Micro-FoundationsofFragility:Concepts,MeasurementandApplication

GhassanBalikib,a,TilmanBrücka,b1,NeilT.N.FergusonaandSinduW.Kebedec,a

aISDC-InternationalSecurityandDevelopmentCenter,Auguststr.89,10117Berlin,Germany

bLeibnizInstituteofVegetableandOrnamentalCrops,Theodor-Echtermeyer-Weg1,14979Großbeeren,Germany

cHumboldt-UniversitätzuBerlin,Invalidenstr.42,10115Berlin,Germany

31July2017Abstract:Inthispaper,weexplorethemicro-foundationsoffragilitybydiscussinghowtomeasure theexposure to fragility at the individual level. Inparticular,we focusontwo important notions that are not covered by existing indicators of fragility at theaggregate level. First, different individualsor societal groupsmayexperience fragilityvery differently. Second, even though a country as a whole may not be “fragile”,individuals living there may be exposed to manifestations of fragility. Thisdifferentiation is particularly important as it suggests that the experience of fragilityvaries not just at national levels but also between sub-national regions and, indeed,between households and individuals. To test this idea, we define the novel concept“exposure to fragility”, which accounts for human security, economic inclusion andsocial cohesion. Building on this definition,we derive a series ofmetrics that can becollected in typical household surveys and test the performance of this “fragilityexposure module” in an on-going survey in Kenya. Analysis of this data shows thatindividualslivinginruralareas,aswellasyoungandsingleindividuals,exhibitgreaterexposure to fragility. These findings demonstrate the importance of understandingfragility at the individual level, particularly as it provides the basis to understandingwhichregionsorpeoplewouldbenefitmostfrompro-stabilityinterventionsandtohowtheseinterventionsperform.

1 Corresponding author. Email: tilman.brueck@isd-center.org. Telephone: +49 151 1117 5462. Address:Auguststr.89,10117Berlin,Germany.Acknowledgement:WearegratefultothecommentsreceivedattwoAERCworkshopsinNairobi,Kenyain2016and2017.Inaddition,wearethankfultoAnkeHoefflerforindepthcommentsonanearlierversionofthisdraftpaper.FinancialsupportfromtheAERCisgratefullyacknowledged.Theusualdisclaimerapplies.

TableofContents1.IntroductionandBackground 1

1.1Introduction 11.2StatusQuo 2

2.Approach 42.1DefiningFragility 52.2TheFragilityExposureIndex 62.3TheFragilityExposureModule 8

3.CaseStudy:FragilityinKenya 103.1CountryandDataBackground 103.2Approach 113.3Results 12

4.Conclusions 14

References 16

Annexes 18Annex1:TablesandFigures 18Annex2:FragilityModule 25

1

1.IntroductionandBackground

1.1IntroductionIn the last twodecades,agrowing literaturehas focusedon thenegativeroleof statefailure for economic growth and development.While it is strongly assumed andwellunderstood that strong institutionsare important, there isno real consensuson theirrole in fostering economic development (North et al., 2007). The causal relationshipbetweentheconstituentcomponentsofstrongstates–suchasgovernance,institutionsand security – and positive economic performance remains complex to disentangle,particularly at the macro-level. On the one hand, weak institutions are considered ahindrancetoeconomicperformance(Acemogluetal.,2005),whileontheotherhand,poorly performing economies are prone to ‘fragility’. Suchdebates have been equallyprevalent among practitioners given that “fragile states” face enormous difficulties inachieving theMillennium Development Goals, particularly as the poor governance ofsuch“fragilestates”oftenledtoreducedinternationalaid(OECD,2013).Nevertheless, it is equally agreed that there remains an ambiguity onwhat defines a“fragilestate”andhowthecriteriaareputtogethertogenerateameaningfulmeasureoffragility.Weposit that the rootof this issue lies in the state-centredapproach that isoftentakento‘measuring’fragility.Instead,wearguethatfragilitycanbetracedtotheindividual-levelbyaccountingforindividuals’exposuretovariousmanifestationsofitsimpacts.Inturn,wearguethataccountingforthemicro-foundations,micro-experiencesand micro-perspectives of fragility will have significant implications for economicresearchonfragilityandforthedesignoffragility-sensitivepoliciesandinterventions.In particular, ourmicro-level approach to fragility focuses on the importance of howdifferent actors experience fragility and how such experiences can aggregate intofragilityatthestate-level.Forexample,adiffuseperceptionofmacro-leveluncertaintyduetofragilitywilltranslateintochangedperceptionsattheindividuallevel,whichcan,subsequently,harmfullyimpactoneconomicdecision-making.Inthisarticle,wegenerateaworkingdefinitionoffragility,focussingonthreedomains:economicinclusion,socialcohesionandhumansecurity,whichinturnactsasthebasisofourproposed‘fragilityexposuremodule’.Althoughthetermsuseddifferslightlyfromthose used in other attempts to empirically understand fragility, such as the FragileStates Index (FSI),2 the concepts included are analogous. Basedon this definition,wegeneratealistofindicatorsthatcompriseeachdomainoffragility,whichwereinsertedinto the HORTINLEA3 survey. Using multidimensional indexing techniques, weaggregate the indicators into a single Fragility Exposure Index (FEI) and compare its2See:http://fundforpeace.org/fsi/3HORTINLEAisanon-goingmicro-levelpanelsurveyconductedinruralKenya.Seewww.hortinlea.orgformoredetails.

2

outcomesacrosskeyregionalanddemographicgroupings.The results indicate notable variations in experiences of fragility across geographiclocations. Individuals living in rural areas are more likely to be exposed to fragility.Moreover,youngandsinglehouseholdsaremoreexposedtofragilitycomparedtootherindividuals in the sample. Religious background also plays an important role on howindividuals experience fragility: Catholics exhibit more fragility compared toprotestants and Muslims. These results provide important support for the idea thatfragility manifests itself differently at the individual level, even if such micro-levelexperiencesarenotnecessarilyitscause.Inafinalstep,wecompareanaggregateversionoftheFEIwithoutcomesfromtheFSI.ThiscomparisonrevealsthattheFSIdrawsableakerpictureofthestateoffragilityinKenya than canbe aggregated from themicro-level.We find themajor source of thisdifferencestems fromthesocialcohesion,as theFSIdoesnot (orcannot)account forthe roleof informal institutionsandnetworks, an importantaspectofKenyansocietythat can mitigate the influence and experience of fragility (Narayan, 2002). Thesefindingsshowtheimportanceofconsideringandmeasuringfragilityatthemicro-level.Futureworkshouldseek toanalysemultiplecasestudiesofnationally representativedata, inorder toprovidemorepreciseand robust findingsacrossmultiple contextualbackgrounds.Therestofthispaperisstructuredasfollows:Section1.2providesaconciseliteraturereviewoncurrentstateof fragility indices.Section2describes indetailourapproach,underscoring the definitions used (2.1), the Fragility Exposure Index (2.2), and thesurveymoduleof fragilityexposure(2.3).Section3presentstheresults fromthecasestudyinKenya.Thelastsectionconcludesanddescribespossiblefuturework.

1.2StatusQuoBeginning inthemid-1990s,a largebodyof literaturehasdevelopedthat focussedontheroleofthestate,statecollapseandstatefailure(BinzelandBrück,2009;Zartman,1995;Milliken,2003;Goldstoneet al., 2004;Francois andSud,2006;Andersonet al.,2007;GhaniandLockhart,2008).Followingthiswork,thedebateonfragilitywidened–particularlyamongpractitioners–toincludecountriesthatemergefromstatuefailureand those thatare threatenedwith futurecollapse.Different terminologieshavebeenproposedandusedtodescribethisphenomenon–“lowincomecountriesunderstress”(LICUS;WorldBank,2005; IEG,2006); “difficultenvironments” (Morenoetal.,2004);“fragilestates”(USAID,2005)and“weakstates”(Rice,2006).Atthesametime,despitesuchtermsbeingincommonuseforoveradecade,thereisstillsignificantdebateaboutthemeaninganddefinitionof fragility–or indeed,ofwhatcharacterisesanon-fragilestate(AsianDevelopmentBank,2006).

3

Definitionsof fragilityare typicallystate-centred,perhapsbecauseof thetermfindingitsoriginsinaliteratureonstatecollapse.Indeed,whatweareeffectivelydiscussingisthe concept of “state fragility” (Khan, 2004; Picciotto et al., 2005; Dibeh, 2008).Consequently, definitions tend to focus on state-level issues like legitimacy,effectiveness,capacitytoimposeaWeberianmonopolyontheuseofviolence,provisionof public goods, etc. Particularly given that some states may exhibit some adversefeatures but not others, however, it is still unclear how a combination of theseadversities,orindeedwhichcombinationoftheseadversities,adduptofragility.Two measures have been proposed to implement these definitions empirically: theFailedStatesIndex(FSI)(FundforPeace,2009)andthePoliticalInstabilityTaskForce(PITF). FSI is updated annually and is composed of 12 state-level indicators ofpresumed drivers of fragility – these include: “mounting demographic pressures”;“uneven economic development along group lines”’ and “progressive deterioration ofpublicservices”.Thesesub-indicatorscanbegroupedinthreedistinguishabledomains,namely political and military, economic, and social - analogous to our micro-levelconceptsofhumansecurity,economicinclusionandsocialcohesion.Basedonthegiventhreshold, the FSI classifies a large number of countries as fragile, affecting not onlylow-income countries, although many low-income countries are deemed fragile. ThePITF also looks at a variety of societal, demographic, economic, political andenvironmental factors that influence the likelihood of state failure (Goldstone et al.,2005). Both approaches, however, suffer potential endogeneity as many of thesefeaturesareasmuchanoutcomeof fragilityastheyareacauseof it.Accordingly, thestrongestpredictoroffragilityinoneperiodisfragilityinthepreviousperiod,creatingaperniciouscycleoffragility.In this context, despite initial promise, definitions andmeasure of fragility remain intheirinfancy.Thereremainsalackofprofoundunderstandingoftheactualmechanismsinvolved.Howdoesfragilityemerge?Whateffectsdoesithaveongrowthandpoverty?Howit ismaintainedorcontained?Andhowitcanbeovercome?Moreover,thereisapaucity of knowledge on how people actually experience fragility, how it impacts ontheir life,howtheycopewith it,andhowtheir livesdiffer fromthose living inanon-fragilesettings.Toovercomethisgap,wedeveloptheFEMandFEI,whichmeasurehowindividuals experience fragility by collecting and aggregating information on howpeople are exposed to various indicators of state or other failures. In turn, whilstproviding understanding of how individuals experience fragility, when the FEM iscollectedinrepresentativehouseholdsurveys,theFEIcanbeaggregatedintoregionalor national measures of fragility. This is akin to recent developments in conflictresearch, which has seen an increased reliance on the measurement of conflict andconflictexposureattheindividuallevel(Brücketal.,2016;Justinoetal.,2016).

4

2.ApproachItisnowcommonlyacceptedinresearchandpolicycommunitiesthatmultidimensionalindicators are importantmeansofmeasuring complexphenomena. Inmany contexts,several factors, rather than single measures, define the extent of what is commonlyconsideredasingleoutcomevariable.Particularlyinrelateddevelopmentfields–suchasthemeasurementandunderstandingofpoverty,measuringfood(in)security,etc.–strongmethodologiesforgeneratingsuchindiceshavebeendeveloped(see:AlkireandFoster, 2011). Until recently, fragility measures did not tend to includemultidimensionality, even though recent work (OECD, 2015) acknowledges itsimportance. The OECD (2015) report suggests that the use of multiple indicators isneeded to measure fragility and that such measures should include: indicators of:violence; access to justice; effective, accountable and inclusive institutions; economicinclusionandstability;andcapacitytopreventandadapttoshocksanddisasters.In typical approaches, all of these indicators - and consequently themeasurement offragilityitself-areidentifiedatthestatelevel.Suchapproaches,however,maybesub-optimalforanumberofreasons.First,whilstastateasawholemaynotbefragile,areaswithin it couldwell surpass given thresholds where those areas, instead, the unit ofanalysis. Second, anumberof important conceptswithin thesemacro-level indicatorsmaybedifficulttomeasureatthestatelevel.Thisisofparticularconcerntoindicatorsthat aim to measure social phenomena, which almost inherently take place at theindividualorgrouplevel.Third,atthemacro-level,anumberofendogeneitiesarelikelypresent between these indicators, inflating (long-term) the level of fragility. Forexample, the link between conflict and macroeconomic growth is uncontroversial(Migueletal.,2004).Inturn,indicatorsthatincludebothconceptsriskmultiplyingsucheffects.Atthemicro-level,however,suchconcernsarenot,immediately,asconcerning.A household that is affected by conflict but experiences no associated change ineconomic status experiences fragility differently to one that experiences both conflictand changing economic fortune. Therefore, whilst we propose to take these ideas ofmultidimensional indexing,wedoso inorder todevelopand identify the indexat theindividuallevel,usingbespokedatacollectedfromhouseholdsurveys.Tothisendweproposeanon-statecentredconceptualisationof fragilitythat looksatthe existence of persistent, systematic and interrelated social, political and economicuncertainties experiencedby individuals. Such an approach allows thesemeasures tovary across even small geographic areas; between different individuals; and betweendifferent socio-economic or demographic groups. More so, the time dynamics of theexperiences of these individuals (say, in response to some counter-fragilityintervention)canalsobestudiedinsituationswherelongitudinaldataiscollected.Thisallowsustoincludespecificfeaturesthatareuniquelyprevalentatthelocalorregionallevel,aswellasatthenationallevel.

5

Inturn,thisboilstheconceptualisationoffragility,evenwhenitisdefinedatthestatelevel, as something that impacts on different groups and different individuals indifferent ways. Even outside of such concerns, however, measurement at this levelovercomessomeoftheendogeneitiesthatmaybepresentinindicesthatbringtogetheraggregatedstate-levelindicators.Ourapproach,therefore,doesnotaimtodisregardorunderminethesignificantandimportantworkthathasbeendoneatthestatelevelbut,rather,torefinetheseconceptsandtoprojecttheirsignificanceattheindividuallevel.In short, this boils down to the notion that fragility, even defined at the state level,impactsondifferentgroupsand individuals indifferentways.Ourapproachdoesnot,therefore,aimtodisregardthesignificantworkthathasbeendoneatthestatelevelbut,rather,toprojecttheirsignificanceontotheindividuallevel.

2.1DefiningFragilityThere is no single shared definition of fragility – indeed, such a lack of a shareddefinition ismuch of the reasonwhymultiple lists of fragile countries exist andwhytheselistsseldom,ifever,fullyoverlap.Inanumberofcases,thedefinitionoffragilityseemstoderivefromthecountriesthatfindthemselvesonagivenlist,ratherthantheotherwayaround.At least threemajor issuesstem fromthiskindofprocess.First, itremains unclearwhy some countries end up being defined as fragile and others not,despite thedifferencesbetweensome fragileandnonfragile statesbeingsmaller thanthosebetweennonfragilepairs.Third,bydefininga countryas ‘fragile’, ormeasuringfragility only at the state level the spatial dimensions within that country arefundamentallyignored.Fourth,thereisasignificantdiversityofreasonswhycountriesmay be considered fragile. This includes experiences and threats that are commonlythought of when considering fragility, like weak governance or conflict risks, but insome cases also includes climate risks, for example, and weak capacity to deal withassociated climate shocks. In turn, “fragile” statesmay be fragile in somedimensionsbuttheymaybenon-fragileinotherdimensions.Inturn,somehighlyfragilestatesmaysharemorecommonalitieswithlessfragilestatesthanwithotherhighlyfragileones.Inresponse,weseekadefinitionoffragilitythatbuildsupfromthemicro-level.Thisisnottosaythatweviewfragility,itself,asstemmingfromtheindividuallevelorasbeingentirely a micro-level phenomenon. Rather, it is an argument that fragility can bemeasured through individuals’ experiences ofmanifestations of fragility. This iswhatwecallinshort“exposuretofragility”.The benefits of such an approach are manifold. First, at the international level, itreduces the need to draw artificial distinctions between fragile and nonfragile states,

6

whichcanbedamaginggiventhediversityofreasonswhyastateisconsideredfragileand the spatial variation in fragilitywithin states. Second, by aggregating up fromanindividual level, we facilitate analysis of differences at individual, group and sub-national levels, as well as across countries. Such an approach not only allows betterunderstanding of the spatial dimensions of fragility but also about which forms offragility affect which groups most. More so, it provides better opportunity tounderstand the feedbacks between fragility and economic development that are notavailable at the state level. Finally, because such an approach aggregates together awide diversity of drivers of fragility, it provides the basis of comparison betweenpeople,sub-nationalregionsandcountries.Inturn,acountrycouldscorewellonsomeaspectsoffragilityandpoorlyonothers,allowingbetter“matching”offragilestatesorregionsby typology. Such comparisonswouldprovide abettermeansof comparativeresearch across fragile countries, areas or regions. This is particularly important inunderstandingtheeffectivenessofvariousfragility-reducinginterventions,particularlythoseeffectiveenough toshape individual livesbutnot largeenough to impactat thenationallevel.A lack of agreed definitions, however, acts as much of a barrier to measuring andanalysingfragilityattheindividuallevelasitdoesatthenationallevel.Asaresult,wegenerate a working definition of fragility that will be used in the generation of anillustrativeFragilityExposureModule.Wedonotpresentthisdefinitionasexhaustiveordefinitive–otherdefinitionscouldwellincludedifferentcomponentsandassociatedmetrics.Atthesametime,webuildthisdefinitionfromarangeofliteratureonfragilityandstability,whichensuresthattheresultingFEMandFEIareproperlycontextualised.

2.2TheFragilityExposureIndexInlinewiththestate-centricmacro-levelliteratureonfragility,weconsiderfragilityinthe context of state functions and institutional capacities. Methodologically, animportantfeatureofanydefinitiontobeusedinthegenerationofamultidimensionalindexisthatateachdomainoftheindex,andthuseachconstituentpartofthedefinitionitself,canbereadilyseparated.Assuch,combiningthisneedforseparationbetweenthedomainswith the fundamental basesof traditional definitionsof fragility yields threedomainsofinterest:humansecurity;economicinclusion;andsocialcohesion.Althoughwetitlethesedomainsdifferentlyforparsimony,theyshareanumberofkeyoverlapswiththosederivedfromotherdefinitionsoffragility(e.g.theFSI)4groundingboththedefinitionandtheoutcomesinthebodyofliteraturetodate.

4Wedonotpresentthisdefinitionaseitherexhaustiveordefinitivebutasausefulbaselinefromwhichtocreateandanalyseamicro-levelindexoffragility.Duetothenatureofthedata,however,theapproachitselfis not sensitive to definitions, as metrics can be included or excluded to match alternative definitions. Afuturesourceofresearchshouldtesthowsensitivetheindexandresultsaretodefinitions.

7

Aswithotherdefinitions,thebasisoffragilityinthisworkisrootedinthefunctioningofa strong and good government, with each domain linking to these functionings.Therefore,theabsenceofconstituentpartsofanyoftheseplaysacontributoryroletofragility.Wefurtherdefineeachsub-componentoftheindexasfollows:Human Security is, at its very base level, a focus on individual protection but isconsidered more broadly than simply individuals being protected from physicalviolence.Wethereforeconsiderhumansecuritytoencompassphysicalsafety,suchasexposuretoarmedactorsandexperienceofviolencebutalsoalackofgroup-orgender-baseddiscrimination,andequal rightsbefore the law. In turn, thisdomainhasstronglinks to political institutions, such as an effective local and national justice system;civilianpoliceforce;etc.Economic Inclusion addresses alleviating extremepoverty and inequality but, again,lacks a single accepted definition. Those definitions that do exist, however, share anumberofimportantoverlaps,onwhichwefocus.Inthatregard,weconsidereconomicinclusionastheprovisionofopportunityandabilityforallpeopletotakeanequalsharein economic opportunity. That is, that no individuals, or groups, should be excludedfromsuchopportunity.Whilstthiscanfocusonpoorpersonaleconomicsituationsandopportunities,itcanalsoincludeunevenaccesstopublicservicesortheexperienceofcorruption.Social Cohesion is based around an idea that members of communities have theopportunitytocooperatewithinandacrossgroups.AsChanandChan(2006)state,thisis a situation that facilitatesvertical andhorizontal interactionsanda setof attitudesandnormsthat includetrust,asenseofbelonging,andawillingnesstoparticipate. Inthis regard, we consider social cohesion to reflect participation in communities andtrustingovernmentandotherinstitutions(bothformalandinformal).TheFragilityExposureIndex(FEI)forindividualiintimetisthenspecifiedas:

where !"#$, &"#$, and '"#$ represent the three domains: human security, economicinclusion, and social cohesion, respectively. Each sub-indicator within the followingdomainsinnormalisedtotakeavaluebetween0and1asfollows:

8

where (, ) and * denote the individual, sub-indicator, and time period (e.g., year),respectively. This is important to ensure that each sub-indicatorwithin each domaintakesthesameweightinthefinalindex,asnotallvariableshavesimilaranswerranges.Hence,foreachdomainwesumupthetotalnumberofthenormalisedsub-indicatorstodevelopanequallyweighteddomainindex.Inthefinalstep,asshowninequation(1),eachdomainisthenalsonormalisedinordertoprovideequalweightsbetweenthesedomains for thegenerationof theFEI.Givena lackof strongpriorsonwhichare themost importantanda lackofbespokedataonthosethat individualsconsiderkey,wearguethatthisequallyweightedindexisthemostappropriatemethodavailableforthisanalysis. These equal weights can be augmented to deal with many of the concernsraisedinarangeofmultidimensionalindexingliterature(GoosandManning,2002).Using equation (1), we are able to classify individual experiences of or exposure tofragilityintooneindexwhichhas,initially,avaluebetween0and3,where0istheleastfragileand3isthemostfragile.However,thenormalisationoftheindexallowsus,orother researchers, to transform the values to suit specific analytical needs withoutjeopardizingtheunderlyingdistributions.HerethefinalFEItakesvalueof0and100.Wediscusstheimplementationofthesefacetsoffragilityinthenextsection.

2.3TheFragilityExposureModuleWefocusourattentiononthreemain“domains”thatweargueconstrueawiderangeoffragility-related issues:humansecurity, economic inclusionandsocial cohesion.Fromthis stems a requirement to discuss which indicators and metrics accurately reflectthesedomainsandwhichdooptimally.Inreality,giventherestrictedspaceinon-goingsurveys,thesedesiresneedtobetraded-offagainstensuringthatthemodulecaneasilybeinsertedinarangeofsurveys.Similarly,theyalsoneedtobetradedoffwiththestyleof the questions asked and the familiarity of these questions to survey designers,statisticalagenciesandenumerationteams.Wefirstboilourthreedomainsdownintotwodistinctivecategories:“Experiences”and“Perceptions” of fragility. The former includes indicators that measure actualexperiences of fragility at the micro level. These include for example political andcommunityengagementandexperiencesofinsecurityandcorruption.5Whilethelatter5 It iswellworthnotingthatallexistinghouseholdsurveyscollect informationonahousehold’seconomicsituationandtheirexpectationsforthefuture.Similarly,questionsabouttrustininstitutionsarecommoninthesesurveys–includingtheHORTINLEAdataweuseinthisarticle,inwhichwehavealreadyinsertedthe

9

include indicators capturing: individual fears and satisfaction on a range of security,economic, and social aspects; individual perceptions of the effectiveness of a numberformalandinformalinstitutions,andtrustintheseinstitutions.Inordertogeneratethequestionnaire,wereviewedarangeofsurveysandgarneredanextensivelistofpotentialquestions.Subsequently,wereducedthislisttoincludethosemost directly linked to these key concepts, whilst also ensuring that these conceptswere now ‘double counted’. Below we describe the indicators that we used in thegenerationofourindex.First,fearandsatisfactioncapturetheextentoffragilityexperiencedatthemicro-levelinallthreedomains.Inregardstohumansecurity,weincludemeasuresonsatisfactionwithpersonal,neighbourhoodandnationalsecurity; fearofcrime,assault,war,ethnicconflict,andpoliceviolence.Asforeconomicinclusionandsocialcohesion,wemeasuresatisfaction with economic and financial situation, education, health, communityintegration,etc.Weregardincreasesinfearasworseningofthefragilitystatusquoandimprovedsatisfactionasabetterment.Second,wemeasureinstitutionalstrengththroughquestionsthataskaboutindividualperception on the effectiveness of a range of formal and informal institutions. Thespecificallynamedinstitutionsincludesomethataregeneric–suchascentralorlocalgovernment,police,orcourts–andothersthatarecontextspecific,suchastribalelders,religiousbodies,etc.Ingeneral,weviewincreasingperceptionsofeffectivenessasthebasis of a lower exposure of fragility. This implies, not only thatmore effective stateinstitutions correspond to lower levels of fragility but that, in the absence of sucheffective institutions, more effective informal institutions still mitigate fragility.Moreover, trust is measured through a range of questions that ask individuals howmuchtheytrustthesevariousformalandinformalinstitutions,aswellastheirfamilies,theirneighboursand theircountrymenandassumes thathigher trust isasignof lessfragility.Third,politicalandcommunityinvolvementisbasedonself-reportedparticipationinarange of secular and religious organisations, political parties, and elections. Wehypothesise thatmoreparticipation isasignof increasedcommunitycohesionandofreducingfragility.Experienceofcorruptionismeasuredthroughquestionsthataskhoweasyitistoobtainassistancefromarangeofinstitutionswithoutpayingabribe,whileexperiencesofinsecurityismeasuredthroughthepresenceofactivecriminalgroups.Table1 listsall thesub-indicatorsusedwithineachdomain.Moreover,anexampleof

fragility exposure module. The fragility exposure module, therefore, should be viewed as an extension ofexistingdatacollectionprocesses,ratherthanaseparateeffort.

10

theaugmentedquestionnairecanbefoundinAnnexAforourworkinKenya.6

[TABLE1ABOUTHERE]

3.CaseStudy:FragilityinKenya

3.1CountryandDataBackgroundKenya is an ethnically, culturally, and economically diverse country. After the post-election violence that took place in 2007, Kenya has been classified by a number ofindicesasafragilestate,andsincethenthecountryhasremainedinthehigh-riskgroupoffragility.Forexample,basedonthe2016FSI,Kenyaliesinthe20mostfragilestates.In the aftermath of the contested 2007 election, Kenya has undergone a number oflegislativeandconstitutionalreforms,whichmainlyreducedthepowerofthepresident,enhanced the role of parliament and citizens, and created an independent judiciary.Most notably the reforms provided a very ambitious decentralization process whichaimed to transfer important governance decision-making to sub-national legislativeunies.Thisgivescountiesfullautonomytoaddresslocalneedsinprovisionofservices(World Bank, 2012). However, these constitutional reforms coupled with impressiveeconomic growth were not sufficient to significantly improve Kenya’s fragilityclassification. Based on the diverse nature of the country, and the newly introduceddecentralisationreforms,exploringtheunderlyingexperiencesoffragilityatthemicro-levelarelikelytoshedlightonwhysuchapparentimprovementshavenotsubstantiallychangedthemacro-levelfragilityenvironment.Weusedatacollectedinthe2016HORTINLEAsurveycollectedinrural,peri-urban,andurbanareasofKenyatotestthevalidityofourfragilityexposureindex.Datacollectionunder theHORTINLEA survey started in September2014and continued in2015and2016inatotalofthreewaves.TheFEMwasintroducedtothesurveyquestionnaireinthe latest wave.7 Even though the main focus of the survey is on agricultural andhorticultural production, it contains comprehensive socio-economic information onhouseholdsandindividuals,whichaugmentthereachoftheFEM.Households for thesurveywereselectedusingamultistagesamplingapproach.Given

6 In each survey, these questionswill be context specific. This ensures that the institutionswe use in ourquestions, the reference to neighbourhoods and areas, etc. are matched to those that our respondentsunderstand.7The2016wavewasconductedfromSeptembertillOctober2016byHumboldtUniversityofBerlinincollaborationwithEgertonUniversityandLeibnizUniversityofHannover.Thedatacollectionisfundedundertheinitiativeforglobalfoodsecurity(GlobE)oftheGermanFederalMinistryofEducationandResearchandtheGermanFederalMinistryforEconomicCooperationandDevelopment.

11

theagriculturalnatureofthesurvey,apurposivesamplingtechniquewasusedtoselectthefivecountieswithinruralandperi-urbanstrata.TheseincludedKisiiandKakamega(rural),Nakuru,Kiambu(peri-urban),andNairobi(urban).Selectionofthesub-countiesanddivisionswasbasedoninformationfromtherespectivedistrictagriculturaloffices.From each division, locations/wardswere randomly selected, and householdswithinlocationswereinturnrandomlyselected.TotalsamplesizeN=1000households:700in rural and peri-urban counties and 300 in Nairobi. It is important to note thatHORTINLEAhouseholdsurveyisnotrepresentativeatthenationallevel.However,thedata provides a comprehensive overview of households engaged in small-scaleagriculturalproductioninruralandperi-urbanareas.Therefore,althoughtheresultsofanalyses conducted using this survey data cannot be fully generalized to representstate-levelfragilityinKenya,itprovidestheopportunitytotestthebaselinehypothesesofthiswork-thatindividualsandgroupsexperiencefragilitydifferently,evenwhenallresidewithinademonstrablyfragilestate.

3.2ApproachTheHORTINLEAsurveyquestionnaireincludesalargesectiononcrimeandfragilityinaddition to general socio-economic and demographic information, ensuring that thesurveyalreadycoversarangeofkeyFEMquestions.Theseinclude:satisfactioninlivingconditions,personalandneighbourhoodsecurity, financialandsocialstanding; fearofcrimeathomeandoutside,andfearofwarandethnicconflict;participationinlocalandcentral elections, as well as membership in political parties and social groups (e.g.,women'sgroups);trustinacomprehensivelistofformalandinformalinstitution(e.g.,central government, police, courts, informal village government, etc); perception ofpowertheseinstitutionsandtheireffectivenessintheprovisionofservices;experiencesofcrime,insecurity(measuredviathepresenceofnon-stategroups),andcorruption.Before delving into the results of the final index, we provide an example on theimportance of using micro-level fragility measures. Figure 1 indicates for a selectednumber of formal and informal institution the mean values of the following fourvariables: (i) Power; (ii) Effectiveness; (iii) Trust; and (iv) Ease of Services WithoutBribes. In all cases, questions are asked on Likert scale running from 1-10, with 1implyingtheworstindicators;and10thebest.Wefindthatreligiousinstitutionshaveameanvalueofabout8forallfourvariables.Inother words, Kenyans perceive religious institutions to be very powerful, effective,trustworthy, and that they can obtain services easily from themwithout bribes. Thisstatictrendalsoappliesforvillagegovernmentsatameanvalueofabout6.However,even thoughKenyansperceive thecentralgovernment tobepowerful (meanvalueofabout7),theydonottrustitwiththesameintensity,andarenotabletogetassistanceeasilywithoutbribes.Thesamediminishingtrendappliestothepoliceandcourts.

12

[FIGURE1ABOUTHERE]

Theseresultsprovidean importantglimpseontheexistenceofa ‘governancegap’ forformal institutions in Kenya between the central and local levels. Central, formalinstitutionsareviewedaslessabletodeliverlegitimateservices,despitetheirperceivedpower.Identifyingthisgapunderscorestheimportanceofusingmicro-levelindicatorsformeasuringfragilityandtounderstandbetterhowdifferentindividualsareaffectedbyit.Apowerfulcentralgovernmentorpoliceforceneednotnecessarilyeffectiveintheprovisionofservicestolocalcommunities,implyingcounteractingimpactsonfragility.Individualstrustlocalinformalvillagegoverningbodiesmore,andarebelievetheyaremore likely to obtain better services from them vis-a-vis the central and countygovernments.Ceterisparibus,suchasituationmaynotbeviewedasdesirable,yetinthecaseofweakdeliveryfromthecentralgovernment,shortfallscanbecompensatedbyaneffective, if informal, formof localgovernance. Insuchacase,ceterisparibus,effective(informal) local institutionsmitigateandreduceexperienceof fragility.Measures thatdo not account for such different experiences across people and across branches ofgovernmentarelikelytomisrepresenttheexperienceoffragilityatthecountrylevel.

3.3ResultsNext,basedonequation(1),weconstructtheFEIusingthenormalisedvalueofthesub-indicatorsaslistedinTable1.Theresultscanbesummarizedasfollows:Inthefirstinstance,wepresentthebasicstatisticsofthethreedomainsthatconstitutetheFEI. Figure2 compares graphically thedistributionof these threedomains in theHORTINLEAsample.Anumberofinterestingfeaturesdevelop.First,Kenyansaremorelikely, on average, to experience fragility via human security and economic inclusion,withmeanvaluesof0.53and0.56,respectively,comparedtosocialcohesion,whichhasmeanvaluesof0.36.Second,themedianisveryclosetothatofthemean,particularlyfor the economic inclusiondomain, suggesting that outliers havenegligible effects onourcomparisons.

[FIGURE2ABOUTHERE]Inordertounderstandtheunderlyingdynamicsbehindthefragilitydomains,Figure3plotsthedistributionoftheFEIseparatingthesub-indicatorscapturingperceptionsoffragilityversusexperiencesofit.Thesameisappliedtoeachoftherespectivedomains.Figure3a. shows thedistributionof the total fragility index separatedbyperceptionsandexperiencessub-indicators.OneclearlyseesthatKenyansaremorefragileintermsofperceptions(meanof55.7)thanintermsofactualexperiences(meanof47.5).Thedifferencesarehighlysignificantwithap-value<0.001.ThisholdstruefortheHuman

13

Securitydomain(Figure3b)andtheSocialCohesiondomain(Figure3d).However,thereverseistruefortheEconomicInclusiondomain.

[FIGURE3ABOUTHERE]Inthenextstage,weruncomparativeanalysesbysplittingthesampleacrossarangeofindividualandhouseholdcharacteristics, includingregion,gender,agegroups,maritalstatus,andreligion.TheFEImeandifferencesbetweenthegroupsandtheirsignificantvaluesareshownintable2,aswellasforthethreedomainsoftheindex.Forcategoricalvariables, such as region,marital status, age, and religion,we report the significancelevelsoftheaveragemeansofthepairwisedifferencesbetweeneachcategory.First,intermsofregionalvariations,wefindthatindividualslivinginruralcountiesinKenyaexperiencesignificantlymorefragilitythanresidentsofperi-urbancounties.Thedifferencesaresignificantatthe5%level.However,thispatterndoesnotalwaysholdifweexamineeachcountyseparately. Individuals residing inKakamega,a rural county,experiencehigherfragilitylevelsincomparisontherestoftheindividualsinthesample.Incontrast,peri-urbanKiamburesidentsexhibitlowerfragilitycomparedtotherestofthe sample. Yet, there are no notable mean differences for Kisii (rural) and Nakuru(peri-urban). Hence, although the differences in fragility exposure between rural andurbanareasaresignificant,theeffectisdrivenmainlybycounty-specificvariations.Second,youngerpeople(aged25andbelow)aremorefragileincomparisontotherestoftheagecohortsinthesample,howeverthesedifferencesareonlysignificantforthesocialcohesiondomainatthe1%level,implyingthatolderpeoplehavestrongersocialnetworks. Given the time taken to build up such networks, such a finding seemsuncontroversial, yet is still important as it implies younger people may lack thenetworkstosuccessfullycopewithmajorshocks.Third,therearenonotabledifferencesinfragilitybetweenmenandwomen,whichismainlydrivenbytheequalmeansinthehumansecuritydomain.Malesexperiencemorefragilitythanfemalesintheeconomicinclusiondomain,withdifferencessignificantatthe5%level,whilefemalesexperiencegreaterfragilitythanmalesinthesocialcohesiondomain.

[TABLE2ABOUTHERE]Fourth, in terms of religion, we find no differences in exposure to fragility forprotestants (the largest group) andmuslims (the smallest group). However, catholicindividualsareonaveragemoreexposed thananyof theothergroups in the sample,including individualswhoreportedhavingotherreligiouspreferencesornone.Lastly,monogamous households experience less fragility than single and polygamoushouseholds. The differences in exposure to fragility between these groups is verynotableandsignificantatthe1%levelandisparticularlystrongforsingleindividuals,despite no noticeable differences in the economic inclusion domain between these

14

groups.Next,wecomparethefindingsfromourcasestudytothatoftheFragileStatesIndexforKenya in2015. Inorder togenerateavalidcomparison,wetransformbothour indexandtheFSItotakeavaluebetween0and30,where30denotesthegreatest fragility.Given that theFSIalsogroups its12 indicators into threesimilarcategories:Political,economic and social, we take the averages of the indicators for each category as anindicator of domain-level fragility. It is important to note that there are indeeddifferencesbetweentheindicatorsfromthetwoindices.Forexample,theFSIincludesdemographicpressureandthepresenceofrefugeesandIDPsasindicatorsinthesocialdomain,whicharenotcoveredinoursocialcohesionindicators.8Table3showstheaveragefragilitypercategoryforthetwoindices.TheFSIforKenyahas a value of 24.4 out of 30,while the FEI index has a value of 14.6 out of 30. Thelargestvariations is,asexpected, stemming fromthedifferences in thesocialdomain.Taking these figures at their face values, the findings indicate that although Kenyaexhibitshigh fragility inall threedomains, fragilityexposureat the individual level islesspronounced.WealsocomparetheresultsoftheFEIwiththatoftheGlobalPeaceIndex, which measures the relative position of states’ peacefulness, taking intoconsiderations,forexample,relationswithneighbouringcountriesandlevelofterroristandcrimeactivity.TheGlobalPeaceIndexhasatotalvalueof14.3 forKenya in2016whichismuchclosertothatourfragilityexposureindex.

[TABLE3ABOUTHERE]

4.ConclusionsIn this article, we seek to inspect the micro-foundations of fragility. In doing so, wedefine fragility along the lines of previous work but apply these concepts to howindividuals experience manifestations of such fragility. In doing so, we open up asignificantly richer research potential than is provided by state-centric, aggregate,and/orbinarymeasuresoffragilitythatarecommonintheliterature.Wegenerateaso-calledFragilityExposureIndex(FEI),whichisbasedonthreedistinctdomains:HumanSecurity,EconomicInclusion,andSocialCohesion.Thesedomainsareinturncomposedofsub-indicatorsfromasetofvariableswhichcanbeinsertedintostandardhouseholdor individual surveys. We propose a comprehensive fragility exposure module thatcapturesthevariousaspectsofourindex.

8WealsonotethattheHORTINLEAdatasetisnotnationallyrepresentative.Althoughwebelieveittobeunlikely,variationsbetweentheFSIandtheFEIcould,technically,bedeterminedbyinconsistenciesarisingfromalackofrepresentativeness.

15

Theanalysisofthisdatanotonlyallowsustodefineifastateis‘fragile’ornotbutalsotoconsiderhowdifferentindividualsinthatstateexperiencefragility.Totesttheideasbehind this work, we conduct a trial using FEI based on our full fragility exposuremodule,whichwasinsertedinasurveyinKenyain2016.Theresultsfromthisanalysisdemonstrate that individuals in Kenya experience fragility differently depending ontheir location (rural areas), age group (youth), religion (catholics), andmarital status(singles). Such findings support our assertions that aggregate measures are, often, ablunt instrument in measuring and understanding fragility. Individuals in differentlocations and of different socio-economic characteristics clearly experience fragilitydifferently.ItisimportanttonotethatthefindingsarebasedonalimitedcasestudyofourfragilityexposuremodulewhichusesdatafromtheHORTINLEAsurveyinKenya.Inthisregard,theresultswepresenthereshouldbetakenasillustrativeofwhatcanbeachievedbythis approach, rather than as a finishedproduct.However, the full FragilityExposureModulehasbeeninsertedalsointheLifeinKyrgyzstan(LiK)Study(Brücketal.,2014).The LiK Study provides a wide range of socio-economic, demographic and otherinformationsimilartothatofHORTINLEA.Itsmainadvantageisthatitisrepresentativeat the national level in Kyrgyzstan. The inclusion of the fragility exposuremodule inothersurveysallowsabetterrobustnessandvalidationtestofourindex.

16

ReferencesAcemoglu,D.,Johnson,S.,&Robinson,J.A.(2005).Institutionsasafundamentalcause

oflong-rungrowth.Handbookofeconomicgrowth,1,385-472.Alkire,S.&Foster,J.(2011).Countingandmultidimensionalpovertymeasurement.

JournalofPublicEconomics,95(7-8),476-487.Andersen,L.,Møller,B.andStepputat,F.(Eds.)(2007).FragileStatesandInsecure

People?:Violence,Security,andStatehoodintheTwenty-FirstCentury(PalgraveMacmillan).

AsianDevelopmentBank(2006).AchievingDevelopmentEffectivenessinWeaklyPerformingCountries(TheAsianDevelopmentBank'sApproachtoEngagingWithWeaklyPerformingCountries),DraftforConsultation,November2006.

Binzel,C.&T.Bruck(2009).“AnalysingConflictandFragilityattheMicroLevel:ExtendingtheVulnerabilityFramework”,unpublishedmanuscript.

Brück,T.,D.Esenaliev,A.Kroeger,A.Kudebayeva,B.MirkasimovandS.Steiner(2014):“HouseholdSurveyDataforResearchonWell-BeingandBehaviorinCentralAsia”.JournalofComparativeEconomics,vol.42,no.3,pp.819-35.

Brück,T.,P.Justino,P.Verwimp,A.AvdeenkoandA.Tedesco(2016):“MeasuringViolentConflictinMicro-LevelSurveys:CurrentPracticesandMethodologicalChallenges”.WorldBankResearchObserver,vol.31,no.1,pp.29-58.

Bourguignon,F.andS.R.Chakravarty(2003).Themeasurementofmultidimensionalpoverty.JournalofEconomicInequality.1:25-49.

Dibeh,G.(2008).ResourcesandthePoliticalEconomyofStateFragilityinConflictStates.UNU/WIDERResearchPaper,2008/35.

Francois,M.andSud,I.(2006).PromotingStabilityandDevelopmentinFragileandFailedStates.DevelopmentPolicyReview,24(2),pp.141-160.

FundforPeace(2009).http://www.fundforpeace.org/web/.Websiteaccessed22June2009.

Ghani,A.andLockhart,C.(2008).FixingFailedStates:AFrameworkforRebuildingaFracturedWorld(Oxford,OxfordUniversityPress).

Goldstone,J.A.,Haughton,J.,Soltan,K.andZinnes,C.(2004).StrategyFrameworkfortheAssessmentandTreatmentofFragileStates.UnpublishedReporttoUSAID.TheIRISCenterattheUniversityofMaryland.

Goldstone,J.A.,Bates,R.H.,Gurr,T.R.,Lustik,M.,Marshall,M.G.,Ulfeder,J.Woodward,M.(2005).AGlobalForecastingModelofPoliticalInstability.PoliticalInstabilityTaskForce(PITF),PaperpreparedforthepresentationattheAnnualMeetingoftheAmericanPoliticalScienceAssociation,Washington,DC,September1-4,2005.

Hulme,D.andA.Shepherd(2003).Conceptualizingchronicpoverty.WorldDevelopment.31(3):403-423.

IEG(2006).EngagingwithFragileStates.AnIEGReviewofWorldBankSupporttoLow-IncomeCountriesUnderStress,IndependentEvaluationGroup,WorldBank.

Justino,P.,T.BrückandP.Verwimp(2016).“NewDevelopmentsinMeasuringthe

17

WelfareEffectsofConflictExposureattheMicro-Level”.In:Backer,D.,R.BhavnaniandP.Huth,eds.Peace&Conflict2016,Routledge.

Khan,M.H.(2004).StateFailureinDevelopingCountriesandInstitutionalReformStrategies,in:Tungodden,B.,Stern,N.andKolstad,I.(Ed.)TowardsPro-PoorPolicies:Aid,InstitutionsandGlobalization(Oxford,TheInternationalBankforReconstructionandDevelopment/TheWorldBank).

KNBS.(2007).Basicreportonwell-beinginKenya.BasedontheKenyaIntegratedHouseholdBudgetSurvey2005/06.KenyaNationalBureauofStatistics,MinistryofPlanningandNationalDevelopment,Kenya.

Miguel,Edward,ShankerSatyanath,andErnestSergenti."Economicshocksandcivilconflict:Aninstrumentalvariablesapproach."JournalofpoliticalEconomy112.4(2004):725-753.

Milliken,J.(Ed.)(2003).StateFailure,Collapse&Reconstruction(Malden,Blackwell).MorenoTorres,M.andAnderson,M.(2004).FragileStates:DefiningDifficult

EnvironmentsforPovertyReduction.PRDEWorkingPaper,1,DepartmentforInternationalDevelopment,UK,1.

MuokBernard,OchiengAndrewAdwera,KariukiJoanKungu,TonuiCharles.(2012)CaseStudyonClimateCompatibleDevelopment(CCD)inAgricultureforFoodSecurityinKenyaNairobi.

Narayan,Deepa."Bondsandbridges:socialcapitalandpoverty."Socialcapitalandeconomicdevelopment:well-beingindevelopingcountries.Northampton,MA:EdwardElgar(2002):58-81.

North,D.C.(2007).Limitedaccessordersinthedevelopingworld:Anewapproachtotheproblemsofdevelopment(Vol.4359).WorldBankPublications.

OECD(2015).StatesofFragility2015:MeetingPost-2015Ambitions,OECDPublishing,Paris.http://dx.doi.org/10.1787/9789264227699-en

Okeno,J.A.,Chebet,D.K.andMathenge,P.W.(2003).StatusofIndigenousVegetableUtilizationinKenya.Proc.XXVIIHC–HorticultureinEmergingEconomies.

Picciotto,R.,Olonisakin,F.,Clarke,M.andFreedman,F.B.S.L.(2005).GlobalDevelopmentStudiesNo.3.GlobalDevelopmentandHumanSecurity:TowardsaPolicyAgenda.ApolicyreviewcommissionedbytheMinistryforForeignAffairs,Sweden,GlobalDevelopmentStudies,EGDISecretariat,MinistryforForeignAffairs,Sweden.

WorldBank(2012)Devolutionwithoutdisruption:pathwaystoasuccessfulnewKenya:Executivesummary.Washington,DC:WorldBank.http://documents.worldbank.org/curated/en/385211468088448074/Executive-summary

18

Annexes

Annex1:TablesandFigures

Table1.Sub-indicatorsanddomainsoftheFragilityExposureIndex(FEI)

Domains IndividualSub-Indicators(Experiences) IndividualSub-Indicators(Perceptions)

HUMAN

SECURITY Presenceofnon-statecriminal

actorsindistrict Previousexperiencesoftheft,

sexualassault,physical

assaultandbriberyhappenedin

thelastyear Preventivemeasuresundertaken

toprotectagainstcrime Easeofservicewithoutbribefrom

formalinstitutions(central

government,countygovernment,

police) Consideringleavingduetosecurity Distance2police

Fearofcrimeathomeandunsafetyin

neighbourhood Satisfactionwithpersonal,neighbourhood,

anddistrictsecurity Importanceofowningpersonalweaponsand

reportingfamilymemberwhocommitted

crime Trustandperceptionofeffectivenessinformal

institutions(centralgovernment,county

government,police,courts) Fearofwar,ethnicconflict,religious/ethnic

conflict,governmentalandlocalauthority

misconduct,andpoliceviolenceandarbitrary

controls,crimeinvillage,banditry,Juvenile

delinquency,andworryaboutlandconflicts

ECONOMIC

INCLUSION Easeofservicewithoutbribefrom

privatesectorandNGOs Satisfactionwithfinancialsituation,food

security,andlivingstandards Fearofcorruptioninvillage,andworryabout

unemployment,loansharking,andfood

insecurityincountry Trustandperceptiononeffectivenessof

privatessectorandNGOs

SOCIAL

COHESION Membershipofwomenandyouth

groupsandpoliticalparties Registrationandparticipationin

previouslocalandcentralelections Stealingifstolenfromandcontact

forsettlingdisputesincaseof

crime EaseofServicewithoutbribefrom

informalvillagegovernments,

courts,religiousinstitutions,and

traditionalinstitutions.

Satisfactionwithleisuretime,socialequalityin

village,communityintegration,andfamilylife. Expectationofregistrationandparticipationin

futurelocalandcentralelections Trustandperceptiononeffectivenessofvillage

govs,courts,religiousinstitutionsand

traditionalinstitutions

19

Figure1.GovernanceGapofFormalInstitutionsinKenya

20

Figure2a.DistributionoftheDomainsofFragilityExposureIndex–Kenya

Figure2b.DistributionoftheFragilityExposureIndex-Kenya

21

Figure3.DifferencesbetweenexperiencesandperceptionsoftheFragilityExposure

IndexanditsDomains-Kenya

22

Table2.Variationsinexposuretofragilityacrossgroupsandlocations

FEI HumanSecurity EconInclusion

SocialCohesion

Gender

Female 52.96(14.03) 0.54(0.17) 0.56(0.15) 0.37(0.13)

Male 52.59(14.24) 0.52(0.17) 0.59(0.15) 0.34(0.13)

p-value 0.721 0.254 0.010** 0.019**

County

Nairobi 55.84(13.32) 0.55(0.16) 0.58(0.14) 0.40(0.14)

Kisii 50.71(14.05) 0.53(0.18) 0.54(0.16) 0.33(0.12)

Kakamega 53.38(14.46) 0.54(0.17) 0.58(0.15) 0.34(0.13)

Nakuru 52.54(14.14) 0.54(0.17) 0.56(0.16) 0.35(0.12)

Kiambu 50.63(14.03) 0.49(0.17) 0.55(0.15) 0.36(0.14)

p-value <0.001*** 0.010** 0.046** <0.001***

AgeGroup

Age<=25 55.44(12.84) 0.55(0.16) 0.53(0.14) 0.44(0.13)

25<Age<=45

53.27(14.03) 0.53(0.17) 0.57(0.15) 0.37(0.14)

45<Age<=65

52.20(14.00) 0.53(0.17) 0.57(0.15) 0.34(0.13)

Age>65 51.47(15.61) 0.51(0.17) 0.57(0.16) 0.34(0.14)

p-value 0.265 0.634 0.322 <0.001***

23

Table2–Continued.Variationsinexposuretofragilityacrossgroupsandlocations

FEI HumanSecurity EconInclusion SocialCohesion

Religion

Protestant 52.59(13.71) 0.52(0.17) 0.57(0.15) 0.36(0.13)

Catholic 55.25(13.76) 0.56(0.16) 0.58(0.14) 0.38(0.13)

Muslim 49.09(13.88) 0.48(0.18) 0.55(0.17) 0.33(0.08)

OtherChristian

51.36(14.95) 0.53(0.16) 0.54(0.15) 0.35(0.14)

Other 55.72(11.34) 0.46(0.13) 0.63(0.08) 0.44(0.14)

p-value 0.033** 0.032** 0.05* 0.044**

MaritalStatus

Single 57.91(13.95) 0.59(0.16) 0.58(0.15) 0.41(0.14)

Married,Poly 53.32(15.60) 0.53(0.18) 0.57(0.15) 0.37(0.14)

Married,Mono

52.03(13.91) 0.53(0.16) 0.56(0.15) 0.35(0.13)

Divorced/Widowed

53.44(13.56) 0.53(0.18) 0.55(0.14) 0.39(0.13)

p-value 0.002*** 0.008*** 0.679 <0.001***

standarddeviationinparentheses.

24

Table3.ComparingmicroandmacrofragilityindicesforKenya

FragileStatesIndex(FSI)* Total

PoliticalandMilitaryIndicators

EconomicIndicators

SocialIndicators

8.2 7.7 8.5 24.4/30

FragilityExposureIndex(FEI)**

HumanSecurityDomain

EconomicInclusionDomain

SocialCohesionDomain

5.3 5.6 3.6 14.5/30

GlobalPeaceIndex*** 14.3/30

*ThemeanoftheindicatorswithineachdomainfromFSIarecalculated.**OurfragilityexposureindexistransformedtomatchthatofFSI,whereeachdomainismultipliedbyten.***TheGPIistransformedtomatchthatofFSI.TherearethreedomainsinGPI:Militarisation,Society&Security,andDomestic&Int.Conflict.Allvaluesfor2016/mostrecentavailable.

25

Annex2:FragilityModulePERSONALSATISFACTION:Pleasechoosethelevel,whichfitstoyourpersonalperception!Onascaleof1=completelydissatisfiedto10=completelysatisfied,howsatisfiedareyouwith…

…yourpersonalsecurity?…thesecurityinyourneighborhood?…thesecuritysituationinyourdistrict?…youcommunityintegration/socialintegration,supportiveinteractionwithneighbours?…yousocialequalityinyourvillage/community…sparetime/leisure…youfamilylife…yourfinancialsituation

Howofteninthelastyearhaveyouoranyoneinyourfamily:

…Feltunsafewalkinginyourneighbourhoodatnight…Fearedcrimeinyourownhome

Intheprevious12months,werethereanyorganizedgroupsposinginsecurityinyourdistrict?

0=No;1=Yes

Onascalefrom1to10,howimportantisittoownaweapontodefendyourselfandyourfamily?

Imaginethataclosefamilymembercommittedaviolentcrime.Onascalefrom1to10,howlikelywouldyoureporthim/hertothepolice?

Doyoubelongtoapoliticalparty? 0=No

1=YesDidyouregistertovoteinthelastnationalelection?Didyouregistertovoteinthelastlocalelections?Doyouthinkyouwillregistertovoteinthenextnationalelection?Doyouthinkyouwillregistertovoteinthenextlocalelection?Pleaseindicatehowmuchyouareafraidofcertainphenomenainyourvillage ... banditry 0=Nofear

5=fearWarReligious/ethnicconflictMisconductofgovern.authoritiesMisconductoflocalauthoritiesPoliceviolenceArbitrarypolicecontrolCrimeCorruption

26

TypeofInstitution HowPowerfulisthe

[institution]inKenyatoday?

Howgooddoes[institution]

deliverservicesinKenyatoday?

Howmuchdoyoutrust

[institution]inKenya?

Howeasycanyouobtain

assistancefrom[institution]

withoutbribeinKenyatoday?

CentralGovernment 1=Extremelyweakto

10=ExtremelyPowerful

1=VeryIneffective

to10=VeryEffective

1=CompletelyUntrustworthy

to10=CompletelyTrustworthy

1=Extremely

difficultto

10=ExtremelyEasy

Local/CountyGovernmentPoliceCourtReligiousInstitutionsTraditionalInstitutionsNGOsPrivateSectorVillagegovernment(informal)