES. These specifications were selected based on the results of a series of goodness‐of‐fit analyses that attempted to determine how to express as parsimoniously as possible the most important features of these variables' distributions.</bibtext> </blist> <blist> <bibl id="bib8" idref="ref21" type="bt">8</bibl> <bibtext> Individuals in the simulation population are randomly extracted with replacement from the 2002 National Survey of Family Growth (NSFG). Because the NSFG is nationally representative and the data set's weights were used during the extraction process, the demographic characteristics of the members of the simulation population are representative of the characteristics of the national population from which the NSFG's sample was drawn.</bibtext> </blist> <blist> <bibl id="bib9" idref="ref22" type="bt">9</bibl> <bibtext> The physical space within which the simulation's agents interact can best be thought of as a social space that represents the full range of schools, workplaces, bars, clubs, networking websites, and other social environments in which individuals might interact with each other. FamilyScape's pairings are based solely on individual demographics; the model does not account in any explicit way for neighborhood characteristics.</bibtext> </blist> <blist> <bibtext> The most recent year for which most of the data needed to parameterize FamilyScape are available is 2012. Thus, the model is designed to reproduce the conditions observed in that year.</bibtext> </blist> <blist> <bibtext> In other words, 50 different randomization processes are used to produce these results, and 10 years of quasi‐steady‐state data are generated for each of these 50 randomization processes. Five hundred years' worth of data are therefore used to produce each set of simulation results reported below. It takes FamilyScape about 10 hours to produce this volume of data.</bibtext> </blist> <blist> <bibtext> The appendix referenced here is available at the end of this article as it appears in JPAM online. Go to the publisher's Web site and use the search engine to locate the article at <ulink href="http://www3.interscience.wiley.com/cgi-bin/jhome/34787">http://www3.interscience.wiley.com/cgi-bin/jhome/34787</ulink>.</bibtext> </blist> <blist> <bibtext> The authors actually calculate the estimates cited here using data from a subset of studies that report results for this particular outcome.</bibtext> </blist> <blist> <bibtext> In fact, the likely direction of any omitted variable bias in the authors' estimate is unclear. In the discussion above, I make the conservative assumption that such bias is positive. To the extent that the opposite is true, the simulated media campaign would be even more cost beneficial than my estimates suggest.</bibtext> </blist> <blist> <bibtext> Because unmarried teenagers have much higher baseline rates of condom use than do unmarried adults, simulating an equivalent percentage‐point increase in the share of males in these two groups who take up condom use produces an implausibly large difference between the campaign's proportional effects on teen and non‐teen pregnancy rates. I therefore assume that the campaign would cause about 3 percent of unmarried adult males to use condoms and that it would cause about 1.5 percent of teenaged males to do the same. Even after I make these assumptions, the campaign's simulated effect on teen birth rates is modestly larger in proportional terms than its effect on birth rates among non‐teens.</bibtext> </blist> <blist> <bibtext> On the fade out of some campaigns' behavioral impacts over time, see for example Zimmerman et al. ([69]).</bibtext> </blist> <blist> <bibtext> For more information on the estimated costs and effects of the <emph>Truth</emph> campaign, see Farrelly et al. ([13]) and Holtgrave et al. ([24]); for more information on the estimated costs and effects of the <emph>VERB</emph> campaign, see Centers for Disease Control and Prevention ([6]), Government Accountability Office ([20]), Huhman et al. ([26]), and Krisberg ([36]); for more information on the estimated costs and effects of NYADMC, see Government Accountability Office ([19]), Hornik et al. ([25]), Orwin et al. ([49]), and Palmgreen et al. ([50]); and, for more information on the estimated effects of the Lexington campaign, see Zimmerman et al. ([69]). I use itemized data on the value of airtime for the Lexington campaign, on the relative sizes of the Lexington and national media markets, and on the shares of other media campaigns' total budgets that are devoted to expenditures on airtime in order to produce a rough estimate of what the cost of a national version of the Lexington campaign might be.</bibtext> </blist> <blist> <bibtext> As is discussed below, however, I also assume that the process of scaling up such a program would reduce its effectiveness.</bibtext> </blist> <blist> <bibtext> The most common methodological flaws that the authors correct are the failure to cluster one's standard errors at the level at which randomization was achieved and the failure to account for selection effects (such as, for instance, estimating the impact of a program on contraceptive use among individuals who were sexually active after baseline without accounting for the fact that the program may have influenced individuals' decisions about whether to become or remain sexually active).</bibtext> </blist> <blist> <bibtext> Before calculating an average of the cost estimates for the programs for which I have the necessary data, I inflate each one to 2008 real dollars using the CPI‐U‐RS.</bibtext> </blist> <blist> <bibtext> This average threshold was calculated using data reported in Kaiser Family Foundation ([30]) and in United States Census Bureau ([65]).</bibtext> </blist> <blist> <bibtext> About a quarter of waiver states could incrementally increase their eligibility levels somewhat further under the new law. For purposes of simplicity, however, I focus here only on the likely effects that would be realized if non‐waiver states were to implement income‐eligibility expansions.</bibtext> </blist> <blist> <bibtext> These average thresholds were calculated using data reported in Guttmacher Institute ([22]), Kaiser Family Foundation ([31]), and United States Census Bureau ([65]).</bibtext> </blist> <blist> <bibtext> For instance, my simulation results suggest that a change of this magnitude in contraceptive use among non‐teenaged women produces a reduction in the birth rate for that group of between about 7 percent and 11 percent. The corresponding differences for teens are also notable but are not as large as for non‐teens.</bibtext> </blist> <blist> <bibtext> According to my tabulations of data from the 2002 NSFG, about 30 percent of women using publicly subsidized contraceptive services are married. As such, an equivalent share of new contraceptors for this simulation is also married.</bibtext> </blist> <blist> <bibtext> Even though the two specifications of this simulation imply that the Medicaid expansion would have different effects on contraceptive use, I assume that its cost would be the same in both instances. In essence, I assume implicitly that the expansion would crowd out private insurance coverage to a lesser extent under the initial specification than under the alternative specification. In Thomas ([62]), I present very rough estimates of the upper bound of the amount of crowd out implied by these assumptions, and I compare those estimates to Cutler and Gruber's ([10]) well‐known finding that about 50 percent of the increases in public coverage associated with earlier Medicaid expansions crowded out private coverage. My calculations suggest that the amount of crowd‐out implied by the parameters for the initial specification is almost certainly lower than the corresponding estimate reported by Cutler and Gruber and that the amount of crowd out implied by the alternative specification's parameters may or may not be lower than Cutler and Gruber's estimate, depending in part on the number of women who take up Medicaid family planning services other than those involving the provision of contraception.</bibtext> </blist> <blist> <bibtext> I refer to these three categories of public spending as "pregnancy care," "infant medical care," and "children's benefits." I estimate taxpayer spending on pregnancy care by synthesizing results reported by Amaral et al. ([1]), Frost, Sonfield, and Gold ([16]), Machlin and Rhode ([39]), and Trussell et al. ([64]). All of these studies estimate the cost of medical services provided to pregnant women. My estimate of public spending on infant medical care is based on results reported in Frost, Sonfield, and Gold ([16]). I measure means‐tested taxpayer spending on children's benefits using findings reported in Macomber et al. ([40]), in which the authors calculate combined federal, state, and local spending on more than 100 programs that serve children and their families. The largest means‐tested federal programs included in Macomber et al.'s estimates are the Child Care and Development Fund, the Earned Income Tax Credit, Medicaid and SCHIP, Section 8 housing, the Supplemental Nutrition Assistance Program (which is the present‐day successor to the Food Stamp Program), Temporary Assistance to Needy Families, and the Women, Infants, and Children program.</bibtext> </blist> <blist> <bibtext> On the limited amount of federal abortion subsidies, see Sonfield, Alrich, and Gold ([58]).</bibtext> </blist> <blist> <bibtext> Not all income‐eligible women and children take up the benefits and services that are available to them. I model take‐up of these services by imputing to each eligible pregnancy an estimated benefit level that is expressed as an average calculated across all members of the eligible population. In Monea and Thomas ([43])—in which my coauthor and I detail our methods for measuring public spending on pregnant women and young children—we estimate that, even after limiting the scope of our analysis to means‐tested spending on individuals who are below 200 percent of poverty and excluding from our analysis public spending on abortion, our estimates still capture about 70 percent of all government expenditures on pregnant women and children under the age of 5.</bibtext> </blist> <blist> <bibtext> Trussell et al. ([64]) make the same assumption in their analysis of the cost effectiveness of various contraceptive methods.</bibtext> </blist> <blist> <bibtext> These estimates are based on unpublished tabulations of data aggregated by the Guttmacher Institute on the relative frequencies of mistimed and unwanted pregnancies.</bibtext> </blist> <blist> <bibtext> Unintended pregnancies are classified as being mistimed or unwanted based on the responses of currently or previously pregnant women in the 2002 NSFG to survey questions about whether and when they had intended to become pregnant before the occurrence of the focal pregnancy. A mistimed pregnancy is defined as one that occurred to a woman who reports that she became pregnant earlier than she had intended, and an unwanted pregnancy is defined as one that occurred to a woman who reports that, when she became pregnant, she did not wish to have a (another) child at all (Finer & Henshaw, [14]).</bibtext> </blist> <blist> <bibtext> To be clear, a 3 percent discount rate is used for two purposes in this analysis. First, it is used to calculate the present value of a 5‐year stream of children's benefits. Second, it is used to account for the fact that the same stream of discounted benefits has a different present value if it starts today than if it starts in the future. Amaral et al. ([1]), Aos et al. ([4]), and Trussell et al. ([64]) use a 3 percent discount rate for similar purposes, and Moore et al. ([44]) recommend the use of a discount rate of 3.5 percent for an analysis of this sort. Other analysts use different discount rates. For example, in a paper in which the authors engage in a somewhat analogous benefit‐cost analysis of a pregnancy prevention program, Wang et al. ([68]) use a 6 percent real discount rate. The Congressional Budget Office, in its estimates of the costs of government programs in out years, uses a 2 percent rate, and the Office of Management and Budget uses a 7 percent rate (Kohyama, [35]). In a series of sensitivity tests, I reestimated the benefit‐cost ratios reported below using discount rates ranging from 2 percent to 7 percent. The use of these alternative rates had only modest effects on my results, and in each sensitivity analysis, all benefit‐cost ratios that were below one using a 3 percent discount rate remained below 1, and all ratios that were above that threshold remained above it.</bibtext> </blist> <blist> <bibtext> Although fetal losses are less expensive than births, the counterfactual upon which my public savings estimates are based is one in which the prevented pregnancies do not occur at all. The prevention of a fetal loss is therefore considered to produce costs savings for taxpayers, even if those savings are smaller than for the prevention of live births. I do not distinguish between the public savings associated with the prevention of fetal losses to teenaged and non‐teenaged females because I assume that there is, in essence, no such thing as a "mistimed fetal loss." As such, there is no need to account for differences by age in the share of pregnancies resulting in fetal losses that are mistimed and unwanted.</bibtext> </blist> <blist> <bibtext> I make these calculations using unpublished tabulations of data gathered by the Guttmacher Institute. These tabulations report the share of unintended pregnancies that result in births, abortions, and fetal losses, and they are disaggregated by age and marital status.</bibtext> </blist> <blist> <bibtext> On the amount of annual federal funding for HIV/AIDS treatment, see Kaiser Family Foundation ([29]).</bibtext> </blist> <blist> <bibtext> Accounting for the public cost savings associated with STI treatment might also have implications for the rank‐ordering of these policies' benefit‐cost ratios: If I were to incorporate estimates of these savings into my analysis, the media campaign's ratios could exceed those of the Medicaid expansion.</bibtext> </blist> <blist> <bibtext> Implicit in equation is an assumption that, if (1−<emph>f<subs>p</subs></emph>) gives the probability of avoiding pregnancy from a single act of sex without contraception, then (<emph>1−f<subs>p</subs></emph>) <ephtml> <math display="inline" altimg="urn:x-wiley:02768739:pam21614:equation:pam21614-math-0001" xmlns="http://www.w3.org/1998/Math/MathML"><msup><mrow /><msub><mi>x</mi><mi>p</mi></msub></msup></math> </ephtml> gives the probability of avoiding pregnancy over <emph>x<subs>p</subs></emph> unprotected sexual encounters. I am therefore also assuming implicitly that the probability of conceiving from a single act of intercourse does not vary by the day in the menstrual cycle. This simplifying assumption is incorrect, but it makes these calculations much more tractable. Were I to account more realistically for variation in female fecundity across days in the menstrual cycle (as FamilyScape does), my estimates of annual conception probabilities would be somewhat lower than is suggested by these calculations. (Any number of algebraic examples can be used to demonstrate why this is the case. For example, X<sups>2</sups> > ((X + ɛ) × (X − ɛ)) for all ɛ.) The rest of equation can be explained as follows: With a contraceptive efficacy rate of <emph>c</emph>, the probability of avoiding pregnancy over <emph>x<subs>p</subs></emph> sexual encounters for women using contraception can be taken to be <emph>(1−f(1−c))</emph><ephtml> <math display="inline" altimg="urn:x-wiley:02768739:pam21614:equation:pam21614-math-0002" xmlns="http://www.w3.org/1998/Math/MathML"><msup><mrow /><msub><mi>x</mi><mi>p</mi></msub></msup></math> </ephtml> under the simplifying assumption that per‐act conception probabilities are independent of the day in the menstrual cycle. The equation gives a weighted average of the probability of avoiding pregnancy over <emph>x<subs>p</subs></emph> sexual encounters, where the weight <emph>a<subs>p</subs></emph> reflects the proportion of women in the target population who use contraception. One minus this weighted average then gives the average probability of conceiving over the course of a year.</bibtext> </blist> <blist> <bibtext> This result is a function of the fact that there are "diminishing marginal returns to intercourse," so to speak, in terms of the risk of becoming pregnant. In other words, because the relationship between the probability of pregnancy and coital frequency is exponential, the incremental reduction to the cumulative probability of avoiding pregnancy is larger if one has sex for the second time than if one has sex for, say, the 22nd time. Also relevant is the fact that individuals who have little or no sex are at virtually no risk of becoming pregnant. Consider an extreme case in which half of the population has no sex and the other half has unprotected sex every day. If one were simply to (incorrectly) distribute sexual frequency evenly over the entire population for the purposes of a back‐of‐the‐envelope calculation, one would conclude that all members of the population are at a very high risk of becoming pregnant, when in fact only half of the population is high‐risk and the other half is at no risk at all. These considerations explain why ignoring the characteristics of the overall distribution of coital frequency can produce overstated estimates of pregnancy rates. For more information on this topic, see Thomas and Monea ([63]).</bibtext> </blist> <blist> <bibtext> For instance, the mass media campaign parameters in Table A1 can be used in conjunction with equation to calculate a predicted baseline (pre‐policy) pregnancy rate for unmarried women. According to this calculation, the annual rate of pregnancy per 1,000 unmarried women is about 173. The actual pregnancy rate within this group, however, is about 89 (Thomas & Monea, [63]). I also used parameters not shown in Table A1 to calculate a predicted pregnancy rate for all women and compared the results of that calculation to the comparable real‐world quantity, and I found that the former was about twice as large as the latter. The disparity between the baseline pregnancy rates produced by FamilyScape and these back‐of‐the‐envelope calculations can also be attributed in part to the fact that—as is discussed in an earlier footnote—the latter rely on an incorrect assumption that conception probabilities do not vary by the day in the menstrual cycle.</bibtext> </blist> <blist> <bibtext> Much of the data and methods used to calculate this rough average are outlined in Thomas and Monea ([63]). Note that this efficacy rate as I have defined it is not the same thing as 1 minus a method's "failure rate," since failure rates tend to be tied to annual pregnancy probabilities among users of a particular method. The quantity that I am estimating here is a proportional reduction in the probability of conception from a single act of intercourse that is brought about by the use of contraception.</bibtext> </blist> <blist> <bibtext> Specifically, α<subs>t</subs> is modeled as follows: α<subs>t</subs> = exp{−(t<subs>ov</subs> − t)/λ<subs>s</subs>} for all t < t<subs>ov</subs>; α<subs>t</subs> = 1 for t = t<subs>ov</subs>; and α<subs>t</subs> = exp{−(t − t<subs>ov</subs>)/λ<subs>e</subs>} for all <emph>t</emph> > t<subs>ov</subs>, where t is an index for the day in the menstrual cycle, t<subs>ov</subs> is the day of ovulation, λ<subs>e</subs> is the average life of the egg in days, and λ<subs>s</subs> is the average life of the sperm in days. Royston ([52]) estimates the value of λ<subs>s</subs> to be 1.47, and he estimates the value of λ<subs>e</subs> to be.7. The mean age of the participants in the author's sample was 32, and he estimates the values of κ<subs>0</subs> and κ<subs>1</subs> to be.48 and.022, respectively. For more information on the way in which FamilyScape models fecundity levels, see Thomas and Monea ([63]).</bibtext> </blist> <blist> <bibtext> Within FamilyScape, however, fecundity is in fact allowed to vary by age and day in the menstrual cycle.</bibtext> </blist> <blist> <bibtext> Technically, the target populations for these two policies are comprised partially (in the case of the teen pregnancy program) or entirely (in the case of the media campaign) of men. However, men in the simulation are paired up with demographically similar women. Thus, unmarried men are paired up with unmarried women, and low‐SES male teens are paired up with low‐SES female teens. Average rates of coital frequency and contraceptive use should therefore be comparable between men and women in these demographic groups (so long as "contraceptive use" is defined in terms of whether either member of a couple uses contraception at a given sexual encounter, which is in fact the definition used for the purposes of these calculations). As such—and because the ultimate goal of these calculations is to estimate pregnancy rates among the women affected by the policies in question—I develop back‐of‐the‐envelope projections using coital‐frequency and contraceptive‐use self‐reports for the women who are paired up with the men affected by these policies.</bibtext> </blist> <blist> <bibtext> The media campaigns whose evaluation results were used to parameterize this simulation encouraged condom use not as a means of avoiding unintended pregnancy but instead as a way of preventing contraction of sexually transmitted infections (STIs). It therefore seems sensible to assume that the simulated campaign would induce condom use among some men whose partners are already using another contraceptive method (given that condoms are the most effective method for preventing STI transmission).</bibtext> </blist> <blist> <bibtext> This dynamic is explicitly modeled in FamilyScape using real‐world contraceptive‐use data.</bibtext> </blist> <blist> <bibtext> Applying the parameters contained in Table 4 to equations and , I calculate the estimated effect of the mass‐media campaign as follows: (1−(1−(.891 × (1−.045 ×.03)<sups>53</sups> +.109 × (1−.045)<sups>53</sups>))/ (1−(.876 × (1−.045 ×.03)<sups>53</sups> +.124 × (1−.045)<sups>53</sups>))) ×.439 ≈.032.</bibtext> </blist> <blist> <bibtext> As is shown in Table 4, I estimate that, among members of this intervention's target population at baseline, 83.1 percent of sexual encounters involve the use of contraception and the average number of acts of intercourse per year is 34. A 16 percent increase in contraceptive use implies that (1.16 ×.831) ≈ 96.4 percent of sexual encounters among members of the target population involve the use of contraception after the intervention's implementation. Similarly, a 7.5 percent reduction in coital frequency implies that members of the target population will, on average, have sex (34 ×.925) ≈ 31 times per year after the intervention is implemented. Following the approach outlined in the previous footnote, I therefore calculate the estimated effect of the teen pregnancy prevention program as follows: (1−(1−(.964 × (1−.065 ×.03)<sups>31</sups> +.036 × (1−.065)<sups>31</sups>))/(1−(.831 × (1−.065 ×.03)<sups>34</sups> +.169 × (1−.065)<sups>34</sups>))) ×.022 ≈.013</bibtext> </blist> <blist> <bibtext> Following the approach outlined in the previous footnotes—and bearing in mind that I allow the Medicaid expansion to have different effects on pregnancy rates among married and unmarried women—I calculate the estimated effect of the expansion as follows: (((1−(1−(0.848 × (1−0.045 × 0.03)<sups>88</sups> + 0.152 × (1−0.045)<sups>88</sups>))/ (1−(0.833 × (1−0.045 × 0.03)<sups>88</sups> + 0.167 × (1−0.045)<sups>88</sups>))) × 0.093) + ((1−(1−(0.889 × (1−0.045 × 0.03)<sups>50</sups> + 0.111 × (1−0.045)<sups>50</sups>))/ (1−(0.854 × (1−0.045 × 0.03)<sups>50</sups> + 0.146 × (1−0.045)<sups>50</sups>))) × 0.078)) ≈.017.</bibtext> </blist> <blist> <bibtext> Since, within the unmarried population, the pool of potential "new contraceptors" is disproportionately comprised of older women.</bibtext> </blist> </ref> <ref id="AN0076170450-20"> <title> REFERENCES </title> <blist> <bibtext> Amaral, G., Foster, D. G., Biggs, M. A., Jasik, C. B., Judd, S., & Brindis, C. D. (2007). Public savings from the prevention of unintended pregnancy: A cost analysis of family planning services in California. Health Services Research, 42, 1960 – 1980.</bibtext> </blist> <blist> <bibtext> Ananat, E. O., & Hungerman, D. M. (2008). The power of the pill for the marginal child: Oral contraception's effects on fertility, abortion, and maternal & child characteristics. Retrieved September 7, 2011, from <ulink href="http://www.irs.princeton.edu/seminars/pdfs/Ananat.pdf">http://www.irs.princeton.edu/seminars/pdfs/Ananat.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Ananat, E. O., Gruber, J., Levine, P. B., & Steiger, D. (2009). Abortion and selection. Review of Economics and Statistics, 91, 124 – 136.</bibtext> </blist> <blist> <bibtext> Aos, S., Lieb, R., Mayfield, J., Miller, M., & Pennucci, A. (2004). Benefits and costs of prevention and early intervention programs for youth. Olympia, WA : Washington State Institute for Policy. Retrieved September 7, 2011, from <ulink href="http://www.wsipp.wa.gov/rptfiles/04-07-3901.pdf">http://www.wsipp.wa.gov/rptfiles/04-07-3901.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Bailey, M. J. (2006). More power to the pill: The impact of contraceptive freedom on women's life cycle labor supply. Quarterly Journal of Economics, 121, 289 – 320.</bibtext> </blist> <blist> <bibtext> Centers for Disease Control and Prevention. (2010). Youth media campaign: VERB, It's What you Do. Atlanta, GA : Author. Retrieved April 1, 2010, from <ulink href="http://www.cdc.gov/youthcampaign/">http://www.cdc.gov/youthcampaign/</ulink>.</bibtext> </blist> <blist> <bibtext> Chandra, A., Martinez, G. M., Mosher, W. D., Abma, J. C., & Jones, J. (2005). Fertility, family planning, and reproductive health of U.S. women: Data from the 2002 National Survey of Family Growth. Report #23‐25. Hyattsville, MD : National Center for Health Statistics, Vital Health and Statistics. Retrieved September 7, 2011, from <ulink href="http://www.cdc.gov/nchs/data/series/sr%5f23/sr23%5f025.pdf">http://www.cdc.gov/nchs/data/series/sr%5f23/sr23%5f025.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Child Trends. (2009). Programs with mixed reviews for delaying the initiation of sexual intercourse: HIV/AIDS education. Washington, DC : Author. Retrieved September 14, 2009, from <ulink href="http://www.childtrends.org/what%5fworks/youth%5fdevelopment/ResponsibleTeen.asp">http://www.childtrends.org/what%5fworks/youth%5fdevelopment/ResponsibleTeen.asp</ulink>.</bibtext> </blist> <blist> <bibtext> Coyle, K., Basin‐Enquist, K., Kirby, D., Parcel, G., Banspach, S., Collins, J., Baumler, E., Carvajal, S., & Harrist, R. (2001). Safer Choices: Reducing teen pregnancy, HIV, and STDs. Public Health Reports, 116, 82 – 93.</bibtext> </blist> <blist> <bibtext> Cutler, D., & Gruber, J. (2006). Does public insurance crowd out private insurance? Quarterly Journal of Economics, 111, 391 – 430.</bibtext> </blist> <blist> <bibtext> Donohue III, J. J., & Levitt, S. D. (2001). The impact of legalized abortion on crime. Quarterly Journal of Economics, 116, 379 – 420.</bibtext> </blist> <blist> <bibtext> Epstein, J. M. (2006). Generative social science: Studies in agent‐based computational modeling. Princeton, NJ : Princeton University Press.</bibtext> </blist> <blist> <bibtext> Farrelly, M. C., Davis, K. C., Haviland, L., Messeri, P., & Healton, C. G. (2005). Evidence of a dose‐dose relationship between "truth" antismoking ads and youth smoking prevalence. American Journal of Public Health, 95, 425 – 431.</bibtext> </blist> <blist> <bibtext> Finer, L. B., & Henshaw, S. (2006). Disparities in rates of unintended pregnancy in the United States, 1994‐2001. Perspectives on Sexual and Reproductive Health, 38, 90 – 96.</bibtext> </blist> <blist> <bibtext> Finer, L. B., & Zolna, M. R. (in press). Unintended pregnancy in the United States: Incidence and disparities, 2006. Contraception.</bibtext> </blist> <blist> <bibtext> Frost, J. J., Sonfield, A., & Gold, R. B. (2006). Estimating the impact of expanding Medicaid eligibility for family planning services. Occasional Report No. 28. New York : Guttmacher Institute. Retrieved September 7, 2011, from <ulink href="http://www.guttmacher.org/pubs/2006/08/16/or28.pdf">http://www.guttmacher.org/pubs/2006/08/16/or28.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Giannarelli, L., Morton, J., & Wheaton, L. (2007). Estimating the anti‐poverty effects of changes in taxes and benefits with the TRIM3 microsimulation model. Washington, DC : The Urban Institute. Retrieved September 21, 2011 from <ulink href="http://www.urban.org/UploadedPDF/411450%5fEstimating%5fEffects.pdf">http://www.urban.org/UploadedPDF/411450%5fEstimating%5fEffects.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Goldin, C., & Katz, L. F. (2002). The power of the pill: Oral contraceptives and women's career and marriage decisions. Journal of Political Economy, 110, 730 – 770.</bibtext> </blist> <blist> <bibtext> Government Accountability Office. (2006a). Contractor's national evaluation did not find that the Youth Anti‐Drug Media Campaign was effective in reducing youth drug use. GAO Publication No. GAO‐06‐818. Washington, DC : Author. Retrieved September 7, 2011, from <ulink href="http://www.gao.gov/new.items/d06818.pdf">http://www.gao.gov/new.items/d06818.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Government Accountability Office. (2006b). Highlights: Media contracts, activities and financial obligations for seven financial departments. Washington, DC : Author. Retrieved September 7. 2011, from <ulink href="http://www.gao.gov/highlights/d06305high.pdf">http://www.gao.gov/highlights/d06305high.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Gruber, J., Levine, P. B., & Staiger, D. (1999). Abortion legalization and child living circumstances: Who is the "marginal child"? Quarterly Journal of Economics, 114, 263 – 291.</bibtext> </blist> <blist> <bibtext> Guttmacher Institute. (2010). State policies in brief: State Medicaid family planning eligibility expansions. New York : Author. Retrieved April 22, 2010, from <ulink href="http://www.guttmacher.org/statecenter/spibs/spib%5fSMFPE.pdf">http://www.guttmacher.org/statecenter/spibs/spib%5fSMFPE.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Harris, A. R., Sabelhaus, J., & Sevilla‐Sanz, A. (2005). Behavioral effects of Social Security reform in a dynamic microsimulation with life‐cycle agents. Working Paper # 2005‐6. Washington, DC : Congressional Budget Office. Retrieved September 21, 2011, from <ulink href="http://www.cbo.gov/ftpdocs/63xx/doc6339/2005-06.pdf">http://www.cbo.gov/ftpdocs/63xx/doc6339/2005-06.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Holtgrave, D. R., Wunderink, K. A., Vallone, D. M., & Healton, C. G. (2009). Cost‐utility analysis of the national truth campaign to prevent youth smoking. American Journal of Preventive Medicine, 36, 385 – 388.</bibtext> </blist> <blist> <bibtext> Hornik, R., Jacobsohn, L., Orwin, R., Piesse, A., & Kalton, G. (2008). Effects of the National Youth Anti‐Drug Media Campaign. American Journal of Public Health, 98, 2229 – 2236.</bibtext> </blist> <blist> <bibtext> Huhman, M. E., Potter, L. D., Duke, J. C., Judkins, D. R., Heitzler, C. D., & Wong, F. L. (2007). Evaluation of a national physical activity intervention for children: VERB campaign, 2002–2004. American Journal of Preventive Medicine, 32, 38 – 43.</bibtext> </blist> <blist> <bibtext> Jemmott III, J. B., Jemmott, L. D., & Fong, G. T. (1992). Reductions in HIV risk‐associated sexual behaviors among black male adolescents: Effects of an AIDS prevention intervention. American Journal of Public Health, 82, 372 – 377.</bibtext> </blist> <blist> <bibtext> Joyce, T. J., Kaestner, R., & Korenman, S. (2000). The effect of pregnancy intention on child development. Demography, 37, 83 – 94.</bibtext> </blist> <blist> <bibtext> Kaiser Family Foundation. (2009a). Fact sheet: Medicaid and HIV/AIDS. Menlo Park, CA : Author. Retrieved September 7, 2011, from <ulink href="http://www.kff.org/hivaids/upload/7172%5f04.pdf">http://www.kff.org/hivaids/upload/7172%5f04.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Kaiser Family Foundation. (2009b). Where are states today? Medicaid and state‐funded eligibility levels for low‐income adults. Menlo Park, CA : Author. Retrieved March 30, 2010, from <ulink href="http://www.kff.org/medicaid/upload/7993.pdf">http://www.kff.org/medicaid/upload/7993.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Kaiser Family Foundation. (2010). Income eligibility levels for children's separate CHIP programs by annual incomes and as a percent of the federal poverty level. Menlo Park, CA : Author. Retrieved March 31, 2010, from http://www.statehealthfacts. org/comparemaptable.jsp?ind=204&cat=4#notes-1.</bibtext> </blist> <blist> <bibtext> Kearney, M. S., & Levine, P. B. (2009). Subsidized contraception, fertility, and sexual behavior. Review of Economics and Statistics, 91, 137 – 151.</bibtext> </blist> <blist> <bibtext> Kirby, D. (2007). Emerging answers 2007: Research findings on programs to reduce teenage pregnancy and sexually transmitted diseases. Washington, DC : The National Campaign to Prevent Teen and Unplanned Pregnancy. Retrieved September 7, 2001, from <ulink href="http://www.thenationalcampaign.org/EA2007/EA2007%5ffull.pdf">http://www.thenationalcampaign.org/EA2007/EA2007%5ffull.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Kirby, D., Baumler, E., Coyle, K., Basen‐Enquist, K., Parcel, G. S., Harrist, R., & Baspach, S. (2004). The "Safer Choices" intervention: Its impact on the sexual behaviors of different subgroups of high school students. Journal of Adolescent Health, 35, 442 – 452.</bibtext> </blist> <blist> <bibtext> Kohyama, H. (2006). Selecting discount rates for budgetary purposes. Briefing Paper No. 29. Cambridge, MA : Harvard Law School Federal Budget Policy Seminar. Retrieved September 7, 2011, from <ulink href="http://law.harvard.edu/faculty/hjackson/DiscountRates%5f29.pdf">http://law.harvard.edu/faculty/hjackson/DiscountRates%5f29.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Krisberg, K. (2005). Successful CDC VERB campaign in danger of losing funding: Budget cuts likely. Nations Health, 35, 18 – 19.</bibtext> </blist> <blist> <bibtext> Levin, H. M., & McEwan, P. J. (2001). Cost‐effective analysis: Methods and applications, 2nd ed. Thousand Oaks, CA : Sage Publications.</bibtext> </blist> <blist> <bibtext> Logan, C., Holcombe, E., Manlove, J., & Ryan, S. (2007). The consequences of unintended childbearing: A white paper. Washington, DC : The National Campaign to Prevent Teen and Unplanned Pregnancy. Retrieved September 7, 2011, from <ulink href="http://www.thenationalcampaign.org/resources/pdf/consequences.pdf">http://www.thenationalcampaign.org/resources/pdf/consequences.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Machlin, S. R., & Rohde, F. (2007). Health care expenditures for uncomplicated pregnancies. MEPS Research Findings Report #27. Rockville, MD : Agency for Healthcare Research and Quality. Retrieved September 7, 2011, from <ulink href="http://www.meps.ahrq.gov/mepsweb/data%5ffiles/publications/rf27/rf27.pdf">http://www.meps.ahrq.gov/mepsweb/data%5ffiles/publications/rf27/rf27.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Macomber, J., Isaacs, J., Vericker, T., Kent, A., & Johnson, P. (2009). Federal expenditures on infants and toddlers in 2007. Washington, DC : The Urban Institute and The Brookings Institution. Retrieved September 7, 2011, from <ulink href="http://www.urban.org/UploadedPDF/411875%5ffederal%5fexpenditures.pdf">http://www.urban.org/UploadedPDF/411875%5ffederal%5fexpenditures.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Manlove, J., Franzetta, K., McKinney, K., Papillo, A. R., & Terry‐Humen, E. (2004). A good time: After‐school programs to reduce teen pregnancy. Washington, DC : The National Campaign to Prevent Teen and Unplanned Pregnancy. Retrieved September 7, 2011, from <ulink href="http://www.thenationalcampaign.org/resources/pdf/pubs/AGoodTime.pdf">http://www.thenationalcampaign.org/resources/pdf/pubs/AGoodTime.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Miller, A. R. (2009). Motherhood delay and the human capital of the next generation. American Economic Review, 99, 154 – 158.</bibtext> </blist> <blist> <bibtext> Monea, E., & Thomas, A. (2011). The public cost of pregnancy. Working Paper. Washington, DC : Brookings Institution. Retrieved March 28, 2011, from <ulink href="http://www.brookings.edu/~/media/Files/rc/papers/2011/03%5fpregancy%5fpublic%5fcost%5fmonea%5fthomas/03%5fpregnancy%5fpublic%5fcost%5fmonea%5fthomas.pdf">http://www.brookings.edu/~/media/Files/rc/papers/2011/03%5fpregancy%5fpublic%5fcost%5fmonea%5fthomas/03%5fpregnancy%5fpublic%5fcost%5fmonea%5fthomas.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Moore, M. A., Boardman, A. E., Vining, A. R., Weimer, D. L., & Greenberg, D. H. (2004). " Just give me a number!": Practical values for the social discount rate. Journal of Policy Analysis and Management, 23, 789 – 812.</bibtext> </blist> <blist> <bibtext> National Campaign to Prevent Teen and Unplanned Pregnancy. (2010). Summary of provisions related to teen and unplanned pregnancy in HR 3590: Patient Protection and Affordable Care Act. Washington, DC : Author. Retrieved April 2, 2010, from <ulink href="http://www.thenationalcampaign.org/policymakers/PDF/SummaryProvisions%5fTUO%5fHealthReform.pdf">http://www.thenationalcampaign.org/policymakers/PDF/SummaryProvisions%5fTUO%5fHealthReform.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> National Campaign to Prevent Teen and Unplanned Pregnancy. (2011). Types of programs funded through the Office of Adolescent Health funding announcement. Washington, DC : Author. Retrieved April 1, 2011, from <ulink href="http://www.thenationalcampaign.org/resources/pdf/Briefly%5fOAH%5fFunded%5fPrograms.pdf">http://www.thenationalcampaign.org/resources/pdf/Briefly%5fOAH%5fFunded%5fPrograms.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Noar, S. A. (2006). A 10‐year retrospective of research in health mass media campaigns: Where do we go from here? Journal of Health Communication, 11, 21 – 42.</bibtext> </blist> <blist> <bibtext> Olaiya, S. T. (2006). Medical cost savings attributable to comprehensive sex education programs that delay coitus and increase condom use among adolescents in the United States. Unpublished doctoral dissertation, The Ohio State University (Document No. OSU1135888323).</bibtext> </blist> <blist> <bibtext> Orwin, R., Cadell, D., Chu, A., Kalton, G., Maklan, D., Morin, C., Piesse, A., Sridharan, S., Steele, D., Taylor, K., & Tracy, E. (2006). Evaluation of the National Youth Anti‐Drug Media Campaign: 2004 report of findings. Rockville, MD : Westat. Retrieved September 7, 2011, from <ulink href="http://archives.drugabuse.gov/initiatives/westat/NSPY2004Report/Vol1/Report.pdf">http://archives.drugabuse.gov/initiatives/westat/NSPY2004Report/Vol1/Report.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Palmgreen, P., Lorch, E. P., Stephenson, M. T., Hoyle, R. H., & Donohue, L. (2007). Effects of the Office of National Drug Control Policy's marijuana initiative campaign on high‐sensation‐seeking adolescents. American Journal of Public Health, 97, 1644 – 1649.</bibtext> </blist> <blist> <bibtext> Patient Protection and Affordable Care Act. H.R. 3590, 111th Cong. (2010). Retrieved September 7, 2011, from <ulink href="http://www.govtrack.us/congress/bill.xpd?bill=h111-3590&tab=reports">http://www.govtrack.us/congress/bill.xpd?bill=h111-3590&tab=reports</ulink>.</bibtext> </blist> <blist> <bibtext> Royston, J. P. (1982). Basal body temperature, ovulation and the risk of conception, with special reference to the lifetimes of sperm and egg. Biometrics, 38, 397 – 406.</bibtext> </blist> <blist> <bibtext> Santelli, J., Rochat, R., Hatfield‐Timajchy, K., Gilbert B. C., Curtis, K., Cabral, R., Hirsch, J. S., Schieve, L., et al. (2003). The measurement and meaning of unintended pregnancy. Perspectives on Sexual and Reproductive Health, 35, 94 – 101.</bibtext> </blist> <blist> <bibtext> Sawhill, I., Thomas, A., & Monea, E. (2010). An ounce of prevention: Policy prescriptions for reducing the prevalence of fragile families. The Future of Children, 20, 133 – 155.</bibtext> </blist> <blist> <bibtext> Scher, L. S., Maynard, R. A., & Stagner, M. (2006). Interventions intended to reduce pregnancy‐related outcomes among adolescents. Report #2006‐12. Oslo, Norway : Campbell Systematic Reviews. Retrieved September 7, 2011, from <ulink href="http://www.sfi.dk/Default.aspx?ID=5084&Action=1&NewsId=1364&PID=11354">http://www.sfi.dk/Default.aspx?ID=5084&Action=1&NewsId=1364&PID=11354</ulink>.</bibtext> </blist> <blist> <bibtext> Sikkema, K. J., Anderson, E. S., Kelly, J. A, Winett, R. A., Gore‐Felton, C., Roffman R. A., Heckman, T. G., Graves, K., Hoffman, R. G., & Brondino, M. J. (2005). Outcomes of a randomized, controlled community‐level HIV prevention intervention for adolescents in low‐income housing developments. AIDS, 19, 1509 – 1516.</bibtext> </blist> <blist> <bibtext> Snyder, L. B., Hamilton, M. A., Mitchell, E. W., Kiwanuka‐Tondo, J., Fleming‐Milici, F., & Proctor, D. (2004). A meta‐analysis of mediated health communication campaigns on behavior change in the United States. Journal of Health Communication, 9, 71 – 96.</bibtext> </blist> <blist> <bibtext> Sonfield, A., Alrich, C., & Gold, R. B. (2008). Public funding for family planning, sterilization and abortion services, FY 1980‐2006. Occasional Report #38. New York : Guttmacher Institute. Retrieved September 7, 2011, from <ulink href="http://www.guttmacher.org/pubs/2008/01/28/or38.pdf">http://www.guttmacher.org/pubs/2008/01/28/or38.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> St. Lawrence, J., Brasfield, T. L., Jefferson, K. W., Allenyne, E., O'Bannon III, R. B., & Shirley, A. (1995). Cognitive‐behavioral intervention to reduce African American adolescents' risk for HIV infection. Journal of Consulting and Clinical Psychology, 63, 221 – 237.</bibtext> </blist> <blist> <bibtext> St. Lawrence, J. S., Crosby, R. A., Brasfield, T. L., &. O'Bannon III, R. E. (2002). Reducing STD and HIV risk behavior of substance‐dependent adolescents: A randomized controlled trial. Journal of Consulting and Clinical Psychology, 70, 1010 – 1021.</bibtext> </blist> <blist> <bibtext> Suellentrop, K. (2009). What works 2009: Curriculum‐based programs that prevent teen pregnancy. Washington, DC : The National Campaign to Prevent Teen and Unplanned Pregnancy. Retrieved September 23, 2009, from <ulink href="http://www.thenationalcampaign.org/resources/pdf/pubs/WhatWorks.pdf">http://www.thenationalcampaign.org/resources/pdf/pubs/WhatWorks.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Thomas, A. (2011). Estimating the effects and costs of three pregnancy‐prevention programs. Working Paper, 2011. Washington, DC : Brookings Institution. Retrieved March 30, 2011, from <ulink href="http://www.brookings.edu/~/media/Files/rc/papers/2011/03%5fpregnancy%5fprevention%5fthomas/03%5fpregnancy%5fprevention%5fthomas.pdf">http://www.brookings.edu/~/media/Files/rc/papers/2011/03%5fpregnancy%5fprevention%5fthomas/03%5fpregnancy%5fprevention%5fthomas.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Thomas, A., & Monea, E. (2009). FamilyScape: A simulation model of family formation. Working Paper. Washington, DC : Brookings Institution. Retrieved March 28, 2009, from <ulink href="http://www.brookings.edu/~/media/Files/rc/papers/2009/05%5ffamilyscape%5fthomas%5fmonea/05%5ffamilyscape%5fthomas%5fmonea.pdf">http://www.brookings.edu/~/media/Files/rc/papers/2009/05%5ffamilyscape%5fthomas%5fmonea/05%5ffamilyscape%5fthomas%5fmonea.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Trussell, J., Lalla, A. M., Doan, Q. V., Reyes, E., Pinto, L., & Gricar, J. (2009). Cost‐effectiveness of contraception in the United States. Contraception, 79, 229 – 230.</bibtext> </blist> <blist> <bibtext> United States Census Bureau. (2010). The 2010 statistical abstract of the United States. Washington, DC : Author.</bibtext> </blist> <blist> <bibtext> United Way of Rochester. (2010). School aged evidence‐based programs. Washington, DC : Author. Retrieved September 7, 2011, from <ulink href="http://uwgeneseecounty.org/uw/Data/Documents/YouthProgramWriteUps.pdf">http://uwgeneseecounty.org/uw/Data/Documents/YouthProgramWriteUps.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Villaruel, A. M., Jemmott III, J. B., & Jemmott, L. S. (2006). A randomized controlled trial testing an HIV prevention intervention for Latino youth. Archives of Pediatrics & Adolescent Medicine, 160, 772 – 777.</bibtext> </blist> <blist> <bibtext> Wang, L. Y., Davis, M., Robin, L., Collins, J., Coyle, K., & Baumler, E. (2000). Economic evaluation of Safer Choices. Archives of Pediatrics & Adolescent Medicine, 154, 1017 – 1024.</bibtext> </blist> <blist> <bibtext> Zimmerman, R. S., Palmgreen, P. M., Noar, S. M., Lustria, M. L. A., Lu, H. Y., & Horosewski, M. L. (2007). Effects of a televised two‐city safer sex mass media campaign targeting high‐sensation‐seeking and impulsive‐decision‐making young adults. Health Education & Behavior, 34, 810 – 826.</bibtext> </blist> </ref> <aug> <p>By Adam Thomas</p> <p>Reported by Author</p> <p></p> <p>ADAM THOMAS is a Visiting Associate Professor at the Georgetown Public Policy Institute, Washington, DC 20057. He conducted the research documented here while he was the Research Director of the Brookings Institution's Center on Children and Families, Washington, DC 20036.</p> </aug> <nolink nlid="nl1" bibid="bib18" firstref="ref4"></nolink> <nolink nlid="nl2" bibid="bib42" firstref="ref6"></nolink> <nolink nlid="nl3" bibid="bib11" firstref="ref8"></nolink> <nolink nlid="nl4" bibid="bib21" firstref="ref9"></nolink> <nolink nlid="nl5" bibid="bib45" firstref="ref14"></nolink> <nolink nlid="nl6" bibid="bib37" firstref="ref17"></nolink> <nolink nlid="nl7" bibid="bib10" firstref="ref23"></nolink> <nolink nlid="nl8" bibid="bib63" firstref="ref25"></nolink> <nolink nlid="nl9" bibid="bib12" firstref="ref27"></nolink> <nolink nlid="nl10" bibid="bib54" firstref="ref28"></nolink> <nolink nlid="nl11" bibid="bib62" firstref="ref29"></nolink> <nolink nlid="nl12" bibid="bib57" firstref="ref30"></nolink> <nolink nlid="nl13" bibid="bib13" firstref="ref32"></nolink> <nolink nlid="nl14" bibid="bib14" firstref="ref33"></nolink> <nolink nlid="nl15" bibid="bib47" firstref="ref34"></nolink> <nolink nlid="nl16" bibid="bib15" firstref="ref35"></nolink> <nolink nlid="nl17" bibid="bib16" firstref="ref36"></nolink> <nolink nlid="nl18" bibid="bib17" firstref="ref37"></nolink> <nolink nlid="nl19" bibid="bib33" firstref="ref38"></nolink> <nolink nlid="nl20" bibid="bib46" firstref="ref40"></nolink> <nolink nlid="nl21" bibid="bib61" firstref="ref43"></nolink> <nolink nlid="nl22" bibid="bib41" firstref="ref45"></nolink> <nolink nlid="nl23" bibid="bib59" firstref="ref46"></nolink> <nolink nlid="nl24" bibid="bib60" firstref="ref47"></nolink> <nolink nlid="nl25" bibid="bib56" firstref="ref48"></nolink> <nolink nlid="nl26" bibid="bib34" firstref="ref50"></nolink> <nolink nlid="nl27" bibid="bib48" firstref="ref51"></nolink> <nolink nlid="nl28" bibid="bib68" firstref="ref52"></nolink> <nolink nlid="nl29" bibid="bib27" firstref="ref53"></nolink> <nolink nlid="nl30" bibid="bib66" firstref="ref55"></nolink> <nolink nlid="nl31" bibid="bib67" firstref="ref56"></nolink> <nolink nlid="nl32" bibid="bib55" firstref="ref57"></nolink> <nolink nlid="nl33" bibid="bib19" firstref="ref58"></nolink> <nolink nlid="nl34" bibid="bib20" firstref="ref59"></nolink> <nolink nlid="nl35" bibid="bib58" firstref="ref60"></nolink> <nolink nlid="nl36" bibid="bib51" firstref="ref63"></nolink> <nolink nlid="nl37" bibid="bib22" firstref="ref64"></nolink> <nolink nlid="nl38" bibid="bib23" firstref="ref65"></nolink> <nolink nlid="nl39" bibid="bib32" firstref="ref66"></nolink> <nolink nlid="nl40" bibid="bib24" firstref="ref67"></nolink> <nolink nlid="nl41" bibid="bib25" firstref="ref68"></nolink> <nolink nlid="nl42" bibid="bib26" firstref="ref69"></nolink> <nolink nlid="nl43" bibid="bib28" firstref="ref71"></nolink> <nolink nlid="nl44" bibid="bib29" firstref="ref72"></nolink> <nolink nlid="nl45" bibid="bib30" firstref="ref73"></nolink> <nolink nlid="nl46" bibid="bib31" firstref="ref75"></nolink> <nolink nlid="nl47" bibid="bib53" firstref="ref77"></nolink> <nolink nlid="nl48" bibid="bib35" firstref="ref81"></nolink> <nolink nlid="nl49" bibid="bib36" firstref="ref83"></nolink> <nolink nlid="nl50" bibid="bib38" firstref="ref86"></nolink> <nolink nlid="nl51" bibid="bib39" firstref="ref87"></nolink> <nolink nlid="nl52" bibid="bib40" firstref="ref88"></nolink> <nolink nlid="nl53" bibid="bib52" firstref="ref90"></nolink> <nolink nlid="nl54" bibid="bib43" firstref="ref94"></nolink> <nolink nlid="nl55" bibid="bib44" firstref="ref95"></nolink> <nolink nlid="nl56" bibid="bib49" firstref="ref100"></nolink> <nolink nlid="nl57" bibid="bib50" firstref="ref101"></nolink>
Header DbId: eric
DbLabel: ERIC
An: EJ969569
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Three Strategies to Prevent Unintended Pregnancy
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Thomas%2C+Adam%22">Thomas, Adam</searchLink>
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="SO" term="%22Journal+of+Policy+Analysis+and+Management%22"><i>Journal of Policy Analysis and Management</i></searchLink>. Spr 2012 31(2):280-311.
– Name: Avail
  Label: Availability
  Group: Avail
  Data: Wiley-Blackwell. 350 Main Street, Malden, MA 02148. Tel: 800-835-6770; Tel: 781-388-8598; Fax: 781-388-8232; e-mail: cs-journals@wiley.com; Web site: http://www.wiley.com/WileyCDA
– Name: PeerReviewed
  Label: Peer Reviewed
  Group: SrcInfo
  Data: Y
– Name: PhysDesc
  Label: Physical Description
  Group: PhysDesc
  Data: PDF
– Name: Pages
  Label: Page Count
  Group: Src
  Data: 32
– Name: DatePubCY
  Label: Publication Date
  Group: Date
  Data: 2012
– Name: TypeDocument
  Label: Document Type
  Group: TypDoc
  Data: Journal Articles<br />Reports - Research
– Name: Audience
  Label: Education Level
  Group: Audnce
  Data: <searchLink fieldCode="EL" term="%22Secondary+Education%22">Secondary Education</searchLink>
– Name: Subject
  Label: Descriptors
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Medical+Services%22">Medical Services</searchLink><br /><searchLink fieldCode="DE" term="%22Family+Planning%22">Family Planning</searchLink><br /><searchLink fieldCode="DE" term="%22Prevention%22">Prevention</searchLink><br /><searchLink fieldCode="DE" term="%22Sexually+Transmitted+Diseases%22">Sexually Transmitted Diseases</searchLink><br /><searchLink fieldCode="DE" term="%22Pregnancy%22">Pregnancy</searchLink><br /><searchLink fieldCode="DE" term="%22Change+Strategies%22">Change Strategies</searchLink><br /><searchLink fieldCode="DE" term="%22At+Risk+Students%22">At Risk Students</searchLink><br /><searchLink fieldCode="DE" term="%22Simulation%22">Simulation</searchLink><br /><searchLink fieldCode="DE" term="%22Public+Policy%22">Public Policy</searchLink><br /><searchLink fieldCode="DE" term="%22Program+Descriptions%22">Program Descriptions</searchLink><br /><searchLink fieldCode="DE" term="%22Cost+Effectiveness%22">Cost Effectiveness</searchLink><br /><searchLink fieldCode="DE" term="%22Contraception%22">Contraception</searchLink><br /><searchLink fieldCode="DE" term="%22Sex+Education%22">Sex Education</searchLink><br /><searchLink fieldCode="DE" term="%22Youth+Opportunities%22">Youth Opportunities</searchLink>
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.1002/pam.21614
– Name: ISSN
  Label: ISSN
  Group: ISSN
  Data: 0276-8739
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: This paper presents results from fiscal impact simulations of three national-level policies designed to prevent unintended pregnancy: A media campaign encouraging condom use, a pregnancy prevention program for at-risk youth, and an expansion in Medicaid family planning services. These simulations were performed using FamilyScape, a recently developed agent-based simulation model of family formation. In some simulation specifications, policies' benefits are monetized by accounting for projected reductions in government expenditures on medical care for pregnant women and infants. In a majority of these specifications, policies' fiscal benefit-cost ratios are less than 1. However, in specifications that account additionally for projected savings to programs that provide a broader range of benefits and services to young children, all three policies have benefit-cost ratios that are comfortably greater than 1. The results from my preferred specifications suggest that the simulated policies would produce returns to taxpayers on each dollar spent of between $2 to $6. On the whole, the results of these simulations imply that all three policies are sound public investments. (Contains 50 footnotes, 5 tables, and 1 figure.)
– Name: AbstractInfo
  Label: Abstractor
  Group: Ab
  Data: As Provided
– Name: Ref
  Label: Number of References
  Group: RefInfo
  Data: 69
– Name: DateEntry
  Label: Entry Date
  Group: Date
  Data: 2012
– Name: AN
  Label: Accession Number
  Group: ID
  Data: EJ969569
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=eric&AN=EJ969569
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1002/pam.21614
    Languages:
      – Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 32
        StartPage: 280
    Subjects:
      – SubjectFull: Medical Services
        Type: general
      – SubjectFull: Family Planning
        Type: general
      – SubjectFull: Prevention
        Type: general
      – SubjectFull: Sexually Transmitted Diseases
        Type: general
      – SubjectFull: Pregnancy
        Type: general
      – SubjectFull: Change Strategies
        Type: general
      – SubjectFull: At Risk Students
        Type: general
      – SubjectFull: Simulation
        Type: general
      – SubjectFull: Public Policy
        Type: general
      – SubjectFull: Program Descriptions
        Type: general
      – SubjectFull: Cost Effectiveness
        Type: general
      – SubjectFull: Contraception
        Type: general
      – SubjectFull: Sex Education
        Type: general
      – SubjectFull: Youth Opportunities
        Type: general
    Titles:
      – TitleFull: Three Strategies to Prevent Unintended Pregnancy
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Thomas, Adam
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 03
              Type: published
              Y: 2012
          Identifiers:
            – Type: issn-print
              Value: 0276-8739
          Numbering:
            – Type: volume
              Value: 31
            – Type: issue
              Value: 2
          Titles:
            – TitleFull: Journal of Policy Analysis and Management
              Type: main
ResultId 1