microsimulation as a tool for the evaluation of public policies potx

They show that, given the complexity of the general in markets for factors and goods, partial equilibrium analysis or the use of a representative households approach would limit the

MICROSIMULATION AS A TOOL FOR THE EVALUATION

OF PUBLIC POLICIES

Methods and Applications

Amedeo Spadaro (Ed)

Sách có bàn quyễn

MICROSIMULATION AS A TOOL

FOR THE EVALUATION

OF PUBLIC POLICIES

Sách có bàn quyễn

Microsimulation

as a Tool for the Evaluation

of Public Policies Methods and Applications

Amedeo Spadaro (cd.)

Fundacién BBVA

‘The BBVA Foundation’s decision to publish this book does ot mph ay responsi fir its content, ut for the inchiden thetein af any supplementary daeuments or information Taibtated by the authors,

[Na part of thi publication, including the cover eesign, may he reproduced, sored in

a tettieval stem a unnsnited ia any form ot by aay mean, electronic, snechanical, photocopying, recording or otherwise, without the prior written permission of the copyright holder

Microsima 1 38.4 too for the evaluation of public

policies : methods and applicstions / Amedeo Spadaro

(ed), — Bilbao z Fundacion BBVA, 2007

387p 4cm

sa 078806512-174

1, Método de evatuacion 2 Simulariin 1 Spark,

Amedeo I, Fundaciin BBVA,

308,001.57

Hicsimuttion a Toler the Bonuaton 6

Patt Pics Mads wt Appicatons

© Fundacion BBVA, 2007

ash i a Action of Coen

Pio in Spin

fine veconeted cello fle (ased paper ter Han th

‘ton Noni àniaròcoilimgi ii Eghiatei

-nhftoanenll lợi eatin aid

1.5.1, Macroeconomic analysis and microsimulation

1.5.2 Inroducing dynamics x0

5, Firms, instnitions and investment climate

‘A Review of Applications in Italy and Norway

Rolf Aaberge and Ugo Colombino

21 Introduction a 2.2 The microcconometrie model a 2.3 Labour supply elasticity 6

24 A simulation of some eform proposals 6 2.5 An empirical exercise in optimal taxation os

5 Integrating the micro andl the CGE model m

2.7 Outofsample predictions a

3 Growth, Distribution and Poverty in Madagascar:

Learning from a Microsimulation Model in a General

Equilibrium Framework

Denis Cognense coud AnneSophie Robiliond

Am

5.2, Modelling income distribution Es

8.2.1, Functional distribution v4, personal distribution T6

5.2.2, The representative agent assumption z

3.3 Microeconomic specifications of the model Es

S31 Prnduction and labour alley

SLL Agricultural households 2 3.3.1.2 Nonageicultural households 81 3.8.2, Disposable income, savings and consumption a

34, Description of the

3.5 An application to Madagascar eral equili 33

3.5.1.1 Agricultural production function, 85 5.5.1.2 Informal and formal wage equations mg 3.5.2, Calibration, parameters and algorithm 89

35.21 Đạ bration sọ 5.5.2.2 Potential wage equation o 3.5.2.8 Heterogeneity we 3.5.2.4 Algorithm and solution

8.8 Impact of growth shocks on poverty and inequality vẽ

3.5.1, Some descriptive elements 98

3.6.2, Description of growth shocks 96

3.6.3, Ex ante and ex post decomposition of the impact

‘of growth shocks 6 3.8.1 Decomposition of microeconomic results by group 1

mm 1

4.24,A, United Kingdom 1

4.4, Empirical application: a microsimutation model

for taxation policies in Spain r 44.1 Objectives of the study 128 44.2, Methodology 128 4.4.5, Coverage type and use of healthcare services 130 4.4.4, An econometric model for the use probability

45 Simulation øFEheahhcare savings under different

double coverage scenarios 138 4.5.1, Evaluation of the cost of visits @ the public network 96 4.5.3, Simulation of healthcare costs and swings under

diverse double coverage scenation na 45.2.1, Simulition of diverse scenarios 10 4.6 Summary and conclusions Mit

Microsimulation in the Analysis of Environmental J

Reforms, An Application for Spain

Navier Labandeiva, José M Labwaga and Miguel Rodvigues

5.1 Intwoduction Me 5.2 Integrating micro and macroeconomic models TH 5⁄3 Dells of the integrated micro-macro model 158 8.1 The Applied General Equilibrium Model 187 5.3.2.A microeconomic model for household energy đetmand 161

54 Results obtained via the integrated micrommacro model 165 5.1, Description of the reform 165 5A.2, Results 165

5 Conchusions tr Appendix: Production and consumption structures 178

6.3 The baseline system Ist

Aa Thee HN ¿6lIEotie ebl a 6.4.2, Empirical specification and identification of the

BA3.A summary yt

8.5.1 The singles model 1s 65.1.1 Missing wages 198 65.1.2 Marginal propensities sử

8.5.2.1 Missing wages 210 6.5.2.2 Preference parameters 21 6.6 The unitary model for couples

6.7, The anasis of ux reform

6.7.2 The consequences of the 1999 tax reform

8.2.3 The consequences of the 1999 tax reform

according to the collective model 231

7 Microsimulation and Macroeconomic Analysis:

An Integrated Approach An Application for Evaluating,

Reforms in the Italian Agricultural Sector

Riccardo Magnani, Eleonora Matieassi and Federico Poali

74.1

72, Dats

7.8 The micromaero experiment 28

FA The macro a

7.5 Distributional impact at the macro level

ZS, Description of household social accounting matrices 268 7.9 Modelling labour market failures B 7.10 An cxample: the impact of changes in offfarm wages

icrosimulation results 77 7.13 Distributional impact at the micro level in the case

8 Child Poverty and Family Transfers in Southern Europe

Manos Matengenis, Cathnl O'Donnghns, Horacio Levy

Manuela Coromaldi, Magda Mevcader-Prats, Carlos Farinha Rodrigues,

‘Stefano Toso and Panos Toakloglow

9 Microsimulation and Normative Analysis of Public Policies

‘Amedeo Spadaro and Xiseo Oliver

9.1 Introduction 338 9.9 The model san 9.3 The data, the microsimulation model and the main

features of redistribution systems 329 9.3.1 The 1998 and 1990 Spanish redistribution systems "30 9.4 Results sz 9.5 Conelusions 436 References 387 List of Figures and Graphs 8

Sách có bàn quyễn

dealing with the e

techniques—based on the representation of individuals’ behaviour ‘ion of public policies Microsimulation when confronted with real or hypothetical changes in their eco- nomic or institutional environment—have become a much used instrument in this context for th ability to provide

a priori assessment of differing scenarios and facilitate decision making,

On the basis of extremely accurate and rich models calibrated on

simulation asa tool for the evaluation

reforms in Italy and Norway They

18)

[1 siterosintutaris a8 4 TOOL FOR Tie EVALUATION OF PusLIC FoLietes

nework for the ex ante evaluation of the

impact of different growth strategies on poverty and inequality in Mada

equilibrium effects and the wide range of household positions

ugascar They show that, given the complexity of the general

in markets for factors and goods, partial equilibrium analysis or

the use of a representative households approach would limit the robustness of the evaluation exer

Microsimutation has re led itself to he @ powerful tool in

many economic areas, and health economics are no excepti

the fourth chapter, Nun

a Badenes and Angel Lopez review some

of the microsimulation models that may be most useful in this sphere After defining the scope of microsimulation in the health economics field and looking at models developed in different countries, the authors give an example of the use of microsimuls- tion models constructed 4 la carte for solving a specific problem:

an assessment of savings generated by dual coverage through the use of private healtheare in pr

nee to the equivalent public

In our fifth chapter, Xavier Labandeira, José M Labeaga and Miguel Rodrig

distibutive and efficiency effeets of the reform of indirect taxes e7 put forward a methodology to evaluate the re~

on energy consi

to gauge the var

model is

ption They propose a microeconomic model

ergy demand This ntegrated throngh prices with a Computable General Equilibrium Model (CGE), whieh can identify the effects of a

aluation of tax reforms through microsimutation

ts from the classic hypothesis that resource alloc: Hon within a household is the work of a benign dictator In the sixth chapt

wier Ruiz Castillo and Raquel Carrasco present

the results of a se of research projects which seck to go beyond

this approach by introducing the collective model in microsinu tion with behavioural r actions The text discusses the impleme tation of collectiv rationality in a microsimulatic model with:

Federico Perali mmploy an integrated microamaero simulation

model to assess the impact of agricultural sector reforms and

trade agreements on Ktaly’s ru

population The analysis they conduct shows how macro approaches based on general economic equilibrium can be compatible with micro ap-

1 the This chapter also provides some general insights into the statistical

ation models

strates the importance of multi-country micro:

in defining supr for the fight against poverty

‘The authors present the results tional pol of EUROMOD, a research project financed by the European Union, aimed at the construction of an arithmetic microsimulation model for the then 15 EU member

countries This tool has since served to detect child poverty probe ems in Europe’s southern countries and to iclentify possible rem- edies at the

Xisco Oliver look at how this technique can also be used for nor

l rosimulation models, they explain, can help

us identify the best possible redistribution policy (

function) The appli nees of social pla

Iso shows how the prefer

be divined through the observation of a tax reform,

Sách có bàn quyễn

1 Microsimulation as a Tool for the

Eyaluation of Public Policies

Amedeo Spaclaro nurdan Sciences Economiques (Paris), FEDEA (Ma and University of the Balearic Islands (Palma de Mall

1.1 Introduction

‘The analysis of public policies (in terms of alternatives and effects)

is one of the major tasks of eco

‘The second and the third properties are necessary to optimise

tative agent approach, commonly used in the analysis of most public pol the accuracy of the analysis, The standard rej

gives more weight to the simplicity factor Without q)

its validity as a powerful approach for economic analysis, the representative agent approach is not useful to evaluate the ef

fects of public policies

the population, Imagine that you are de:

reform that determines chai

ng into account the heterogeneity of

g with an income tax,

in the consumption or Tabour

supply patterns of the population, These be ral efficets dite

' More specifically, we woint to highlight the importance of the representative agent approach

tì

[181 siterosiuLarion a8 4 TOOL FOR THE EVALUATION oF PusLIC FOLICiES

lidual

fer from one agent to another depending on his or her

characteristics andl preferences: a robust analysis of the effects of such a measure cannot be conducted without a model taking into

vidual behaviour This necessity

account the he ngencity of

has pushed applied economists towards the use of microsimula- tion models (MSMs)

MSMs are tools that allow simulution of the efficets of a policy

dividual

nple of agents (individuals or households) at

level The microsimulation approach is based on the represen

when agents face different economic and institutional frameworks

tion of individual behaviow

widely used in sciences

The simulation approach has be

like mathematics or physics, Lis use as a tool for the analysis of and support for public decision-making processes is more cent, Although it was as carly as 1957 when the seminal paper

by Oreute planted the seed of microsimulation as an instrument for economic analysis, it was only in the 1980s that the use of microsimulation tools increased substantially This was duc to the

an increasingly powerful instrument for evaluating redisuibution

and social policies

The importance of microsimulation in the analysis of public policies owes to several of its qualities

For a detailed description of the “history” and developments af microsimatation

ân economic analysis, xe Atkinson and Sutherland (1988), Merz (1091), Citra and Hanusheck (1991), Harding (1886), Gupea and Kapur (2000)

policies I certainly gives usa general idea about the performance

of the new institutional policy framework, but may also conceal

important effects of the new system depending on certain char: istics that are not so fre!

rminant of the net effects of a policy The richness of information contained in the micro dataset should be completely exploited in the s

is normally performed when running

a simulation of a reform Reforms of fiseal or social policies do not affect all agents in the same way It is not easy, for example,

to

ticipate, without a detailed micro analysis, the impact of a small change in the progressivity of income tax given that the net effect on disposable income

th sults from the interaction of ‘ome tax mechanism with other redistribution inst

such as social contributions or minimum income schemes, Even if

no behavioural responses are considered, the knowledge of who wins or loses as a result of a reform gives us a fist approximation

to the welfare effects of the measure simulated, and helps policy makers have an idea about its political feasibility

‘The third qual to its ability to fully eh

distribution mechanisms, The equity ® ney trade-offis at the acterize re

MSM should be able

3 The reduction (increase) in inequality produced by a reform of the

core of redistibution policy design And a

to provide a clear and detailed picture of its functic

vedisuibnu nm mechanism can he assessed by simply looking at the difference in the disposable income distribution of the popu- lation before and after the reform, The efficiency (inef

ency) ioural changes

effects ¢

be assessed directly by measuring beh:

) a model inchiding behavioural reactions) or indirectly by looking at changes in the disuibution of effective marginal tax

rates after reform (in a model without behavioural reactions) The size of the inefficiency will also depend on the number of

120.1 MiEROeiMiLATIBX 48 4 TOOL FOR TH EVALUATION OF rusLIC FOLICiES

ca in the datasets where necessary) at

ed with the MSM at the individu

and the partial yersus general equilibrium focus,

An MSM that replicates the institutional framework without simulating the behavic

al responses of the agent is called arith: metical These types of model simply reproduce the budget con

nt that agents face and are often used to simulate changes

in tax-benefits policies, They compute, st from the gross

income and sociodemographic eh cteristies of an agent, the disposable income of that agent under a given tax-benefit system With such models, the analysis of reform is limited to first order effects The models that go beyond arithmetical analysis inelude the simulation of age

the optimal consui

agent, With beh

can be measured and a more detailed welfare analysis can be p

you are interested in the shortierm redistribution effects of such a sure, you will simply need an MSM simulating the new budget constraint for households with children to be able to €

the new distribution of disposable income Ifyou

the long-term effects of thị

impact on fertility decisions of s

form, you will need to simulate the

hy a measure This means dat

your MSM must contain an algorithm that computes for cach year,

ach household as an endogenous va anporal beh:

then

able MSMs containing i

marriage, fertility,

ings, retirement decisions, ete

ioural decisions such,

as ageing te mporal consumption and sav-

are called dynamic MSMs in oppo-

prices and quantities are determined by the equilibrium heeveen

vane and supply in each market, has inspired the construc tion of simulation models reproducing the mechanics of the instantaneous equilibrium underlying the Walra

al Models ( have been widely used in taxation, redistribution and inte

world They

are called Computable C quilibr

m with many agents and many goods is enormous and not

ally speaking) Basie ve

MSMs, on the contrary,

'e based on many agents, but frequi

do not take into account general equilibrium effects: gross prices

and wages are fixed and changes in net prices and wages are de term ned by changes in taxation and redistribution mechanisms and are fully shifted to consumers and workers They work in a

J McROsICULATION sẽ 4 TOOL FOR THE EVALUATION oF PusLiC FOLietEs

MSN

were built in the US {models have been used in many fields The first models

and Europe for the analysis of direct and

ly, for thee : recently, given the increasing availability of household income surveys, the use

ation of redistribution and social policies Mo!

tues has been extended to the analysis of social cloped countries (LDC) The debate on the

distributional, poverty and oth:

ing public policies implemented by v 1 governments

international institutions has made the & nte andl ex post e ation of such policies a fundamental objective for economist and policy makers For this reasing number of MSMs

simulat 1g the social policies and/or the fiscal systems of LDG

It both at national (gover

cl international (multilateral and aid agencies) level The ment services, universi-

ing lists, ete.).7

Independently from the field of application and from the

equipment, physician teams, waite

nature of the MSM used, a good microsimulation policy analysis

jon of aggregate financial ef

going h

fects needs to be supported by an economic background (even yond the simple identifi

if very simple) For this reason, instead of focusing on technical

ction of an MSM," in this chapter I

details related to the constr

1 10 discuss microsimulation techniques and their economic theoretical background as @ tool for the analysis of public policies rent etal UH), Atkinson ane Seternd (1858), Mere (2901), Citra amd Hianashieck (1891), Semons and Warten (2996), Harting (1990), Redmond et al (1908), Suthevland (190, 2101), Gupta and Kapur (240), Blundiel and MaCurdy (1989) and Crccdy ao Duan (202) among others, provide detited descriptions of

the MSM built in developed countries or dircet and indice ical reforimsand

es)

redisuibution and social pokey analysis (under both sate and dynamic appro

003) present a detailed description of the

Bourguignon and Pereira da Siva

application of MSM for

See Grnher (8000), Zabinsi (199), Klein overt ailing aly it LDC

m {fora detited survey on the application of MMB tn health economies

The inerested reader can see Merz (1991), Setherland (1998) or Redmond et

si (1988)

The structure of the chapter is the following In section 1.2,

I discuss the use of arithmetical MSMs for tax incidence a

as an archetypical example of microsimulation analysis 1 will analyse the underlying theoretical f

applications to the analysis of direct

14

application to the ex ante marginal incidence of redistribution policies Section 1

mework and discuss some

axation reforms In section

present a discussion on directions for future research,

‘en over to general conclusions,

7 What are the aggregate fi

Identifying the winners and the losers of a reform and its

cial cost is the aim of tax incidence analysis, and one of the main

med with MSMs A comprehensive analy: sis would require an in-depth knowledge of individual behaviour

tasks that cant be per

terms of consumption and labour supply, and also

a good understanding of the general equilibrium mechanies of

the economy This information is not often available or can only

ed ata high cost, In such a case, could we still perform idence analysis of a given public policy? When the rẻ

form we want to evaluate is marginal.’ the answer is affirmative;

A margdmal van coma ceded sta sition difring fh he

reference one only in sal cangerin the stactural parameters An ine Vein tte UR marginal inoue bax sale 4 marginal reform, H compared (he complete replacement of income ts with, for example, # hase incomeat tax redistribution

4 1 MiEROeiMiLATIBX sẽ 4 TOOL FOR THE EVALUATION OF PusLIC FOLIeiEs

otherwise, it is negative, Regardless of the object of the reform (ai

a good theoretical framework alloy eet or in -ct taxation), the starting point of such analysis is rpret the results,

in terms of welfare

The basic underlying theory of incidence analysis (for both indirect and direet taxation) is the duality con theory To

measure household welfare gains and losses trom a reo

ne Vip, y,) and B(p, 0) as the indirect utility function and the n Ter’s

del

function of houschold ire

‘ing from the following optimisation problems:

where pris the price veetor, 9, is the household #3 income, U(x) is

0 is an exogenous utility level, 2,3) and x'(p,0) are respectively the Marshallian and Hicks

jons of the problem (1.1) and (1.2)]

The marginal incidence analysis of public policies affecting

the direct uility function

mand functions [solu

household income can be performed by using the Marshallian framework (problem 1) By differen!

the pri

malize V, to one without loss of gener

ing V{ ) holding constant

es p, we obtain that A= VA, and, as we ean always nor

to just keep the household utility constant at th

level (situation 0) given the postreform price vector (situation 7) Analytically, CV is defined as follows

sed induces @ change in prices (for

function, we can directly calculate the C¥'and thus evaluate how much real income declines/increases because of tax reform,

When we do not have an estim: jon of the expenditure fanction,

as is often the case, we can compute an approximation of the CV

where Ap, is the change in price j caused by the reform

The first term of equation (1.5) is the change in household

p household utility constant at the

expenditure necessary to

initial level without changing consumption patter

is the first order approximation of the compensating variation not considering behavioural responses (they

as in (1.5) induced by each price change:

St ferm of equation 15), we ¢ d why we

nal reforms The CV

120.1 MiEROSiMALATIBX sẽ 4 TOOL FOR THE EVALUATION OF rusLiC FoLICtEs

rectangle bapyp, The second term of equation (15) is the wiangle

‘abe Higher order terms in the Taylor expansion would capture

other words, the first order ana

good approach only when a

such a case

ysis approach of policy reform is a

nal tax reforms In

expenditure) a

Such comparison can be performed on an individual basis us-

ranking households by quantiles of gross income or

and after reform,

.l by compating them befor

nce The attractiveness of this

ing the theory of welfare domin:

approach is that it provides us with general crite!

be used to decide if an income (or expenditure) distribution a that can

is socially preferable to another under a broad class of social

welfare functions.” A comprehensive redistribution analysis of tax reforms (limited to first order effects) will be rounded off by computing inequality, poverty and polarization indexes (Esteban and Ray 1994) All the:

casily be computed by using arithmetical MSM Most of them

social welfare comparison measures can incorporate routines th automatically compute the standard

measures (Gini, Atk

poverty headcount, pov son, Theil, Kakwani, ReynoldsSmolensky, ty gap, etc.) and give # picture of the

tion curves before and nerging body of literature applying arithmed- cal MSM techniques in the analysis of tax reforms AU

Sutherland (1988), Merz (1991), Citro and Hanusheck (1991), Harding (1996), jand (1998)

present a detailed revision of MSMs and their use

ical MSMs have also been used extensively to review indirect tax incidence (

la Silva (2003) present a de-

1 description of the application of MSMs for poverty al inequality analysis in LDC

Particular attention bas been given in Ensope to the analysis

of policy reforms at domestic and European level

Atkinson, Bourguignon and Chiappori (1988), for example, analyse the redistribution

hasten the convergence of social polici

pact of a reform in which, for a give

replaced by the UK's tax system

De Lathouwer (1996) simulates the implementation of the un- employment bene

sample of French house- holds, the French tax system

scheme enforced in the Netherlands on a

sample of Belgian households, thus reflecting the importance

ge the resulting effects, Call

of the sociode phic characteristics of the population on

lan and Sutherland (1997) compare the effects of different types of fiscal and social policies on the

¢ of households in cert

well EEC countries Bourguignon,

O Donoghue, Sastre-Deseals, Spadaro and Utili (1997) use an Foe a cumplete survey on welfare dominance theory, se Lambert (1909) See also Atkinson (1970), Sen (1073), Kolm (1969), Shorrocks (1985),

Shorrocks (1988), Bourguignon (1072), Atkinson and Bourguignon (1887), Sen (1992), Bourguignon aad Eiclds (1997), Boanguignon and Chakravarty (2003) and 2henhg (1907)

8T MEROSiMULATIBX sẽ 4 TOOL FOR THE EVALUATION oF rusLIC FOLICiES

crosimulation model to

Me the effects of the enforcement

of the same child benefit scheme on the populations of France, the U

impact on the reduction of poverty in those countries

iced EUROMOD, a involving researchers from the EU 15 co

wide MSM used in several papers providing esti-

and Italy y show that this policy can have a strong

Recendy, the European Union fin

research proje

tries!! with the objective of building @ Eu pea

This model has be

mates of the distributional impact of cha ges to personal tax and transfer policy taking place at either the domestic or the Enropean level."

tical MSMs to

ns performed in

As an example of the application of arith

tax reforms, we present the results of simulati

Spadare (2005) consisting of applying 19:

redistributio

French and British

social assistance mechanisms The results are shown in table

nily benefits and

1.1, which shows the p ‘centage cha

social insurance contributions, income tax and family benefits ges in disposable income,

observed by deciles of reference households’ equivalent gross

income Enforcing the UK system on the French population

leads to a reducti nin disposable income for the lower

+ The reason for the negative effects on poor households is the redue-

deciles and an increase income for the top five de

tion of means-tested benefits On the other side of the income distribution scale rich households perfo better because of the reduction in social security contributions In the scenario based on the enforcement of the Erench tax-benelits system on

he effects are the just opposite

ar a detailed description, see Suhertanel (2001)

* Down

able at itp /wwwiserestex ack ms /emtod,

TABLE ‘1: Redistribution performance of replacing the two tax benefits

systems on domestic samples without behaviour reactions

Beces of ge Dipole etyeige ine, Mec tn/grine | MASTS

‘UK system on French sample (in French Francs)

dies, All the deciles pay less income tax but thei

the domestic instance contibutions paid by all The amount of

ns of money) than Income

pport and Child Benefits in the UK The second ( nd prob-

our simulations, aver \ge gross income of those at the bottom end

of income distribution was lower in the UK th nin France I

indices computed on the dis

posable income before and 1g in the case of the

Atkinson measure, two alternative val es for parameter a The results show that enforcement of the French tax-benefits system always reduces distribution inequality

Sample T/Benefits Gi Atkinson Index Atkinson Index

System Coefficient (e=0.1) (

UK UK mì 00816 0.1955

UK French 6 00131 0.1180 French French, thế 0/0186

French ok ov wn

Sines Spada (2005)

As shown in this ¢ mple, the simplicity of social welfare

ical MSM makes this appre

to a broad public int

analysis by aril ich appealing also

‘ested in economic policy, Unfortunately This parameter teptesents the ineality aver he nalts the tanger ơ

ig the more important for de analyst are the lowest income brackets The Atkin ilar weiies the fraction of ircomne dt can be series! who long gotta

sweats ime were equally disteihuted, Atkinson's index varies between zero and tne, For values close toon the amownt of equality isvery Inge (Atkinson 1970)

there are various potential sources of inaccuracy (see Sahn and Younger 2003), The first comes from the

retical MSMs for changes work through completely (© consumer prices This would

ssumption, often made ineide

when using a ysis, that tas,

be true only in the case of perfectly competitive markets (which

assistance, In reality, we know that tax evasion is common prac

more than in others) We also know that, for multiple reasons (lack of information, social stigma, com-

tice (in some counties

plexity of administrative procedures, etc.), some households do

ce even though they are entitled to it by

when evaluating the welfare effects of a marginal tax reform,

On the contrary, when the reform to be evaluated is specifically

J vesponses may be ignored as a first appro

designed to induce changes in agent behaviour, when reform

"About the takenp problem sec Hancock, P

sing Econametiic Mats ofBenait The byBtsh Pensioners in Mierniation Maas

a paper presented at the International Microsimulation € 1» Popalation, Ageing ad Heald: Modelling Our Fatune, held ip Canberra, Australia, ia December

ney and Sathenlanl (2003)

nference

142.1 sMterosituLarion 48 4 TOOL FOR THE EVALUATION oF rsLIC FOLIeiEs

1.3, Behavioural microsimulation

T

in particular, their application

‘sis of redistribution poli

in a different way The point is not to count how much eve

is receiving or paying but to generate a model representing the

likely behaviour of agents as a function of variables directly ak

fected by the policies being evaluated This may be done through

the estimation ofa structural ecomometric model on the cross-section

of households provided by the household survey and/or through, the calibra on of a model with a given structure so as to make

it consistent with what is obs:

corresponding to the status quo

Tax benefit models with labour supply response in developed

ved in the survey and supposedly

‘of Iabour—they also modify labour supply decisions By how much is determi: ed through a behavioural model, whieh is gen- erally estimated econometrically across households observed in the status quo

The whole behavioural MSM approach comprises three steps’

The logical economic structure is that of the testhook utility

eter

maximizing consumption An economic agent with chai

istics + chooses his/her volume of consumption ¢and his/her

Max u (e, Li258, €) st Sy, +l + NT(uh, Ly Yi 21)

In the budget constraint, 5» stands for (exogenous)

hour income, w for the wage rate and NT} } for the tx benefit schedule Taxes and benetis

ve

cteristies of the

depend on the ch

agent, his/her non-abour income and his/h + Labour income

wl They may also depend directly on the quantity of labour be: ing supplied, as in workfare programmes y stands for the pat

Suppose that a sample of agents fis observed in some houschold survey The problem now is to estimate the function Ff) above or, equivalently, the prefer nce parameters B and e, since all the other ables or tax-henefit paramet

tional difference: in labour supply This ues to a set of estimates

f, for the common preference p

cratic preference terms By defini meters and & for the idiosyn-

on of the latter, it is true For each observation in the sample that

4.1 siterosicuLarion sẽ 4 TOOL FOR THE EVALUATION OF PusLIC FoLIeiEs

*) — Fs, 110.5 yy 5 BE, 77) 10)

The change in disposable income may also be computed for eve

agent, Itis given by:

al drawbacks of the preceding model must be empha-

n is not that easy It is highly non-

nt and NT) and corner solutions at £=0, Functional forms must be chosen

for preferences, which may introduce some arbitrariness in the

whole procedure Finally, it may be feared that imposing full

‘out that simpler and less restrictive specifi

be used that considerably weaken the preceding critiques In par-

asa discrete variable that may take only

and evaluate the utility of the agent for each of these values and

the corresponding disposable income given by the budget con:

As hefore, the behavioural rule is then simply that agents

Assuming structural specification ofthe ATL) function general enon forall reforms to he tepreschted by a change in parameters

choose the value that leads to the highest level of mility However,

ed in a very general way In par

y he allowed to vary with the different

“ Such a representation is therefore the utility fimetion may be speci

quantities of labour that may be supplied, no restriction b

posed on these coefficients

1s close as possible to what is revealed by data,

Formally, a specification that generalizes what is most often found in the recent tax and supply-supply literature is the follow- ing:

L.= Dit Us f (56 acts Bas) 2 f (5s teh: Bio!) for all kaj, (1.18)

where D, is the duration of work in the jth alternative and Ujthe utility associated with that alternative and e the disposable come given by the budget constraint in (1.7)

6) = yo4 wil + NT{wD, D, ys 14) When the function /f ) is linear with respect to its common preference parameters and idiosyncratic terms are assumed to

be fd with a double exponential distribution, this model is the standard multinomial logit, It may also be noted that it encom: passes the initial model (1.7) Itis sufficient to make the following substitution

q15) This speci volves restrictions across the vari: ous supply-supply alternatives, is actually the one that

often used,

Even under its more general form, the preceding specification might stil be found t be restietive because it relies on some utility maximizing assumption Two remarks can be made in this,

respect First, it must be clear that ex ante incidence analysis can- not dispense with such a basie a

the analysis requires assumptions to be made about the way agents choose

Detowen alternatives, Assuming that agents maxi

16 1 stterosiuLarion 48 4 TOOL FOR THE EVALUATION oF PusLIC FoLIetes

wed to individual char-

supply choices observed in a survey are

acteristics, including the wage rate and the disposable

SN) That model (1.18) can be interpreted as representing utility

maximizing behaviour is to some extent secondary,

of course permits counterfacttal simulations to be implemented

in a simple way, What is more important is that this model fits the

estingly enough, the only restric:

data as closely as possible, Inte

tion with respect to this objective in the gen expression (1.13)

come effect in each all native

is the assumption that th K the ¿ÿ argMmentin /[J—đepends on đisposable income as given

chenefit schedule,

by the bi get constraint and the

The economic structure of this model thus lies essentially in the income effect Ifit were not for that property it would simply be a reduced form model aimed at fitting the data as well as possible,

In effect, the restriction that income effect must be propor:

nption tation of crosssectional differences

tional to disposable income seems to be a minimal assu

to ensure that this repres

in supply-supply behaviour may represent at the same time

various supply:supply alternatives This fectly clear that, within this framework, the

rational choice among

remark also makes p

simulated effect of a reform of the tax benefit system, NT), on individual labour supply is estimated on the basis of the eross-sec-

tional disposable income fect in the status quo

The role of idiosyncratic terms &, or &) in the whole approach

must not be downplayed They represent the unobse fed heteroge- neity of agents’ labour supply behaviour: Thus, they may be respon:

nả

al

‘ome, despite the

sible for some heterogeneity in responses to reform of taxes

might react differently toa change in disposable i

faet that these change:

ed is for the idioy

the same for all of them, All that is need-

wg thy Estimates of idiosyncratic terms result directly from the

neters, por 1." Note, however, that it is possible to use a calibration rather

They woul be stand

residualsin the discrete formulation (1.18) residuatswith specification (18) and sos likely pseudo

thân 1 estimation approach W h the for 5 some of the cork ficients B or / would not be estimated but given arbitrary values, deemed reasonable by the at lyst Then, as in the st vrd cstic mation procedure, estimates of the idiosyncratic terms would be obtained by imposing the coincidence of predicted choices under the status quo and actual choices

Iris important to emphasize that there is som ambigui about

who the agents behind the labour supply model (1.7) should be

the welfare implications of the analysis concern households,

age This makes analysis more complex

Examples of the applic

fous, A survey is provided in Blundell and MaCurdy (1999)

sly the labour supply decisions of all members of working ion of the preceding model are nu-

trates the potential of this approach, is the work of Blundell ct

al, (2000) evaluating the likely effect of the introduction of the Working Families Tax Credit (WTEC) in the UK They estimate separately discrete labour supply model for married couples

ing « sample of UK households drawn from

Resources Survey The particularity of

and single parents us

18.1 siterosiuLarion 48 4 TOOL FOR THE EVALUATION oF rusLIC FoLIetEs

public policies, Demand for schooling or health care are among them Progrsa in Mexico, Bolsa Escola in Brazil and similar eondi- tional cash transfer programmes in several other countries offer a clear example of policies that can be evaluated ex ante by behay- ioural MSMs,

MSMs to this type of policies, consider the Bolsa Escola pro- idea about the possible application of behavio

amme in Brazil, It consists of a wansfer to households whose

ely USS 45) per month, on condition that they send all their children between 6

income per capita is below 90 Reais (approxi

and L5 to school The monthly wansfer is equal to 15 Reais per child going to school but is timited to 45 RB

This may be considered as a conditional cash transfer programme becanse it combines cash transfers based on a means test and

per household

some additional conditionality—ic., having children of school

age actually going to school As the m

to school is work, this really is a labour supply problem similar to

sand Leite (2003) estimate a child labour supply discrete model on all children aged 10 to 15 in households surveyed in the Brazilian sample, PNAD Mtter estimating all coelficients of the discret

the labour supply-schooling decision, the Bolsa Escola programme has been simulated on each of the households in the PNAD The

in occupational akernative the one analysed above, Bourguignon, Ferrei

model of

results show that the programme is inde effective in reducing the number of poor children not going to school Their propor: Hion in the population of poor 10-15 children goes down from 8.9 pe

programme, Interestingly enough, the proportion of children

cent without the programme to 3.7 under the simulated

both going to school L engaging in some labou arket activ= ity tends to increase, which suggests that the programme has little effect on child labour when children are alte

Another useful result of the MSM a ly going to school

the paper is that the

Despite the appeal of this methodology, surprisingly few ap- plications are available in developing countries In most cases,

dimension

applying it only requires a structural model of sos

of he chold behaviour that per its a change (o be simulated in

Younger (9003) tases this kind of approach to analyse the consequences of reduce

Ị Volpin (2002) apply similar techniques to evaluate the effects of

one or several policy parameters, For instanc

ing uniformly the distance to school in rural Peru, Todd ai

the child schooling prog: ne Progresa in Mexico

Some limitation of the preceding approach, which has not been mentioned explicitly before, must be stressed First, this ap- proach is difficult to implement because it generally requires the estimation ofan original behavioural model that fits the policy to

be evaluated or designed, and of course the corresponding micro, data, Because of this, it is

a given county for a particular policy can be applied without ikcly that an analysis conducted in

substantial modi ation to other counuy or another wpe of policy, The methodological investment behind this approach may thus be considerable, Because of this, it must be preceded by a

rithmetical n

icrosimulation based on simpler assumptions,

Secondly, we have the fact that the behavioural approach rests

ily on a structural model that requires a minimal set of assumptions In general, there is no way these assumptions may

be tested, In the labour supply model with a discrete choice rep- resentation, the basic assumption is that net disposable income,

as given by the taxcbe efit system, is what matters for occupational decisions A reduced form model would say that the exogenous idiosyneratic determ nants of the budget constraint are what matters Econometically, the difference may be tenuous but the

plications in te ns of microsimulation are huge, Finally, the

come effects, as estic

strongest hypothesis is that cross-sectional

mated on the basis of a standard household survey, coincide with

the income effects that will be produced by the programme under study or reforms in the same In other words, time income effects for a given agent should coincide with the effect of cross-sectional

income differences Here again, this is a hypothesis that is hard

to test and yet absolutely necessary for ex ante analysis Nothing

is possible without think of would be to

The only test one