DEPARTMENT OF ECONOMICS

YALE UNIVERSITY P.O. Box 208268

New Haven, CT 06520-8268

http://www.econ.yale.edu/

Economics Department Working Paper No. 100

Economic Growth Center Discussion Paper No. 1010

The Impact of Consulting Services on Small and

Medium Enterprises: Evidence from a Randomized

Trial in Mexico

Miriam Bruhn The World Bank

Dean Karlan

Yale University

Antoinette Schoar MIT Sloan; NBER; Ideas42

February 2012

This paper can be downloaded without charge from the Social Science Research Network Electronic Paper Collection:

http://ssrn.com/abstract=2009582

TheImpactofConsultingServicesonSmallandMediumEnterprises:EvidencefromaRandomizedTrialinMexico1

MiriamBruhn(mbruhn@worldbank.org)

DeanKarlan(dean.karlan@yale.edu)

AntoinetteSchoar(aschoar@mit.edu)

Abstract

Wetestwhethermanagerialhumancapitalhasafirstordereffectontheperformanceandgrowthofsmallenterprisesinemergingmarkets.InarandomizedcontroltrialinPuebla,Mexico,werandomlyassigned150outof432smallandmediumsizeenterprisestoreceivesubsidizedconsultingservices,whiletheremaining267enterprisesservedasacontrolgroupthatdidnotreceiveanysubsidizedtraining.Treatmententerpriseswerematchedwithoneofninelocalconsultingfirmsandmetwiththeirconsultantsonceaweekforfourhoursoveraoneyearperiod.Resultsfromafollowupsurvey,conductedaftertheintervention,showthattheconsultingserviceshadalargeimpactontheperformanceoftheenterprisesinthetreatmentgroup:monthlysaleswentupbyabout80percent;similarly,profitsandproductivityincreasedby120percentcomparedtothecontrolgroup.Wealsoseeasignificantincreaseintheentrepreneurialspiritindexforthetreatmentgroup,asetofquestionsdesignedtoillicittheSMEownersconfidenceintheirabilitytomanagetheirbusinessanddealwithanyfuturedifficulties.However,wedonotfindanysignificantincreaseinthenumberofworkersemployedinthetreatmentgroup.Keywords:Enterprisegrowth;entrepreneurship;managerialcapitalJELCodes:D21,D24,L20,M13,O12

1WearegratefultothemanagementandstaffofIPPCandwouldliketothankthefieldresearchmanagementteamatInnovationsforPovertyAction,includingAlissaFishbane,JavierGutirrez,AshleyPierson,DouglasRandallandAnnaYork,andXimenaCadenaatideas42,forexcellentresearchassistance.FinancialsupportforthisprojectwasprovidedbytheGovernmentoftheStateofPuebla,viatheConsejoparaelDesarrolloIndustrial,ComercialydeServicios;theKnowledgeforChangetrustfundoftheWorldBank;andtheBillandMelindaGatesFoundationviatheFinancialAccessInitiativeforfunding.AllopinionsanderrorsinthispaperarethoseoftheauthorsandnotofanyofthedonorsoroftheWorldBank.

1

1.IntroductionAlargeliteratureindevelopmenteconomicsandentrepreneurshipaimstounderstandtheimpediments

tofirmgrowth,especiallyforsmallandmediumsizeenterprises.Financialconstraintsareoftenputforwardasacentralobstacle to firmgrowth.Theempirical literaturehasdocumented theseconstraintsat themicro level(seeBanerjeeetal2010,deMeletal2008,andKarlanandZinman2011)aswellasatthemacrolevel(seeforexampleKingandLevine1993orRajanandZingales1998).

However,capitalalonecannotgenerategrowth;onemustalsoknowhowtouse it.Bruhn,KarlanandSchoar (2010) discusses atmore length the role of managerial capital as a key component for enterprisedevelopment, distinct from human capital.We argue thatmanagerial capital can directly affect the firm byimproving thestrategicandoperationaldecisions,but italsoaffects theproductivityofother factorssuchasphysicalcapitalandlaborbyhelpingtousethemmoreefficiently.RecentworkbyBertrandandSchoar(2003),Bloom and Van Reenen (2007 and 2010), and Bennedson et al (2007) shows enormous heterogeneity inmanagement practices and CEO styles across firms. But a central question remains: is this observedheterogeneityareflectionofanoptimalmatchbetweentheunderlyingfundamentalsofdifferentfirmsandtheadequatelevelofmanagementgiventhefirmsstateofdevelopment?Oraredifferencesinmanagementstyleandlackofmanagerialcapitalafirstorderimpedimenttofirmgrowthandprofitability?Managersindevelopingcountriesmightbeconstrained in theacquisitionof theseskills, ifsuchskills requireeither formal trainingorexperience inotherwellrunenterprises,orboth (see forexampleGompers,Lerner,andScharfstein2005orCaselliandGennaioli2005).

We test if lackofmanagerialcapitalhasa firstordereffecton theperformanceandgrowthof smallenterprises inemergingmarkets.We focuson smallbusinesses since in this setting theowner/manager caneasilybedeterminedasthemaindecisionmaker,andinadditiontheseenterprisesareoftenseenashavingthemost potential for scale up if bottlenecks to their growth can be removed. In order to test this ideamoresystematically we set up a randomized control trial in Puebla, Mexico, where 432 small and medium size

2

enterprisesappliedtoreceivesubsidizedconsultingservices,and150outofthe432wererandomlychosentoreceive the treatment. The remaining 267 enterprises served as a control group that did not receive anysubsidized consulting services. Treatment enterpriseswerematchedwith one of nine local consulting firmsbasedonthespecializedservicestheyneeded.Onaverage,enterprisesmetwiththeirconsultantsonceaweekforfourhoursoveraoneyearperiod.Theenterpriseownerandconsultingfirmdecidedjointlyonthefocusandscopeoftheconsultingservices.

Thisinterventionisajointtestoftwocloselyrelateddimensions:Ontheonehandweaimtoestablishifmanagerialcapital isa limiting factor in thegrowthofenterprises.But the testat thesame timedependsonwhether this knowledge can be conveyed via a consulting intervention in the first place. We cannot testseparatelytheabovetwoquestions.Forexample, itcouldbethatmanagerialcapital is indeedahindrancetogrowth, but it might not be possible to transfer this knowledge by simply providing consulting services.Therefore,failuretofindaresultherewouldnotprovethatmanagerialcapitaldoesnotmatter,sincefailuretofindaresultmaysimplymean thatthisprogramwasnoteffective in thetransmissionofmanagerialskills (orthat managerial skills are innate skills and simply not teachable). Our results can only be interpreted as areducedformoftheeffectsofmanagerialcapitalasprovidedthroughaconsultingintervention.

Wefindthatourconsultinginterventionhadalargeimpactontheperformanceoftheenterprisesinthetreatmentgroup:monthlysaleswentupbyabout80percent;similarly,profitsandproductivity increasedby120percentandonefifthofastandarddeviation,respectively,comparedtothecontrolgroup.However,wedonotfindanysignificantincreaseinthenumberofworkersemployedintreatmententerprises.Theresultsoftheinterventionarequitelargebutwebelievethattheyarereasonablegiventhecontextoftheintervention:Theenterprisesinoursamplewerestartedbypeoplewhoarenotprofessionalmanagersandmanyofthemhadnotreceivedanyformalmanagementtrainingatallpriortoourintervention.

Whenlookingattheprocessbywhichthesechangesarebroughtaboutwefindthatenterprisesinthetreatmentgroupshowasignificantincreaseintheirlikelihoodtoengageinmarketingeffortsandaremorelikely

3

tokeepformalaccountsabouttheirfirms.Butwealsotestforseveralotherchangesinbusinessprocessesanddo not find any consistent pattern.We do however find a pronounced impact on an index ofwhatwe callentrepreneurialspirit.This index isacombinedmeasureofanswers toasetofquestionson theenterpriseownersbeliefsabouttheirabilitytocontrolthesuccessoftheirbusiness(orwhethertheyaremerelysubjecttoexternalforcesoutsideoftheircontrol)andontheownersdriveforsuccess2.Theincreaseinthisindexmightreflect the fact thatenterpriseowners setnewgoalsaspartof theprogramand that consultantshelped toprovide motivation and strategy for how to achieve these goals. In addition, enterprise owners increasedconfidenceintheirabilitytocontrolthesuccessoftheirbusinesscouldbedrivenbyhavingbettercommandofmanagement tools such asmarketing and bookkeeping. Additional support for this interpretationmight bederivedfromthefactthatbusinessesinthetreatmentgroupreportnosignificantdropinsalesbutproactivelyadjusttothe2008economiccrisisbycuttingcosts.

Researchandpracticehaverecentlyseenaflurryofprogramsfocusedondevelopingmanagerialcapitalformicroenterprises(i.e.enterprisestypicallywithzeroemployees,orunderfiveatthemost).Theinterventionsvarywidely in the scopeof themanagement skills thatare transmittedand the typeofenterprises thataretargeted. The training is typically provided as inclass training, often linkedwith amicrocredit program. Forexample,KarlanandValdivia (2010)andDrexler,FischerandSchoar (2010)evaluate inclassprograms.Thesepapersshowthattraditionalmicroenterprisetrainingseemstoaffectthecommandofaccountingpracticesformicroenterprises,buthaslimitedtonoeffectsonactualfirmoutcomesandperformance.Drexler,FischerandSchoar finds that a rule of thumb based training program that focuses strictly on separating business fromhouseholdmoneyappearstohaveamoresignificantimpactonfirmsalesandsavings.Morerecently,BruhnandZia (2011) and Gine and Mansuri (2011) also find that inclass training for microentrepreneurs leads toimprovementsinbusinesspracticesbuthasonlylimitedeffectsonbusinessperformanceandsales.

2SeeAppendix1forthelistofquestionsusedtoconstructtheentrepreneurialspiritindex.

4

Bloom et al (2010) ismore closely rated to our study in that they evaluate the impact of intensiveconsultingservicesfromaninternationalmanagementconsultingfirmonthebusinesspracticesoflargeIndiantextile firms. The average firm in their sample has about 270 employees, whereas the average number ofemployees inour study is14.Bloomet al find thateven these larger firmswereunawareofmanymodernmanagementpractices,andtreatedplants improvedtheirmanagementpracticesduringthe intervention.TheapproachesofBloometalandthisstudyarecomplementary innature:Bloometal focusesonasmallsetoflarge firms inone industry, textilemanufacturing,witha tightlydefined interventionbyamajor internationalconsultingfirm.Thecurrentstudy includesa largersetoffirmsand industries(closeto400firmscomparedto20experimentalplantsinBloometalandemploysaheterogeneoussetoflocalconsultingfirms.

Theremainderofthispaperisstructuredasfollows:InSection2,wedescribethesubsidizedconsultingprogram.Section3discussestheexperimentalsetup,datacollection,andcharacteristicsofoursample.Section4givestheresults,examiningbothbusinessoutcomesandbusinessprocessvariables.Section5askswhymoreenterprisesdonotuseconsultingservices,i.e.,giventheseresults,whatarethepossiblemarketfailuresintheconsultingservicesindustry?Section6concludes.

2.ConsultingProgram

We conducted a randomized control trial in collaborationwith the Puebla Institute for CompetitiveProductivity(knownasIPPC,afteritsSpanishacronym),atraininginstitutesetupbytheMinistryofLaboroftheMexican State of Puebla. IPPC implemented a business development program to provide participatingenterpriseswith subsidizedconsulting services fromoneofanumberof localconsulting firms.Theprogram,which started in March 2008 and ended in February 2009, aimed to include 100 micro, 40 small, and 10mediumsizedenterprises3andactuallyincluded108microenterprises,34smallenterprisesand8mediumsize

3 As defined by the Mexican Ministry of the Economy, micro enterprises have up to 10 fulltime employees. Smallenterpriseshavebetween11and50fulltimeemployeesinthemanufacturingandservicessectorsandbetween11and30

5

enterprises.Theprimarygoalwastohelpenterprisesreachthenextsizecategorybytheendoftheprogramandthuscontributetojobcreationandeconomicgrowthoftheregion.

Consultantswereaskedto(1)diagnosetheproblemsthatpreventedtheenterprisesfromgrowing,(2)suggestsolutionsthatwouldhelptosolvetheproblemsand(3)assistenterprisesinimplementingthesolutions.The consultants dedicated four hours perweek to each enterprise on average. The programwas originallyintendedto lasttwoyearsbutendedprematurelyafteroneyearduetogovernmentfunding issues.Thustheimplementationphasewasshortened.

Theconsulting serviceswerehighly subsidizedby theStateofPuebla.Microenterprisespaidonly10percent of the market cost of the consulting services, small enterprises 20 percent, and medium sizedenterprises about 30 percent. The unsubsidized cost of the consulting services varied by firm size butwasequivalenttoaboutUS$574perhouronaverage,amountingtoUS$11,856perfirmforoneyear(4hoursfor52weeks).

Consultingfirmswereselectedthroughacompetitivebiddingprocess.Inresponsetoacallforproposalsputoutby IPPC,eleven consulting firms submittedproposals toparticipate in theprogram.Two firmswereeliminatedbasedon inadequate references from former clients.Themajorityof theparticipating firmswereprivatelocalconsultingfirmsthatusuallyworkwithmicro,small,andmediumsizeenterprises.

Atthebeginningoftheprogram,principaldecisionmakers,aswellasmostemployees,fromallprogramenterprisescompletedacomputerizedtestthatdeterminedtheirindividualstrengthsandtalents.ThistestwasbasedonGallupsStrengthFindermethodandIPPCwaslicensedtoconductthistestinPuebla.IPPCencouragedenterprisestousetheresultsofthistesttohelpassignemployeestoresponsibilitiesbasedontheirstrengthsasidentifiedbytheStrengthFindermethod.Theconsultantsweretrainedinhelpingtheenterprisesinterpretandapplytheresultstotheir labordecisions.Forexample,onetalentwascommunicationwhereasanotherwas

fulltimeemployees inthecommercesector.Mediumsizeenterpriseshaveup to100 fulltimeemployees in theserviceandcommercesectorsandupto250fulltimeemployeesinthemanufacturingsector.4700MexicanPesos(MXP)

6

operations. Employeeswith the communication talentwere particularly suited to interactingwith clients,whileemployeeswiththeoperationstalentwoulddowellatrecordkeepingandaccounting.

Apart from the employee talent diagnostic, the content of the consulting varied across enterprisesdependingontheirneeds.Inordertogainanunderstandingoftheissuesthatenterprisesworkedonwiththeirmentors,weconducted indepthcasestudiesofeighttreatmententerprises.Table1 liststheareasthattheseeightenterprises coveredwith their consultants,alongwith thenumberofenterprises thatworkedoneachtopic.Almostallenterprisesstartedbyestablishingmissionandvisionstatementswiththeirconsultants,settingspecificgoalsforwhattheywantedtoachieveinthefutureandthroughouttheprogram.Mostenterprisesalsoworked on improving accounting and record keeping (through training and/or use of new software), clearlyassigning staff responsibilities, and sales strategy and advertising. Apart from these common topics, theremaining topics coveredarediverse, includingoptimizing thenumberand locationofpointsof sale,qualitycontrol,accesstocreditoralternativefinancingsolutions,pricingstrategy,teamworkand leadershiptraining.Thisreflectsthefactthattheconsultantstailoredtheiradvicetoeachenterprisesindividualchallenges,leadingthemtoworkondifferentareaswitheachenterprise.

7

3.ExperimentalSetupandDataFortheimplementationoftheprogram,IPPCadvertisedtheprogramthroughouttheStateofPueblavia

variousmediaoutletsinordertoattractaninitialsampleofinterestedmicro,small,andmediumenterprises.Inresponsetotheadvertising,432enterprisesexpressedinterestintheprogramandsignedaletterofinterest.Abaseline survey of these interested enterprises was conducted between October and December 2007. Thissurveycollectedinformationonenterprisecharacteristicsandperformance,aswellasonbusinesspracticesandcharacteristicsoftheenterprisesprincipaldecisionmaker(typicallytheownerormanager).

Using data from the baseline survey, 150 enterprises were randomly selected to participate in theprogram5.Therandomizationwasstratifiedbysector(manufacturing,services,andcommerce)andenterprisesize (micro, small,andmediumsized)6,conducted throughaStataprogram thatwas runon thepremisesofIPPC in the presence of government officials and a public notary,who certified that the assignment to thetreatmentgroupwasrandom,i.e.,notrerundependingonanyparticularassignments.

Outof the150enterprises in the treatmentgroup,80chose to takeup theconsulting services7.Theremainingtreatmentgroupenterprisesdeclinedtoparticipateintheprogramalthoughtheyhadinitiallysigneda letter of interest saying that theywould participate if offered a spot.Most enterprises that chose not toparticipatesaidtheirfinancialsituationhadchangedsincetheysignedtheletterofinterestandtheynolongerhadsufficientfundstopaythefee(albeitsubsidized)fortheconsultingservices.IPPCpairedthe80treatment

5Weoriginallyhad434observationsintherandomizationandassigned150ofthemtotreatment,butwelaterdiscoveredthattwofirmshadexpressedinterestintheprogramtwiceunderseparatenames.Forthisreason,wehadtodroptwoobservations,givingus432uniquefirms.Inoneofthecases,bothseparatenameswereinthecontrolgroup,andwedroppedoneofthese.Intheothercase,onenamewasassignedtothetreatmentgroupandtheothertothecontrolgroup.Here,wehadtokeepthefirminthetreatmentgroupsincetheyhadalreadybeennotifiedthattheyhadbeenrandomlyselectedtoparticipateintheprogram.6Withinstrata,wererandomizeduntilthemaximumtstatisticonthedifferences inaveragesacrossthetreatmentandcontrolgroupsof the followingvariableswasabove1.25and theaverage tstatisticwasabove0.35:WithinPueblaCitydummy,businessage,totalassetvalue,profitmargin,measuredriskaversion,entrepreneurialspiritindex,currentlyhasaloanfromafinancialinstitutiondummy,principaldecisionmakershoursworked,principaldecisionmakersage,principaldecision makers gender, principal decision makers years of schooling, principal decision maker is of indigenousbackgrounddummy,aswellastwodummiesindicatingwhetherthefirmhasparticipatedinotherIPPCprograms.7Due toanadministrativeerror, therewasalsoone controlgroup firm thatwas invited toparticipate,anddid, in theprogram.Foranalysispurposes,weadheretotherandomassignmentandthisenterpriseisincludedinthecontrolgroup.

8

group enterprises that tookup the program with consulting firms according to the consultants sector andenterprisesizeexpertise,aswellasgeographicrestrictions.

Table 2 provides summary statistics of baseline characteristics for enterprises and their principaldecisionmakersinthetreatmentandcontrolgroups.About30percentofenterprisesineachgroupoperatedinthemanufacturingsector,25percentinthecommercesector,and45percentintheservicessector.Onaverage,theenterprises in the studyhadabout14 fulltimepaidemployeesandwere slightlyover10yearsold.Theenterprisesprincipaldecisionmakerswereonaverage43yearsold,72percentofthemweremen,andtheyhadcompleted16yearsofschooling.

PanelCofTable2displaysourmainmeasuresofbusinessperformance,startingwithlastmonthssales.Thisvariablevarieswidely inoursample.Atbaseline,average lastmonthssales in the treatmentgroupwasUS$76,343withastandarddeviationof283,025,andUS$50,844inthecontrolgroup,withastandarddeviationof119,987.Toreducethenoise inthisvariable,wedropthetoponepercentofoutliers,aftercontrolling forenterprisesizegroup(micro,small,andmediumsize,asdefinedabove).Thisreducesthestandarddeviationbya factorofabout three in the treatmentgroupandby1.3 in thecontrolgroup.Theaveragesof the trimmedvariablesaremore similaracross the treatmentand controlgroups (US$40,479andUS$46,113, respectively)thanfortheuntrimmedvariable.

Weusetwodifferentmeasuresofprofitsinthispaper.Thefirstoneiscalculatedaslastmonthssalesminuscosts.Thesecondone iscalculatedbasedon lastmonthssalesand reportedprofitmargin,where theprofitmarginwasobtainedbyaskingForevery100pesosofsales/revenue,howmanypesosofcostsdoesthebusinesshave?8DeMel,McKenzie, andWoodruff (2009)use a similarquestionwhenmeasuringprofits toadjustfordifferencesinthetimingofinputsandsalesofoutputs9.Oneimportantdifferencebetweenourtwomeasuresofprofitsisthatprofitscalculatedfromprofitmarginandsalesdonothavenegativevaluessincethe8Theformulausedtocalculatetheseprofitsis:Profits=Sales*(1Pesosofcostsper100pesosinsales/100).9 De Mel, McKenzie and Woodruff (2009) also suggest asking business owners directly what their profits are as analternativetocalculatingprofitsbasedonunderlyingvariables.Wetriedthisapproachbuthadaveryhighnonresponseratetothisquestion.

9

respondentsdidnotreportcostshigherthan100pesosintheprofitmarginquestion.Profitscalculatedassalesminuscosts,ontheotherhand,dohavenegativevalues.Thuswealsoreporttreatmenteffectsforthelogofthesecondmethod,butnotthefirst.

We calculate two separate measures of enterprise productivity. The first is the residual from aregressionoflogsalesonlogemployeesandlogbusinessassets.Thesecondisreturnonassets(ROA),definedasprofits(calculatedassalesminuscosts)dividedbybusinessassets.

Similarly to sales, thevariancesofprofits,assets,productivityandROAare large.For this reason,weincludetheaveragesoftheonepercenttrimmedvariablesinTable2,droppingthetopandbottomonepercentofoutliersforprofitscalculatedassalesminuscosts,productivityandROA.Forassetsandforprofitscalculatedbasedonprofitmarginandsales,weonlydropthetoponepercentofoutlierssincethesevariablesareboundedbelowbyzero.Overall,theaveragesoftheonepercenttrimmedvariablesareverysimilaracrossthetreatmentandcontrolgroups.

In Table 3, we examine whether the baseline characteristics summarized above predict whichenterprises in the treatmentgroupdecided toparticipate in the consultingprogram.Column1 includesonlyenterpriseandprincipaldecisionmakerbackgroundcharacteristics.InColumn2,weaddbusinessperformancevariablestotheestimation.Theresultsrevealthreestrongpredictorsoftakeup.First,enterpriseswithamaleprincipal decision maker were about 20 percentage points more likely to participate in the program thanenterpriseswithafemaleprincipaldecisionmaker.Second,enterpriseswithhighersales,aswellasenterpriseswithhigherproductivity,havehigher takeup rates.Wealso findweakevidence thatolderprincipaldecisionmakerswerelesslikelytotakeuptheprogram.

Weconducteda followupsurveybetweenMarchandMay2009 (i.e.,one to threemonthsafter theinterventionended,whichis1215monthsaftertheinterventionbegan),reinterviewing378enterprisesor88percentof the432enterprises interviewedatbaseline, tomeasure the impactof the consulting servicesonbusiness outcomes. Out of the 54 enterprises that could not be reinterviewed, eleven enterprises were

10

confirmed closed10,31declined toparticipate in the interview11and sevenenterprises couldnotbe trackeddowndespiterepeatedcontactattempts.Theremainingfiveenterpriseshadmergedwithanotherenterpriseoneofthemwithanenterpriseoutsideoursampleandtwowithtwootherenterprisesinthesample.Forthesefiveenterprises,wewerenotabletoobtainseparatedatafortheunitcorrespondingtotheoriginalenterprise,and thus they arenot included in the analysis.Weprovide an analysisof attrition rates and correlateswithbaseline information in theappendix.Weshow that therearenodifferentialattrition ratesacross treatmentandcontrolgroups;neitherdoweseecompositionalshifts.

4.ResultsandDiscussion OurmainspecificationusesOLSregressionsforthevariousoutcomemeasuresonanindicatorvariableforwhethertheenterprisewasassignedtothetreatmentratherthanthecontrolgroup.Thecoefficientsonthisindicator variable are displayed in Tables 4 through 6 and represent the intentiontotreat effect of theconsulting program. In all regressions, we control for strata dummies and rerandomization variables12, assuggestedinBruhnandMcKenzie(2009).Wealsoincludethebaselineoutcomevariableasanadditionalcontrolvariableandshowtheresultswithandwithoutthiscontrol.Forobservationswherethebaselinevalueoftheoutcome ismissing,wereplacethisvaluewithzeroand includeadummyvariable indicatingthatthevalue ismissing,inordertokeeptheobservationinthesample.

4.1BusinessPerformance

10Weverifiedwiththeformerprincipaldecisionmakerand/orneighborsthattheseenterpriseshadindeedclosed.Intheempiricalanalysis,wereplacedemployees,sales,costs,andassetsfortheseenterpriseswithzero.Thepercentageofclosedenterpriseswaslowerinthetreatmentgroup(1.4percent)thaninthecontrolgroup(3.3percent).However,thedifferenceisnotstatisticallysignificant.11Thepercentageofenterprisesthatrefusedtheinterviewwasslightlyhigherinthecontrolgroup(8.7percent)thaninthetreatmentgroup(5.6percent),butthedifferencebetweenthesetwonumbersisnotstatisticallysignificant.12Duetobaselinedataentrytyposthatwerediscoveredandcorrectedaftertherandomizationtookplace,afewvaluesofthevariablesincludedintherandomizationproceduredonotcorrespondtothetruebaselinevalues.Thestratadummiesandrerandomizationcontrolsincludedintheregressions,containthevaluesoriginallyusedintherandomizationprocedure.Allotherbaselinedatausedinthesummarystatisticsandregressions,,containsthetruebaselinevalues.

11

TheresultsinColumns1and2ofTable4suggestthattheconsultingprogramincreasedmonthlyprofits,calculatedbasedonsalesandprofitmargin,byabout120percent.Profits,calculatedassalesminuscosts,arealsohigher in the treatmentgroup than in thecontrolgroupbyaboutUS$5,400.Thiseffect is,however,notstatisticallysignificant,possiblydue to thenoise in thismeasure.Unlikeprofitscalculatedbasedonsalesandprofitmargin,wecannotusethe logofsalesminuscoststoreducethe influenceofoutlierssincethisvariablehasnegativevalues. Inorder tocheckwhether resultsare robust tooutliers,weestimate the results foronepercent trimmedvariables (Columns3and4)and for twopercent trimmedvariables (Columns5and6).Theeffectoftheconsultingprogramonprofitscalculatedassalesminuscostsispositiveinallthreesamplesbutisonly statistically significant in theonepercent trimmed sample,and themagnitudeof theeffectvaries fromUS$4,300toUS$7,750.Theeffectonlogprofitscalculatedbasedonsalesandprofitmargin,ontheotherhand,isrobustinsizeandisstatisticallysignificantatthefivepercentlevelinallthreesamples. Table1alsoshowsapositive impactoftheconsultingonsales,ofabout80percent,but theeffect isonlymarginally statistically significant in the full sample,at the12.1percent levelwithoutcontrolling for thebaselinesales,andatthe10.3percentlevel,aftercontrollingforbaselinesales. TheresultsinTable1furthersuggestthattheconsultingimprovedenterpriseproductivityasmeasuredbytheresidualfromaproductivityregressionandalsoasmeasuredbyROA,byaboutone fifthofastandarddeviation. Incontrast,Table1doesnotshowanystatisticallysignificanteffectonemployment.Thiscouldbeduetothefactthattheconsultingendedafteroneyearinsteadoftwoasinitiallyplanned.Aoneyearfollowupperiodmayalsobe tooshort toexpectaneffectonslowmovingvariables,suchasemployment;havingsaidthat, itmakes it clear that the consultingwasnotdirectly attending to labormarket failures, i.e., inhelpingenterpriseslearnhowtohireandtrainworkers. Overall,theresultssuggestthattheconsultingdidimprovebusinessoutcomesfortheenterprisesinthetreatment group. However, the effects on sales and productivity are only statistically significant at the 10

12

percentlevel,likelybecausethedataarequitenoisyandthesamplesizeisrelativelysmall.Manyenterprisesintreatmentandcontrolgroupdonotreportsales,costs,andassets,furtherreducingthesamplesize. Ourestimatedeffectsof the consultingon salesandprofitsof80and120percent, respectively,arelarge compared to the estimated impacts of improved access to capital and of business training for smallbusinessesfound inthe literature.DeMel,McKenzie,andWoodruff(2008)estimateareturntocapitaloffivepercentforSriLankanmicroenterprises,McKenzieandWoodruff(2008)find2033percentreturntocapitalinMexico,andUdryandAnagol(2006)find60percentforGhana.BanerjeeandDuflo(2004)estimateanelasticityofsaleswithrespecttobankcreditof0.75forlargeIndianenterprises.BruhnandZia(2011)detectanincreaseinprofitsof53percent forentrepreneurswithexantehigh levelsof financial literacywhoparticipated inaclassroombased business training program. One concern could be that there might be potential demandeffects inthesurveyresponse iftreatmententerprisesoverstatetheiroutcomesto justifyreceivingsubsidizedconsultingservices.Thiscouldpossibly inflatetheeffectsontheoutcomesforthetreatmentgroup.Whilewetriedtoprovidepositive incentivestoanswerthesurveysevenforthefirms inthecontrolgroupbypromisingthempreferentialaccesstofutureconsultingprograms,therestillmighthavebeenadifferentialbiasbetweentreatmentandcontrolgroups.4.2ProcessVariables

Inorderto investigatethechannelsthatdrivetheobservedeffectsonsales,productivity,andprofits,we now study how the consulting program changed processes within the enterprise. We measure theseprocesses as follows. First, the surveys asked enterprise owners whether or not they implemented certainchangesduringthepastyear,suchasdevelopingnewproducts,attractingnewinvestors,andlaunchinganewmarketingcampaign.Notethat iftreatmententerprisesbelievedtheyshouldpleasetheprogrambyreportingprocesschangesthatdidnotactuallyoccur,theseestimateswillbeupwardlybiased.

Second,weconstructedanentrepreneurialspiritindex,developedincollaborationwithIPPC.Thisindexisbasedon theanswers to theeightquestions listed inAppendix1,which intend tocaptureentrepreneurial

13

attitudesoftheprincipaldecisionmaker,andisgeneratedusingPrincipalComponentsAnalysis(PCA).Third,weuse aPCAhuman resourcesmanagement indexbasedon the sixquestions listed inAppendix1. Fourth,weaskedenterpriseownershow theykeep theiraccountsandclassifyaccountingpracticesas formal if theyuseeitheranaccountantoracomputerizedsystemasopposedtokeepinghandwrittenornonotesatall.

Table5displaysthetreatmenteffectsonbusinessprocessvariables.Weonlyfindstatisticallysignificantimprovements in three processes: made a new marketing effort (13 percentage points increase),entrepreneurialspirit index (17percentofastandarddeviation increase),and thepercentofenterprises thatkeepformalaccounts(7percentagepointsincrease).OtherprocessesexaminedinTable5,suchasregisteringapatent,developingnewproductsorattractingnew investors,donotappeartobechangedsignificantly.Thesecouldbemoredifficulttodetectsincetheyaremoreheterogeneousacrossenterprises,orrequirealongertimeto change than is observable in the treatment period. In addition, since the content of the consultingwastailored to each firms needs it is perhaps not surprising thatwe do not see improvements in some of theprocessesinTable5,onaverage.Butitispuzzlingthattreatmententerprisesdonotreportacquiringnewclientsat a higher rate than enterprises in the control group. Either the estimated increase in sales and profits isentirelydrivenbysalestoexistingclients,orthedataonnewclientsaremisreportedortoonoisytodetectaneffect.

Thefindingthattheprogramincreasedmarketingeffortsandtheuseofformalaccountingpracticesisconsistentwiththecasestudyevidencementioned inSection2,whichsuggeststhatmanyenterprisesworkedonaccountingand record keeping,aswellas sales strategyandadverting,with theirmentors. Similarly, theincreaseintheentrepreneurialspiritindexmightreflectthefactthatenterpriseownerssetnewgoalsaspartoftheprogramandthatconsultantshelpedtoprovidemotivationandstrategyforhowtoachievethesegoals.Apotentiallimitationtointerpretingtheentrepreneurialspiritindexasaprocessvariableisthatitmightimproveasa resultofbetterenterpriseperformance insteadof theotherwayaround.Twoof thequestionsused toconstructthe indexareparticularlysubjecttothiscriticism(Questionsdande inAppendix1).Asarobustness

14

check,weconstructthe indexwithoutthesetwoquestions.Asshown inTable5,theresultsforthismodifiedentrepreneurialspiritindexarebasicallyunchanged.4.3ResponsetoEconomicShocks

Theprogramcouldhavealsoimprovedenterpriseperformancebyhelpingenterprisescopebetterwiththe2008economiccrisis.Inthefollowupsurvey,about89percentofenterprisesbothinthetreatmentandcontrolgroupreportedthattheyhadbeenaffectedbythecrisis.Weaskedtheseenterpriseswhatchangestheymadeinresponsetothecrisis.Table6displaystheanswerstothesequestionsandexamineswhethertheresponsesdifferedacross the treatmentandcontrolgroups.The results show that treatmententerprisesareeightpercentagepoints(standarderroroffourpercentagepoints)lesslikelythancontrolenterprisestoreportthattheyhadtocutproductioninresponsetothecrisis.Theabilitytoweathershocksmoreeffectivelycouldbearesultofbeingabletomoreproactivelyengageinmarketingactivitiesandbettercontrolfinances,asshownintheprevious section. Enterprises that are lesswell trained in these skillsmight experience economic shocksmorepassivelyanddonothavetoolstocounteractashortfallindemand.

Otherchangesinresponsetothecrisiswerenotstatisticallysignificantacrossthetreatmentandcontrolgroups, but one of magnitude (but not statistical significance) to note is a positive impact on seekinggovernmentassistance(a5.6percentagepointsincrease,standarderrorof4.4percentagepoints,relativetoanaverageof12.8percentinthecontrolgroup).Forenterprisesthatreportedseekinggovernmentassistance,weasked which program or agency they contacted and most answers indicate state or federal programs thatprovidefundingorsubsidiestomicro,small,andmediumsizedenterprises.

5.CostEffectiveness:WhyDontMoreEnterprisesUseConsultingServices?

GiventhelargeincreasesinsalesandprofitsobservedinSection4.2,weaskwhymorefirmsdonotuseconsultingservices.Inparticular,acosteffectivenesscalculationsuggeststhatthereturnstohiringaconsultant

15

are well worth the cost. The average increase in monthly profits lies between US$7,600, and US$11,000(dependingon themeasureofprofits), compared toanannual costof the consulting servicesofUS$11,856.Sincetheprogramwashighlysubsidized,participatingenterprisesonlyhadtopaybetween10and30percentofthiscost(dependingonfirmsize).Yetamongtheenterprisesinthetreatmentgroup,only53percentchosetoparticipateinthesubsidizedconsultingprogramonceofferedaspot.

Several issuesmayhinderthemarket inconsultingservices.First,theremaybenofailureatall:thosewhooptinmaybetheoneswhocanbenefit,andthosewhodonotoptinwouldnotbenefit.Naturallywedonotobservewhattheimpactwouldhavebeenonthosewhodonotoptin,butgiventheexcessivegapbetweentheaverageincreaseinprofitsandthecostoftheservice,thereseemstoremainafailureforthosewhodidoptin,inthattheyhadnottakenuptheservicesbefore,evenattheunsubsidizedrate.Itisimportanttoemphasizethatallenterprises inourstudyhad initiallyexpressed interest inthesubsidizedconsultingprogram,andthattheir views are thus not representative of enterprises that do not have a preexisting interest in consultingservices.Itcouldbethatfirmsexpressedaninterest,learnedmoreabouttheservice,andthendecidedthatthiswas unlikely to yield profitable results for them, and thus failure to takeup remains a rational and correctdecision.

Second, theremaybeacreditmarket failure. In fact,mostof theenterprises in the treatmentgroupthatdeclinedparticipation intheprogramonceofferedaspotgave liquidityconstraintsasthereason.This isconsistentwiththefindinginSection3thattreatmentgroupenterpriseswithhighersalesandhigherROAweremorelikelytotakeuptheprogram.However,itstillbegsthequestion:whydowenotobserveconsultingfirmsacceptingdelayedpayment,orworkingwith financialservices firms toprovidecredit tocover theirservices?Eitherway,itsuggestsacreditmarketfailureisthesourceoftheproblemforsomeenterprises.

Third,entrepreneursmayberiskorambiguityaversewithrespecttothepotentialreturnsfromhiringaconsultant.Thiscouldbeperpetuatedby lackof information inthemarketonthereturnstoconsultingadvice(andwhichconsultingfirmshavedifficultycrediblysignaling).

16

Toexamine this issue, in the followupsurveywe includedsomequalitativequestions for thecontrolgrouponwhethertheywereusinganyconsultingormentoringservices,andifnotwhynot.About21percentofcontrol group enterprises said that they were indeed using some services, and provided the name of theconsulting firm they were using. Examining these names reveals that only about half of these firms offermanagement consulting services similar to the consulting firms that worked with the treatment groupenterprises.Theotherfirmsmentionedbythecontrolgroupprovidespecializedservices,suchasaccountingortechnical assistance. Overall, the incidence of using management consulting services in the control groupappearstobearound10percent.Table7 liststheselfreportedreasonswhycontrolgroupenterprisesdonotuse consulting services. By far the most frequently mentioned reason is lack of funds (46.3 percent ofenterprises mention this reason), followed by uncertainty about the benefits of consulting services (22.2percent)andsimplynothavingconsideredhiringaconsultant(18.5percent).

Ourfindings indicatethatmanagementconsultingservicescanhavehighreturnsformicro,small,andmedium enterprises, andwe consider funding constraints to be the most likely explanation for the lack ofmarkettransactionsinconsultingservices.6.Conclusion

Ourresultssuggestthat lackofmanagerialskillsconstitutesasignificantconstrainttofirmgrowthandtheabilityofmicro,small,andmediumenterprisestowithstandeconomicshocks.Theeffectsofthestudyarelarge.Onaveragewefindanincreaseinsalesandprofitsof80and120percent,respectively,forthetreatmentgroup compared to the control group. However, we believe that the magnitude of the impact is notunreasonablegiventhatmanyenterprisesinthesamplehadnotreceivedanyformalmanagementtrainingpriorto our intervention. The sales and productivity improvements seem to be brought about primarily byimprovements inmarketingand financialcontrols.Consultantsalsoappear tohavehelpedenterprises to setcleargoalsanddefineastrategyforhowtoachievethesegoals.

17

In contrast, we do not see any significant impact on employment generation or the number ofemployees.Onecanonlyspeculatewhetherthescopeoftheinterventionwasnotlongorsignificantenoughtoaffect employment,orwhether thedecision tohire additionalworkerswouldhave tobeprecededby evenlarger ormore sustained increases in output. Alternatively, some recent studies suggest that there is largeheterogeneity in thewillingnessof smallbusinesses toexpandwhichmaybedue tovariation in theownersobjectivefunction(seeforexampleHurstandPugsley2011).

Overall our results suggest that managerial inputs have a large and important impact on firmperformance.However,thereisstillmuchtolearnaboutthewaythisinformationaffectsfirmperformanceasawholeandmorespecificallyhowitinteractswiththemarginalproductivityofinputssuchaslaborandcapital.Inaddition,whiletheremaybealotofheterogeneityineffectsoursampleisnotlargeenoughtoallowustolookatallthefirm level interactionsthatmightbeof interest,suchascompetitivenatureofthe industry,ageandgenderoftheowner,ownersambitionlevel,risktakingability,orgeneralskilllevels.Webelievethisisacriticalareaforfurtherresearch.

18

ReferencesBanerjee,AbhijitandEstherDuflo,2004,DoFirmsWanttoBorrowMore?TestingCreditConstraintsUsingaDirected Lending Program, M.I.T. working paper, Accessed February 22, 2012, http://econwww.mit.edu/files/791Banerjee,Abhijit,EstherDuflo,RachelGlennersterandCynthiaKinnan.2010., Themiracleofmicrofinance?Evidence from a randomized evaluation, 2010working paper,Accessed September 10th, 2011,http://econwww.mit.edu/files/5993Bennedsen,Morten,KasperMeisnerNielsen,FranciscoPerezGonzalez,andDanielWolfenzon.2007.InsidetheFamilyFirm:TheRoleofFamilies inSuccessionDecisionsandPerformance.Quarterly JournalofEconomics,122(2):647691.Bertrand, Marianne, and Antoinette Schoar. 2003. Managing with Style: The Effect of Managers on FirmPolicies.QuarterlyJournalofEconomics,118(4):11691208.Bloom,Nicholas,BennEifert,AprajitMahajan,DavidMcKenzie,and JohnRoberts.2011. Doesmanagementmatter?EvidencefromIndia.WorldBankPolicyResearchWorkingPaperNo.5573.Bloom,Nicholas,and JohnVanReenen.2007.MeasuringandExplainingManagementPracticesacrossFirmsandCountries,QuarterlyJournalofEconomics,122(4):13411408.Bloom, Nicholas, and John Van Reenen. 2010. Why do management practices differ across firms andcountries?JournalofEconomicPerspectives,24(1):203224.Bruhn,MiriamandDavidMcKenzie.2009. InPursuitofBalance:Randomization inPractice inDevelopmentFieldExperiments.AmericanEconomicJournal:AppliedEconomics,1(4):200232.Bruhn,Miriam,DeanKarlanandAntoinetteSchoar.2010.WhatCapitalisMissinginDevelopingCountries.AmericanEconomicReviewPapers&Proceedings100(2):629633.Bruhn,MiriamandBilalZia.2011.StimulatingManagerialCapitalinEmergingMarkets:TheImpactofBusinessandFinancialLiteracyforYoungEntrepreneurs.WorldBankPolicyResearchWorkingPaperNo.5642.Caselli,Francesco,andNicolaGennaioli.2005.DynasticManagement.EconomicInquiry,forthcoming.DeMel,Suresh,DavidMcKenzie,andChristopherWoodruff.2008.ReturnstoCapital:ResultsfromaRandomizedExperiment.QuarterlyJournalofEconomics,123(4):13291372.DeMel,Suresh,DavidMcKenzie,andChristopherWoodruff.2009.MeasuringMicroenterpriseProfits:MustWeAskHowtheSausageIsMade?JournalofDevelopmentEconomics,88(1):1931.Drexler,Alejandro,GregFischer,andAntoinetteSchoar.2010.FinancialLiteracyTrainingandRuleofThumbs:EvidencefromaFieldExperiment.Unpublished.

19

Gin,XavierandGhazalaMansuri.2011.MoneyorIdeas?AFieldExperimentonConstraintstoEntrepreneurshipinRuralPakistan.Mimeo.WorldBank.Gompers,Paul,JoshLerner,andDavidScharfstein.2005.EntrepreneurialSpawning:PublicCorporationsandtheGenesisofNewVentures,1986to1999.JournalofFinance,60(2):577614.Hurst,EirkandBenjaminWildPugsley.2011.WhatDoSmallBusinessesDo?BrookingsPapersonEconomicActivity,forthcoming.Karlan,Dean,andMartinValdivia.Forthcoming.TeachingEntrepreneurship:ImpactofBusinessTrainingonMicrofinanceInstitutionsandClients.ReviewofEconomicsandStatistics.Karlan,DeanandJonathanZinman.2011.MicrocreditinTheoryandPractice:UsingRandomizedCreditScoringforImpactEvaluation,Science,332(6035):12781284.King,RobertG.,andRossLevine.1993.FinanceandGrowth:SchumpeterMightBeRight.QuarterlyJournalofEconomics,108(3):717737.McKenzie,DavidandChristopherWoodruff.2008.ExperimentalEvidenceonReturnstoCapitalandAccesstoFinanceinMexico.WorldBankEconomicReview,22(3):457482.Rajan,RaghuramG.,andLuigiZingales.1998.FinancialDependenceandGrowth.AmericanEconomicReview,88(3):559586.Udry,Christopher,andSantoshAnagol.2006.TheReturntoCapitalinGhana.AmericanEconomicReview,96(2):388393.

20

Appendix1:SurveyQuestionsforEntrepreneurialSpiritand

HumanResourcesManagementIndicesSurveyQuestionsforEntrepreneurialSpiritIndex

Stronglydisagree Disagree Neutral Agree

Stronglyagree

a. Ihaveprofessionalgoals. 1 2 3 4 5b. Irevisemygoalsperiodically. 1 2 3 4 5c. IfIdontreachagoalinthewayIwanted

toItryagain. 1 2 3 4 5d. Icantmotivatemybusinesspartners.* 1 2 3 4 5e. EverythingIneedforsuccessliesinmyself. 1 2 3 4 5f. Iprefertalkingaboutsolutions,not

problems.* 1 2 3 4 5

g. Ithinkthegovernmentshouldgivemeopportunities.* 1 2 3 4 5

h. Ihavetoreachsomegoalseverydaytofeelsatisfied. 1 2 3 4 5

*ReversecodedSurveyQuestionsforHumanResourcesManagementIndex

Stronglydisagree Disagree Neutral Agree

Stronglyagree

a. Theemployeesidentifywiththeobjectivesofthecompany. 1 2 3 4 5

b. Thefirmletsitsemployeesknowiftheyhavedonesomethingwrong. 1 2 3 4 5

c. Allresponsibilitiesareclearlyassignedforeachofthemembersofthefirm. 1 2 3 4 5

d. Alldecisionsaremadebythesameperson. 1 2 3 4 5e. Thefirmgivespositiverecognitiontoits

employees. 1 2 3 4 5

f. Thereislowturnoverofemployeesinthefirm. 1 2 3 4 5

21

Table1:TopicsthatFirmsWorkedonwithTheirConsultantBasedonEightCaseStudies

Topic #offirmsthatcoveredthistopicDefinemissionandvisionstatements 6Accountingandrecordkeeping(trainingand/ornewsoftware) 5Clarifyorganizationalstructure,clearlyassignresponsibilities 5Salesstrategyandadvertising(marketing) 4Strategicallyselectlocationandnumberofsalespoints 2Qualitycontrol 2Accesstocreditoralternativefinancingsolutions 2Humanresourcesmanagementandhiringpractices 2Mediatefamilyproblemsinfamilyfirms 1Pricingstrategy 1Reducecosts(negotiatewithsuppliers,findalternativesuppliers) 1Figureoutwhichproductsaremostprofitableandfocusonthese 1Teamworkandcommunicationstrainingforemployees 1Leadershiptrainingforfirmowners 1

22

Table2:BaselineSummaryStatistics

TreatmentMean

(StdDev)ControlMean

(StdDev)

OrthogonalityVerification

(1)(2)tstat

(pvalue)PanelA:Stratificationvariables (1) (2) (3)Manufacturingsectordummy 0.300 0.323 0.023 (0.460) (0.468) (0.630)Commercesectordummy 0.253 0.230 0.023 (0.436) (0.422) (0.597)Servicessectordummy 0.447 0.447 0.000 (0.499) (0.498) (0.998)Fulltimepaidemployees 14.400 13.684 0.716 (30.887) (31.479) (0.821)PanelB:Rerandomizationvariables Ageofprincipaldecisionmaker(years) 42.561 42.876 0.315 (10.212) (9.878) (0.756)Maleprincipaldecisionmakerdummy 0.727 0.720 0.007 (0.447) (0.450) (0.881)Yearsofschoolingofprincipaldecisionmaker 15.630 15.932 0.302 (4.919) (5.196) (0.559)Businessage(years) 11.053 13.652 2.599 (10.330) (28.120) (0.275)N 150 282 432Note:Columns1,2,4and5displaymeansandstandarddeviations(inparentheses).Columns3and 6 show the difference in means across the treatment and control group with thecorrespondingpvalueinparentheses.Column7showsthedifferenceindifferenceinmeanswiththe corresponding pvalue in parentheses. For the 1% trimmed variables, last month's sales,assets,aswellasprofitscalculatedbasedonprofitmarginandsales,areonlytrimmedatthetop1%sincetheyareboundedbelowbyzero.Allothervariablesaretrimmedatthetopandbottom1%.Significancelevels:*10percent,**5percent,***1percent.

23

Table2:BaselineSummaryStatistics(continued)

TreatmentMean

(StdDev)ControlMean

(StdDev)

OrthogonalityVerification

(1)(2)tstat

(pvalue)PanelC:Othervariablesbusinessoutcomes (1) (2) (3)Lastmonth'ssales(1000sUSD) 76.343 50.844 25.499 (283.025) (119.987) (0.222)Lastmonth'ssales(1000sUSD),1%trimmed 40.479 46.113 5.634 (97.605) (94.238) (0.591)Profits(salesminuscosts,1000sUSD) 12.498 3.713 16.211 (111.446) (202.489) (0.426)Profits(salesminuscosts,1000sUSD),1%trimmed 9.251 9.590 0.339 (58.564) (48.851) (0.955)Profits(profitmarginandsales,1000sUSD) 13.659 12.253 1.407 (42.718) (36.997) (0.753)Profits(profitmarginandsales,1000sUSD),1%trimmed 8.797 9.522 0.725 (22.443) (20.361) (0.766)Businessassets(1000sUSD) 296.963 945.842 648.879 (767.969) (7822.005) (0.376)Businessassets(1000sUSD),1%trimmed 251.358 259.471 8.113 (594.726) (503.203) (0.898)Productivityresidual 0.028 0.016 0.045 (1.349) (1.253) (0.787)Productivityresidual1%trimmed 0.003 0.009 0.006 (1.189) (1.155) (0.968)Returnonassets(ROA) 0.026 0.151 0.177 (0.956) (0.808) (0.119)Returnonassets(ROA)1%trimmed 0.110 0.095 0.015 (0.318) (0.487) (0.802)N 150 282 432Note:Columns1,2,4and5displaymeansandstandarddeviations(inparentheses).Columns3and6showthedifferenceinmeansacrossthetreatmentandcontrolgroupwiththecorrespondingpvalueinparentheses.Column7showsthedifferenceindifference inmeanswiththecorrespondingpvalue inparentheses.Forthe1%trimmedvariables,lastmonth'ssales,assets,aswellasprofitscalculatedbasedonprofitmarginandsales,areonlytrimmedatthetop1%sincetheyareboundedbelowbyzero.Allothervariablesare trimmedat the topandbottom1%.Significance levels:*10percent,**5percent,***1percent.

24

Table3:PredictorsofProgramTakeUp

OLS Dependentvariable:

Binary=1iftreatmentgroupfirmtookupthe

program (1) (2)Commercesectordummy 0.106 0.132 (0.109) (0.110)Servicessectordummy 0.035 0.031 (0.097) (0.097)Log(fulltimepaidemployees+1) 0.099** 0.046 (0.040) (0.047)Ageofprincipaldecisionmaker(years) 0.006 0.008* (0.004) (0.004)Maleprincipaldecisionmakerdummy 0.189** 0.212** (0.090) (0.090)Businessage(years) 0.005 0.005 (0.004) (0.004)Log(lastmonth'ssales+1) 0.047** (0.023)Log(profits+1): 0.021Profitscalculatedbymultiplyingprofitmarginbysales (0.019)Returnonassets(ROA) 0.088*** (0.021)Constant 0.429** 0.334 (0.216) (0.251)Rsquared 0.122 0.172N 148 148Averageofdependentvariable 0.533 0.533Note: All explanatory variables are measured at baseline. Binary controlvariablesincludedforwhencovariateismissing,andthenmissingcovariatecodedaszero.Significancelevels:*10percent,**5percent,***1percent.

25

Table4:TreatmentEffectEstimates,BusinessOutcomesOLS

Outcomevariable

Fullsample

Fullsample

1%trimmed

1%trimmed

2%trimmed

2%trimmed

Fullsamplecontrolgroupmean

(std.dev.) (1) (2) (3) (4) (5) (6) (7)Log(fulltimepaidemployees+1) 0.079 0.096 0.063 0.092 0.077 0.111 1.835 (0.090) (0.080) (0.091) (0.081) (0.091) (0.080) (1.189) 389 389 385 385 381 381 Log(lastmonth'ssales+1) 0.562 0.563* 0.573 0.585* 0.625* 0.623* 7.994 (0.365) (0.335) (0.369) (0.340) (0.369) (0.342) (3.640) 323 323 319 319 315 315 Profits: 5.921 5.426 7.609* 7.751* 4.362 4.335 9.707Calculatedassalesminuscosts (5.178) (5.071) (4.247) (4.179) (3.283) (3.202) (93.057) 288 288 282 282 276 276 Profits: 0.687 0.544 0.415 0.202 0.128 0.017 8.733

(2.728) (2.435) (1.560) (1.396) (1.250) (1.142) (23.170)Calculatedbymultiplyingprofitmarginbysales 309 309 306 306 303 303 Log(profits+1): 0.802** 0.820** 0.761** 0.794** 0.797** 0.828** 6.261

(0.352) (0.336) (0.349) (0.334) (0.349) (0.335) (3.451)Profitscalculatedbymultiplyingprofitmarginbysales 309 309 306 306 303 303 Log(businessassets) 0.052 0.107 0.098 0.129 0.107 0.168 11.215 (0.176) (0.156) (0.172) (0.156) (0.174) (0.156) (1.699) 319 319 315 315 312 312 Productivityresidual 0.266* 0.249* 0.199 0.193* 0.242** 0.232** 0.095

(0.140) (0.129) (0.122) (0.116) (0.116) (0.111) (1.272)Residualfromregressionoflogprofitsonlogemployeesandlogbusinessassets 250 250 244 244 240 240 Returnonassets(ROA) 0.120* 0.114* 0.061 0.057 0.061* 0.064* 0.015 (0.061) (0.065) (0.040) (0.045) (0.034) (0.036) (0.541) 247 247 241 241 237 237 Controlsforbaselinevalueofoutcome No Yes No Yes No Yes Note:Eachcellcontainsthetreatmenteffectpointestimate,robuststandarderror,andnumberofobservations,foraseparateOLSestimation.Fortheregressionsthatcontrolfortheoutcomevariablemeasuredatbaseline(Columns2,4,and6),whenthebaselineoutcomevariableismissing,themissingvalueisfilledinwithzeroandadummyvariableindicatingthatthebaselineobservationismissingisaddedtothemodel.Allregressionsincludecontrolsforstratadummiesandrerandomizationvariables.Inthex%trimmedsamples,fulltimepaidemployees,lastmonth'ssales,assets,aswellasprofitscalculatedbasedonprofitmarginandsales,areonlytrimmedatthetopx%sincetheyareboundedbelowbyzero.Allothervariablesaretrimmedatthetopandbottomx%.Weuselog(profits+1)insteadoflog(profits)forprofitscalculatedbymultiplyingprofitmarginbysalessincesomeprincipaldecisionmakersreportedzerosalesforthelastmonth.Significancelevels:*10percent,**5percent,***1percent.

26

Table5:TreatmentEffectEstimates,BusinessProcesses

OLS

OutcomevariableFull

sampleFull

sample ObservationsControlgroup

mean(std.dev.)

(1) (2) (3) (4)0.039 0.038 378 0.531Developednewproductsduringlastyeardummy(0.056) (0.054) (0.500)0.017 0.030 376 0.789Attractednewclientsduringlastyeardummy(0.046) (0.045) (0.409)0.056 0.066 378 0.617Implementednewprocessduringlastyear

dummy (0.053) (0.052) (0.487)0.026 0.024 378 0.074Attractednewinvestorsduringlastyeardummy(0.032) (0.031) (0.262)0.054 376 0.079Beganprocesstoregisterapatentduringlast

yeardummy (0.034) (0.270)0.022 378 0.156Begancertificationprocessforaninternational

standard(e.g.ISO) (0.035) (0.364)0.132** 378 0.440Madenewmarketingeffortduringlastyear

dummy (0.055) (0.497)0.024 377 0.240Expandedinstallationsduringlastyeardummy(0.045) (0.428)0.025 377 0.459Remodeledinstallationsduringlastyeardummy(0.054) (0.499)

Entrepreneurialspiritindex 0.242* 0.227 373 0.094 (0.140) (0.139) (1.371)

0.245* 0.210 373 0.095Entrepreneurialspiritindexw/oquestionsdande(seeAppendix1) (0.140) (0.138) (1.343)

0.053 0.050 363 0.022Humanresourcesmanagementindex(0.152) (0.147) (1.450)

Keepsformalaccountsdummy 0.072** 0.065** 378 0.852 (0.030) (0.028) (0.356)Controlsforbaselinevalueofoutcome No Yes Note:Columns1and2containthetreatmenteffectpointestimates,robuststandarderrors,andnumberofobservations, for separate OLS estimations. All regressions include controls for strata dummies and rerandomizationvariables.Somevariablesarenotavailableatbaseline,whichiswhythecorrespondingcellsinColum2areempty.Column4 includes thecontrolgroupmeanandstandarddeviationofeachvariableatfollowup.Significancelevels:*10percent,**5percent,***1percent.

27

Table6:TreatmentEffectEstimates,ChangesinResponsetoCrisisOLS

Outcomevariable Fullsample Controlgroupmean&std.dev. (1) (2)Laidoffstafforcutdownonhiring 0.036 0.257 (0.053) (0.438)Loweredemployeesalaries 0.023 0.092 (0.032) (0.289)Cutproduction 0.080** 0.206 (0.040) (0.406)Diversifiedbusinessactivities 0.015 0.431 (0.057) (0.496)Soughtgovernmentassistance 0.056 0.128 (0.044) (0.335)None 0.005 0.115 (0.037) (0.319)Other 0.045 0.216 (0.050) (0.412)Numberofchangesmade 0.020 1.330 (0.093) (0.810)N 340 218

Note: Column 1 contains the treatment effect point estimates and robust standarderrors for separate OLS estimations. All outcome variables, except for "number ofchangesmade",arebinaryvariablesfortheresponsestothequestion"Whichchangeshasyour firmmade inresponsetothecurrenteconomicsituation?" (multipleanswerswereallowed).Thisquestionwasaskedatfollowupinreferencetotherecenteconomiccrisis. "Number of changes made" is a count of the number of changes reported inresponsetothequestionabove.Allregressionsincludecontrolsforstratadummiesandrerandomization variables. Column 2 includes the control groupmean and standarddeviationofeachvariable.Significancelevels:*10percent,**5percent,***1percent.

28

29

Table7:SelfReportedReasonsforNotUsingConsultingServices

Reasonsfornotusingconsultingservices%ofenterprises

mentioningthisreason(multiplemention)

Wouldbeagoodinvestment,butdon'thavefunds 46.3Don'tknowwhatthebenefitswouldbe 22.2Simplyhadn'tconsideredit 18.5Didn'tneedtheservices 13.9Other 11.1Didn'tknowtheseservicesexisted 7.4Notworththecost 5.6

DP 1010 cover.pdfDP 1010