plan.be Recent Developments in Dynamic Microsimulation Modeling for Policy Support at the Federal Planning Bureau It takes two to tango Gijs Dekkers FPB, CESO KU Leuven and CEPS/INSTEAD Guest Lecture, Université de Liège, April 19 th, 2012
plan.be A discussion and classification of microsimulation models The dynamic microsimulation model MIDAS: ready to tango The institutional setting in Belgium: what is MIDAS being used for anyway? It takes two to tango: MIDAS, MALTESE (and S3BE). Maltese Hypotheses, assumptions, and other relevant stuff Base simulation results Overview of this presentation
plan.be Some examples highlighting the consistency between MIDAS and MALTESE through various channels variant 1: an increase of the “guaranteed income for the elderly” (GRAPA/IGO) by 13.7% in 2006 variant 2: a change of the assumed growth rate of GDP variant 3 : an increased employment rate among the older active population Opening the black box (a.k.a. what determines long-run developments in income inequality anyway?) Overview of this presentation, part two
plan.be A discussion and classification of microsimulation models The essential function of (dynamic/static) microsimulation models......is the imputation of (prospective/alternative) microdata n i
plan.be A classification of microsimulation models Longitudinal ageing Cross- sectional ageing Dynamic ageing Microsimulation models Static Dynamic Static ageing
plan.be A classification of microsimulation models Advantages of static microsimulation overnight changes relatively cost efficient: less cumbersome in development in maintenance (limitations to) use are easily explained to policy makers Advantages of dynamic microsimulation how will demographic ageing affect current simulation results? how do economic and socio-demographic developments change the current simulation results? how about policy measures that have a gradual implementation? what about the impact of (partial) indexation of benefits to wages? How about longitudinal redistribution?
plan.be MIDAS - Microsimulation for the Development of Adequacy and Sustainability Starting dataset: Previously: PSBH survey sample 2002: ± 8K individiduals Currently: MIMOSIS 2001 administrative sample, expanded: ± 2.2K 2 individuals DEMOGRAPHIC MODULE LABOUR MARKET MODULE PENSION & BENEFITS MODULE Mortality, fertility, education, ‘marriage market’ Employment, public and private sector, civil servants, unemployment, disability, CELS, retirement Hours of work per month, months of work per year, earnings per hour 1st pillar pension benefits - employees’ scheme - civil servants scheme - independent workers’ scheme CELS benefits disability pension benefits unemployment benefits welfare benefits 7 CONTRIBUTIONS AND TAXATION MODULE REDISTRIBUTION, POVERTY, INEQUALITY
plan.be MIDAS – the demographic module Mortality, fertility: AWG- projections (2005) Education Step 1: Monte Carlo-routine ‘assigns’ an educational attainment level to every 10-year old. Step 2: The individual enters the labour market at an age determined by the level of eduation. Partnership formation : the ‘marriage market’ module
plan.be MIDAS – the marriage market module (I) person is ‘selected’ to find a partner “marriage market” -link individuals -create a new household Marriage or cohabitation? Divorce?Separation? DivorcedMarried Single Marriage? Cohabiting MarriageCohabitation NoYes Married Cohabiting Yes No Yes No
plan.be MIDAS – the marriage market module (II) … 1 p(1,1)p(1,2)p(1,3) … 2 p(2,1)p(2,2)p(2,3) … 3 p(3,1)p(3,2)p(3,3) … 4………… 5 6 … Males selected for the marriage market Females selected for the marriage market age, age difference dummy of working dummies for eduational attainment levels
plan.be MIDAS – the labour market module (I) IN WORK EMPLOYEE PUBLIC SECTOR CIVIL SERVANT PUBLIC SECTOR PENSION SCHEME PRIVATE SECTOR PENSION SCHEME SELF-EMPLOYED PENSION SCHEME Yes No Yes MONTHS OF WORK HOURS OF WORK BY MONTH HOURLY WAGE XX WIDOW(ER)S PENSION BENEFIT
plan.be MIDAS – the labour market module (II) IN WORK No UNEMPLOYMENT DISABILITY EARLY-RETIREMENT, UNEMPLOYMENT FOR OLDER WORKERS, … RETIREMENT OTHER INACTIVE Yes PUBLIC SECTOR PENSION BENEFIT PRIVATE SECTOR PENSION BENEFIT SELF-EMPLOYED PENSION BENEFIT Disability benefit Early retirement benefit, unemployment benefit for older workers, … WIDOW(ER)S PENSION BENEFIT
plan.be Alignment of state variables: Procedure to have the model respect or ‘mimic’ exogenous aggregates while respecting individual probabilities in the occurrence of the event Behavioral equation determining the probability of the transition Individuals are ranked depending on the obtained probability (from the highest to the lowest) The number of selected individuals reproduces targeted aggregates Monetary alignment or ‘amount alignment’: Proportional adjustment of first-run values of earnings to match exogenous macroeconomic productivity growth rates Uprating of social security benefits (see later) MIDAS – ready to tango
plan.be Can I have this dance? ‘Channels of consistency’ of MIDAS with MALTESE 1.State alignment 2.Monetary alignment 3.Joint social hypotheses MALTESE (meso) MIDAS (micro) S3BE (macro)
plan.be Can I have this dance? State alignment (for each simulation year) – mortality (by age, gender and year of simulation) – fertility (idem) – employment rate (by age classes, gender and year of simulation) – unemployment rate (idem) – proportion of self-employed (idem) – proportion of public sector employees (idem) – proportion of civil-servants (idem) – proportion of disabled (idem) – proportion of CELS recipients (idem) Monetary alignment (for each simulation year) - earnings, to gender
plan.be Because the sustainability and adequacy of pensions are two sides of the same coin Assumptions and projections underlying the assessment of sustainability affect adequacy productivity growth, wages, employment, the link between wages and benefits Not all aspects of the adequacy of pensions are reflected by the replacement rate (re)distributional impact, poverty, the link between wages and benefits so… An assessment of the sustainability of pension systems should take into account the adequacy of pension benefits....and why would we be interested in this joint assessment via MIDAS and MALTESE anyway?
plan.be And policy makers are aware of this! (“gee whizz!”) Gouvernement : -Note sur le Vieillissement -Programme de stabilité, dont volet « long terme » Conseil Supérieur des Finances, section « Besoin de financement » : -Avis annuel (financement du coût budgétaire du vieillissement), dont objectifs budgétaires de moyen terme – allocations du fonds du vieillissement Comité d’Etude du Vieillissement (C.S.F.): rapport annuel : -Conséquences budgétaires (coût budgétaire du vieillissement) et sociales du vieillissement - Études spécifiques (2ème pilier, …) C.C.E
plan.be Some underlying hypotheses and assumptions of the Study Committee for Ageing (report 2010) Key demographic hypotheses Fertility Life expectancy at birth Men women Key macro hypothesesUp to ≥ 2015 Yearly productivity0.01%1.28%1.50% Unemployment rate14.75 in 2014Decreasing towards 8% Social policy hypotheses ≥ 2015 Wage ceilingCurrent legislation1.25% Minimum right per working year1.25% Welfare adjustment non-lump-sum benefits Employed and self-employed 0.50% Welfare adjustment of lump-sum benefits1.00%
plan.be Simulation results (Finally!)
plan.be Simulation results Base scenario (1,5% increase of productivity per year) Pensions 5.3 Health care 4.2 Other -1.4 Total 8.2 Inequality of income to state (Gini)
plan.be Some base results of MALTESE and MIDAS (finally!) Table 1Budgetary costs of ageing: rate of increase in % of GDP, 2008 to 2060 Source: Annual Report of the Study Committee for Ageing (High Council of Finances, 2009) June 2009 Base scenario (1,5% increase of productivity per year) Pensions 5.3 Health care 4.2 Other -1.4 Total 8.2 Poverty risk to state
plan.be Simulation variant 1: an increase of the “guaranteed income for the elderly” (GRAPA° by 13.7% in 2006 An illustration of the benefits of alignment Budgetary impact: % GDP
plan.be Simulation variant 2: a change of the assumed growth rate of GDP Base scenario (1,5% increase of productivity per year) Variant (1,75% increase per year) Pensions Health care 4.2 Other Total Table 2Budgetary costs of ageing: rate of increase in % of GDP, 2008 to 2060 Source: Annual Report of the Study Committee for Ageing (High Council of Finances, 2009) June 2009
plan.be Simulation variant 3 : an increased employment rate among the older active population Base scenario (1,5% increase of productivity per year) Productivity variant (1,75% increase per year) Employment variant Pensions5,34,54,7 Health Care4,2 Other-1,4-1,7-1,9 Total8,27,0 Table 3Budgetary costs of ageing: rate of increase in % of GDP, 2008 to 2060 Source: Annual Report of the Study Committee for Ageing (High Council of Finances, 2009) June 2009
plan.be An assessment of sustainability is difficult – if not meaningless - without taking into account adequacy MIDAS is a dynamic microsimulation model designed to align to the output of a macroeconomic model, thus integrating the approach to sustainability and adequacy Poverty among the elderly is going to decrease This decrease is partially the result of a considerable increase of the “guaranteed income for the elderly” (GRAPA) – and of the lower growth rate of wages. Working harder or working longer has the same impact on the budgetary costs of ageing, but the former comes with an increase of the risk of poverty among the elderly while the latter does not. It takes two to tango: intermediate conclusions
plan.be And now for something completely different: we focus on the “black box problem”
plan.be The black box problem: validating a dynamic model Starting point validation: “the comparison of the model’s results to counterpart values that are known or believed to be correct, or that are consistent with one’s assumptions, [or] other trustworthy models’ results” (Morrison, 2007, 5) Theses 1. the long run development of inequality of pensions is driven by just a few factors 2. A very simple stylized model can therefore be used to validate the simulation results of a dynamic microsimulation model.
plan.be A base stylized model for the inequality of pensions Simulate a change of the indexation parameter, and the impact of the retirement age Apply two ‘forms’ of demographic ageing - A ‘baby boom’ generation - Increasing longevity Validation of the Belgian MIDAS model The black box problem: validating a dynamic model
plan.be The base model Suppose individuals in time t≥0, each of a different age (so, age t =[0,…, 100], t=[0, …, 100]). 2. everybody retires at 60 and dies at 100, 3. the pension benefit at 60 equals € The model is expressed relative to wage growth, and pensions lag behind the development of wages with a constant fraction Ψ. Then p 0,age =100(1- Ψ) age-60 when t=0
plan.be The base model
plan.be A change of the indexation parameter ψ in the period cht
plan.be A change of the indexation parameter ψ in the period cht: ψ decreases from 1.8 to 1.25 in t=20
plan.be Demographic ageing I: a baby boom generation Write the base model as A special case is with w 0,age0 ~N(43,23) and w t,age =w 0,(age-t)
plan.be Demographic ageing I: a baby boom generation KDE of age at t≤50 KDE of age at t≤90
plan.be Demographic ageing I: a baby boom generation
plan.be Demographic ageing II: increasing longevity age of death x increases by 10 years, from 90 in period 0 to 100 in period 100.
plan.be Demographic ageing I and II: impact of ageing on pension inequality: the compound effect of fertility shock and increasing life expectancy
plan.be And now we put everything in our stylized model and compare it with MIDAS’s results...
plan.be MIDAS – Key simulation techniques for behavioural equations: logits State 1 State 2 Individual i U i < P i P i =logit- 1 (βX) U i < 1-P i ‘standard’ Monte Carlo simulationAligned simulation to target x% State 1 State 2 Rangk 1...n First xN individuals Rangk i =logit- 1 (βX+ε i ) Other (100-x)N individuals Run the model y i =βX i + u i (2001, t) Sum to aggregate ; derive aggregate growth rate gY Monetary alignment target growth rate gx t Derive corrected growth rates Run the model y i =βX i + u i (2001) Apply corrected growth rate -> t
plan.be Gini Proportion of population or sample 100% 0% Proportion of total income Gini=A/(A+B)
plan.be Poverty risk / at-risk-of-poverty-rate / headcount- ratio Suppose an exogenous poverty line z y F(y) z Then the headcount ratio HC is or HC=p/n, for which p=p(y z;z), p n