SURVEYDESIGNANDMETHODS statistics. When aggregatevariableswereconstructedforalloftheplans,theimputationratewasusuallymuchlower. There areafewvariablesthatKaiser/HREThasdecidednottoimpute;thesearetypicallyvariableswhere“don’tknow”isconsidereda valid responseoption(forexample,thepercentageofworkersrespondentsbelievearecoveredbyMedicaid). Somevariables areimputedbasedontheirrelationshiptoeachother. Forexample,ifafirmprovidedaworkercontributionforfamilycoverage butnopremiuminformation,aratiobetweenthefamilypremiumandfamilycontributionwasimputedandthenthefamily premiumwascalculated. Inaddition,thereareseveralvariablesinwhichmissingdataarecalculatedbasedonrespondents’ answerstootherquestions(forexample,employercontributionstopremiumsarecalculatedfromtherespondent’spremium andtheworkercontributiontopremiums). Since2014,weestimateseparatesingleandfamilycoveragepremiumsforfirmsthatprovidepremiumamountsastheaverage costforallcoveredworkers,insteadofdifferentiatingbetweensingleandfamilycoverage. Thismethodmoreaccurately accountsfortheportionthateachtypeofcoveragecontributestothetotalcostfortheonepercentofcoveredworkerswho areenrolledatfirmsaffectedbythisadjustment. SAMPLEDESIGN WedeterminedthesamplerequirementsbasedontheuniverseoffirmsobtainedfromtheU.S.CensusBureau. Priortothe 2010survey,thesamplerequirementswerebasedonthetotalcountsprovidedbySurveySamplingIncorporated(SSI)(which obtainsdatafromDunandBradstreet). Since2010,wedefineEducationasaseparatesamplingcategoryforthepurposes of sampling,ratherthanasasubgroupoftheServicecategory. Inthepast,Educationfirmswereadisproportionatelylarge shareofServicefirms. Educationiscontrolledforduringpost-stratification,andadjustingthesamplingframetoalsocontrolfor EducationallowsforamoreaccuraterepresentationofboththeEducationandServiceindustries. In past years, both private and governmentfirmsweresampledfromtheDunandBradstreetdatabase. Beginningin2009, Governmentfirmsweresampledfromthe2007CensusofGovernments. Thischangewasmadetoeliminatetheoverlapof state agencies that were frequently sampledfromtheDunandBradstreetdatabase. Thesampleofprivatefirmsisscreened for firmsthatarerelatedtostate/localgovernments,andifthesefirmsareidentifiedintheCensusofGovernments,theyare reclassified as governmentfirmsandaprivatefirmisrandomlydrawntoreplacethereclassifiedfirm. Thefederalgovernment is not included in the sample frame. Finally, the data used to determine the 2017 EmployerHealthBenefitsSurveysampleframeincludetheU.S.Census’2013 Statistics of U.S. Businesses and the 2012 Census of Governments. At the time of the sample design (December2016),these datarepresentedthemostcurrentinformationonthenumberofpublicandprivatefirmsnationwidewiththreeormore workers. Asinthepast,thepost-stratificationisbasedonthemostup-to-dateCensusdataavailable(the2014updatetothe CensusofU.S.Businesseswaspurchasedduringthesurveyfieldingperiod). WEIGHTINGANDSTATISTICALSIGNIFICANCE BecauseKaiser/HRETselectsfirmsrandomly,itispossiblethroughtheuseofstatisticalweightstoextrapolatetheresults tonational(aswellasfirmsize,regional,andindustry)averages. Theseweightsallowustopresentfindingsbasedonthe numberofworkerscoveredbyhealthplans,thenumberoftotalworkers,andthenumberoffirms. Ingeneral,findingsindollar amounts(suchaspremiums,workercontributions,andcostsharing)areweightedbycoveredworkers. Otherestimates,such astheofferrate,areweightedbyfirms. SpecificweightswerecreatedtoanalyzetheHDHP/SOplansthatareofferedwitha HealthReimbursementArrangement(HRA)orthatareHealthSavingsAccount(HSA)-qualified. Theseweightsrepresentthe proportionofemployeesenrolledineachofthesearrangements. Calculationoftheweightsfollowsacommonapproach. Wetrimmedtheweightsinordertoreducetheinfluenceofweight outliers. First, we grouped firms into size and offer categories of observations. Within each strata, we identified the median and theinterquartile rangeoftheweightsandcalculatedthetrimmingcutpointasthemedianplussixtimestheinterquartile range(M+[6*IQR]).Weightvalueslargerthanthiscutpointaretrimmedtothecutpoint. Inallinstances,veryfewweight valuesweretrimmed. Finally,wecalibratedtheweightstoU.S.CensusBureau’s2014StatisticsofU.S.Businessesforfirmsin theprivatesector,andthe2012CensusofGovernmentsasthebasisforcalibration/post-stratificationforpublicsectorfirms. Historic employer-weightedstatisticswereupdatedin2011. The Kaiser Family Foundation and Health Research & Educational Trust / Page 21