The effect of abortion restrictions on the timing of abortions

22 Pages·2001·0.38 MB·English

JournalofHealthEconomics20(2001)1011–1032
The effect of abortion restrictions on the
timing of abortions
Marianne Bitlera,∗, Madeline Zavodnyb
aNICHDPost-DoctoralFellow,RAND,1700MainStreet,SantaMonica,CA90407,USA
bNICHD,RAND,Post-DoctoralFellow,Atlanta,GA,USA
Received1January1999;accepted14July2001
Abstract
Thispaperusesdataonthedistributionofabortionsbyweeksofgestationtoexaminethere-
lationship between abortion restrictions and the timing of abortions. State-level data from 1974
to1997indicatethatadoptionofparentalinvolvementlawsforminorsorenforcementofmanda-
torywaitingperiodsispositivelyassociatedwiththepost-firsttrimesterpercentageofabortions.
However,autocorrelation-correctedspecificationsindicatethatenforcedparentalinvolvementlaws
increasetheshareoflater-termabortionsbyloweringthefirsttrimesterabortionrateratherthanby
delayingabortions.Medicaidfundingrestrictionsgenerallydonothaveasignificanteffectonthe
timingofabortionsinourresults.©2001ElsevierScienceB.V.Allrightsreserved.
JELclassification:I10;I18
Keywords:Abortion;Medicaid;Parentalconsent
1. Introduction
Sincethe1973SupremeCourtdecisioninRoev.WademadeabortionlegalacrosstheUS,
anumberoffederalandstatelawshaverestrictedwomen’saccesstoabortion,particularly
forlow-incomewomenandminors.Thefederalgovernmentandmanystateshavestopped
funding most abortions for Medicaid recipients, and some states have begun requiring
minors to notify their parents or to obtain parental consent before having an abortion.
Several states have also imposed mandatory waiting periods before women may obtain
abortions. Such laws may lower the number of abortions in a state by making it more
difficultforwomentoobtainabortions,andtheymayalsohavemoresubtleeffects,such
asdelayingthetimingofabortionsuntillaterinthepregnancy.
∗Correspondingauthor.Tel.:[phone];fax:[phone].
E-mailaddress:[email protected](M.Bitler).
0167-6296/01/$–seefrontmatter©2001ElsevierScienceB.V.Allrightsreserved.
PII:S0167-6296(01)00106-0
1012 M.Bitler,M.Zavodny/JournalofHealthEconomics20(2001)1011–1032
ThisstudyexamineswhetherMedicaidfundingrestrictions,parentalnotificationorcon-
sentlaws,andmandatorywaitingperiodsaffectthetimingofabortionsbyweeksofges-
tation. Medicaid restrictions may delay abortions among low-income women who need
additionaltimetogatherfundstogetherintheabsenceofMedicaidcoverage,andparental
involvement laws may delay or prevent abortions among teens reluctant to inform their
parents.AlthoughtheselawsdirectlyaffectonlyMedicaidrecipientsandminors,theymay
impactthetimingofabortionsamongotherwomeniftherestrictionsresultinfewerabor-
tion providers in an area. Mandatory waiting periods may delay abortions by requiring
womentomakemorethanonevisittoaprovider.Adoptionofarestrictionmayalsolead
to confusion or other information problems that lead to delay. Some women may not be
awareoftherestrictionsuntiltheyattempttogetanabortionandarethendelayed;these
womenmighthaveadaptedtheirbehaviorandsoughtanabortionearlieriftheyknewabout
therestrictions.1
Laterabortionsareofconcerntowomen,healthpractitioners,andpolicymakersforsev-
eralreasons.Theriskofdeathormajorcomplicationsisatleasttwiceashighforapost-first
trimesterabortionasforafirsttrimesterabortion(Atrashetal.,1990).Inaddition,thelike-
lihoodofmajorcomplicationsisatleastthreetimeshigherforsecondtrimesterabortions
thanforabortionsperformedat8weeksofgestationorearlier(TietzeandHenshaw,1986).
Fewer than one-half of abortion facilities surveyed by the Alan Guttmacher Institute in
1993offeredservicesat13weeks,withtheproportiondecliningrapidlyathigherweeksof
gestation(Henshaw,1995a).Theaverageclinicchargeforafirsttrimesterabortionisabout
one-halftheaveragechargeforanabortionat16weeks(Henshaw,1982,1995a).Womenare
alsomorelikelytoexperienceanegativeemotionalreactiontoasecondtrimesterabortion
thantoafirsttrimesterabortion(CouncilonScientificAffairs,1992).
We use annual state-level data to examine the association between the timing of abor-
tions,theabortionrate,andthepresenceofabortionrestrictionsinastateduring1974–1997.
Becausewefindevidenceofautocorrelationwithinstates,wepresentbothordinaryleast
squares(OLS)andautocorrelation-correctedresults.Bothspecificationsindicatethaten-
forcedandenjoinedparentalinvolvementlawsarepositivelyassociatedwiththepercent-
ageofabortionsoccurringafterthefirsttrimester.Ingeneral,specificationscorrectingfor
autocorrelationshowsmallerdirecteffectsofabortionrestrictionsthandotheOLSspeci-
fications.Specificationscorrectingforautocorrelationwithinstatesindicatethatenforced
parentalinvolvementlawslowertheoverallabortionrate,suggestingthatthelawsdiscour-
agesomewomenfromhavingabortionsbutdonotdelayabortions,whereasOLSresults
suggestthatthelawsaffectthetimingofabortions.Enforcedwaitingperiods,incontrast,do
notaffectthetotalabortionrateineithersetofresults.Medicaidfundingrestrictionshave
littleeffectonthenumberortimingofabortionswhencorrectingforautocorrelation,al-
thoughsomeoftheOLSspecificationssuggestthatMedicaidfundingrestrictionsdecrease
thepercentageofpost-firsttrimesterabortions.
Webrieflydiscussthehistoryofabortionrestrictionsandthepreviousliteratureonthe
effectoftheserestrictionsonabortionratesinSection2.Section3discussesourestimation
1Utilitymaximizingmodelswithnon-exponentialdiscountingalsoimplythatsomewomenmightwaitlonger
ifnewrestrictionsareviewedasraisingthecostofgettinganearlyabortionrelativetogettingalaterone.Insome
models,peopleavoidmakingadecisionlongerwhenthecostsofanearlydecisionarehigher(Laibson,1998).
M.Bitler,M.Zavodny/JournalofHealthEconomics20(2001)1011–1032 1013
strategy.Section4discussesthedataonabortions,abortionrestrictionsandothercontrols.
Theeffectofabortionrestrictionsonthetimingofabortionsandtheabortionrateisdiscussed
inSection5.TherobustnessofourresultsisexaminedinSection6,andSection7concludes
thestudy.
2. Background
Rather than settling the issue of abortion availability, Roe v. Wade spawned numerous
federal and state laws regulating abortion access and court injunctions concerning those
laws. Some of the most contentious issues involve insurance coverage of abortions for
Medicaidrecipients,parentalnotificationorconsentrequirementsforminorsandmandatory
waitingperiodsforwomenseekingabortions.
Thefederalgovernmentandmanystatesrestrictpublicfundingofmostabortionsunder
Medicaid, the public health insurance program for low-income families. Congressional
legislation in 1976 cut off federal Medicaid funds for almost all abortions, but a court
injunctiondelayedimplementationofthelawuntilAugust1977.Thefederallawwasagain
enjoinedfor7monthsin1980buthasbeencontinuouslyineffectsinceSeptember1980.2
StatescanusetheirownfundstopayforabortionsundertheMedicaidprogram,butfew
have opted to do so. Only 17 states and the District of Columbia funded most abortions
under the Medicaid program in 1997, for example, with some required to do so by court
order.
Parentalnotificationorconsentlawsareanotherrestrictioncommonlyadoptedbystates.
Such laws require that a woman under the age of 18 years notify her parents or obtain
theirconsentbeforeanabortioncanbeperformed.Thecaselawonparentalinvolvement
lawsiscomplex,butcourtshavegenerallyupheldlawsthatincorporateajudicialbypass
mechanism, which allows a minor to petition a court for permission to have an abortion
instead of involving her parents.3 In 1997, 27 states enforced parental involvement laws
foratleastpartoftheyear.Inaddition,courtshaveenjoinedasubstantialnumberofstate
parentalinvolvementlaws.
Somestateshavealsoadoptedmandatorywaitingperiods.Suchlawstypicallyrequire
womentoreceiveinformationaboutabortionproceduresandalternativestoabortionand
then wait a certain number of hours before the procedure can be performed. In 1992,
Mississippibecamethefirststatetoenforcesuchapolicy;mandatorydelaylawswerein
effectin10statesin1997andwereenjoinedinfivestatesforatleastpartoftheyear.
Severalstudieshaveexaminedtheeffectofabortionrestrictionsonabortionrates.Blank
et al. (1996), Haas-Wilson (1993), and Levine et al. (1996) find that the abortion rate in
a state, defined as the number of abortions per 1000 women aged between 15 and 44
years, is negatively associated with the presence of a Medicaid funding restriction. Cook
etal.(1999)concludethatabortionratesfellwhenNorthCarolinadidnotprovidepublic
2Federalfundinghascontinuouslycoveredabortionsnecessarytosavethelifeofthewoman.Federalfunding
hasalsocoveredabortionsforpregnanciesresultingfromrapeorincestsinceOctober1993.However,somestates
havenotextendedMedicaidcoveragetoabortionsforpregnanciesresultingfromrapeandincest.
3Merzetal.(1995)provideadetailedhistoryofcourtdecisionsonstateparentalinvolvementlawsandMedicaid
fundingrestrictions.
1014 M.Bitler,M.Zavodny/JournalofHealthEconomics20(2001)1011–1032
funding for all abortions eligible for Medicaid funding. Matthews et al. (1997) report a
negativeassociationbetweentheabortionrateamongstateresidentsandMedicaidfunding
restrictionsandparentalinvolvementlawsinsomespecifications;however,theyfindthat
theestimatedrelationshipsarenotsignificantwhenstate-specifictimetrendsareincluded.
Blanketal.(1996)reportthatparentalinvolvementlawsdonotappeartoreduceabortion
rates among women aged between 15 and 44 years, whereas Haas-Wilson (1996) finds
that the abortion rate among minors is lower in states that enforce parental involvement
laws.
A few studies have examined the effect of restrictions on the timing of abortions. En-
forcementofparentalinvolvementlawsinMississippiandMinnesotaledtoanincreasein
thefractionofabortionstominorsoccurringafterthefirsttrimesterrelativetothefraction
among older women (Henshaw, 1995b; Rogers et al., 1991).4 Adoption of a mandatory
delay law in Mississippi in 1992 also appears to have increased the fraction of abortions
occurringafter12weeksofgestation(AlthausandHenshaw,1994;Joyceetal.,1997;Joyce
andKaestner,2000,2001).AveragegestationatabortionishigheramongMedicaid-eligible
womenthanamongnon-eligiblewomeninstateswithMedicaidrestrictions,whereasges-
tations are similar for the two groups in states without restrictions (Henshaw and Wal-
lisch, 1984; Trussell et al., 1980). Henshaw and Wallisch (1984) estimate that 22% of
Medicaid-eligiblewomenwhohadsecondtrimesterabortionswouldhavehadfirsttrimester
abortions if the lack of public funds had not resulted in delay as women tried to raise
funds.
Ourpaperaddstothisliteratureacomprehensiveexaminationoftheeffectonthetimingof
abortionsofthethreemaintypesofabortionrestrictions.Previousstudieshaveexamined
the effect on abortion timing when one or a few states adopted an abortion restriction.
Our findings, based on panel data over a period of up to 24 years from 48 states and the
District of Columbia, indicate whether the findings of these earlier studies hold across
most of the country. We also examine the effect of both enforced and enjoined Medicaid
funding restrictions, parental involvement laws and mandatory delay laws, whereas most
previousstudiesfocusedonasingleenforcedrestriction.Asdiscussedbelow,manystates
haveadoptedmorethanonetypeofrestrictionandhavehadrestrictionsenjoinedatsome
point.5 In addition, we examine the relationship between abortion restrictions and the
abortionrateinordertoconcludewhetherrestrictionschangethetimingofabortions,the
number,orboth.Thisstudyalsoinvestigatestherobustnessoftheresultstoassumptions
aboutidentification.
4However,Rogersetal.(1991)emphasizethatthelateabortionrateamongwomenaged15–17yearsfellafter
Minnesota’sparentalnotificationlawwentintoeffect;thedeclinewassmallerthanthedeclineinthefirsttrimester
abortionrateamongminors,causingthefractionofabortionsafterthefirsttrimestertoincrease.
5Forexample,theMississippimandatorydelaylawwasenjoinedforayearbeforegoingintoeffectinAugust
1992.AlthausandHenshaw(1994),Joyceetal.(1997),andJoyceandKaestner(2000,2001)includetheperiod
whenthelawwasenjoinedinthe“before”periodwhencomparingabortionratesbeforeandafterthelawwent
intoeffect,buttheinjunctionmayhaveaffectedthebehaviorofwomenand/orproviders.Inaddition,aparental
consentlawthathadbeenenjoinedwentintoeffectinMississippiin1993.JoyceandKaestner(2001)conclude
thatenforcementofMississippi’sparentalconsentlawledtolaterabortionsamongminorsbutdonotexamine
whetherthelawhadaneffectwhileitwasenjoined.AlthausandHenshaw(1994),andJoyceetal.(1997)include
minorsintheirsampleswithoutaddressingthepossibleeffectsoftheparentalconsentlaw.
M.Bitler,M.Zavodny/JournalofHealthEconomics20(2001)1011–1032 1015
3. Estimationmethodology
In a simple rational choice model, women’s decision whether and when to have an
abortioninvolvesseveralsteps.First,womendecidewhethertohavesex.Conditionalon
having chosen to have sex, women choose contraceptive behavior, and if they become
pregnant, women choose whether to abort the pregnancy or carry it to term. Changes in
lawscanaffectwomendifferentlyatvariousstagesinthefertilitydecisionprocess.Women
whoarealreadypregnantwhenarestrictionisadoptedcanonlyadjusttheirabortion/birth
decision.Thepassageofalawmightcausethesewomentodelaytheprocedure(becausethe
costofanabortionisnowhigher)ortoinsteadcarrythepregnancytoterm.Similarly,ifa
restrictionisenjoined,somepregnantwomenmighthaveanabortioninsteadofgivingbirth,
orwomenwhohavealreadydecidedtoabortmightdosoearlierinthepregnancy.Women
who are not yet pregnant when an abortion restriction is adopted have more dimensions
onwhichtoadjusttheirbehaviorthanwomenwhoarealreadypregnant.Theycanchange
theirlevelofsexualactivityorcontraceptivebehavioraswellastheirabortion/birthdecision
(KaneandStaiger,1996).
Werelyonasimpleunderlyingassumptioncommontothisliterature,namelythatlaws
restricting access to abortion increase the cost of obtaining an abortion, leading to fewer
abortionsandmorebirthsortochangesinthetimingofabortionsorboth.6 Someofthese
lawsshouldonlyincreasethecostforcertainsub-populations.Forexample,lawsrestricting
minors’accesstoabortionsshouldonlyaffectminors.However,parentalinvolvementlaws
mayaffectfertilityamongolderwomenaswellifthelawsleadtoreducedaccesstoabortion
providersforallwomen.Weregressvariousmeasuresofthenumberofabortionsandwhen
theyoccuronmeasuresofabortionrestrictionstoestimatetheeffectofrestrictionsamong
allwomenagedbetween15and44years.
We estimate the relationship between measures of abortions and abortion restrictions
usingregressionsoftheform:
A =P β +X β +δ +ω +ε (1)
st st 1 st 2 s t st
or
A =P β +X β +δ +γ∗trend+ω +ε (2)
st st 1 st 2 s s t st
where ‘s’ indexes states and ‘t’ indexes years. The dependent variable A is one of three
measures of abortions: the percentage of abortions after the first trimester, the post-first
trimester abortion rate, or the overall abortion rate. The percentage of abortions after the
firsttrimestervariableisinlevels,andtheabortionratevariablesareinlogs.7 Thevector
Pincludesthemeasuresofstateabortionrestrictions.ThevectorXincludesothervariables
thatmaybedeterminantsofabortionratesorthetimingofabortions,suchasdemographic
characteristicsofwomeninthestate,economicconditionsinthestate,andpoliticalclimate
of the state. Observations are weighted using the population of women aged between 15
and44yearsineachstateandyear.Somespecificationscorrectforautocorrelationusing
6SeeKlerman(1998)foralongertheoreticaldiscussionoftheeffectofabortionrestrictionsonfertility.
7Intheregressionanalysis,statesthatreportzeropost-firsttrimesterabortionsareassignedratesof0.1per
thousandwomen.Resultsavailablefromtheauthorsshowthattheresultsarenotsensitivetothetreatmentofzero.
1016 M.Bitler,M.Zavodny/JournalofHealthEconomics20(2001)1011–1032
thePrais–WinstenmethodasoutlinedinBhargavaetal.(1982)becausetestsoverwhelming
rejectedthehypothesisofnoautocorrelation,asdiscussedbelow,whileotherspecifications
reportOLSestimatesnotcorrectedforautocorrelation.8
Usingstate-levelpaneldataandafixedeffectsmethodologyhasseveraladvantages.The
statefixedeffectsδcaptureunobservabledifferencesthatareconstantovertimeacrossstates.
Regressionswithstatefixedeffectsmeasuretherelationshipbetweenthedependentvariable
andthecovariateswithinastateratherthanacrossstates.Theyearfixedeffectsωcapture
time-varyingfactorscommontoallstatesinagivenyear,suchasthenationalbusinesscycle.
In effect, the regressions measure the relationship between the abortion variable and the
presenceofanabortionrestrictioninthatstateandyear.Somespecifications,asrepresented
byEq.(2),alsoincludestate-specificlineartimetrends.
However,thefixedeffectsapproachalsohasdisadvantages.Theuseoffixedeffectscan
increase the bias associated with measurement error in the right-hand-side variables in a
paneldatasetting(Hsiao,1986).Inaddition,aspointedoutbyMatthewsetal.(1997),this
identificationstrategymakesitdifficulttoidentifytheeffectsofslowlychangingvariables.
Thisproblemisexacerbatedinregressionsthatincludestate-specifictimetrends.
4. Data
Thedatausedinthispaperaredescribedindetailbelowbecausesomeofthevariables
we use differ slightly from previous research. We first summarize the variables used to
measurethetimingandnumberofabortionsandthepresenceofabortionrestrictions.We
thenbrieflydiscusstheothercontrolsincludedontheright-hand-sideofEqs.(1)and(2).
Completedataareavailableforallofthevariablesusedinouranalysisfor855observations
from1974to1997.9
4.1. Measuresofabortionsandabortionrestrictions
ThenumberoflegalabortionsintheUSrosesteadilyduringthe1970sbeforeleveling
offatslightlymorethan1millionperyearinthe1980sand1990s.TheCentersforDisease
Control (CDC) publishes annual data on the number of abortions and the distribution by
weeksofgestation.Thedataarebasedonreportsfromstatepublichealthagencies,butnotall
statesreportweeksofgestationdataeveryyear.Inaddition,thedataareincompleteinstates
8WecorrectforautocorrelationwithinstatesbyassumingthattheautocorrelationisanAR(1)processbecause
thisisthesimplestformofcorrectionandtheonemostcommonlyusedwhenthereisnostrongprioraboutthe
data-generatingprocess.ThePrais–Winstenmethodinvolvesfirstestimatingaweightedleastsquaresregression,
wheretheweightsarethefemalepopulationagedbetween15and44yearsinastate,estimatingtheautocorrelation
parameterforanAR(1)process,transformingthedatausingtheestimatedautocorrelationparameter,andthen
re-estimatingtheregressionusingthetransformeddata.ThestandarderrorsareWhite-correctedforheteroscedas-
ticity,whichisalsopresentinthedata.DonohueandLevitt(2001)useasimilarmethod.Thestandarderrorsin
theOLSregressionsareWhite-corrected.
9Likemostotherresearchers,weomit1973becauseofconcernsaboutthequalityofabortiondatareportedfor
thatyearandthereliabilityofinformationaboutstates’implementationofRoev.Wade.Alistofthestateandyear
pairsincludedinthedataisavailablefromtheauthors.
M.Bitler,M.Zavodny/JournalofHealthEconomics20(2001)1011–1032 1017
inwhichnotallprovidersreporttothepublichealthagency.Blanketal.(1996)findthatthe
CDCdataconsistentlyincludefewerabortionsthandatafromtheAlanGuttmacherInstitute
(AGI),theotherprimarysourceofstate-levelabortiondata.Despitethesedrawbacks,the
CDCistheonlysourceofinformationonthedistributionofabortionsbyweeksofgestation
foralargesampleofstates.10 Furthermore,thecorrelationbetweenthetwomeasuresof
thenumberofabortionsisabove0.98.Wefocusonabortionsoccurringafter12weeksof
gestation(abortionsoccurringafterthefirsttrimester).
We use several measures to examine the impact of abortion restrictions on the
timing and number of abortions. First, we investigate the relationship between restric-
tions and the percentage of abortions occurring after the first trimester. Even if abor-
tion restrictions have no effect on whether women have abortions, they could still affect
the timing of the abortions by making it more costly for women to receive abortions.
This would imply that abortion restrictions lead to an increase in the percentage of later
abortions.
Ifrestrictionsalsoaffectthepoolofwomenobtainingabortions,thentheexpectedeffect
ofrestrictionsontimingismoreambiguous.Ifsomewomenwhowouldhavehadarelatively
late abortion in the absence of abortion restrictions instead give birth or do not become
pregnantinthepresenceofabortionrestrictionswhereasthebehaviorofwomenwhohave
relativelyearlyabortionsisunaffectedbyrestrictions,thepercentageofabortionsoccurring
after the first trimester may fall when a restriction is imposed.11 On the other hand, the
percentageofpost-firsttrimesterabortionscouldriseeveniftherateofsuchlaterabortionsis
unchanged.Iftheimpositionoftheserestrictionsonlyreducesabortionsamongwomenwho
wouldotherwisegetfirsttrimesterabortions,thepercentageofpost-firsttrimesterabortions
couldbepositivelyassociatedwithabortionrestrictionswithoutanyincreaseinthepost-first
trimester abortion rate. We therefore also examine the relationship between restrictions
andthenumberofabortionsafterthefirsttrimesterper1000womenofchildbearingage
(between 15 and 44 years) in a state. We also report the association between abortion
restrictionsandtheoverallabortionrate.
A finding that abortion restrictions reduce the overall abortion rate could imply that
these restrictions cause some women to give birth instead of having abortions, condi-
tional on their having become pregnant. Alternatively, it would be consistent with the
possibility that fewer women become pregnant because of the restrictions. Either could
cause the percentage of post-first trimester abortions to be positively related to the im-
position of restrictions without the rate of post-first trimester abortions increasing. If,
however, there is no relationship between abortion restrictions and the overall abortion
rate, then a finding that the percentage of post-first trimester abortions increases when
restrictions are imposed implies that restrictions are associated with an increase in the
rate of post-first trimester abortions, which has implications for women’s health and
finances.
10ThenumberofstatesreportingweeksofgestationdatatotheCDCrangesfromalowof29in1974toahighof
40in1993.DelawareandFloridadonothaveCDCdataongestation.TheAGIdataonthenumberofabortions
arenotavailablefor7ofthe24yearsforallofthestates.
11Studiesoftherelationshipbetweenabortionrestrictionsandbirthratesreportmixedeffects(Levineetal.,1996;
Matthewsetal.,1997;Klerman,1998).
1018 M.Bitler,M.Zavodny/JournalofHealthEconomics20(2001)1011–1032
DataonabortionrestrictionswereobtainedfromMerzetal.(1995)andtheaddendato
Blank et al. (1996).12 As in Levine et al. (1996), our restrictions variables measure the
fractionofagivenyearthatagivenabortionrestrictionwasineffect.Weusethiscodingfor
severalreasons.Ifarestrictionisliftedorenjoined,womenwhoarealreadypregnantmay
beabletoobtainanabortion,buttheabortionmaybelaterthanintheabsenceofthelaw.
Passageofalawmayalsoaffectabortionprovidersandpatientsevenbeforeimplementation.
Usingthefractionoftheyearthatarestrictionwasineffectinsteadofwhetherarestriction
wasenforcedatanypointduringayearhaslittleeffectonourresults,asdiscussedbelow.
Wealsoincludevariablesthatmeasurethefractionofayearthateachtypeofrestrictionwas
enjoined.13 Population-weightedaveragesofenforcedabortionrestrictionsinneighboring
statesarealsoincludedintheregressions.14
Table 1 displays summary statistics for the abortion and policy variables used in our
analysis. The first column shows means and standard errors for all state and year combi-
nations in the sample. The second column shows summary statistics for states without a
Medicaidrestriction,aparentalinvolvementlaw,orawaitingperiodineffectatanypoint
duringagivenyear.SummarystatisticsarealsoshownseparatelyforstateswithaMedicaid
restriction,aparentalinvolvementlaw,orawaitingperiodineffectatsomepointduring
ayear.ThelastcolumninTable1reportssummarystatisticsonrestrictionsforstateand
yearcombinationsnotincludedinoursamplebecauseofmissingCDCdata.15
Inoursample,about11%ofabortionsoccurafter12weeksofgestation.Slightlymore
than one-half of these post-first trimester abortions occur between 13 and 15 weeks of
gestation, and about one-third occur between 16 and 20 weeks of gestation. Very few
abortionsoccurafter20weeksofgestation,whenthehealthrisksarehighestandprovider
accessisthemostlimited.Inoursample,theincidenceofpost-firsttrimesterabortionsis
slightlymorethan2per1000womenagedbetween15and44years.Thetotalabortionrate
isabout21per1000womenagedbetween15and44years.
Medicaidrestrictionsorparentalinvolvementlawswereenforcedorenjoinedforasub-
stantialfractionofoursample.AMedicaidfundingrestrictionwasineffectfor52%ofour
sample,andaparentalinvolvementlawwasineffectfor21%ofthesample.About15%of
statesinthesampleprovidedMedicaidfundingbecauseofacourtorderorinjunction,and
15%ofstateswereenjoinedfromenforcingaparentalinvolvementlaw.Fewstatesinthe
samplehadawaitingperiodlawineffect,reflectingtherelativelyrecentadoptionofsuch
laws.
12InformationonabortionrestrictionsisalsoavailableintheNationalAbortionRightsActionLeagueannual
publicationWhoDecides?after1988andsporadicallyinFamilyPlanningPerspectives.Weusedthesesourcesto
obtaininformationonmandatorywaitingperiodsandtoreconciledifferencesinthechronologiesofBlanketal.
(1996)andMerzetal.(1995).ForNorthCarolina,wereliedonthechronologyinCooketal.(1999).
13Analternativeistouseavariablethatmeasuresonlywhetheralawrestrictingabortionswasadopted.This
wouldimposetherestrictionthatthecoefficientsonenjoinedlawsarethesameasthoseonenforcedlaws.
14Theweightsarethepopulationofwomenagedbetween15and44years.Ifallborderingstatesenforceda
Medicaidfundingrestrictionallyear,forexample,theborderstatesMedicaidvariablewouldequalone.Results
inregressionsusingadistance-weightedaverageorasimpleaveragearequalitativelythesame.
15Thestatesperyearsincludedandnotincludedinourdataaresimilarexceptthatthoseexcludedarelesslikely
thantheincludedstateandyearcombinationstohaveenforcedaparentalinvolvementlawandmorelikelytohave
hadaMedicaidfundingrestrictionandanenjoinedparentalinvolvementlaw.
M
.
B
itle
r,
M
.
Z
a
v
o
d
n
y
/
J
o
u
r
n
a
l
o
f
H
e
a
lth
E
c
o
n
o
m
ic
s
2
0
(2
0
0
1
)
1
0
1
1
–
1
0
3
2
1
0
1
9
1020 M.Bitler,M.Zavodny/JournalofHealthEconomics20(2001)1011–1032
ManystatesborderonstatesthatenforcedaMedicaidfundingrestriction.About56%of
womenagedbetween15and44yearslivednexttostatesthatrestrictedMedicaidfunding
forabortions.MoststatesthatenforcedaparentalinvolvementlawalsorestrictedMedicaid
fundingforabortions,andalmostallstatesthathadamandatorywaitingperiodineffect
alsorestrictedMedicaidfunding.
The descriptive statistics generally do not suggest that states with abortion restrictions
have more post-first trimester abortions. The percentage of abortions occurring after 12
weeks of gestation and the post-first trimester abortion rate are higher in states without
any restrictions than in states that enforced a Medicaid funding restriction or a parental
involvement.However,thisdifferenceinabortiontimingmaybeduetootherdifferences
betweenstateswithrestrictionsandstateswithoutrestrictions.Thetotalnumberofabortions
per1000womenagedbetween15and44yearsislowerinstatesthatenforcedrestrictions
thaninstatesthatenforcednoneofthethreerestrictions.
4.2. Othercontrols
Theregressionsincludemedical,demographic,economic,andpoliticalvariablestocon-
trol for other factors that affect the timing and number of abortions, as well as state and
yearfixedeffectsinallspecificationsandstatespecifictimetrendsinsomespecifications.
DatasourcesarelistedinAppendixA,andsamplemeansareavailablefromtheauthors.
The percentage of post-first trimester abortions and the abortion rate are likely to be
affected by the number of abortion providers in a state. There are drawbacks to directly
controllingforthenumberofabortionproviders.BecausetheAGIistheonlysourceofdata
onthenumberofabortionproviders,dataarenotavailableforseveralyearsinoursample.
Inaddition,thenumberofprovidersinastatemaybeendogenousbecausethesupplyof
providersispartiallydeterminedbythedemandforabortion(Blanketal.,1996).Following
Levine et al. (1996), we control for the availability of medical services by including the
numberofphysiciansinastatewhoarenotobstetriciansorgynecologistsper1000state
residentsandthenumberofhospitalbedspermillionpeople.16
Demographicsmayaffectboththepercentageofabortionsinastateoccurringafterthe
firsttrimesterandtheabortionrateforthatstate.Wecontrolforthepercentageofastate’s
female population aged between 15 and 44 years that is under age 20, over age 34, and
black.17 Wealsocontrolforthenumberofmarriagesper1000womenagedbetween15
and44yearsinastateandthepercentageofthepopulationlivinginnon-metropolitanareas.
Economic conditions and the political climate within a state may affect the timing of
abortions.Ourregressionsincludetheannualaverageunemploymentrate,realpercapita
income, and female labor force participation rate in a state. We also control for the real
valueofthemaximumAidtoFamilieswithDependentChildren(AFDC)benefitsavailable
to a four-person family with one adult in a state because welfare generosity may affect
16Ourapproachisthereducedformofthetwo-stagemodelusedbyBlanketal.(1996).
17For1974–1979,weuseone-halfofthetotalpopulationagedbetween15and44yearsforthepopulationof
womeninthatagegroupbecausepopulationbreakdownsbybothageandsexarenotavailable.For1974–1979,
wealsousethepercentageofthetotalpopulationagedbetween15and44yearsthatisunder20orover34years
andthepercentageofastate’spopulationthatisblack.