London, 2009. Microsimulation in decision support The latest news about our results József Csicsman [email protected].

London, 2009.

Microsimulationin decision support

The latest news about our results

Microsimulationin decision support

The latest news about our results

József [email protected]

ContentsContents

IntroductionMicrosimulation Research Group in BUTE

Formal presentation at IFIP WS-s

Microsimulation theory

Microsimulation in practiceResearch data sets from 2004

Problems of modelling demographic changes

Application in Student Loan forecast

Applications in bank sector and telcos

IntroductionIntroduction

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)

Calculus and BUTE cooperationCalculus and BUTE cooperation

Microsimulation Research GroupMicrosimulation 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-sFormal Presentation in IFIP WS-s

Microsimulation Service System, Statistical Matchingpresented by Péter Baranyai in Budapest

Microsimulation Servise System based on SAS and Application of Microsimulation in Decision Supportpresented by Balázs Látó in Cork

MicrosimulationMicrosimulation

MicrosimulationMicrosimulation

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 MicrosimulationWorkflow of Microsimulation

Microsimulation Service SystemMicrosimulation Service System

Statistical MatchingStatistical 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 possibilitiesApplication possibilities

Demographic, social and economic impacts of various measures

Improving the quality of statistical surveys

Aging of datasets (bringing data of former surveys up-to-date)

More accurate forecast of probable events

International comparisons (competitiveness, tax and subsidy systems…)

Research data sets from 2004

Research data sets from 2004

Research data sets from 2004Research 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 changesProblems of modelling demographic changes

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 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 Population, vital events in Hungary

Denomination 2004 2005 2006 2007 2008 2009

Population, 1 January 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 1 106 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

Death number 132 492 135 732 131 603 132 938 130 000

Natural increase, decrease (–)

number -37 355 -38 236 -31 732 -35 325 -30 800

Application in Student Loan forecast

Application in Student Loan forecast

Student Loan forecastStudent 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 forecastStudent Loan forecast

Estimation of paid and unpaid loansproblems 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

Applications in Bank sectorApplications in Bank sector

As we discussed before:Conducting stress test analysis and creating reportsReplacement of missing data

merging simulated research data sets with real client data

Predicting success of new business productsSupport for credit scoring

Applications in Bank sectorApplications 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 sectorApplications in Bank sector

Take changes of consumption, income, etc. into consideraion

data used for modelling had been registered in different termsproblem with compatibilityapplying microsimulation in order to age data

Stress testStress test

Examines the probability and possible effects of unforseen eventsStress test for Hungarian banks (2006)

BASEL2mandatory for banksconcentrating on the extreme values of unexpected events

increased inflationunemployment and exchange ratethe drastic effect of these matters on the credit system

Stress testStress test

Done on the research dataset The test examines the possible outcomes and

effects of an unforeseen eventConditions:

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

1-3 100%4-6 50%7-10 25%

Another 3% cannot make repayments because of the rise in unemployment

Applications in TelcosApplications 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!

[email protected]

Thank you for your attention!

[email protected]