A goalprogrammingapproachtoestimatingperformanceweightsfor ranking firms

This paperproposesseveralgoalprogramming(GP)modelsforestimatingtheperformancemeasure weightsoffirmsbymeansofconstrainedregression.Sincesomesingle-criterionperformancemeasures are usuallyinconflict,weproposetwoopposedalternativesfordeterminingmultiple-criterion performance:thefirstistocalculateaconsensusperformancethatreflectsthemajoritytrendofthe single-criterionmeasuresandtheotheristocalculateaperformancethatisbiasedtowardsthe measuresthatshowthemostdiscrepancywiththerest.GPmakesitpossibletomodelbothapproaches as wellasacompromisebetweenthetwoextremes.Usingtwocasestudiesreportedintheliterature and introducinganotheroneexaminingnon-financialcompanieslistedinIbex-35,wecompareour proposalwithothermethodssuchasCRITICandamodifiedversionofTOPSIS.Inordertoimprovethe comparisonsaMontecarlosimulationhasbeenperformedinallthreecasestudies. Scopeandpurpose: The studyfallsintotheareaofmultiple-criteriaanalysisofbusinessperformance. Firms areobligedtoreportavastamountoffinancialinformationatregularintervals,andforthisthere is awiderangeofperformancemeasures.Multicriteriaperformanceiscalculatedfromthesingle- criterionmeasuresandisthenusedtodrawuprankingsoffirms.Asacomplementtotheother multicriteriamethodsdescribedintheliterature,weproposetheuseofGPforimplementingtwoquite differentstrategies:overweightingthemeasuresinlinewiththegeneraltrendoroverweightingthe measuresthatconflictwiththerest.BesidestheuseofSpearman’scorrelation,weintroducetwoother measuresforcomparingthesolutionsobtained.