The Efficacy of Medicaid Family Planning Waivers On Young Women: Difference-in-Difference Evidence from National Microdata Amy M. Wolaver Bucknell University.

The Efficacy of Medicaid Family The Efficacy of Medicaid Family Planning Waivers On Young Women: Planning Waivers On Young Women:

Difference-in-Difference Evidence Difference-in-Difference Evidence from National Microdatafrom National Microdata

Amy M. WolaverAmy M. WolaverBucknell UniversityBucknell University

Health Economics Interest Group Meeting, ARMHealth Economics Interest Group Meeting, ARMJune 7, 2008June 7, 2008

Alternative Views on Alternative Views on ContraceptionContraception

When the history of civilization is written, it will be a biological When the history of civilization is written, it will be a biological history and Margaret Sanger will be its heroine." history and Margaret Sanger will be its heroine." –   H.G. WellsH.G. Wells

Chastity: The most unnatural of the sexual perversions.Chastity: The most unnatural of the sexual perversions.– Aldous HuxleyAldous Huxley

For most women, including women who want to have children, For most women, including women who want to have children, contraception is not an option; it is a basic health care necessity.contraception is not an option; it is a basic health care necessity.– Louise Slaughter Louise Slaughter

You must strive to multiply bread so that it suffices for the tables You must strive to multiply bread so that it suffices for the tables of mankind, and not rather favor an artificial control of birth, which of mankind, and not rather favor an artificial control of birth, which would be irrational, in order to diminish the number of guests at would be irrational, in order to diminish the number of guests at the banquet of life.the banquet of life.– POPE PAUL VI, speech, Oct. 4, 1965POPE PAUL VI, speech, Oct. 4, 1965

The best contraceptive is the word The best contraceptive is the word nono - repeated frequently.  - repeated frequently.  – MMargaret Smith argaret Smith

The best contraceptive is a glass of cold water:  not before or The best contraceptive is a glass of cold water:  not before or after, but instead.  after, but instead.  – Author Unknown Author Unknown

IntroductionIntroduction

Half of pregnancies in US are Half of pregnancies in US are unintended (Guttmacher Institute)unintended (Guttmacher Institute)

Teen pregnancy has fallen in US but Teen pregnancy has fallen in US but remains higher than other countriesremains higher than other countries

Problems with teen childbearingProblems with teen childbearing– Public costs (Medicaid, welfare, Public costs (Medicaid, welfare,

education)education)– Health of mother & infant (low Health of mother & infant (low

birthweight, premature birth etc…)birthweight, premature birth etc…)

Family Planning Coverage for Family Planning Coverage for Poor WomenPoor Women

Medicaid accounts for over ½ of Medicaid accounts for over ½ of Federal funds for contraceptive Federal funds for contraceptive servicesservices– More than Title X fundsMore than Title X funds– Copays are prohibitedCopays are prohibited– Federal matching rate more generous Federal matching rate more generous

than for other Medicaid servicesthan for other Medicaid services

Medicaid Section 1115 Family Medicaid Section 1115 Family Planning WaiversPlanning Waivers

Provides limited (contraceptive, STD Provides limited (contraceptive, STD testing, counseling) benefits to additional testing, counseling) benefits to additional persons not on regular Medicaidpersons not on regular Medicaid

Must be budget neutral (to Medicaid) over Must be budget neutral (to Medicaid) over 5 year period5 year period– 90% federal matching rate for BC90% federal matching rate for BC– Higher than other servicesHigher than other services– Justification: reduces more costly, but lower-Justification: reduces more costly, but lower-

matched Medicaid birthsmatched Medicaid births– Additional public savings from related avoided Additional public savings from related avoided

costscosts

Political ConsiderationsPolitical Considerations

Most estimates find FP waivers cost-Most estimates find FP waivers cost-neutral or saving from federal neutral or saving from federal perspectivesperspectives

Attractive to states because of generous Attractive to states because of generous match ratematch rate

Public funding of contraception remains Public funding of contraception remains controversialcontroversial– Encourages teen sex?Encourages teen sex?– Unintended consequences?Unintended consequences?– Religious objections to any contraceptionReligious objections to any contraception

Waiver HistoryWaiver History

Two strategies:Two strategies:– Extending FP services after regular (post-Extending FP services after regular (post-

partum) Medicaid loss: 1994 Rhode Island & partum) Medicaid loss: 1994 Rhode Island & SC post-partum extensionSC post-partum extension

– Raise Income cut-off for FP services: California Raise Income cut-off for FP services: California PACT 1997PACT 1997

As of 2/1/08 Twenty-seven states have As of 2/1/08 Twenty-seven states have implementedimplemented

Variation in timing, eligibility rules, Variation in timing, eligibility rules, coverage of teens/malescoverage of teens/males

Previous Research on Public Previous Research on Public Contraceptive CoverageContraceptive Coverage

May increase provider availability (Frost et May increase provider availability (Frost et al. 2004)al. 2004)

Increases use, more effective BC methods Increases use, more effective BC methods (Forrest & Samara 1996)(Forrest & Samara 1996)

Inattention to endogeneity may lead to Inattention to endogeneity may lead to underestimates of policy efficacy (Mellor underestimates of policy efficacy (Mellor 1998)1998)

Income-related waivers reduce state birth Income-related waivers reduce state birth rates (Lindrooth & McCullough 2007)rates (Lindrooth & McCullough 2007)

MethodsMethods

Difference-in-difference-in-differenceDifference-in-difference-in-difference– Create treatment group (eligible/would Create treatment group (eligible/would

be eligible) based on waiver rules in be eligible) based on waiver rules in policy & matched statespolicy & matched states

– Two control groups: Medicaid eligible, Two control groups: Medicaid eligible, ineligible for both FP & regular Medicaidineligible for both FP & regular Medicaid

Methods, continuedMethods, continued

Stage 1: Difference-in-difference (DD)Stage 1: Difference-in-difference (DD)– Compare pre- & post-waiver outcomes Compare pre- & post-waiver outcomes

of treatment & control groups within of treatment & control groups within waiver states (DD1)waiver states (DD1)

– Repeat with treatment & control in Repeat with treatment & control in matched/comparison states (DD2)matched/comparison states (DD2)

Stage 2: Difference-in-difference-in-Stage 2: Difference-in-difference-in-difference (DDD)difference (DDD)– Compare first stage results = DD1-DD2Compare first stage results = DD1-DD2

Regression FrameworkRegression Framework

Because data are panelBecause data are panel– same women in pre- & post-, tx & same women in pre- & post-, tx &

control groups; control groups; – policy variation also occurring as cohort policy variation also occurring as cohort

ages, experiences life cycle fertility ages, experiences life cycle fertility changeschanges

itijtitijtijtijt

ijtijitijt

ijtijtijtijtititijt

StateYearEligiblePolicyYeareWaiverStat

EligibleeWaiverStatEligiblePolicyYear

PolicyYeareWaiverStatEligibleeWaiverStatPolicyYearXS

1098

76

54321

)(

)()(

)(

Regression Framework, Regression Framework, cont’dcont’d

OLS & Fixed effectsOLS & Fixed effects– Also includes time & state dummiesAlso includes time & state dummies– Individual fixed effectsIndividual fixed effects– Includes controls for age, menses y/n, Medicaid eligible, Includes controls for age, menses y/n, Medicaid eligible,

urbanurban Linear probability modelsLinear probability models

– Fixed effects complex in nonlinear models, can introduce Fixed effects complex in nonlinear models, can introduce biases (Greene 2004)biases (Greene 2004)

– Interaction effects even more complex in nonlinear Interaction effects even more complex in nonlinear models (Ai & Norton 2003)models (Ai & Norton 2003)

– But, heteroskedasticity, predictions outside 0/1 boundsBut, heteroskedasticity, predictions outside 0/1 bounds– Use LPM, correct standard errors for heteroskedasticity, Use LPM, correct standard errors for heteroskedasticity,

check against WLS estimatescheck against WLS estimates

DataData

1997 National Longitudinal Survey of 1997 National Longitudinal Survey of YouthYouth– Women aged 12-18 in 1997Women aged 12-18 in 1997– Annual waves available from 1997-2005Annual waves available from 1997-2005– Only women 14 or olderOnly women 14 or older

Policy information from Guttmacher Policy information from Guttmacher Institute, cross checked with CMS Institute, cross checked with CMS

OutcomesOutcomes

ChildbearingChildbearing– Pregnant since last interviewPregnant since last interview– Gave birth since last interviewGave birth since last interview– Pregnant w/out live birth (abortion, Pregnant w/out live birth (abortion,

miscarriages & still births combined) since last miscarriages & still births combined) since last interviewinterview

Contraceptive useContraceptive use– At last intercourseAt last intercourse– Typical pregnancy risk w/ usual BC methodTypical pregnancy risk w/ usual BC method– Percent of time use BCPercent of time use BC

Sexually active since last interviewSexually active since last interview

Policy VariationPolicy VariationVariable Fraction of Sample

In a Waiver state (pre or post waiver) 55.9 (49.7)

In a state with income eligibility waiver (pre or post waiver)

36.4 (48.1)

In any Medicaid loss extension waiver state (pre or post waiver)

4.48 (20.7)

In a post-natal extension state (pre or post waiver) 20.5 (40.3)

Simulated Income Eligibility, in waiver state 24.0 (42.7)

Simulated Income Eligibility, in comparison states 6.62 (24.8)

Simulated Extension Eligibility, in waiver states 12.12 (24.9)

Simulated Extension Eligibility, in comparison statesSimulated Extension Eligibility, in comparison states 22.6 (41.8)

Fraction of Sample Post-waiver years, all statesFraction of Sample Post-waiver years, all states 77.4 (41.8)77.4 (41.8)

N 23,583

Waiver/Comparison States CharacteristicsWaiver/Comparison States CharacteristicsVariable Waiver States Comparison States

Father’s High Grade Completed

12.5 (3.18) 12.8 (3.06)

Ratio Gross Income to Federal Poverty Level

294.6 (3.00) 305.0 (3.22)

Black 31.6 (46.49) 21.7 (41.25)

Hispanic 23.3 (42.47) 16.2 (36.82)

Year of First Sex 1999 (3.08) 1999 (3.15)

Ever had Sex 74.1 (43.73) 74.1 (44.74)

Sex since last survey 49.25 (50.00) 48.81 (49.99)

Used BC, last sex 73.1 (44.37) 73.2 (44.28)

Pregnant, last year 15.7 (36.38) 15.9 (36.57)

Pregnancy, no live birth 9.8 (29.74) 9.8 (29.70)

Pregnancy risk, usual BC 13.79 (25.3) 13.78 (28.7)

Pregnancy risk, usual BC, sexually active

21.6 (36.2) 21.5 (40.9)

Age 19.3 (3.12) 19.2 (3.15)

Urban 84.2 (43.50) 80.2 (47.92)

Medicaid Eligible 23.3 (42.25) 21.1 (40.78)

N 13,180 10,413

DDD ResultsDDD Results

OLS Fixed Effects

Outcome

DDD Std Err R2 DDD Std Err R2 N

# person

s

Any Type of Family Planning Waiver

Pregnant in past yeara -0.148

0.022

* 0.26 -0.132

0.024

* 0.19

23437

4314

Gave Birth in past yeara -0.0505

0.013

* 0.12 -0.0462

0.016

* 0.08

23437

4314

Pregnancy & non-live birtha

-0.097

0.17 * 0.16 -0.086

0.019

* 0.16

23437

4314

Sexually Active in past yeara

-0.068

0.033

* 0.30 -0.050

0.029

† 0.30

18955

4274

Pregnancy Risk, usual BC method

0.724 1.87 0.069

0.0036

0.0024

0.07

17785

4141

Used BC last sexa 0.020 0.063

0.034

-0.017

0.084

.025

6073 2903

Percent sexual intercourse used BC

5.13 4.11 0.083

9.88 4.31 † 0.06

11479

3377

Source: Author’s calculations from 1997 NLSY. Regressions also control for state, sample year, age, menses, Source: Author’s calculations from 1997 NLSY. Regressions also control for state, sample year, age, menses, Medicaid eligibility, and urbanicity. Standard errors corrected for clustering at individual level. *, † Statistically Medicaid eligibility, and urbanicity. Standard errors corrected for clustering at individual level. *, † Statistically significant at the 1%, 5% level. a. Standard errors corrected for heteroskedasticity in linear probability models.significant at the 1%, 5% level. a. Standard errors corrected for heteroskedasticity in linear probability models.

DDD Results, cont’dDDD Results, cont’d

Source: Author’s calculations from 1997 NLSY. Regressions also control for state, sample year, age, menses, Source: Author’s calculations from 1997 NLSY. Regressions also control for state, sample year, age, menses, Medicaid eligibility, and urbanicity. Standard errors corrected for clustering at individual level. *, † Statistically Medicaid eligibility, and urbanicity. Standard errors corrected for clustering at individual level. *, † Statistically significant at the 1%, 5% level. a. Standard errors corrected for heteroskedasticity in linear probability models.significant at the 1%, 5% level. a. Standard errors corrected for heteroskedasticity in linear probability models.

OLS Fixed Effects Outcome DDD Std Err R2 DDD Std Err R2 N #

persons

Income Eligibility Waivers Pregnant in past yeara 0.0326 0.025 0.29 0.0371 0.028 0.22 15189 3403

Gave Birth in past yeara -0.009 0.015 0.13 -0.0059 0.018 0.11 15189 3403

Pregnancy & non-live birtha 0.042 0.020 † 0.15 0.043 0.022 0.12 15189 3403

Sexually Active in past yeara 0.054 0.042 0.32 0.073 0.037 0.30 12455 3332

Pregnancy Risk, usual BC method

0.802 2.20 0.07 -0.140 2.46 0.062 11423 3118

Used BC last sexa 0.0766 0.080 .04 0.032 0.11 0.04 3612 1910

Percent sexual intercourse used BC

2.76 5.05 0.09 7.52 5.31 0.07 7076 2384

DDD Results, cont’dDDD Results, cont’d

Source: Author’s calculations from 1997 NLSY. Regressions also control for state, sample year, age, menses, Source: Author’s calculations from 1997 NLSY. Regressions also control for state, sample year, age, menses, Medicaid eligibility, and urbanicity. Standard errors corrected for clustering at individual level. *, † Statistically Medicaid eligibility, and urbanicity. Standard errors corrected for clustering at individual level. *, † Statistically significant at the 1%, 5% level. a. Standard errors corrected for heteroskedasticity in linear probability models.significant at the 1%, 5% level. a. Standard errors corrected for heteroskedasticity in linear probability models.

OLS Fixed Eff ects Outcome

DDD Std Err R2 DDD Std Err R2 N #

persons

Extension Waivers Pregnant in past yeara -0.229 0.060 * 0.27 -0.282 0.07 * 0.19 11911 2825

Gave Birth in past yeara -0.0445 0.042 0.12 -0.0586 0.05 0.08 11911 2825 Pregnancy & non-live

birtha -0.18 0.053 * 0.14 -0.22 0.06 * 0.17 11911 2825

Sexually Active in past yeara

0.0088 0.091 0.34 -0.041 0.08 0.33 9825 2762

Pregnancy Risk, usual BC method

2.15 4.74 0.068

2.42 4.92 0.07 9082 2554

Used BC last sexa -0.066 0.13 0.05 -0.17 0.21 0.04 2870 1529 Percent sexual

intercourse used BC 1.13 8.89 0.07 6.94 9.10 0.05 5635 1922

General ResultsGeneral Results

Decreases sexual activityDecreases sexual activity Decreases probability of pregnancy, giving Decreases probability of pregnancy, giving

birth, & combined abortion, miscarriages & birth, & combined abortion, miscarriages & stillbirthstillbirth– Large, statistically significant effectsLarge, statistically significant effects– Greater relative impact on combined abortion, Greater relative impact on combined abortion,

miscarriages & stillbirth than on giving birthmiscarriages & stillbirth than on giving birth Extension waivers have larger impactExtension waivers have larger impact No measured impact on contraceptive No measured impact on contraceptive

outcomesoutcomes

Robustness ChecksRobustness Checks Dropping pre-1997 waiver statesDropping pre-1997 waiver states

– Income eligibility waivers have negative, statistically Income eligibility waivers have negative, statistically significant impact on pregnancy & giving birthsignificant impact on pregnancy & giving birth

– Extension waivers impact same magnitude except in FE Extension waivers impact same magnitude except in FE (drops to match OLS results)(drops to match OLS results)

Dropping nonwhites increases estimates of Dropping nonwhites increases estimates of efficacyefficacy

Separate examination compared to Medicaid Separate examination compared to Medicaid eligible, other control groupeligible, other control group– More effective relative to Medicaid eligible control, More effective relative to Medicaid eligible control,

stronger impactsstronger impacts FE similar to OLS, except for extension waiversFE similar to OLS, except for extension waivers WLS estimates slightly smaller than OLS/FE WLS estimates slightly smaller than OLS/FE

TeensTeens

No statistically significant impact on No statistically significant impact on sexual activitysexual activity

Any-type waiver decreases teen Any-type waiver decreases teen pregnancy, motherhood, combined pregnancy, motherhood, combined abortion, miscarriage & still birthsabortion, miscarriage & still births

Income eligibility waivers decrease Income eligibility waivers decrease pregnancy, teen mother hoodpregnancy, teen mother hood

Extension waivers decrease teen Extension waivers decrease teen pregnancypregnancy

Rural/Urban DifferencesRural/Urban DifferencesTable 7: Difference-In-Difference-In-Difference, Multivariate Results Rural Urban Differences

Rural Urban Outcome DDD S.E. DDD S.E.

Income Eligibility Waivers Pregnant 0.0116 0.056 0.028 0.033

Gave Birth 0.0099 0.039 -0.130 0.214 Pregnant, no live

birth 0.003 0.038 0.040 0.028

Sexually Active 0.13 0.07 0.061 0.044 Pregnancy Risk 2.91 4.32 -0.111 3.18 Use BC last sex 0.26 0.16 ‡ 0.064 0.13

Extension Waivers Pregnant -0.647 0.14 * -0.216 0.082 *

Gave Birth -0.415 0.17 * -0.0304 0.059 Pregnancy, no live

birth -0.23 0.18 -0.18 0.065 *

Sexually Active 0.12 0.17 -0.042 0.083 Pregnancy Risk -7.79 9.34 4.69 6.08 Use BC last sex 0.30 0.25 -0.21 0.29

Source: Author’s calculations from 1997 NLSY. Regressions also control for state, sample year, age, menses, and Medicaid eligibility. Standard errors corrected for clustering at individual level. Bolded entries are statistically significantly diff erent from urban set. *, † Statistically significant at the 1%, 5% level.

Future DirectionsFuture Directions BC consistency of use sensitive to outliers?BC consistency of use sensitive to outliers? More work on unplanned/unwanted pregnancyMore work on unplanned/unwanted pregnancy Other pregnancy outcomes (spacing, prenatal Other pregnancy outcomes (spacing, prenatal

care, low birth weight / premature birth)care, low birth weight / premature birth) MalesMales Other aspects of policy (enrollment practices Other aspects of policy (enrollment practices

etc…)etc…) Older womenOlder women Other reproductive health policies, provider Other reproductive health policies, provider

availability (addition of Guttmacher Institute data)availability (addition of Guttmacher Institute data) Cost-benefit ratiosCost-benefit ratios

ConclusionsConclusions

Effective at reducing pregnancies, Effective at reducing pregnancies, combined abortions, miscarriages & combined abortions, miscarriages & stillbirths, bigger effects for teensstillbirths, bigger effects for teens

No increased sexual activityNo increased sexual activity Results apply to all eligible young Results apply to all eligible young

women, not just participantswomen, not just participants Robust to sampling assumptionsRobust to sampling assumptions DDD + fixed effects provide strong DDD + fixed effects provide strong

support for waiverssupport for waivers