7.2 Operación del Módulo Harmonizer
7.2.5 Criterios de Valoración del Algoritmo
The two major types of microsimulation models which are widely used for social welfare
policies are static and dynamic modelling (Citro & Hanushek, 1991; Creedy et al., 2002;
Lloyd, 2003; Martini & Trivellato, 1997; Merz, 1994). Static models operate on cross-
sectional databases that provide a snapshot of the population at one time (Creedy et al,
2002:8). They typically include static “ageing” routines to bring their databases up to date or
project them into the future. Such routines re-weight the individual records to match outside
control totals for key demographic characteristics and make other adjustments for changes in
income and employment. Dynamic models operate on longitudinal databases that contain
individual histories. They “grow” their databases forward in time by applying transition
probabilities to each record for such events as birth, death, marriage, labour force status
change, and so on.
Within these two distinct model types, there are variations in handling common functions that
result from such factors as differences in client needs and in styles of the model developers.
115 The general ideas and principles of these typical models will be discussed in the sections to
follow.
Static microsimulation model
Static microsimulation models usually take a cross-section of the population at a specified
point in time and apply programme rules to the individual units to measure the instantaneous
effects of policy changes. The starting point for most static microsimulation models is a unit
record file, which provides comprehensive information on such things as earnings, size,
structure of kinship networks, age of an individual, number of children in a family, income
and transfers of a household and/or a respective tax unit, labour force status, education, and
housing status for every individual on the file.
Lambert et al. (1994) argued that static microsimulation models usually, although not always, show the so-called “first-round/morning after” effects of policy changes, before individuals
have had time to adjust their behaviour to the changes. Baroni and Richiardi (2007) noted
that these models will allow the researchers to vary the rules of eligibility or liability and
produce output showing the gains or losses (both to individuals and in aggregate) from the
policy change. However, some analysts (Baekgaard, 1996; Klevmarken, 1997; Klevmarken
& Olovsson, 1996; Symons & Warren, 1996) argued that static microsimulation models have
a comparative advantage when it comes to modelling to policy decisions.
Static microsimulation models are normally based on household micro-data to estimate the
revenue cost distributional effects of tax and benefit policy changes. They are invaluable for
the design and evaluation of policy reforms. Static models allow holding constant many
variables so that a study can focus on the aspects of interest. In addition, they also help to
separate the direct effects of changing tax and social security policy on incomes from all the
116 underlying influences on income and from the other characteristics and behavioural patterns
of a particular population (Sutherland, 2000).
Static ageing: Static microsimulation models use static ageing techniques, which include changing certain variables on the original micro-data files to produce a file with the
demographic and economic characteristics expected in the future year. According to Merz
(1994), technically, static ageing procedure involves applying adjustment factors to account
for changes in the population structure, inflation, the distribution of income and changes in
policy rules. Static ageing accounts for changes in the population structure by assigning
weights in such a way that the external control totals represent forecasts rather than
describing the situation in the year in which the survey was conducted. Merz further argued
that in static ageing, the temporal adjustment of the demographic arrangement is reached
mainly by re-weighting the existing information according to exogenous, given aggregate
data of another time period. After re-weighting a sample, one micro-unit will then represent
a certain number of particular units in the whole population. According to him, in static
ageing, the relations among the variables of each micro-unit are generally maintained. An
overall structural amendment is expressed by a changed weight of each micro-unit,
respectively, of each association of micro-units (e.g., families, households). Thus, the cross-
section after simulation (t + v, v=0, 1, 2...) contains the equal number of micro-units (n) as
the cross-section before simulation (t).
Merz (1994:6) maintained that
a static ageing procedure is relatively well-suited for short- and medium-range forecasts provided it can be assumed that the characteristics of the population under examination do not change rapidly. If the demographic structure essentially changes,
117
which is particularly likely in the longer run, the use of 'dynamic ageing' in the framework of dynamic microsimulation will be more appropriate.
However, Baroni and Richiardi, (2007) argued that, in some cases, static microsimulation
models are used to make short-term forecasts (one or two years ahead, for instance), under
the assumption that only small changes to the fundamental structure of the population of the
economy or individual behaviours would occur within such a short time span.
A number of authors (Creedy et al, 2002; Merz, 1994; Redmond et al., 1998) agreed that static microsimulation models have relative advantages over the dynamic microsimulation
models. For example, Redmond et al. (1998) noted that, in static microsimulation, using the
extensive information about the characteristics of age, sex, income, wealth and so on, and the
current legislation regarding tax and benefits, it is possible to simulate the budgetary and
distributional effects in different policy regimes. Creedy et al. (2002) showed that static
microsimulation models are easy to build and maintain. They added that they are easy to use
and quick to run and, therefore, can be accessed by a wide range of users. For Merz (1994),
the primary advantage of using static microsimulation models is that they are less expensive.
Even if substantial modules (such as education, labour force, income, and expenditure
modules) are integrated into a static simulator in the same way as into a dynamic model, the
static approach is less expensive because time-consuming simulation of demographic
processes with interactions among members of different micro-unit associations such as
marriages and market processes are omitted. Baroni and Richiardi (2007) observed that
microsimulation models are simple to create and can offer a cost-efficient tool for certain
types of policy analysis, and the model can answer the question regarding what the variation
in variable and for household h at time t + 1 would be if policy rules r were applied,
everything else remaining the same.
118 A large number of static microsimulation models are currently used across the world by
government departments and research institutes, such as, for example, FAMSIM (Austria),
GLADHISPANIA (Spain), SPSD/M, (Canada), STINMOD (Australia), TAXMOD, PSM and
POLIMOD (UK), TRIM, MATH, STATS, OTA and TAXSIM (US) and EUROMOD in
Europe.
It is argued here that static microsimulation is appropriate and applicable in terms of
quantifying the financial impact on child poverty of current government policies, to illustrate
the impact of various taxes and benefit policies that could be implemented. For example,
static microsimulation can clearly help to show how an increase in the Child Support Grant
affects income distribution, as well as the budget. It is also further argued that the strength of
micro-models, as opposed to macro-models, is the possibility of getting detailed information
about the distributional effects of policy changes. All child-related policy reforms can be
measured in terms of "winners" and "losers" and changes in measurements like the Gini co-
efficient can easily be calculated.
Dynamic microsimulation model
Dynamic microsimulation models are primarily used to analyse economic, social and public
policy that involves a time dimension. These models help to project the population forward
in time so that the sustainability and performance of public policies like pensions, long-term
care and education financing can be evaluated. Moreover, it also takes into account methods
and results from other kinds of analyses, such as behavioural micro-econometrics.
Dynamic ageing simulates transitions at the individual level and thus produces hypothetical
sets of panel data. Simulated transitions include changes in demographic characteristics,
educational qualifications, labour market patterns and income mobility. Merz, (1994) noted
that in dynamic microsimulation, individually based ageing by survival probabilities is
119 applied for each micro-unit of the whole sample. In addition to a micro-unit's survival, a
child (or children) could be born within a simulation period or a family and household
situation might be altered by marriage, divorce or other occurrences. Thus, individual
dynamic demographic ageing will change the size of the cross-section under study; in
general, the cross-section after simulation (t+1) does not contain the equal number of micro-
units (n and associations thereof) as the cross-section before simulation (t). According to
Lambert et al. (1994) and Merz (1994), the distinguishing feature of dynamic
microsimulationmodels is the ageing of the original unit records on the basis of probabilities
of different real-life events occurring. This allows the original population to be projected
forward in time, while maintaining detailed information on the individuals within the
simulation. In this regard, Merz (1994) stated that static microsimulation ageing is done by
re-weighting. In a dynamic microsimulation model, each micro-unit is aged individually by
an empirically based survival probability.
Baroni and Richiardi (2007) pointed out that the most standard process included in dynamic
models used for economic policy evaluation are demographic changes, marriage and
household formation, educational path, health status, labour market status, taxes and benefits,
as well as savings and wealth.
The overall advantage of dynamic microsimulation is the presence of information on the
entire life cycle of each “sample” member. This is of interest for life-cycle analyses of
earning patterns with both income from former rewarding employment and, afterwards,
income from pension security systems.
The number of micro-units involved in the life cycle (dynamic microsimulation) is reduced to
those micro-units of real full life-cycle interest. Correspondingly, the expenses are only a
fraction of a comparable full cross-section simulation (Merz, 1994). He added that dynamic
120 microsimulation is individually forecasting all micro-units of a given sample, and dynamic
life-cycle or longitudinal microsimulation creates a cohort of “synthetic” micro-units with
complete life cycles from birth to death. Thus, a dynamic microsimulation does not forecast
the characteristics of real sample units but the assigned characteristics of synthetic micro-
units. All characteristics of a synthetic micro-unit are determined by the behavioural and
institutional modules. A complete life cycle of a synthetic micro-unit is simulated period by
period and may include information on spouse and children if the micro-unit has married and
brought up children during its lifetime. A number of simulated life cycles then constitute a
sample of a certain cohort.
Examples of dynamic microsimulation models include DESTINE (France), DYNACAN
(Canada), DYNAMITE (Italy), DYNASYM and PENSIM (US), LIAM (Ireland), MIDAS
(New Zealand), MOSART (Norway), PENSIM II (UK), and SESIM (Sweden).
The advantages of using dynamic microsimulation models include helping to simulate inter-
temporal issues requiring historical information and allowing the inclusion of future
behavioural adjustments of the population to either policy reforms or to changing economic,
demographic or social scenarios. However, their limitations include large data requirements,
large building and maintenance costs and lack of an agreed validation methodology.
Similarities and differences
Both static and dynamic microsimulation types have relevance to the application of income
distribution analysis, and the benefits of any of them cannot be ruled out. However, the
conceptual difference between static and dynamic microsimulation is that the static model
will take a cross-section of the population at a specified point in time and apply the tax-
benefit rules, for example, to see the effects of policy changes. Usually, the impact shown by
these models is also known as the “first round” effect, showing the gainers and losers,
121 whereas the dynamic models differ in the technique by which they simulate the effects of
time on population. They can model an individual’s transition based upon the occurrence
probabilities of real-life events, thus allowing the previous population to be projected in a
future time period.