University of DaytoneCommons

Political Science Faculty Publications Department of Political Science

9-2016

Sticky Legacies: Persistence of State ConstitutionalProvisionsNancy Martorano MillerUniversity of Dayton, nmiller1@udayton.edu

Maria ArocaRice University, mp.arocav2@gmail.com

Keith E. HammRice University

Ronald D. HedlundNortheastern University

Follow this and additional works at: http://ecommons.udayton.edu/pol_fac_pub

Part of the Political Theory Commons, and the Public Affairs, Public Policy and PublicAdministration Commons

This Conference Paper is brought to you for free and open access by the Department of Political Science at eCommons. It has been accepted forinclusion in Political Science Faculty Publications by an authorized administrator of eCommons. For more information, please contactfrice1@udayton.edu, mschlangen1@udayton.edu.

eCommons CitationMiller, Nancy Martorano; Aroca, Maria; Hamm, Keith E.; and Hedlund, Ronald D., "Sticky Legacies: Persistence of StateConstitutional Provisions" (2016). Political Science Faculty Publications. 108.http://ecommons.udayton.edu/pol_fac_pub/108

StickyLegacies:PersistenceofStateConstitutionalProvisions

NancyMartoranoMillerUniversityofDaytonNmiller1@udayton.edu

MariaArocaRiceUniversity

mp.arocav2@gmail.com

KeithE.HammRiceUniversityhamm@rice.edu

RonaldD.Hedlund

NortheasternUniversityr.hedlund@neu.edu

Abstract: In this paper, we assess the evolution of 32 state constitutions and the U.S.Constitutionovera100+yeartimeperiod(1776-1907).Weconstructanoriginalsection-level dataset containing the text of every section within a constitution for every yearbetweentheadoptionofthestate’sfirstconstitutionand1907.Weclassifyeachsectionbytopic and compare the content of each new constitution as well as the impact ofamendments. Withasubsetof thesedata,weanalyze theextent towhichsectionswereadded, deleted,modified and remained the same over time using a novel approach thatrelies on an edit distance measure to quantify the similarity between sections of twoconstitutionaldocuments.Wearealsoabletoempiricallyevaluatethelevelofsimilarityofmodified sections as new constitutions were adopted or alterations were made to anexistingconstitution.Finally,wedeterminewhichtopicareasweresubjectedtothelargestamount of change. We demonstrate that it is possible to systematically assess a largecorpus of constitutional documents to test theories of institutional change, provideempirical support to existing qualitative accounts, and create operationalizations ofconceptssuchas“stickiness”thatarecomparableacrossstatesandovertime.Paperprepared forpresentationat theAnnualMeetingof theAmericanPolitical ScienceAssociation,Philadelphia,PA,September1-5,2016.

2

StickyLegacies:PersistenceofStateConstitutionalProvisions

“Thus,thestatesremainsignificantdeterminantsofthequalityoflifeoftheAmericanpeople.Thewayinwhicheachstateframesandallocatespowersthrough its constitution reflects certain conceptions of government andunderstandingsofthetwofacesofpolitics–powerandjustice.Thatis,stateconstitutionsareimportantdeterminantsofwhogetswhat,whenandhowinAmericabecausetheyareconceptualandat times,veryspecificstatementsofwhoshouldgetwhat,whenandhow(Elazar1982:17).”

WhyStudyStateConstitutions?

The activities of state governments have a profound impact on the quality of thelivesofthoselivingwithintheirborders.Inadditiontoestablishingthebasicstructureofgovernance,thesedocumentsalsosettheruleswithregardstocitizenparticipationinthegoverningprocess;guaranteetherightsandlibertiesofcitizens;outlinethefunctionsandpowersoflocalgovernments;andinmanycases,directpublicofficialstoenact/implementspecificpublicpoliciesaswellasprovidespecificservicestocitizens.Thus,theimpactofthesedocumentsonoureverydaylivesisfar-reaching.

Historically,stategovernmentshavebeenfrontandcenterinthegoverningprocess.IntheUnitedStates’earliestdays,thesovereigntyoftheabilityofthestatestogovernasthey saw fit was a core tenet of the nation’s founding. Today, the U.S. states are oftencharacterizedas“laboratoriesofdemocracy”fortheirabilitytoinnovateandtoformulatepublic policies aimed at solving some of themost pressing public policy issues. Electedofficials in the states have often had both the public’s support and willingness toexperimentwithinnovativesolutionstopolicythatisabsentatthefederallevel.Thepolicyareas of same-sexmarriage,marijuana legalization, and immigration are all areaswherestateshaveadoptedpoliciesinthefaceofinactionbythefederalgovernment. Ofcourse,thepolicyadoptedineachareavariessignificantlyacrossstateswiththepoliciesadoptedreflectingthepoliticalcultureandvaluesofthestates’populations.

A society’s constitution represents a fundamental statement of that society’s corecultural,politicalandsocialvaluesat the timeof itsadoption,andweshouldexpect thatthesedocumentshavesignificantimpactonhowpolicyisadoptedaswellasonthecontentof the policy adopted. Donald Lutz has studied state constitutions extensively, and hebelievesthatawrittenconstitutionisa“politicaltechnology”or“theveryembodimentofthetechnologyforachievingthegoodlife(Lutz1982).Further,Lutzargueseloquentlyfortheneedtostudythecontentandeffectofthesedocuments:

“Constitutional variables do not matter, we are told, because politicalvariables do not seem to matter. However, studies of political variablesusuallyignoreconstitutionalvariables. Forinstance,thepresenceofaone-partysystemasopposedtoa two-partysystemisapoliticalvariable,notaconstitutional one. The presence of a unicameral versus a bicameral

3

legislature,ontheotherhandisaconstitutionalvariable.Hereinliespartofthe problem. Many important constitutional provisions do not vary muchfrom state to state, and can hardly be used to explain variance in publicpolicy.Ontheotherhand,thetendencyforcertainconstitutionalfeaturestobe similar over most states must itself be subject to investigation, just asthesesimilaritiesmustbeconsideredaspossiblereasons forsimilaritiesorlack of diversity in certain types of public policy. Put another way, theseconstitutionalsimilaritiesmaysostructurethepoliticalvariablesthatpublicpolicy will tend to respond to environmental conditions in a differentialmannersimplybecauseconstitutionalvariableshavealreadyhadtheireffectbylimitingtherangeofpossiblepolicyoutcomestoarelativelynarrowset–a set of possibilitieswithinwhich environmental factors are determinative(Lutz1982:28).”

Ahandfulofotherpolitical scientistsand legal scholarswritingat thesame timeasLutzargued for more systematic study of state constitutions (Elazar 1982; Kincaid 1988,Friedman 1988). For the most part, this call has gone largely unheeded by politicalscientists,andlegalscholarshavechosentofocusmoreoninterpretationsandanalysisofisolatedprovisionsfoundinaspecificstateconstitutionorondeepconstitutionalhistoriesinsinglestates.1

Webelievethatthisinattentionisdrivenbyaperceptionthatstateconstitutionsaresimilar to the U.S. Constitution as well as the challenge of determining how best tosystematicallyevaluatethesedocuments. Inpreviousresearch,weconvincinglyillustratethat state constitutions are in fact very different from the federal document (MartoranoMiller, Hamm and Hedlund 2009; 2010; 2011; 2014a; 2014b, 2015, 2016), and in thispaperweprovideasystematicanalysisofpre-20thCenturyconstitutions. Currenteventsalsoprovidesuswithanexampleofwhyweshouldcareaboutthepolitical“playingfields”that state constitutional documents create. A recent session of the Texas Legislatureprovidesevidence to this claim. RossRamsey,of theTexasTribunepublisheda storyonMay27,2015titled,“Analysis:VotesCount,buttheRulesCanCountMore,”regardingthepassageof legislation in theTexas legislativeassembly. TheTexasConstitution containssome very specific rules for the consideration of bills and a well-schooled member orminorityofmemberscanusethatknowledgetostoplegislationfrompassage.Thisinfactoccurredwithbillsandamendmentsregardingabortionpolicy,same-sexmarriagepolicyandchildwelfarethatsession(Ramsey2015a,2015b).

TheScopeofThisStudy

1ExceptionsareHammons(1999),Gardner(2014)andCayton(2015)whoprovidesomeveryabstractquantitativeanalysisofstateconstitutions.IntheOxfordCommentariesontheStateConstitutionsoftheUnitedStatesseries,eachstate’sconstitutionalhistoryisgivenanin-depthbook-lengthtreatment.

4

Inthispaper,wearepresentingourfirstattemptstoleveragethepowerofmachinereadingandanalysisofstateconstitutionsinordertobetterdescribeandunderstandhowthesedocumentshaveevolvedovertime.Wehavechosentofocusonthedevelopmentofthe first state constitutions or the timeperiod of 1776-1907. For this roughly 100 yeartime period we have collected and processed each constitution and constitutionalamendmentforthirty-twostatesaswellastheU.S.Constitution.2First,wewillusethesedata to illustrate some basic descriptions about the evolution of state constitutions.Second, we will use the cases of Alabama, Colorado, Illinois, Minnesota, New Jersey,Wyomingand theU.S.Constitution to illustrate thepowerofapproaching theanalysisoftextsinthismanner. The data for this project comes frommainly two sources: 1) theNBER/MarylandStateConstitutionsproject(NBER)3and2)FrancisNewtonThorpe’s,TheFederalandStateConstitutions,ColonialCharters,andtheOrganicLawsoftheState,Territories,andColonies;NoworheretoforeFormingtheUnitedStatesofAmerica(Thorpe)4. TheNBERproject is afederally funded project that created a searchable web-based database of stateconstitutionsandtheiramendments.TheThorperesourceisaseven-volumecompendiumofimportantstatedocumentsthatwascommissionedbytheU.S.Congressandpublishedin1909. NBER relied heavily on the Thorpework and two of the co-authors of this studycrosscheckedbothsourcesforaccuracyasweprocessedthedata. AppendixAcontainsadetaileddiscussionofhowweusedthedatafoundintheNBERdatabase5. WealsoadoptedthecodingschemeofNBER.Inthisresearchourunitofanalysisisthesectionlevelofaconstitution. MoststateconstitutionsareorganizedbyArticle, thenSectionandsubsection.6TheNBER treatedeach sectionof a constitutionseparatelyandcodedeachasbelongingtooneof91topicareas7.Wethencollapsedthosecategoriesinto

2ThesestatesareAlabama,California,Colorado,Delaware,Georgia,Idaho,Illinois,Indiana,Iowa,Kansas,Kentucky,Michigan,Minnesota,Mississippi,Montana,Nebraska,Nevada,NewJersey,NorthDakota,Ohio,Oklahoma,Oregon,Pennsylvania,RhodeIsland,SouthCarolina,Tennessee,Utah,Virginia,Washington,WestVirginia,WisconsinandWyoming.3http://www.stateconstitutions.umd.edu/index.aspx4SeeAppendixAforinformationoncomplimentarysourcesused.5NotethattheNBERhadcompleteinformationfor55constitutions,andcompleteamendmentsfor45constitutions.TheremainingconstitutionsandamendmentswerecodedbytheauthorsusingtheNBERcodingscheme.Formoredetails,seeAppendixA.6ForstatesthatwerenotorganizedusingtheArticle,Sectionformat,wetreatedeachitemasdistinctsection.Sectionscorrespondtothesmallestunitofanalysisthatdividesaconstitution.Insomecases,clausesandsubsectionswereusedassections.7NotethattheNBERhadacompletetopicclassificationfor55constitutions.TherestoftheinformationwascodedbytheauthorsusingtheNBERtopiccodingscheme.Formoredetails,seeAppendixA.

5

sevenbroadtopicareas found inmoststateconstitutions:1)BillofRights/Minorities,2)ConstitutionalAmendmentProcess,3)SuffrageandElections4)FinanceandTaxation,5)Policy, 6) Structure of Government and 7) Federal Government. We excluded from ouranalysisthefollowingtypesofsections:preambles,schedules,certificatesandordinancesas these sectionshavenodirect impacton theprocessesofgovernanceof thegoverningstructuresthemselves. UsingR,asoftwareenvironmentforstatisticalcomputing,wetransformedthetextfilesoftheseconstitutionsintoasection-leveldatasetwhereeachunitisasectionwithinaconstitution. A dataset was created for every year since the adoption of the firstconstitutionofthestateupuntil1907.Thedatasetsincludethetopicandfulltextofeachsection.Ifamendmentsweremadeoranewconstitutionadopted,wereplacedthetextofeach sectionwith the text that resulted from anymodification. If sectionswere deleted,theyweredroppedforthedatasetofthatyearandiftheywereadded,theywereincludedin the dataset of that year. Our final datasets are then a reflection of the way eachconstitutionlookedasofthatparticularyear.Thisdatastructureallowsustomakesectionlevelcomparisonsofaconstitutioninagivenpointintime,tothatsameconstitutionatalaterorpreviouspointintime,aswellastoempiricallytracktheevolutionofeachstate’sconstitutionaldocument(s).Withthesedataweareabletoassesschangeintheaggregateacrossmanystatesovertimeaswellasexplorethatchange insomedetailwithinastate.Moreover,weareabletocomparethetextofeverysectionofastate’sconstitutionatpointt to the text at point tn. To assess the similarity of two sections and after properlyprocessingandcleaningthetext,weusedamodifiednormalizedversionoftheLevenshteinedit distance, which calculates the minimum number of insertions, deletions andsubstitutionsofwordsthatwouldberequiredtotransformthetextinanygivensectionattimet,tothetextofthatsamesectionattimetn.Thescoretakesvaluesbetween0and1,where1indicatesmaximumsimilarity,and0indicatesnullsimilarity8.

Usingthisapproach,wehavetracedtheevolutionofeachsectionofasub-sampleofstatesovertime.Wewillfirstpresentdatafromourthirty-twostateconstitutionsplustheU.S.Constitutionsamplethatprovidesabasicoverviewofthesedocuments.WewillthenproceedtouseoursmallersixstateconstitutionsplusU.S.Constitutionsampletoillustratethepowerofutilizingthesetechniquestoleveragethesetextsources.

WhatdoWeKnowAboutStateConstitutions?

Regardingconstitutionalpolitics,theU.S.statesdiffersignificantlyfromthenationalgovernment. Collectively, the states have adopted 145 constitutions since the nation’sfounding9,andthecurrentstateconstitutionshavebeenamendedover10,000times(Tarr

8SeeAppendixBforinformationonthecalculationoftextsimilarity.9ThisnumberisactuallyslightlyhigheriftheconstitutionsofsouthernstatesformingtheConfederacyduringtheCivilWarareincludedinthetally.

6

2014). During the time period of this study (1776-1907), the thirty-two states and theUnitedStatesadopted80constitutionsand614amendments(SeeFigure1). Incontrast,theUnitedStateshashadasinglefederalconstitutionthathasbeenamendedonlytwenty-seven times in 229 yearswith twenty-one of those amendments occurring before 1908.Figure 1 illustrates that during the first 100 years, the statesweremuchmore able andwilling to modify their constitutional documents with some states like California,MinnesotaandMichiganchoosingapaththatreliedmoreontheamendmentprocesswhilemostSouthernstatesadoptedfiveormoreconstitutionsduringthisperiod.10

[Figure1abouthere]

Further,stateconstitutionsonaveragetendtobelongerthantheU.S.Constitutionandaremore likely to containexplicitprovisionsoutlining the structuresofgovernmentanddirectingthestategovernmenttoadoptspecificpoliciesorprovidespecificservicestocitizens (see recent work by the authors: Martorano Miller, Hamm and Hedlund 2009;2010;2011;2014a;2014b,2015,2016).

Additionally, scholars such as Lutz (1982) and Kincaid (1988) have argued thatwhile the scopeof influenceof thenationalgovernmenthasexpandedsignificantly sincethe Nation’s founding, the true basis of governance in the United States most certainlyexists in the individual state governments. These scholars assert that the FramersintentionallydraftedtheU.S.Constitutionasanincompletedocument–purposelyleavingtothestatesallresidualpowersnotenumeratedinthefederaldocument.Thus,itwasleftto the states to “define and implement many provisions of the U.S. Constitution andstructurethepotentiallyvastdomesticpowersreservedtothestatesandtothepeoplebytheU.S.Constitution (Kincaid1988:12)”and to “develop thedescriptionof awayof life(Lutz1982:41)”inthenewformednation.

Figure 1 also provides additional empirical support for John Dinan’s (2009)extensive research on the state constitutional convention proceedings. He convincinglyargues and provides evidence that reinforces the contention of earlier scholars that thefounders intended the states to “flesh out” governance in their respective constitutionaldocuments.Inhisbook,Dinan(2009)makesastrongcaseforthebeliefthatthestateshavebeenbetter at revisiting their constitutions and revising their institutions andgoverningprinciplesbasedonpastexperiencesorfundamentalshiftsinculture,etc.Heassertsthatconstitutional revision and amendmentprocesses at the state level are easier relative tothe process of revising or amending the U.S. Constitution, and thus have allowed stategovernmentstoevolveinwaysthatmakethemmoreresponsivetomodernproblems.

State constitutions, while containingmany of the provisions found in the federal

document, also differ markedly from our federal constitution in the manner in whichgovernmentstructuresandrelativepowerwasoriginallyperceived.Forexample,theU.S.Constitutionclearlydelineates theCongress’ legislativepowers inArticle I,Section1and10ThesecessionofthesestatesfromtheUnionaccountsforsomeofthesedifferencesinnumberofconstitutionadoptedthatexistsbetweenSouthernandnon-Southernstates.

7

limits the Congress to only those “legislative Powers herein granted.” The TenthAmendment further limits the federal government by granting the states the power notdelegatedtothefederalgovernmentorprohibitedtothestates.Giventherelativebrevityof the national document, this granted substantial, non-delineated power to the stategovernments.TheconstitutionsadoptedbythestatesreflectedthevastnatureofwhatwasleftunwritteninthefederalConstitutionandlargelyconceivedofstategovernmentpoweras plenary and granted significant general powers to the legislative branch in particular(Tarr1998;Elazar1982).Tarrexplains,

“…state governments have historically been understood to possess plenarylegislative powers – that is, those residual legislative powers not ceded to thenational government or prohibited to them by the federal Constitution. As theKansas Supreme Court has observed: ‘When the constitutionality of a statute isinvolved,thequestionpresentedis,therefore,notwhethertheactisauthorizedbytheconstitution,butwhetheritisprohibitedthereby(Tarr1998:7).’”

Elazar also writing about the plenary nature of state governments explains that incomparisontotheU.S.Constitution,thestateconstitutionsneedtobemorecomprehensiveandexplicitaboutlimitinganddefiningthescopeofgovernmentalpowerstopreventtheirgrowthandexpansion(Elazar1982).Thus,thesedescriptionsofstateconstitutionsbyTarr(1998)andElazar(1982)echoandsupportLutz’sperspectiveontheU.S.Constitutionasrather “incomplete” in its “dosanddon’ts”of thegovernanceof theUnitedStatesand itscitizens.Asaresult,theindividualstatesreallyhavehadnochoicebuttodraftandadoptmoredetaileddirectivesregardinghowsocietywoulddeterminewhogetswhat,whenandhowaswellasestablishingthebalanceofpowerandjusticewithinthatsociety. Stateconstitutionalscholarslargelyagreethataconstitutionrepresentsapeoples’belief of how their society should operate at a distinct point in time. Therefore, we cananticipate that the contentof thesedocumentsmaybeheavily influencedby thehistoriceraswithinwhichtheyhavebeenadopted. AlbertSturm(1982)identifiesfiveperiodsofstateconstitutionaldevelopment:

1. TheFirstStateConstitutions(1776-1780)• Briefdocuments• Focused on outlining basic government structures and protections for

citizens(e.g.BillofRightstypeitems)• Marked by establishing of strong legislatures with significant plenary

powers.

2. Early19thCenturyDevelopments(1800-1860)• ThisperiodmarkedtheriseofJacksonianDemocracy.• Increaseinsuffragerights,butnomajorprovisionsofaBillofRightsnature• Morelimitationsonlegislativepower• Clarifications/furtherdelineationofjudicialpower• Increaseinexecutivepower

8

• Provisionsforconstitutionalamendmentandrevision

3. CivilWar,ReconstructionandItsAftermath(1860-1900)• SignificantimpactonSouthernstates• AdditionofJimCrowprovisionsinSouthernconstitutions• Provisionsthatauthorizedthegovernmenttocontrolandregulateeconomic

activity• Increaseintheestablishmentofagenciesnotcontrolledbytheexecutive• Moredetaileddirectivesforthelegislature,judiciaryandexecutive

4. BeginningsofReform(1900-1950)

• Pressuretoreformstategovernmentaftermassivecorruptionuncovered• Extensionofpopularcontrolofgovernment–riseofinitiativeandreferenda• InfluenceoftheNationalMunicipalLeague’sModelStateConstitution• Additionallimitationsonlegislatures

5. ConstitutionalModernization(1950forward)

• Revisions to state government needed to meet the expansion of stategovernmentfunctions

Figure2empiricallyillustratesSturm’sevolutionarypathofstateconstitutions.The

figurepresentsa simple countof thenumberof sectionsdevoted to the seven topicsweidentifiedintheprevioussection.GiventhetimeframeofourdatawewillfocusonthefirstthreeerasidentifiedbySturm.Sturmdescribesanevolutionarypathinwhichtheearlieststate constitutional documents were brief documents that outlined only the basicgovernment structures. Over time, these documents become more complex furtherdelineatingthestructuresofgovernmentaswellasbetterdefiningthepowerrelationshipsbetweenthem.Furthermore,astheneedarosemoresectionswereaddedgivingthestategovernmentthepowertoregulatefinanceandtaxationaswellprovideclearerinstructionsregarding public policy responsibilities. Figure 2 clearly shows that between 1776 and1908 the number of sections found in state constitutions increased at a steep ratewithincreasingnumberofsectionsintheareasoffinanceandtaxationaswellaspublicpolicypost-1850.

[Figure2abouthere]

Figure2onlyrevealsanaggregatepictureofthisevolution.Coulditbethecasethatthe increase in sections is simplyanartifactof additional states entering theUnionovertime? Figure 3 offers two bubble charts. The first is a bubble illustration of the 1819Alabama Constitution. In this constitution there were just four topics, with all but twosections falling into either the structure of government topic (76 sections) or the bill ofrightsandminoritiestopic(39sections).Onesectionconcernedamendingtheconstitutionand theotherpolicy. Thesectionsof theAlabamaConstitutionof1901arespreadmorebroadlyacrosssixtopics,withasignificantincreaseinthesectionsconcerningpolicy(from

9

1 section to 58 sections), suffrage and elections (from 0 sections to 26 sections) andstructureofgovernment(from76to138sections).

[Figure3abouthere]

StickinessandSimilarityinConstitutions

Studentsofinstitutionshaveoftenconsideredthestickinessofinstitutions–thatisoncea structure, rule,procedureor informalnormhasbecomeadopted it isnotoriouslydifficult to repeal, change, etc. In this section, we will explore stickiness as well as thenotionof similarity. Forourpurposes,weconsidera constitutional sectionas sticky if itenduresunchangedovertimesinceitsadoption.Wearealsointerestedinthesimilarityofeach a section over time andwill employ Levenshtein distance scores to determine theextent to which the adoption of new constitutions and amendments modified existingsections.

Figure4providesaseriesofVenndiagrams for theU.S.Constitutionand6states.

Thecircletotheleftrepresentsthetextoftheinitialconstitutionaldocumentandthecircletotherightrepresentsthetextoftheconstitutionaldocumentasitexistedin1907.11TheVenn diagrams simply compare to see if a section either: 1) exists only in the initialdocument (deleted), 2) in both documents (remained) or 3) only in 1907 (added). TheVenndiagramsdonottakeintoaccountanymodificationsmadetosectionsthatenduredatboth timeperiods. ThepointofFigure4 ismerely to illustrate that thereappears tobesignificant variation in the stickiness of sections across states. As expected the U.S.Constitution appears very stickywith 99%of the sections in 1787 remaining in 190712.Minnesota issimilarwith96%of thesectionsexisting in1857carryingoverto1907. Incontrast, only 47% of the sections in the 1819 Alabama Constitution still existed in the11Forexample,fortheU.S.Constitution,theleftcirclerepresentsthe1787constitutionasdraftedandratified.TherightcircleisthetextoftheU.S.Constitutionwithallamendmentsthrough1907included.ForAlabama,theleftcirclerepresentsthe1819Constitutionandtherightcircle,thetextofthe1901AlabamaConstitutionwithanyamendmentsthrough1907.TheVenndiagramdoesnottakeintoaccountthe1861,1865,1868or1875constitutionsandanyamendmentstothem.TheVenndiagramssimplycomparetheconstitutionatentryintotheUnitedStateswiththeconstitutionat1907.12NotethatfortheU.Sconstitutionclausesweretreatedassections.Between1787and1907Amendments1through10added10sectionsonthetopicofBillofRights,Amendment11addedasectiononthetopicofJudicialpolitics,andmodifiedanexistingsection(Article3Section2Clause1)onthatsametopic.Amendment12modifiedanexistingsection(Article2Section1Clause3)onthetopicoftheexecutive.Amendment13addedasectiononslavery(whichwetreatedasbelongingtothetopicofBillofRightsandminorities).Amendment14added3sectionsoncitizenship(treatedasbelongingtothetopicofBillofRightsandminorities),modifiedasection(Article1Section2Clause3),andanadditionalsectionwassupersededbythisamendment(Article2Section1Clause3),whichwastreatedasadeletion.Finally,Amendment15addedtwosectionsoncitizenshipandsuffragerespectively.

10

1907document. InColoradoandWyomingtherewereveryfewdeletionsoradditions ifany. Ofcourse,giventhetimeperiodofthisstudy,thisresultisnottoosurprisingsinceColoradoachievedstatehoodin1876andWyomingin1889leavinglittletimetoengageinmajorconstitutionalalterations.

[Figure4abouthere]

Using the Levenshtein distance technique discussed in Appendix B,we can easilycomparethetextbysectionofastate’sconstitutionatanytwopointsintime.13Inessence,thismeasureprovidesuswithasimilarityscoreforeachsectionbetweentwotime-points.Inthisanalysis,wearecomparingthesectionsofthefirstconstitutionwiththeconstitutionin1907.Ascoreofoneindicatesthatthesectionisidenticalin1907asitwasinthefirstconstitution. A score of zero indicates that the sectionwas added or deleted sometimebetweentheadoptionof the firstconstitutionand1907. Therangeofscores inbetweenrepresent a relative similarity in text between first constitution and 1907. What thesescoresdonotrevealisthesubstanceofthesesections–dotheygrantagovernmentpowerordotheymandateorrestrictgovernmentpower?Ourhopeisthatinthefuture,wewillabletocombinethisanalyticaltechniquewithourpreviousworkonthesubstanceofthesedocumentstomoreefficientlyandaccuratelyempiricallyanalyzethesubstanceandimpactofthesechanges. The similarity scores between the first constitution and 1907 in our sub-samplerevealsthatthereisquiteabitofvariationintheextenttowhichthewrittentextofthesedocumentsendureovertimeacrossstates.Table1providesadescriptivesummaryofthepercentage of sections that were unchanged, modified, added and deleted in our sub-sample. In some cases (United States, Colorado, Minnesota and Wyoming), a largepercentage of sections endured unchanged from the first constitution though 1907. Inothers (Alabama, Illinois and New Jersey) very few sections went unchanged. Similarpatterns emerge for modifications, additions and deletions. The point is that there istremendousvariationacrosscasesinthesebehaviors.

[Table1abouthere]

Figure 5 presents the distribution of similarity scores for six states and the U.S.Constitution.Eachdotontheplotrepresentsasection.Attheverytopofeachbox-plotisadarklineofdots(inblue)–thesearethesectionsthatpossessperfectsimilaritybetweenthe first document and 1907. For all but one case, there are quite a few caseswhere asectionenduresvirtuallyunchangedover time. Theoneexception isNew Jersey,wherenot a single section remained intact between1776and1907. This is not too surprisingsincethe1776NewJerseyConstitutionismoresimilartoacolonialcharterinitslanguagethantheconstitutionsofmoststatesuponjoiningtheUnitedStates.

[Figure5abouthere]13Wecanalsocomparethetextbysectionoftheconstitutionsoftwodifferentstatesaswell,butthatisatopicandanalysisthatwewilltackleinafuturepaper.

11

MovingtotheotherextremeinFigure5,arethelinesofdots(ingreenandyellow)running along the bottom of each box-plot. These dots represent the sections of theconstitution that have been added or deleted since the first constitution was adopted.Acrossourcases,itisnotsurprisingtofindthatinmostinstancesquiteafewsectionshavebeenaddedordeletedtothetextasitevolvedovertime.TheoneexceptionisWyoming,wherenonewsectionswereaddednoranysectionsdeletedbetween1889and1907. Great change in a constitutional text occurs through the modification of existingsections. Overtime,structuresbecomemoredelineatedandcitizensmaydecidethatthegovernmentneedstobemandatedtoactonitsbehalforlimitedinitsaction.Themiddlesectionofdots (in red) ineachbox-plot represents thesections thathavebeenmodifiedbetween the first constitution and 1907. As is evident, there is quite a wide range ofmodificationoccurringinthesesections. Insomecases,themodificationtothesectionisrelatively minor with similarity score greater than .75. In others the modification isextremewithasimilarityscorelessthan.25.InAlabama,Colorado,Illinois,andMinnesotathereisextensivevariationinsimilarityscoresacrossmodifiedsections.Thedarkredlinewithintheboxplotisthemediansimilarityscore.Thefactthatthereisagreatamountofdistance between the median and the upper and lower quartiles as well as fairly longwhiskers to the minimum and maximum similarity scores indicate there is substantialvariationripeforfutureanalysis.

Adifferentpatternofmodificationemergesineachofthethreeremainingcases.InNewJersey, thebox-plotrevealsthatmodifiedsectionsof itsconstitutionin1907exhibitlittlesimilaritytothetextofthesectionsadoptedin1776withamediansimilarityscoreoflessthan.25andaverysquatquartileboxwithtinywhiskers. Only3sectionsoftheU.S.Constitutionweremodifiedwithtwoofthemodifiedsectionspossessingsimilarityscoreofless than .5 and onewith a score in excess of .75. Finally, inWyoming therewas not asinglesectionmodifiedbetween1889and1907. We have also produced this same set of box-plot breaking the sections out byconstitutional topic. Thesearepresented inFigure6.What this figure shows is that thesectionsthatweremodifieddealtlargelywithgovernmentstructureandbillofrightsandminorities sections. While there were some modifications to sections dealing with theother topics, the change between 1776 and 1907 in areas such as finance and taxation,policyandsuffrageandelectionswasduemoretotheadditionofnewsections(seelowerleft box of Figure6) than changes in existing sections. Thepatterns of section addition,deletionandmodificationexhibitedinthetablefurthersupportSturm’s(1982)accountoftheerasofstateconstitutionaldevelopment.

[Figure6abouthere] Inthissectionwehaveeffectivelyillustratedthatthereseemstobesomeevidencethat constitutional provisions are somewhat sticky – once included in a document, it islikely that theywill remain. However,wehavealso shown, especially in the states, that

12

modifications of these provisionswill occur. Whatwe cannot illustrate in this paper iswhether or not these modifications resulted in significant substantive changes in thesesections.Tacklingthatquestionwillbethefocusofafuturepaper.

ConstitutionalChangeOverTimeand

TheChoicetoAmendorAdoptNewDocuments

Many scholars have asserted that the manner in which political institutions andprocesses evolve are by their nature path dependent. That is the initial choices anddecisionsmaderegardingthose institutionsandprocessstarta trajectory inwhichthoseearly choices influence all future decision-making and changes in those institutions andprocesses. Thus, any attempt to account for institutional evolutionmust take this pathdependencyintoaccount(Pierson2000a,2000b;Jervis2000;Thelen2000;Bridges2000).Comparative politics scholars studying democratization and the adoption of othergoverning institutions aswell as scholars studyingAmerican political development havelong acknowledged the importance of path dependency in their work (e.g., Lipset andRokkan1967;North1990;Schickler2001;Skocpol1992;CollierandCollier1991;Ertman1996;Hacker1998). Thesesameauthorsassertthatevents–bothsmallandlarge–thatoccur as the path is travelled create “forks” in the road that may lead to differentinstitutionalchangesandevolutiondependingontheturntaken.

Inthissectionwewilldiscusshowinitialproceduresforchangingaconstitutionaldocumentmayinfluencehowadocumentisalteredovertime.Figure7isabox-plotofU.S.Constitutional changeover the timeperiodof study. TheU.S.hasonlyeverhada singleconstitutionandasisevidentfromtheplotduringitsfirstroughly100yearswasmodifiedonlyminimallymostlythroughtheadditionofahandfulofamendments. Sincetheinitialconstitutionalconvention, therehavebeennosuccessfulattempts tocallaconvention topropose amendments to the constitution nor have most of the over 10,000 proposedamendmentssuccessfullynavigatedtheamendmentprocess.

[Figure7abouthere]

The state of Alabama provides a stark contrast to our national constitutionalexperience.Between1819and1901,Alabamaadoptedsixseparateconstitutions.Figure8displaysabox-plotof thesimilarityscorestrackingconstitutionalchangeovertime. Theboxplotsinthegray(shaded)areasrepresentthedistributionofsimilarityscoreswhenwecompare the text of a new a constitution to the text of that constitution in the year justpriortotheadoptionofanewconstitution.Thebox-plotsintheunshadedareasrepresentacomparisonofthetextoftheoldconstitutioninitslastyearofexistencetothetextofthenew constitution that succeeded it.Wewanted tomake this distinction to separate thechangesthatoccurduringtheamendmentprocessthattakesplacebetweentheadoptionof two constitutions, from those that take place as a result of the adoption of a newconstitution.Forexample,Alabamaadopteditsfirstconstitutionin1819anditssecondin1861whenitsecededfromtheUnitedStatesatthestartoftheCivilWar.Thefirstbox-plot

13

inthegrayareaconveysthesimilarityofthesectionsofthe1819constitutionasadopted,towhatthetextreadin1860afterithadbeenamendedseveraltimesover41years.Thesecondbox-plotinthewhiteareacomparesthesectionsofthe1819Constitutionasitreadin1860tothetextofthenewConstitutionadoptedin1861.

[Figure8abouthere] InAlabama,despiteprovisionsineachofthesixconstitutionsthatprovidedfortheamendingoftheexistingdocument,itisclearmajorconstitutionalchangeswerereservedfor when new documents were adopted. In fact, prior to 1907, the only Alabamaconstitution that was ever amended was the 1819 Constitution. Recall the earlierdiscussionofpathdependencyandtheimpactofsmallvs.largeevents.Alabama’sdecisionto rely on the adoption of newdocuments to alter its constitution is likely driven by itslocationinthedeep-South.

TheexperienceoftheSouthernstatesduringtheCivilWarandReconstructioneraillustrates this point beautifully. Prior to their secession from the Union each of thesestateshadadoptedconstitutionsthatweresimilartothosefoundintheNorthernstates–save provisions that furthered and protected the institution of slavery. Somewhatinterestingly, these states did little to alter their constitution as they created theConfederateStatesofAmerica(CSA)–mostofthestatessimplyjustamendedtheirexistingdocuments by changing United States of America to Confederate States of America andadoptedprovisions that furtherentrenchedandprotected the institutionof slavery. ThesignificantchangesdidnotoccuruntilafterthewarwaslostandtheCSAwasnomore.Theprice of reentering the Union was the drafting and establishment of a new stateconstitution that met the specific conditions set by the federal government for reentry.Once the federal government returned governance to the defeated Southerners, theyadopted another set of constitutions that institutionalized Jim Crow laws aimeddiminishingthecivilandpoliticalrightsofAfrican-Americans.Figure1clearlyshowedthattheSouthernstatesinourstudyhadadoptedsignificantlymorenewconstitutionsversusstatesinotherregions.ThisisdueinlargeparttotheCivilWarexperience.ThreedistinctconstitutionswereadoptedbySouthernstatesduringthisperiodinhistory:1)ConfederateStatesofAmerica(CivilWar)constitution;2)Reconstructionconstitution(neededforre-admittancetotheUnitedStates)and3)redeemerconstitutions(reestablishmentoftheoldorderoncereconstructionended).Thus,inthiscasealargeevent–theCivilWar–createda “fork in the road” that caused these states to choose the patch of adopting newconstitutionsversusamendinganexistingone.

Additionally, theeventsof theCivilWar created conditions that led to substantial

change in thesedocuments. Let’sreturntoourAlabamaexample. Figure9usesmediansimilarityscorestoillustratetheimpactoftheadoptionofanewconstitution.Inthiscaseweareplottingthemediansimilarityscoreintheconstitutionforthetimeperiodof1819-1907.Wearecomparingthesectionsofadocumentinagivenyear,tothesectionsofthedocumentthefollowingyear.Thelineremainsflatatone(perfectmatchingofsections)formost years. The exceptions are 1865 and 1868 (Reconstruction constitutions), 1875(redeemer constitution) and 1901 where the median similarity score was less than .5

14

indicating that the sections thatweremodifiedexhibiteda lowdegreeof similarity fromtheir previous version. In fact, the median similarity score for the 1875 redeemerconstitutioniszeroindicatingthatmanysectionsweredeletedentirelyand/orcompletelynewsectionswereadded.TheonlytimeanewconstitutionwasadoptedinAlabamaandthemediansimilarityscoreremainedatonewaswhenthe1861CivilWarconstitutionwasadopted. Histories of the Alabama constitution indicate that the state made minimalchangestothedocumentuponsecession(Winkle2014).

[Figure9abouthere]Thisnotion thata largeevent can lead to significant changealsoechoes thework

Baumgartner and Jones (1993) who developed a theory of punctuated equilibrium toexplainpolicychange.Intheirworktheyshowthatpolicydoesnotchangeincrementally.Rather,policyisstableforalongperiodoftimeandthenaneventoccursthatcausesthepublic and elected officials to call for policy change that leads to significant policyalteration,whichthenleadsintoanotherextendedperiodofstability.ThisbriefanalysisofthecaseofAlabamaprovidessomeevidencethatpunctuatedequilibriummayhavesomeabilitytoaccountforconstitutionalchangeundercertainhistoricalconditions.

Discussion

The description of the historical evolution of state constitutions in this paperrepresentsinmanywaystheviewfrom50,000feet. Themostgeneralconclusionwecandrawisthatatthestateleveltherewastremendouschangeinconstitutionsbetween1776and 1907. Specifically, we have also shown that number of sections increased and thetopics addressed in a constitution increased over time and are correlatedwith the timeperiodoftheadoptionoftheconstitution.Ingeneral,astimepasses,constitutionsbecomemorecomplex–thedocumentcontainsmoreprovisionsoverawiderrangeoftopics. Wealso, in very general terms,were able to address thenotionsof stickinessandsimilarity. Once a section is included in a constitution, it is likely that it will remain(stickiness). However, there appears to be a wide range in variance regarding thesimilarity of the text of these sections over time in some states, and at this timewe areunabletodeterminethetrueimpactofthemodificationsrevealedbythesimilarityscores.In future research we intend to dig deeper into the tension between stickiness andsimilaritywithinsectionsofaconstitutionovertime.

Theothercontributionofthispaperisthetechniquethatwehaveusedtoanalyzeand compare the constitutions in our sample. The techniques employed here haveprovidedleverageoverthedatawehadbeenunabletopreviouslyachieve.Humanreadingandcodingof thesetexts is timeconsuming. Theuseofmachine-readingandautomatedtextanalyticshaveallowedustobeabletocompareover350,000sectionsofconstitutionsover a 100-year period. Our hope is that we will be able to further develop thesetechniquestotacklethecodingofconstitutionalcontent inthefuture.Ataminimum,thesimilarityscoremakestheanalysisofchangemoreefficientandlessresourcedemanding

15

byhelpingusquicklyidentifywhichsectionshavechangedsothatwecanlaterassessthedepthandmeaningofthatchange.

Inafutureanalysisofstateconstitutions,weplantotacklethequestionofcontent.Whatdothesedocumentssayspecificallyregardinggovernanceinastate?Pastresearchwehaveconductedrevealsthattheprovisionsfoundinstateconstitutionsrestrict,exclude,mandateorempowerstructuressuchasthelegislature.Whatisthebalancebetweenthesetypesofprovisionswithinstatesovertimeaswellasacrossstatesindifferenteras?Whatdoesthismeanforpolicy-makingandthecontentofthatpolicy?

Additionally, towhat extent do states learn and/or borrow from one another? In

thisanalysisweclearlyshowthatstatesthatenteredtheUnitedStateslaterintheperiodofstudybeganstatehoodwithmorecomplexconstitutions–moreprovisionsinmoreareas.Howmuchofwhat exists inone state’s constitution canbe found in the constitutionsofotherstates?ItislikelythatasstatesenteredtheUnionorchosetoengageinwholescalerevisionofanexistingconstitution,thattheyconsultedtheconstitutionsofotherstatesinanattempttoidentifythebestpracticesoftheday.Squire(2012),inhishistoricalstudyofthe evolution of American legislatures provides extensive evidence that states modeledtheir legislatures on already existing legislatures. Therefore, it is likely that particularprovisionsexperienceadiffusionintothestatesovertime.Withmoreelaboratedmeasuresoftextsimilaritythatallowustocompare,not justtheliteral,butthesemanticsimilaritybetween two texts,whicharecurrentlybeingdeveloped in the fieldofNaturalLanguageProcessing,wecanbegintotacklequestionsofdiffusiononlargerscalesthaneverbefore. Finally,theapproachandmethodswehaveusedinthispapercaneasilybeusedtostudy other phenomena in the states. For example, ourmethod ofmachine reading andcomparingastate’sconstitutionovertimeaswellasconstitutionsacrossstatescaneasilybe translated to studies of public policies, the development of legislation or a state’sstatutorycode.

16

APPENDIXA

To conduct a quantitative analysis of state constitutions the texts of theconstitutionsmust be in a readable format (.txt) for processing, andwemust be able toeasily search and capture fragments of the text (such as articles or sections) to conducttargeted comparisons. The NBER/Maryland State Constitutions project website has apublicrepositoryofconstitutionssavedin .txt format.Additionally, thecodingrulesusedbytheNBERprojectallowustoretrieveitemswithinaconstitutioninanautomatedway.Each of the constitutions in the repository has been coded using the same scheme. Ourinitialanalysiswasconductedusing80constitutionsbetween1776and1907.Outofthese80,NBERhadcompleteinformationon55constitutions,andhadrecordedamendmentsfor45 of these. Themissing constitutions include: California 1849, California 1879, Georgia1777, Georgia 1789, Georgia 1798, Georgia 1861, Georgia 1865, Georgia 1868, Georgia1877, Iowa 1844, Nebraska 1866, Nebraska 1875, New Jersey 1776, New Jersey 1844,Oregon1857,Tennessee1796,Tennessee1835,Tennessee1870,Virginia1776,Virginia1830,Virginia1851,Virginia1869,Virginia1902,Wisconsin1848,UnitedStates1787.

These missing constitutions were coded by the authors. Amendments were alsocodedbytheauthorsfortheseconstitutionsaswellasfortheconstitutionsofAlabamaandDelaware, which were not completed by the NBER. These missing constitutions andamendmentsweremostlyretrievedfromFrancisNewtonThorpe’s,TheFederalandStateConstitutions, Colonial Charters, and the Organic Laws of the State, Territories, andColonies; Now or heretofore Forming the United States of America (Thorpe). The scansfromthevolumeswereconvertedinto.txtformatusingABBYY,anOCRandtextscanningsoftware.Aftertheconstitutionswereconvertedinto.txtformattheywerecodedfollowingNBER coding rules. It is important to note that NBER texts were all revised and crosschecked by the authors also using Thorpe. In some cases, complimentary sources wereused. These include the Delaware Journal of Corporate Law14, the GeorgiaInfo onlinealmanac15,ActsandResolutionsoftheGeneralAssemblyoftheStateofGeorgiaforvariousyears16,theCodeoftheStateofGeorgiabyHowardVanEpps189517,documentsfromtheTNGenWeb Project18, documents from the For Virginians: GovernmentMatterswebsite

14http://www.djcl.org/wp-content/uploads/2014/07/A-History-of-Delaware-Corporation-Law1.pdf15http://georgiainfo.galileo.usg.edu/topics/government/related_article/constitutions/georgia-constitution-of-186116https://play.google.com/store/books/details/Acts_and_Resolutions_of_the_General_Assembly_of_th?id=deYXAAAAYAAJ,https://books.google.com/books?id=-RI4AAAAIAAJ&pg=PA19&dq=georgia+acts+and+resolutions+1890+1891&hl=en&sa=X&ved=0ahUKEwjVgtfJzKHOAhVISCYKHSawBigQ6AEIRzAH#v=onepage&q=georgia%20acts%20and%20resolutions%201890%201891&f=false17http://digitalcommons.law.uga.edu/ga_code/25/18http://tngenweb.org/law/

17

ownedbyTheCenter forHistory andNewMedia (CHNM) atGeorgeMasonUniversity19andinformationfromtheconstitutionsuswebsite20.

Oncethesetextwereinreadable.txtformatweusedRtotransformthetextfilesofthese constitutions into a section-level dataset where each unit is a section within aconstitution. A dataset was created for every year since the adoption of the firstconstitutionofthestateupuntil1907.Thedatasetsincludethetopicandfulltextofeachsection.Ifamendmentsweremadeoranewconstitutionadopted,wereplacedthetextofeach sectionwith the text that resulted from anymodification. If sectionswere deleted,theyweredroppedforthedatasetofthatyearandiftheywereadded,theywereincludedinthedatasetofthatyear.Ourfinaldatasetsconsistofthesectionsofeachconstitutionforthe32statesandtheU.S.asofeachyearbetweenthefirstconstitutionand1907.

19http://vagovernmentmatters.org/primary-sources/51820http://constitutionus.com/

18

APPENDIXBBeforewewereable to compare sectionsof twoconstitutions the textof sections

neededtobepre-processed,cleanedanddecodedtogetanaccuratesimilarityassessment.Firstwedecodedthedatasothatallofitwouldbeinastandardencodingformat.InthiscaseUTF-8,whichiswidelyacceptedandrecommendedformat.Second,weremovedstop-words,orcommonlyoccurringwordsintheEnglishlanguageusingthetmpackage21.Wedidnotremovestopwordsthat,althoughcommonintheEnglishlanguageweconsideredto be reflective of substantive change such as “may” and “shall”. Third, we removedpunctuation marks and other non-character symbols. Finally, we standardized words,replacingwordsthatusedtheBritishspelling(i.e.“Defence”),withtheAmericanspellingsothatwordsthathadthesamemeaningbutadifferentspellingwouldnotdrivedownthesimilarityscore.Weusedavarietyofsourcesonthewebtogetalistofsuchwords22.

After the sections textshadbeenprocessedweproceeded toassess the similaritybetween sections of constitutions. For this purpose, we used a modified version of theLevenshteineditdistance.TheoriginalLevenshteinalgorithmcalculatestheleastnumberofeditoperations(deletions,insertionsandsubstitutions)thatarenecessarytomodifyonestring to obtain another string. The algorithm assigns a value of 1 to each characterinsertion,deletionandsubstitution.Thenormalizedversionofthealgorithmthendividesthisminimumnumberofinsertions,deletionsandsubstitutionsbythelengthofthelongerstring and subtracts the resulting value from1. TheLevenshteinmeasurewas originallycreated to compare characters inwords, instead of sentences,where a small number ofdifferencesistobeexpected,andistypicallyusedbyspellcheckerandcorrectionsystems.Since we were interested in comparing words within sentences, instead of characterswithinwordsweadaptedthealgorithmtousewordsinsteadofcharacterstocalculatethedistance and used the standard dynamic programming approach for computing theLevenshteindistancewiththesemodifications.Wethendividedthatdistancebythelength(amountofwords)of the longersentenceandsubtractedtheresultingvalue form1.Thevalues of themodified Levenshtein distance range from 0 to 1, with 1 denoting perfectsimilarityand0denotingthelowestdegreeofsimilarity.

In the case of amendments knowing which sections to compare with another wasstraightforward, since they shared a common article, section and subsection (whenapplicable)number, sosectionswith thesame identifiers (article, sectionandsubsectionnumbers)werecompared.Inthecaseoftheadoptionofnewconstitutionsthisishardertoassess since sections might no longer share the same identifiers as constitutions getrenumberedandreorganized.Tosolvethisproblem,wecomparedeverypossiblesectionofagiventopicinthefirstdocumentwitheverypossiblesectionofthatsametopicinthe

21Seethedocumentationofthepackageforafulllistofstopwordshttps://cran.r-project.org/web/packages/tm/tm.pdf22Seeforexamplehttp://www.tysto.com/uk-us-spelling-list.html

19

seconddocument.The result is amatrix of similarity scoresbetween the sectionsof thesame topic.We then calculated themaximum similarity score for each section, workingunder the assumption that thiswould increase the probability that a section in the firstdocument would be matched with its true equivalent in the second document. Weproceededtodoavisualcheckofmatches,toensurethateachsectionwasmatchedwithanequivalentsectioninthesecondconstitution.Wefoundthatthealgorithmwasproducingacorrect match whenever the similarity score was above .25. Typically, scores of theLevenshteinrarelyfalltozerosincesomecommonwordsbetweensentences,forexample,theword“legislator”allalwayspresentwhenyouarecomparingsectionsthatsharetopic.WhatwewantisforaddedordeletedsectionstobecomparedtoablankspacesothattheLevenshteindistancewilltakeavalueofzero.Wemadesuretoreviseeverycase,ensuringthat sections that were deleted from the first document, or added from the seconddocument,werenotmatchedwithanothersection,andthateachsectionwasmatchedwithitstrueequivalent.Afterthisprocesswerecalculatedthesimilarityscore.

OneofourconcernswhenusingtheLevenshteinmeasuretocalculatesimilaritysectionby section is that sections vary a great deal in the amount of words they contain, andconstitutions vary in the average number of words they include in each section. If thedistance is sensitive to the number ofwords,meaning sectionswith a lower number ofwordswillhavelowersimilarityscores,comparisonsacrosssectionsandstateswouldnotbeaccurate.Toassesswhethersuchbiasexistsweplottedthesimilarityscoresperpairofsection over the total number of words in each section for all of our constitutioncombinations and there appears to be no bias. Similarity scores are distributedwith nodiscernablepatternacrossdifferentvaluesoftheloggednumberofwordspersection.

Finally, it is important to note some of the limitations of this approach to comparesimilarity of sections. The methods we have discussed to classify and measure thesimilarity of texts are part of what is known as a statistical approach to InformationRetrieval.Fromastatisticalapproach,documentscanbedescribedbyarepresentativesetof keywords (BaezaYates andRibeiro-Neto 1999)23. A problemof this approach, versuswhat is known as a natural language approach, is word sense ambiguity, or “TheVocabularyProblem”(Furnasetal.1987)24,whichreferstosituationsinwhichreformersofoneconstitutiondonotuseexactlythesamewordsasreformersofanotherconstitution,even though the intention of both groups of reformers is the same. They might usesynonymsorequivalentphrases.Anothersideof thisproblem iswhen twoconstitutionsusethesameterms,sobothdocumentsincludethesamesearchterms,buttheyareusedwithdifferentintentions,orinanothersense.

23Baeza-YatesandRibeiro-Neto.1999.ModernInformationRetrieval.http://people.ischool.berkeley.edu/~hearst/irbook/print/chap10.pdf24Furnas,G.W.,Landauer,T.K.,Gomez,L.M.,&Dumais,S.T.1987.Thevocabularyprobleminhuman-systemcommunications.CommunicationsoftheACM,30.s

20

Natural language texts contain sentences that have a grammatical structure and usevocabularyrichinsynonyms.Theseelementscontributetoatextísmeaning,whichisnotconsidered in statistical information retrieval. Automated information retrieval with astatisticalapproachwillalwaysfacethesetypesoflimitations,whichiswhytheymightnotachievehumanlevelsofperformance,andarelikelytoidentifysomedegreeoferroneousinformation,orignorerelevantinformation.

Becauseweareanalyzingconstitutionswithinthesamestateandwithinalimitedtimeperiod we considered and were able to determine through visual checks that thevocabulary problem was not giving bias to our results. However, moving forward intocomparisonswithinlongertimeframesandcomparisonofthetextsofdifferentstatesthisproblemneeds to be addressed. Oneway of proceeding is to look into natural languageapproachesthatalreadyexistandusemethodssuchasdomainspecificthesaurallyderivedlexicalchainsandtextcasebasedreasoningtoassesssimilarityofdocumentsthatdonotsharethesamewordsbuthavethesameunderlyingmeaning.

21

References

Baumgartner, Frank R. and Bryan D. Jones. 1993. Agendas and Instability in AmericanPolitics.Chicago,IL:UniversityofChicagoPress.Bridges, Amy. 2000. “Path Dependence, Sequence, History, Theory,” Studies in AmericanPoliticalDevelopment.14:109-112.Cayton,Adam.2015.“WhyAreSomeInstitutionsReplacedwhileOthersPersist?EvidencefromStateConstitutions,”StatePoliticsandPolicyQuarterly.16:1-23.Collier,RuthBerinsandDavidCollier.1991.ShapingthePoliticalArena:CriticalJunctures,the Labor Movement and Regime Dynamics in Latin America. Princeton, NJ: PrincetonUniversityPress.Dinan,John.2009.TheAmericanStateConstitutionalTradition.Lawrence,KS:UniversityofKansasPress.Elazar, Daniel J. 1982. “The Principles and Traditions Underlying State Constitutions,”Publius.12:11-25.Ertman,Thomas.1996.BirthoftheLeviathan:BuildingStatesandRegimesinMedievalandEarlyModernEurope.Cambridge:CambridgeUniversityPress.Friedman,LawrenceM.1988.“StateConstitutionsinHistoricalPerspective,”AnnalsoftheAmericanAcademyofPoliticalandSocialScience.496:33-42.Gardner, James A. 2014. “Autonomy and Isomorphism: The Unfulfilled Promise ofStructuralAutonomyinAmericanStateConstitutions,”WayneLawReview.60:31-67.Hacker, Jacob. 1998. “The Historical Logic of National Health Insurance: Structure andSequence in the Development of British, Canadian, and U.S. Medical Policy.” Studies inAmericanPoliticalDevelopment.12:57-130.Hammons, ChristopherW. 1999. “Was JamesMadisonWrong? Rethinking the AmericanPreference for Short, Framework-Oriented Constitutions,” American Political ScienceReview.93:837-49.Jervis,Robert.2000.“TimingandInteractioninPolitics:ACommentonPierson,”StudiesinAmericanPoliticalDevelopment.14:93-100.Kincaid, John.1988. “StateConstitutionsand theFederalSystem,”AnnalsoftheAmericanAcademyofPoliticalandSocialScience.496:12-22.

22

Lipset,SeymourMartinandSteinRokkan.1967“CleavageStructures,PartySystems,andVoter Alignments: An Introduction.” InPartySystemsandVoterAlignments, ed. SeymourMartinLipsetandSteinRokkan.NewYork:FreePress.Pp.1-64.Lutz, Donald. 1982. “The Purpose of American Constitutions,” Publius: The Journal ofFederalism.12:27-44.Martorano,Nancy,KeithE.Hamm,RonaldD.Hedlund.2009. “Paths,EventsandStrategicChoices: The Evolution of American State Legislatures,” Paper presented at the annualmeetingoftheAmericanPoliticalScienceAssociation,Toronto,Ontario,Canada..2010.“StateLegislativeEvolution:PathDependency,OrganizationTheoryandBoundedRationalChoice,”Paperpresentedatthe10thannualStatePoliticsandPolicyConference,Springfield,Illinois..2011.“LimitingtheEvolutionaryPath:ExplainingConstitutionalandStatutoryEntrenchment in State Legislatures,” Paper presented at the annual meeting of theAmericanPoliticalScienceAssociation,Seattle,Washington..2014a.“WithOneHandTiedBehindTheirBacks:ConstitutionalEntrenchmentinStateLegislatures,”PaperpresentedattheannualmeetingoftheMidwestPoliticalScienceAssociation,ChicagoIllinoisandStatePoliticalandPolicyConferenceIndianaUniversity..2014b.“ConstrainedBehavior:UnderstandingtheEntrenchmentofLegislativeProcedureinAmericanStateConstitutionalLaw,”PaperpresentedattheannualmeetingoftheAmericanPoliticalScienceAssociation,Washington,DC..2015.“WorkinginaCage:TheEvolutionofConstitutionalRestrictivenessinU.S.StateLegislatures,”PaperpresentedattheannualStatePoliticsandPolicyConference,Sacramento,California..2016.“Powers,Mandates,RestrictionsandExclusions,OhMy...ExploringtheInfluenceofStateConstitutionsontheLegislativeBranch,”PaperpresentedattheannualMidwesternPoliticalScienceAssociationconference,Chicago,Illinois.Pierson,Paul.2000a.“IncreasingReturns,PathDependence,andtheStudyofPolitics,”AmericanPoliticalScienceReview.94:251-267..2000b.“NotJustWhat,butWhen:TimingandSequenceinPoliticalProcesses,”StudiesinAmericanPoliticalDevelopment.14:72-92.North,DouglassC.1990.Institutions,InstitutionalChangeandEconomicPerformance.Cambridge:CambridgeUniversityPress.

23

Ramsey,Ross.2015a.“Analysis:VotesCount,buttheRulesCanCountMore,”May27,2015.TheTexasTribune..2015b.“Lawmakers:BillSlaughterNotaPartisanDeal,”May27,2015.TheTexasTribune.Schickler,Eric.2001.DisjointedPluralism:InstitutionalInnovationandtheDevelopmentoftheU.S.Congress.Princeton,NJ:PrincetonUniversityPress.Skocpol,Theda.1992.ProtectingSoldiersandMothers:ThePoliticalOriginofSocialPoliciesintheUnitedStates.Cambridge,MA:BelknapPressofHarvard.Squire,Peverill.2012.TheEvolutionofAmericanLegislatures:Colonies,Territories,andStates,1619-2009.AnnArbor,MI:UniversityofMichiganPress.Sturm,AlbertL.1982.“TheDevelopmentofAmericanStateConstitutions,”Publius.12:57-

98.Tarr,G.Allan.1998.UnderstandingStateConstitutions.Princeton,NJ:PrincetonUniversityPress..2014.“ExplainingStateConstitutionalChange,”TheWayneLawReview.60(Spring2014).Thelen,Kathleen.2000.“TimingandTemporalityintheAnalysisofInstitutionalEvolutionandChange,”StudiesinAmericanPoliticalDevelopment.14:101-108.Thorpe,FrancisNewton.1909.TheFederalandStateConstitutions,ColonialCharters,andtheOrganicLawsoftheState,Territories,andColonies;NoworheretoforeFormingtheUnitedStatesofAmerica.Washington,DC:U.S.Congress.Winkle, John W. 2014. The Mississippi State Constitution, second edition. The OxfordCommentarieson theStateConstitutionsof theUnitedStates,G.AlanTarr, serieseditor.Oxford,UK:OxfordUniversityPress.

Table 1. Similarity of Constitutional Sections, First Constitution-1907

StateNot Modified

[Similarity=1]Modified

[Similarity<1 & >0 ]Added

[Similarity=0]Deleted

[Similarity=0]4.07% 11.92% 65.99% 18.02%

[14] [41] [227] [62]

89.02% 7.84% 3.14% 0%[227] [20] [8] [0]

4.48% 20.63% 56.50% 18.39%[10] [46] [126] [41]

0% 13.11% 83.61% 3.28%[0] [16] [102] [4]

68.42% 20.30% 8.27% 3.01%[91] [27] [11] [4]

79.21% 2.97% 16.83% 0.99%[80] [3] [17] [1]

100% 0% 0% 0%[261] [0] [0] [0]

47.46% 10.63% 34.12% 7.78%[683] [153] [491] [112]

*Number of sections in between brackets

United States

Wyoming

Totals

Alabama

Colorado

Illinois

New Jersey

Minnesota

Figure 1. New Constitutions and Total Amendments per State 1776−1907

10 0 0 10 20 30 40 50 60

OK

WY

WA

OR

MT

UT

ND

ID

NV

US

WI

KS

RI

CO

MN

NE

IN

OH

NJ

WV

MI

CA

TN

IL

IA

KY

DE

PA

VA

MS

AL

SC

GA

State Amendments

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

2

2

2

2

2

2

2

3

3

3

4

4

4

5

5

6

7

7

0

0

4

6

7

7

8

10

13

21

22

23

25

28

60

1

12

17

30

38

55

62

6

7

14

1

6

21

14

18

9

24

45

Constitutions

0

2000

4000

6000

1780 1790 1800 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900Year

Tota

l Sec

tions

Bill of Rights and MinoritiesConstitutional AmendmentElections and ApportionmentFederal GovernmentFinance and TaxationLocalPolicyStructure of Government

Figure 2. Evolution of Constitutional Topics, 1776−1907

�'DVKHG�OLQHV�UHSUHVHQW�WKH�DGRSWLRQ�RI�QHZ�FRQVWLWXWLRQV

FIGURE 3.1 SECTIONS BY TOPIC, ALABAMA 1819-1907

17 [Added]

1 [Deleted]

United States

NewJersey1907

102

[Added]

NewJersey1776

4

[Deleted]

New Jersey

Alabama1907

227

[Added]

Alabama1819

62

[Deleted]

Alabama

Illinois1907

126

[Added]

Illinois1818

41

[Deleted]

Illinois

Minnesota1907

11

[Added]

Minnesota1857

4

[Deleted]

Minnesota

&RORUDGR1907

8 >$dded]

ColorDGR�����

0

[DeleteG@

Colorado

:\RPLQJ�����

Wyoming

)LJXUH����What Remains of First Constitutions?

8QLWHG�6WDWHV��7�� 8QLWHG�6WDWHV�����

:\RPLQJ�����

��[5HPDLQHG@

��

[5HPDLQHG@

��

[5HPDLQHG@

��

[5HPDLQHG@

���

[5HPDLQHG@

���

[5HPDLQHG@

���

[5HPDLQHG@

Alabama Colorado Illinois Minnesota New Jersey United States Wyoming

0.00

0.25

0.50

0.75

1.00

1819−1907 1876−1907 1818−1907 1857−1907 1776−1907 1787−1907 1889−1907

Sim

ilarit

y Sc

ore

Amendment TypeAddedDeletedModifiedNot Modified

)LJXUH����Similarity of Constitutional Sections��First Constitution − 1907

+LJKHVW�6LPLODULW\

/RZHVW�6LPLODULW\

Not Modified Modified

Added Deleted

0.0

0.5

1.0

0.0

0.5

1.0

Sim

ilarit

y Sc

ore

TopicBill of Rights and MinoritiesConstitutional AmendmentsFederal GovernmentFinance and TaxationLocalPolicyStructure of GovernmentSuffrage and Elections

Figure 6. Similarity of Constitutional Sections by Topic, First Constitution − 1907

+LJKHVW�6LPLODULW\

/RZHVW6LPLODULW\

+LJKHVW�6LPLODULW\

/RZHVW6LPLODULW\

,QFOXGHV�VLPLODULW\�VFRUHV�EHWZHHQ�WKHILUVW�FRQVWLWXWLRQ�DQG������RI�$ODEDPD��&RORUDGR��,OOLQRLV�0LQQHVRWD��1HZ�-HUVH\��WKH�8QLWHG6WDWHV�DQG�:\RPLQJ�

0.00

0.25

0.50

0.75

1.00

1787−1907Period

Sect

ion

Sim

ilarit

y Sc

ores

Amendment TypeaddeddeletedmodifiedNo changes

)LJXUH����United States Constitutional Changes Over Time

0.00

0.25

0.50

0.75

1.00

1819−1860 1860−1861 1861−1864 1864−1865 1865−1867 1867−1868 1868−1874 1874−1875 1875−1900 1900−1901Period

Sect

ion

Sim

ilarit

y Sc

ores

Amendment TypeaddeddeletedmodifiedNo changes

)LJXUH����Alabama Constitutional Changes Over Time

0.00

0.25

0.50

0.75

1.00

1819−1

820

1830−1

831

1840−1

841

1850−1

851

1860−1

861

1864−1

865

1867−1

868

1874−1

875

1880−1

881

1890−1

891

1900−1

901

1906−1

907

Year

Med

ian

Sim

ilarit

y Sc

ore

)LJXUH����<HDU�E\�\HDU�VLPLODULW\��Alabama 1819−1907+LJKHVW�6LPLODULW\

/RZHVW�6LPLODULW\