Regression Modeling Practice
lanquetuit.cyril@gmail.com, Université de Cergy Pontoise
Abstract
We will lead a study in function of multiple variables such as Socio-Professionnal Category (csp) or capital or date (number of month since january 2000 past un- til people answer questionnary) on a condential data set and try to answer the following problematic :
"How and Why people are risk averse ?"
Our data were collected on a representative sample of french population (3319 students) but condentiality restriction does not allow us to publish all the experimental procedure set in place to collect our data.
About the study...
http://cyrillanquetuit.free.fr/PDF/DataAnalysisInShort.pdf
1 About the Data
1.1 Sample
3319 french superior students who have answer online to a survey lead to in- dividual level and group outlook :θ, the risk aversion, depends on individual but also on it's position relatively to the whole population. The data base consists in 23233 lines, 7 series of "choice" for each 3319 students. Each of 287 variables are represented by a column.
The response variable which should be explain by the other one areθ, λ, γ respectives aversion to risk and loss and probability deformation. Those three parameters should be computed thank's to "choice" and others variables, posi- tionning an individual in a 3D space.
Histogram of age and choice (safe=1,risky=8) repartition 1
1.2 Survey
Lead in french high school and university based on an online questionnary which main purpose is to better know and understand why and how people are risk averse. A typical application is that risk aversion prol of indiviual can be used by Wealth Management Advisor. Each student receive a login/password to log on a website and have a specic timed session to answer (between 15 to 30 minutes). Data where collected between 2006 and now by RiskDynametrics.
1.3 Variables
Each student answers personnal questions about his individual caractéristics, his wealth level, investment objectif and nance market knowledge, we collect in this way 287 variable in which a special one called "choice" could be seen as the response variable and could be partially expplained by the others. There was 7 séries of choice per students and each choice value represents 1 + an answer to a "3-binary-questions-tree" coded on three bits where a "1" signify that student has taken the risky option instead than a safer one, for example if a student chose three time a risky option in the three alternative choice proposed in the serie 1, his choice value for this serie will be8=1+7 where7=111in base 2), a safe-safe-safe choice value is therefor1=1+000.
That is a rst basic approach to consider "choice" value as the response variable which should be explain by the others. In reallity "choice" is a comb of 8 oating slice in a 3D risk aversion space, each slice determine a plane inθ, λ, γ space. Below is represented a "2-binary-questions-tree" which is in fact a comb with 3 branches.
Figure 1 3D representation of a comb inθ, λ, γspace
2
Code Book
V ariableLabel academic agec age2c age3c
retireageage de la reraite reg_domDOM
reg11IDF
reg21Champagne-Ardenne reg22Picardie
reg23Haute-Normandie reg24Centre
reg25Basse-Normandie reg26Bourgogne
reg31Nord-Pas-de-Calais reg41Lorraine
reg42Alsace
reg43Franche-Comte reg52Pays-de-la-Loire reg53Bretagne
reg54Poitou-Charentes reg72Aquitaine reg73Midi-Pyrenees reg74limousin reg82Rhone-Alpes reg83auvergne
reg91Languedoc-Roussillon reg93PACA
reg94Corse reg_miss codetuu0
codetuu1moins de 5 000 hab codetuu25 000 à 9 999 hab.
codetuu310 000 à 19 999 hab.
codetuu420 000 à 49 999 hab.
codetuu550 000 à 99 999 hab.
codetuu6100 000 à 199 999 hab.
codetuu7200 000 à 1 999 999 hab.
codetuu8Unité urbaine de Paris codetuu_miss
dateNb mois a partir de janv 2000 date_f romaug2008
date_f romnov2010 min(0,date- date QE de FED)
date_f romjanv2015 min(0,date- date menace Grexit)
educ1 Aucun diplôme educ2 Brevet des colleges
educ7Diplôme professionnel : CAP et BEP
educ3Baccalaureat
educ4Bac +1 à 3, type Licence educ5_6Bac +4 et +, type master, ecole de commerce, ecole d ingenieur, doctorat
educ_miss
matri1Marie avec contrat matri2marie sans contrat matri3PACS
matri4Vie maritale matri5Divorce matri6Veuf
matri1_2_3_4 en couple, sans precision matri5_6_8 Separe, sans precision
matri5_6_8 matri7Celibataire matri_miss gender2 Femme
dependentenb personnes a charge csp1( Etudiant )
csp2(Recherche_emp ) csp3( Agriculteur) csp4( Ouvrier ) csp5(Employe) csp6( Technicien ) csp7(Agent_de_Mait ) csp8( Prof_intermed ) csp9( Artisan ) csp10( Cadre_sup ) csp11( Prof_liberale ) csp12( Chef_entreprise ) csp13(Retraite )
csp14(Sans_activite ) csp_miss
f publicfonctionnaire
product1Connaissance Actions product2Connaissance Obligations product3Connaissance Livrets product4 Connaissance Produits structures
product5 Connaissance FC- PIFCPR
product6Connaissance Aucun product7Connaissance OPCVM 3
product8 Connaissance Assurance vie
product9Connaissance PEA productown1 Detention Actions productown2 Detention Obliga- tions
productown3 Detention Livrets productown4 Detention Produits structures
productown5 Detention FC- PI/FCPR
productown6 Detention Aucun productown7 Detention OPCVM productown8 Detention Assurance vie
productown9 Detention PEA logCreditEstateln(credit immobi- lier)
CreditEstate_miss
logAssetEstate ln(patrimoine im- mobilier total)
AssetEstate_miss
logAssetF inancial ln(patrimoine nancier total)
AssetF inancial_miss
logamountln(montant investi) logincome
logincome_miss
logcapitalln(montant loterie) durationhorizon
duration2horizon2 duration3horizon3
experobj1 Experience<1 an experobj2_3_4 Experience 1 a 5 ans
experobj5_6Experience=4 a 5 ans experobj7 Experience =6-10 ans (5-10 ans pour AGORA)
experobj8 Experience=11-15 ans (10-15 pour pour AGORA)
experobj9 Experience=16-20 ans (15-20 ans pour AGORA)
experobj10 Experience=21-25 ans (20-25 ans pour AGORA)
experobj11Experience>25 ans
experobj_miss expersubj1Novice
expersubj2Plutôt Experimente expersubj3Experimente expersubj4Tres Experimente expersubj_miss
ecoclim1Nettement meilleur ecoclim2Un peu meilleur ecoclim3A peu pres le meme ecoclim4Moins bon
ecoclim5Nettement moins bon ecoclim_miss
f luctu1 Intolerance aux uctua- tions
f luctu2 Toleranceaux uctuations faibles
f luctu3Tolerance aux uctuations moderees
f luctu4Tolerance aux uctuations importantes
gainloss1Liquide immediatement gainloss2 Liquide seulement apres une perte signicative
gainloss3Ne liquide pas, conserve gainloss4Re-investis
motivmain1Transmission motivmain2Retraite
motivmain3Frais de formation motivmain4Matelas de securite motivmain5 Revenus supplemen- taires
motivmain6Revenus reguliers motivmain7 Acheter/renover une maison
motivmain8Acheter des biens du- rables
motivmain9 Fructier mon patri- moine
motivmain10Autre motivmain_miss
remove1 liquide, avant T/2 remove2liquide mais après les T/2 remove3 Ne liquide pas
remove_miss
4