Microsimulation in decision support The latest news about

Microsimulation in decision support The latest news about our results József Csicsman csicsman@itm. bme. hu London, 2009.

Contents Introduction Microsimulation Research Group in BUTE Formal presentation at IFIP WS-s Microsimulation theory Microsimulation in practice Research data sets from 2004 Problems of modelling demographic changes Application in Student Loan forecast Applications in bank sector and telcos

Introduction

Calculus and BUTE cooperation Microsimulation research group at the Information and Knowledge Management Department of Budapest University of Technology and Economics was founded in 2001. Cooperation with the Hungarian Central Statistical Office (KSH) International cooperation (EU) Custom economic applications Students graduates with practical SAS knowledge (more than 50 former students work in the field of SAS application in financial sector)

Microsimulation Research Group Models simulate large representative populations of these low-level entities (using probabilities, laws, rules or empirical facts) 2001 -2003: Common development group for the technical background of Microsimulation Real applications from 2003

Formal Presentation in IFIP WS-s Microsimulation Service System, Statistical Matching presented by Péter Baranyai in Budapest Microsimulation Servise System based on SAS and Application of Microsimulation in Decision Support presented by Balázs Látó in Cork

Microsimulation

Microsimulation has been used for decades in economics and other areas. The microsimulation procedure examines social and economic changes by assessing the effect of each provision with small units and the description of the overall effects is derived from these assessments. It has essential role in decision support.

Workflow of Microsimulation

Microsimulation Service System

Statistical Matching A new function of the Microsimulation Service System How to merge the records of two (or more) datasets having no key variable Based on statistical analysis, and distribution of other variables Example: Simulation of marriages Replacement of missing or corrupt data from other surveys

Application possibilities Demographic, social and economic impacts of various measures Improving the quality of statistical surveys Aging of datasets (bringing data of former surveys up-todate) More accurate forecast of probable events International comparisons (competitiveness, tax and subsidy systems…)

Research data sets from 2004

Research data sets from 2004 Microsensus at HCSO, 2004 and recording income data correction of data with Microsimulation Service System of Calculus Household statistic survey at HCSO, 2004 Creation of research data set with statistical matching relatively good data about consumption and income

Problems of modelling demographic changes

Problems of modelling demographic changes We cannot use weighted data for demographical simulation (because of small sample size) Multiplication to the complete population Marriage and devorce simulation models the most complicated method

Problems of modelling demographic changes Birth and death simulation models Migration in Hungary is too big (the hungarian population hasn’t decreased, however, birth rate is too small and death rate is higher than other European countries)

Population, vital events in Hungary Denomination 2004 2005 2006 2007 2008 2009 10 116 742 10 097 549 10 076 581 10 066 158 10 045 401 10 031 000 male 4 804 113 4 793 115 4 784 579 4 779 078 4 769 562 4 761 000 female 5 312 629 5 304 434 5 292 002 5 287 080 5 275 839 5 270 000 Number of females per thousand males 1 106 1 107 1 106 1 107 Density per km² 108. 7 108. 5 108. 3 108. 2 108. 0 107. 8 Marriges number 43 791 44 234 44 528 40 842 40 100 Divorces number 24 638 24 804 24 869 25 160 25 300 Live births number 95 137 97 496 99 871 97 613 99 200 132 492 135 732 131 603 132 938 130 000 -37 355 -38 236 -31 732 -35 325 -30 800 Population, 1 January Death number Natural increase, decrease (–) number

Application in Student Loan forecast

Student Loan forecast Hardly predictable number of persons who require Student Loan Simulation of demographical changes till 2007 Merging real student loan data with simulated population Simulation till 2010

Student Loan forecast Estimation of paid and unpaid loans problems of high level intrest rate in Hungary pay-backs are too frequent, thus traditional bank estimations aren’t usable

Applications in Bank Sector

Applications in Bank sector As we discussed before: Conducting stress test analysis and creating reports Replacement of missing data merging simulated research data sets with real client data Predicting success of new business products Support for credit scoring

Applications in Bank sector Income is not an efficient enough indicator in Hungarian bank sector Income isn’t considered as determinant information about the financial background of individuals Consumption is closely connected with financial background, thus it provides more relevant information

Applications in Bank sector Take changes of consumption, income, etc. into consideraion data used for modelling had been registered in different terms problem with compatibility applying microsimulation in order to age data

Stress test Examines the probability and possible effects of unforseen events Stress test for Hungarian banks (2006) BASEL 2 mandatory for banks concentrating on the extreme values of unexpected events increased inflation unemployment and exchange rate the drastic effect of these matters on the credit system

Stress test Done on the research dataset The test examines the possible outcomes and effects of an unforeseen event Conditions: Exchange reate of CHF rises from 160 HUF to 200 HUF Those who have an amortization instalment greater than their income’s 30%: Failure rate in the income decimals 1 -3 4 -6 7 -10 100% 50% 25% Another 3% cannot make repayments because of the rise in unemployment

Applications in Telcos Replace missing demographic data by using statistical matching Correction of corrupt marketing survey data Forecast of marketing strategies aimed at avoiding attrition Fraud protection

Thank you for your attention! csicsman@itm. bme. hu