Professional Documents
Culture Documents
Economic Modelling: P Eter Benczúr, G Abor Katay, Aron Kiss
Economic Modelling: P Eter Benczúr, G Abor Katay, Aron Kiss
Economic Modelling: P Eter Benczúr, G Abor Katay, Aron Kiss
Economic Modelling
journal homepage: www.journals.elsevier.com/economic-modelling
Assessing the economic and social impact of tax and benefit reforms: A
general-equilibrium microsimulation approach applied to Hungary☆
Peter Benczúr a, Gabor Katay a, *, Aron Kiss b
a
European Commission, Joint Research Centre (JRC), Italy
b
European Commission, Directorate-General for Employment, Social Affairs and Inclusion (DG EMPL), Belgium
A R T I C L E I N F O A B S T R A C T
JEL Codes: We present a general-equilibrium behavioural microsimulation model designed to assess long-run macroeco-
H22 nomic, fiscal and distributional consequences of tax and benefit reforms in a small open economy. We apply the
H31 model to reforms adopted in Hungary between 2008 and 2013, a period characterised by intense reform activity
C63 in the wake of the financial crisis. Innovations of the model include the treatment of labour supply responses of
Keywords: individuals, affecting both participation (extensive margin) and work intensity (intensive margin), and the
Behavioural microsimulation incorporation of general-equilibrium effects by a parsimonious macroeconomic model. Our results suggest that
Linked micro-macro model Hungary's flat tax reform may have boosted output by improving work incentives of high earners (along the
Flat tax intensive margin), but reduced work incentives and employment levels of low earners. We find that policy
Structural reforms measures introduced since 2008 substantially increase income inequality. Moreover, the contribution of changes
after 2010 is about four times that of the changes before 2010.
☆
Part of the work was done when Benczúr, Katay and Kiss were at Magyar Nemzeti Bank (the central bank of Hungary. This research has been conducted before Kiss started to work
at the European Commission. The views expressed here are of the authors and do not necessarily reflect the official view of MNB or the European Commission.
* Corresponding author. European Commission, Joint Research Centre (JRC), Via E. Fermi 2749, 21027 Ispra, Italy.
E-mail address: gabor.katay@ec.europa.eu (G. Katay).
1
First simulation results using a preliminary version of the model are presented in a non-technical paper assessing the 2011 tax changes in Hungary (Benczúr et al.,
2011).
https://doi.org/10.1016/j.econmod.2018.06.016
Received 5 December 2017; Received in revised form 20 April 2018; Accepted 17 June 2018
Available online 27 July 2018
0264-9993/© 2018 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
P. Benczúr et al. Economic Modelling 75 (2018) 441–457
margins by merging the elasticity of taxable-income literature (see, for It is important to note that we find quantitatively large behavioural
example: Gruber and Saez, 2002; Saez et al., 2012) and the discrete responses and fiscal feedbacks. For example, different personal income
choice labour supply literature (van Soest, 1995; Aaberge et al., 1995; tax cut scenarios with a static fiscal cost (decrease in fiscal revenues) of
Bargain et al., 2014). In particular, the intensive margin is interpreted as 0.3% of GDP can lead to a 0.3–3.2% increase of GDP. Similarly, the quasi-
work intensity or work effort, and the labour supply response at this neutral static fiscal cost of the 2010–2013 reform package can turn to a
margin is calibrated by using estimates of the taxable-income elasticity sizable gain in the long run, ranging up to 5.18% of GDP. This is not
for Hungary by Benczúr et al. (2013), and Kiss and Mosberger (2015). because of particularly large labour supply elasticities, though the results
The extensive margin is the participation decision, and the labour supply are clearly sensitive to the underlying estimates of the elasticity of
response at this margin is implemented using the discrete choice esti- taxable income. Instead, it is partly because of the high starting level of
mations of Galuscak and Katay (2017).2 non-linearity in the Hungarian tax system, and partly because the reforms
The second important feature and contribution is that labour supply we consider are not marginal: the 2010–13 tax reforms, for example,
shocks, resulting from individual behavioural responses to tax and have cut the top marginal tax rate from 40 to 20%.
transfer reforms, are fed into a simple model of a small open economy. The rest of the paper is structured as follows. Section 2 reviews some
This linked micro-macro modelling approach has important advantages. closely related literature. In Section 3, we give a detailed description of
Shifts in labour supply will, over time, lead to changes in wages, corpo- the principles of the model: we describe the data, the main elements of
rate profits and thus a change in the demand for capital. By embedding the Hungarian tax-benefit system, our approach to modelling labour
the labour supply model in a macro model, we can introduce labour supply adjustment, and the main elements and some limitations of the
demand, capital demand and capital supply (not necessarily perfectly macro model in which the microsimulation is embedded. Section 4
elastic) and take account of these general-equilibrium feedback effects. presents results of simulations for hypothetical reform scenarios, while
More importantly, this approach enables us to assess how changes in the Section 5 presents and evaluates the actual Hungarian reforms between
corporate side of taxation affect the macroeconomy and the labour 2008 and 2013. Section 6 discusses various robustness checks. The final
market. Still, the macro model is parsimonious and transparent: it con- section offers some concluding remarks. Some details are presented in
sists of an aggregate production function, ensuring that input prices equal the Appendix.
their marginal products, and a capital supply curve.
Both key features of the model fit into recent tendencies in micro- 2. Related literature
simulation modelling (see, e.g., Bourguignon and Spadaro, 2005;
Aaberge and Colombino, 2014; for recent surveys). Section 2 explains in 2.1. Modelling labour supply responses
more detail how these features of the model relate to the literature.
Using our model, we perform two sets of simulations. The first one Early work on incorporating behavioural responses includes work by
explores the impact of personal income tax reductions that are identical Aaberge et al. (2000), Blundell et al. (2000), and Creedy and Duncan
in their static cost (under no behavioural change) but different in their (2002). In a seminal paper, Immervoll et al. (2007) simulated the effects
structure: an across-the-board rate cut, a plain flat tax, a flat tax with a of two hypothetical welfare-reform scenarios in 15 European countries
zero rate at the bottom, and a flat tax with an employee tax credit scheme based on the EUROMOD microsimulation model. They stress the
at the bottom. The second one evaluates actual Hungarian policy mea- importance of considering the behavioural labour-supply response not
sures between 2008 and 2013. just at the intensive but also at the extensive margin, since the evaluation
It is important to stress that we do not model the dynamic response of of a welfare reform hinges crucially at the extensive-margin response.
the economy. All our exercises are comparative statics. Still, working By now, the extensive margin adjustment has become a standard
with different elasticities of capital supply, one can get a sense of a short ingredient of labour supply estimations, assuming a discrete choice from
run (perfectly inelastic supply, thus fixed capital) and a long run response a small set of possible hours worked (including inactivity, part-time, full-
(elastic supply). Our baseline calibration features a quite elastic supply of time and overtime, or sometimes a more refined set). The most frequently
capital; thus we usually refer to our results as long run effects. adopted methodology is based on Bargain et al. (2014). Aaberge and
The results from our hypothetical policy simulations show that while Colombino (2014) presents a large but still admittedly incomplete review
a cut in the marginal tax rate of high-income individuals may boost of many analyses following similar approaches (Table 1 of Aaberge and
output, it does not have a significant employment effect. On the other Colombino, 2014).
hand, programs like the Employee Tax Credit (ETC) do have a significant To extend this approach, we want to incorporate also the taxable
employment effect. income elasticity approach to the intensive margin adjustment, on top of
Simulating the effects of recent actual policy changes, we find that the the possibility of full-time or overtime work (this is similar to the
measures passed in the two years before the 2010 elections increase long- approach taken by the Congressional Budget Office, see for example
run employment and GDP, but there is no significant adjustment at the Harris and Mok, 2015). As argued by Feldstein (2000), concentrating
intensive margin (effective labour). In contrast, measures passed in the only on hours worked has some practical and conceptual limitations.
two years after the 2010 elections, including the introduction of the flat Data on hours worked might be imprecisely recalled and thus mis-
personal income tax, produce a large gain at the intensive margin, but reported by survey respondents. It can also miss effort and work in-
employment increases only due to cuts in unemployment benefits. Both tensity, or career concerns. For example, more effort might lead to a
policy packages substantially increase income inequality, the contribu- faster career development, higher wages and thus income, without
tion of the changes after 2010 being about four times that of the changes necessarily observing higher hours worked. And unlike for hours worked
before 2010. The cumulative change has the potential to place Hungary for primary earners, there is evidence of a positive, though moderate,
at the median of the EU-27, as a marked change from its original ranking response to marginal tax rates (for example, Gruber and Saez, 2002; Saez
as the country with the 6th most equal income distribution. This pre- et al., 2012; or Kiss and Mosberger, 2015 and Benczúr et al., 2013 for
diction based on the model has since been broadly borne out by official Hungary). At the same time, some caution is warranted when inter-
data. preting taxable income elasticities as labour supply elasticities, since they
may also reflect tax evasion. Section 3.4 discusses the relevance of the tax
evasion margin for our Hungarian estimates.
2
We also performed similar simulation exercises based on the gains to work Taxable income elasticity estimations typically require administrative
approach of Benczúr et al. (2012) and Benczúr and Katay (2017). The results, tax data. These are less easily accessible than household budget surveys,
which are very close to the ones presented in the paper, are available upon and the two data sources cannot be merged because of data protection
request. issues. We resolve this limitation by adopting a two-step approach. The
442
P. Benczúr et al. Economic Modelling 75 (2018) 441–457
first step is the extensive margin adjustment, which is interpreted as a guides the calibration of the DGSE model, and pins down the size of the
shift in the probability of a full-time job3 (part-time work is very limited shock entered into the model. The implied evolution of wages and
in Hungary, as discussed in Galuscak and Katay, 2017). The intensive employment is then fed back to microsimulation to calculate the distri-
margin, on the other hand, represents work effort, conditional on butional and fiscal impacts, but the labour supply response is not
working. This way, a shock to the aggregate effective labour supply may reassessed.
come from intensive adjustment if, with all employment probabilities The parsimony of our macro model has the advantage that it is easy to
held constant, some individuals change their work effort (conditional on see which assumptions and parameters are responsible for the nature of
employment) or extensive adjustment if, work effort held constant, the general-equilibrium feedback effects. This is in sharp contrast with the
probability of work increases for some individuals. complexity of most CGE or DSGE models. The parsimony of the macro
model also minimizes potential theoretical and practical inconsistencies
between the micro and macro models, without the need for aggregating
2.2. Linking micro and macro models
heterogeneous agents into a representative one (as proposed by Magnani
and Mercenier, 2009; and also adopted by Barrios et al., 2016). Finally, it
Linking micro and macro models for policy analysis also fits into
allows the complete integration of the two components: they are
current trends in microsimulation modelling. These approaches typically
repeatedly run in an iterative process until full convergence.
use Computable General Equilibrium (CGE) models to take account of the
One important cost of parsimony is that shifts among sectors are
general-equilibrium effects (see Davies, 2009, for a recent survey).4 CGE
ignored. Such shifts, however, are not central for assessing the reaction of
models are complex tools allowing the researcher to model the con-
labour supply to changes in the tax and transfer system, and its re-
sumption and labour supply behaviour of one or more representative
percussions for the whole economy. The full integration of the Hungarian
types of households, and model the complex interrelations of wages and
economy into the European Economic Area reinforces the point that a
prices in several sectors of the economy. In most cases, the linked
simple small open economy macro model is appropriate.6
CGE-microsimulation method is used to assess the effects of trade
openings, or other large macroeconomic shocks, on the income distri-
3. Description of the model
bution in developing countries. There is by now also a growing number of
applications of micro-macro models to questions of tax policy.5
3.1. Data
There are recent works on the interactions of labour supply and
general-equilibrium feedbacks, which are very close to our focus.
The microsimulation model partly builds on a tax-benefit model
Aaberge et al. (2004) analyse how endogenous labour supply interacts
created by Benedek et al. (2009). It runs on the 2008 wave of the
with long-term fiscal sustainability in Norway; Arntz et al. (2008) use
Household Budget Survey (HBS) compiled by Hungary's Central Statis-
such an approach to evaluate a hypothetical welfare reform in Germany;
tical Office. Though we have access to more recent waves as well, we
while Fuest et al. (2008) and Peichl (2009) simulate the effects of a hy-
decided to stick to the last pre-crisis year, where an assumption of the
pothetical German flat-tax reform. Jongen et al. (2014) develop a
economy being “in steady state” is plausible. The data set provides
detailed behavioural microsimulation tool for analysing Dutch tax and
detailed information on nearly 20,000 individuals (including informa-
benefit reforms. Barrios et al. (2016) extend the EUROMOD model with a
tion on their labour market status and income) living in nearly 8000
behavioural and macroeconomic part, and apply it to various policy
households. Our analysis relies strongly on household characteristics
changes in Italy, Belgium and Poland. Ayala and Paniagua (2016) adopt a
when modelling labour supply and eligibility for social transfers. For this
similar approach (with no macroeconomic part) to Spain.
reason, we could not base our analysis on tax return data, since these do
Our approach differs from these studies by using a fully integrated,
not include information about household characteristics (not even the
simple, single-sector macro model to calculate general-equilibrium ef-
number of children).
fects. It was outlined in Benczúr et al. (2011), then further developed in
While the HBS is a representative survey of households living in
Benczúr et al. (2012), and eventually extended to the Czech Republic
Hungary, the income distribution of individuals observed in the data set
(Galusc
ak and Katay, 2017) and Slovakia (Horvath et al., 2017). Besides
does not exactly match the official tax data. As is typical of survey data,
contributing to the scant literature on linked micro-macro modelling in
the top of the income distribution is underrepresented. We solve this
transition economies, our approach to micro-macro modelling keeps the
issue by including a wage-correction stage before the simulation takes
macroeconomic model parsimonious and computationally easy.
place. This is done by multiplying the wage income of individuals in the
In a closely related, yet markedly different approach, Barrios et al.
HBS by a percentile-specific factor to match the wage income distribution
(2016) establish a link between the EUROMOD microsimulation model
in administrative data – except for the bottom quintile, in which most
and QUEST, the European Commission's DSGE modelling suit. In their
individuals have no or very little wage income.7 For most of the income
approach, however, integration is incomplete. The micro model first
distribution, the differences between the two data sets are not large (less
than 10% for the 2/3 of the individuals). The difference, however, grows
3 bigger in the top 20% of the income distribution, particularly in the
We do not address the issue of unemployment: our implicit assumption is
that once an individual has decided to be active on the labour market, the
highest percentiles (see Fig. A1 in Appendix 1).
incidence of unemployment is an independent random event, with a fixed After the correction of wage income, the static (no behavioural
exogenous, group-specific probability. This consideration makes it important to response) fiscal implications of the model were validated by tax experts
use pre-crisis data for our extensive margin estimation and model simulation. In of the Hungarian Central Bank.8 In particular, we compared the fiscal
particular, we use 2008 household data, the last pre-crisis year, where the predictions of the model and experts for a number of reform scenarios,
assumption of the economy being "in steady state" is more plausible and hence
the simplification of an exogenous unemployment probability seems acceptable.
Horvath et al. (2017) extends a similar modelling framework with endogenous
6
unemployment (search and matching), but their results are not affected strongly As Davies (2009, p. 60) puts it: "In the case of national subregions, or
by this modification. countries embedded in free-trade areas, it can be argued that microsimulation
4
There are a few papers that do not employ a CGE framework to close the may adequately be combined with pure macro models. That is, CGE modelling
model. As one example, Cameron and Ezzeddin (2000) use a regional may not be necessary."
7
input-output model to assess indirect effects of regional and federal tax and The effect of this procedure is similar to a procedure in which survey data
transfer policies in Canada. are statistically matched to individuals from a database of tax returns.
5 8
An early example is Slemrod (1985), who focused on the incidence and the We thank Zsolt Lovas and Gabor P. Kiss for their help in validating the
effects on portfolio choice of a hypothetical flat-rate income tax in the US. model.
443
P. Benczúr et al. Economic Modelling 75 (2018) 441–457
including increases in social security contributions and adjustments of Appendix 3 reports the conditional marginal effects relevant for the
the personal income tax system. The predictions were quite similar. present study (Table A1), and briefly enlists the set of individual char-
acteristics employed in the estimation.9
3.2. The main elements of the Hungarian tax and benefit system To assess the order of magnitudes of these extensive margin effects, it
is instructive to compare them to the “consensus” 0.25 value of aggregate
The main elements of the Hungarian tax and transfer system are the (steady state) net wage elasticity reported by Chetty et al. (2012). The net
following (see Appendix 2 for a more detailed description). Hungary has wage marginal effects in Table A1 are somewhat larger, but they are not
an individual, rather than family-based, personal income tax (PIT) sys- directly comparable: the reported marginal effects indicate the effect of
tem. In 2008, the PIT was progressive, with a top marginal rate of 40% one percent increase in net wage on the probability of being active (or on
with a number of tax brackets. There was no basic allowance (income the participation rate) in percentage points, as opposed to the elasticity
exempt from taxation), but in 2008 workers could benefit from a rela- measures in Chetty et al. (2012) indicating the percentage change in
tively generous Employee Tax Credit (ETC), which reduced the PIT on employment to the same shock. To produce the equivalent of the exercise
the minimum wage to nearly zero. The tax credit was gradually with- by Chetty et al. (2012), one needs to increase the net wage of all in-
drawn as the taxable income increased, increasing the marginal effective dividuals by 1% and simulate its employment or participation effect. The
tax rate of medium income earners. Besides this tax credit, and some resulting 0.28% increase in participation implies an elasticity of 0.28,
other minor ones (e.g. for charitable giving), the main tax credit in the quite in line with the consensus value.
PIT system was the family tax credit. To validate the impact of transfer reforms on the extensive margin, we
Social contributions, used to cover a pay-as-you-go pension system report the result of an additional labour supply simulation exercise. We
and the costs of a national healthcare insurance, are paid by both em- simulate the effects of the cut in maternity benefits comparable to the one
ployers and employees. The employers' rates are higher than the em- that took place in 1995. K€ oll}
o (2009) found a positive, though often
ployees' rates. Given the comparatively high contribution rates, the tax insignificant, effect of this measure on the activity of mothers with in-
wedge, especially that of low-income earners, has been high. fants, while Szab o-Morvai (2013) found a negative, though delayed, ef-
The benefit system includes unemployment benefits, family benefits fect of the reversal of the reform. Our estimates imply a 0.09% increase in
and other welfare benefits. In 2008, the insurance-based job-seekers’ total employment, which corresponds to a roughly 1.02% increase in
benefit had a maximum duration of nine months and was followed by an employment of the target group (women with children).10 This is
additional three months of job-seekers' assistance. Family benefits consistent with a positive but statistically not always significant treat-
include a universal benefit, depending on the number of children. ment effect.
Hungary also has a system of generous parental leave benefits, both
insurance-based and universal. Finally, the Hungarian welfare system
had a means-tested welfare benefit of last resort, and several smaller 3.4. Modelling labour supply at the intensive margin
benefits.
Labour supply response at the intensive margin means that in-
dividuals change their work intensity (hours, work effort, etc.) after a
3.3. Modelling labour supply at the extensive margin change in their marginal or average tax rate. The general view is that
such a behavioural response exists for high-income earners (mostly in
Labour supply response at the extensive margin means that an indi- response to marginal rate changes) but less in the lower ranges of the
vidual may exit the labour force if the financial gains to market work income distribution.
decrease (and vice versa). According to the literature, this type of In our model, the intensive margin response is calibrated based on
behavioural adjustment is more significant in the case of secondary estimations by Kiss and Mosberger (2015) of the elasticity of taxable
earners, low-income earners, women with children, young workers and income with respect to the tax rates. They employ administrative tax
the elderly (see, e.g., Meghir and Phillips (2010) for an overview). data11 and estimate the compensated elasticity of taxable income with
In our model, the labour supply response at the extensive margin is respect to the marginal net-of-tax rate to be approximately 0.2 for high
modelled based on the estimations of Galuscak and Katay (2017). They earners. We apply this elasticity to the top fifth of wage earners;
pool eleven consecutive waves of the HBS to estimate a standard discrete lower-income households are assumed to have no labour supply response
choice structural labour supply model for Hungary (and also for the
Czech Republic). In this framework, originally proposed by van Soest
(1995), individuals choose between a discrete set of hours of work 9
Galuscak and Katay (2017) also report results for a "reduced form" or
yielding different utilities. As part-time work is relatively rare in Hungary "gains-to-work model", in which the probability of being active depends on the
(less than 5% before the crisis), only two labour-market states are net income an individual can achieve when out of work and the "financial gains
considered: full time work (or seeking full time work) and inactive. to work", i.e. the change in disposable income due to taking up a full-time job
Assuming that the utility Uij that individual i derives from choosing (which equals the net wage minus lost transfers). Such an approach also requires
full-time work (j ¼ 1) or inactivity (j ¼ 0) is represented by a random an integrated microsimulation tool, to assess the counterfactual values of in-
utility model of the form Uij ¼ Vij þ εij , where Vij is a deterministic term come: the transfers that the currently active would get if they quit or lose their
job, and the transfers the currently inactive would lose if they took up work. The
and εij is an i.i.d. type I extreme value distributed random term, the
estimated elasticities for this model are very similar to those obtained with the
probability of being economically active is given by: standard structural model. We performed all simulations presented in this paper
using this alternative labour supply model. The simulation results are very close
Pi;j¼1 ¼ P½Vðci1 ; 1 l1 ; Zi ; θ; η1 Þ Vðci0 ; 1; Zi ; θÞ > εi0 εi1 :
to our baseline findings. For the sake of brevity, these results are not presented
Here Vij is conditional on the individual's net disposable income if he in the paper, but are available from the authors upon request.
10
works full time ðci1 Þ or if he stays out of the labour force ðci0 Þ, leisure According to 2008 values in our database, total employment is 4,080,525. A
0.09% increase adds 3672 to the employed population, which is 1.02% of all the
ðð1 l1 Þ or 1, with time endowment normalised to 1), a vector of indi-
mothers with infants (a group of about 360,000).
vidual characteristics Zi and preference parameters θ, and fixed costs of 11
In fact Kiss and Mosberger (2015) use the same tax returns data that we use
work η1 . With only two labour market states, the participation decision is for correcting the wage income distribution in the HBS dataset (see Section 3.1).
a simple logit model that can be estimated using standard maximum This means that the extensive and intensive margin labour supply responses are
likelihood techniques. simulated on nearly identical income distributions: the intensive margin on the
Galusc
ak and Katay (2017) find that the labour supply response at the tax data, while the extensive margin on an adjusted HBS data, matching the
extensive margin is strongest for low-wage groups and married women. distribution of the tax data.
444
P. Benczúr et al. Economic Modelling 75 (2018) 441–457
at the intensive margin (Bakos et al., 2008 and its updated estimates exogenous probability (see footnote 3 for more discussion).
reported in Benczúr et al., 2013 provide estimations that support this Finally, the shock to the aggregate effective employment is fed into
type of dependence of the elasticity on income). Robustness of the results the macro model, which calculates general-equilibrium effects on wages
with respect to these elasticities is investigated in Section 6. and the capital stock. Based on the general-equilibrium change of the
A note is in order about the interpretation of changes in taxable in- wage level, the microsimulation is repeated. This iterative process is
come as a labour-supply response. Some studies (especially in the U.S.) repeated until convergence; that is, until the general equilibrium of the
found that part of the response in taxable income to taxation is due to tax economy is consistent with the reform-induced labour supply shock.
optimisation (through itemised deductions) and has, therefore, little to
do with additional real economic activity. For Hungary, however, there 3.6. General equilibrium feedbacks
are good reasons to view changes in taxable income as a labour supply
response. First, itemised cost deductions are negligible in the Hungarian The general-equilibrium macro model is a simple model of a small
personal income tax system. Correspondingly, the existing estimations on open economy. Capital and labour are paid their marginal products, ac-
Hungarian data are lower than taxable-income elasticities estimated in cording to a constant-returns-to-scale production function. This mini-
the US. Second, Kiss and Mosberger (2015) present additional indirect malistic, very general structure makes the model tractable, transparent,
evidence that supports this interpretation. Firstly, their estimated elas- nesting the steady state of many complex DSGE models as well. Since the
ticity does not differ significantly between those individuals who have labour supply shock comes from microsimulation and we are not inter-
wage income only and those who have multiple sources of income ested in the change in sectoral consumption patterns, the general-
(dividend, entrepreneurial income, etc.). Plausibly, the former group has equilibrium model does not detail the household side. In the following,
less opportunity for tax avoidance and evasion. Secondly, they find no we sketch the main ingredients of the model, with further details pre-
support for income shifting in the tax-reform episode they investigate. sented in Appendix 4.
The production function of the representative firm exhibits constant
elasticity of substitution (CES).14 The profit-maximisation problem of
3.5. The aggregate labour supply shock firms can be formulated as:15
445
P. Benczúr et al. Economic Modelling 75 (2018) 441–457
The first equation represents labour demand, ensuring that wages are increase somewhat after an increase in aggregate labour supply. While
equal to the marginal product of labour. The second equation is capital capital accumulation will mitigate these effects, it will not neutralise
demand, ensuring that the return on capital is equal to its marginal them completely.
product. The third equation defines the capital-labour ratio, where the The model is calibrated using data from national accounts, fiscal
labour supply shock is the result of microsimulation (reflecting both the statistics and previous estimations (Katay and Wolf, 2004 in particular).
exogenous labour supply response to the policy shock and the endoge- Appendix 5 contains all the details of the procedure.
nous response to the change in wages). The last equation is capital
supply.
3.7. Limitations
Fig. 1 illustrates the same mechanism graphically. It depicts a simple
labour and capital market equilibrium schedule. The microsimulation
Based on a limited set of ingredients (the estimated behavioural
model pins down the slope of labour supply ðLs Þ in general, and its shift
elasticities and the small open economy framework), the model is able to
after a tax or benefit reform in particular. The production function pa-
give an assessment of the long-term effects of changes of the tax and
rametrizes the demand for labour and capital, and the shift of the latter
transfer system to the macroeconomy, the income distribution and the
due to potential corporate tax changes. Finally, the small open economy
government budget. However, to ensure its tractability and transparency,
assumption adds the supply of capital.
one has to accept some simplifications.
To interpret the comparative statics implied by this stylized macro
model, let us first look at the case of perfectly elastic capital supply ðη ¼
(1) The model is suitable for comparative statics exercises. The dy-
∞Þ. In this case, the domestic rate of return is pinned down by the in-
namics of the adjustment path from pre-reform to post-reform
ternational rate and is thus constant ðbr ¼ 0; r ¼ rÞ. The capital-labour
steady state equilibrium is not modelled.
ratio must also stay constant which, in turn, implies that wages will
(2) Since the model is supply-driven, the consumption-savings deci-
also stay constant (perfectly elastic labour demand). We thus get the
sion of households is not modelled. The consumption decision
usual result that with perfectly elastic capital supply the capital stock
affects our results in only one way: it affects the fiscal effects
adjusts to shocks, so that the capital-labour ratio and factor prices can
through the VAT. Our simplified assumption that all disposable
return to their previous equilibrium values. For example, if there is a
income is consumed by households admittedly results in the
positive labour-supply shock following a tax cut, capital accumulation
(short-run) overestimation of the VAT effect of policy measures.
will follow in identical proportion, so that a new equilibrium is reached
(3) The model is not closed on the side of government. Budget balance
with an unchanged capital-labour ratio.
is not enforced either directly or by an assumption that higher
The graphical analysis of Fig. 1 gives some indications of the under-
debt results in higher interest rates paid on government debt. This
lying dynamic adjustments (without implying any quantification).
0 simplification is innocuous if the policy measures analysed are
Following a labour tax decrease (from τw to τw ), the labour supply curve
approximately budget neutral or small in magnitude; or the
ðL Þ is shifted to the right, but the labour demand curve stays the same.
s
objective is to perform a technical projection (no other policy
Transfers and other non-labour incomes ðTÞ, as well as capital taxes ðτK Þ change). Otherwise, in case of a fiscal loosening (a worsening of
remain unchanged. As a consequence of the tax cut, gross wages ðwÞ start the government budget), which typically produces a positive
to decrease, the rental rate of capital goes above the required rate and macroeconomic effect, balancing the budget would require a fiscal
capital ðKÞ starts to flow in: the demand for capital ðK d Þ shifts to the right. tightening, partly undoing the positive effects. Hence, the model
This, in turn, shifts the short-run labour demand curve ðLd Þ to the right predictions are overly optimistic in the case of measures that
and the gross wages start to reverse (dashed lines). In the long-run, the weaken the position of the government budget, and vice versa.
0
equilibrium goes from ðw* ; L* Þ to ðw* ; L* Þ on the labour market and from (4) The search-and-matching mechanisms on the labour market are
0
ðr; K * Þ to ðr; K * Þ on the capital market. not modelled. If a policy measure changes the equilibrium un-
If capital is calibrated to be imperfectly elastic (which will be the case employment rate of any demographic or skill group, the model
in our baseline), wages will decrease and the return on capital will will not take this into account. It might be, for instance, that
shortened eligibility for unemployment benefits, in addition to
446
P. Benczúr et al. Economic Modelling 75 (2018) 441–457
Table 1
Hypothetical personal income tax scenarios.
Across-the-board PIT cut Flat tax (20.1%) 2 tax rates (0% þ 31%) 1 tax rate (26.2%) þ tax credit
(1) (2) (3) (4)
Baseline
Gini 27.18 27.34 32.31 28.39 29.13
P90/P10 3.12 3.15 3.80 3.25 3.28
P90/P50 1.76 1.76 2.03 1.82 1.85
P50/P10 1.78 1.78 1.87 1.78 1.77
Notes: The upper panel of the table shows percentage changes of macroeconomic variables in levels and the dynamic fiscal effects expressed as a percentage of the GDP
(positive numbers indicate an improvement of the government balance). The bottom panel shows how various income inequality measures is expected to change in the
long-run following the reform. The baseline inequality measures correspond to 2008. *Dynamic effects include labour supply reaction of individuals as well as general-
equilibrium macroeconomic effects.
strengthening job-search incentives, reduces the success rate and labour supply at the extensive margin, the results show that the highest
average quality of matches between job openings and the unem- positive impact on employment can be obtained when the PIT cut is
ployed. In this case the long-run employment and the output effect concentrated at low-skilled, low incomes groups.
of the transfer cut will be overestimated by the model. Horvath On the other hand, the three other tax reforms create better incentives
et al. (2017) extends a similar modelling framework with a search for top earners, especially the pure flat-tax regime with the lowest top
and matching block, and estimates the model for Slovakia. Over- marginal tax rate. As a result of the significantly lowered top marginal tax
all, their results are not affected strongly by the incorporation of a rate, the top 20% individuals increase their labour intensity so that
job search mechanism. aggregate effective labour increases by 3.38% (flat-tax scenario, column
2), 1% (two-rate scenario, column 3) and 1.55% (single-rate scenario
4. Simulation results: hypothetical labour tax reforms with ETC, column 4). As a comparison, a one-percentage point across-
the-board tax cut of the first scenario creates little additional incentive
The key features of the model are best shown through a set of simple, for top-earners: the 0.39% increase of effective labour comes almost
easily tractable hypothetical simulation exercises. Table 1 shows the ef- exclusively from the adjustment at the extensive margin.
fects of four hypothetical personal income tax (PIT) cut scenarios with Since the only macroeconomic shock here is the labour supply shock,
the same direct, short-run fiscal cost of about HUF 100 billion (or 0.37% GDP and the capital stock adjusts almost perfectly in proportion to the
of the GDP). All four scenarios are defined as changes relative to the 2008 effective labour supply. The increased GDP also improves public finances
PIT system (see Section 5.1 and Appendix 2 for the description of the tax in the long run relative to the static effect (0.37% of GDP).
and transfer system in 2008 in Hungary). In sum, the social planner faces a trade-off between seeking the
In the first scenario (“across-the-board PIT cut”, column 1), all three highest possible employment effect (in which case, the PIT cut should
PIT rates are reduced by 1 percentage point. In the second scenario mostly target low-skilled, low incomes groups) and the highest impact on
(column 2), an entirely flat tax system with a 20.1% rate is introduced GDP (in which case, the marginal tax rates for the highest earners matter
(the employee tax credit, ETC is also eliminated). In the third scenario the most). Among the four scenarios presented, the two extreme cases are
(column 3), the PIT system is replaced by a 2 tax rates system: a 0% tax represented by the across-the-board PIT cut (scenario 1) and the intro-
rate applies to income up to the minimum wage, and a rate of 31% above duction of a flat-tax system (scenario 2).
that. In the fourth scenario (column 4), a single basic tax rate applies to Moreover, this trade-off is complemented by social concerns: while the
all taxpayers (26.2%), but there is an ETC that makes the minimum wage across-the-board PIT cut (scenario 1) has little effect on income inequality,
PIT-free and is phased out roughly similarly to the actual 2008 ETC. The the transition from a progressive tax system towards a pure flat-tax system
exact parameters of the PIT scenarios are summarised in Table A2 in (scenario 2) increases considerably inequality (see the last bloc of
Appendix 6. Table 1). These inequality measures, resulting from a dynamic simulation,
Results from the four simulations show that different ways of correspond to equivalised disposable household income per capita. It is
reducing the PIT burden have starkly different aggregate effects. The important to note that our reported inequality measures for the 2008
ranking of scenarios depends on the criteria used. On the one hand, an baseline are obtained using predicted transfers and – for non-employed
across-the-board tax cut (scenario 1) has the highest positive impact on individuals – wages, to ensure their comparability with the simulated
employment (a gain of 0.27%).16 Unlike the three other scenarios, the values for the hypothetical scenarios and (in the next section) with the
across-the-board PIT cut decreases the tax burden of low- and middle simulated 2010 and 2013 measures. This leads to an overestimation of the
income individuals, thereby influencing their financial gains to work 2008 Gini coefficient (which was 25.2, according to Eurostat data).
positively. Consistently with previous findings from the literature on the The employment, GDP and the social impact of the two other sce-
narios are between the two extreme cases. It is interesting that the two-
rate system without ETC (column 3) is dominated by the single-rate
system with ETC (column 4) in practically every aspect. Both the two-
16
In fact, the employment effect is only positive in the first scenario. For a rate system and the ETC keep the average tax rate zero at the mini-
larger tax rate reduction package, the employment effects would turn positive mum wage, but the ETC is less costly. This relative budget surplus can be
for the other scenarios as well, but the ranking would remain the same.
447
P. Benczúr et al. Economic Modelling 75 (2018) 441–457
used for lowering marginal tax rates for high earners, which creates 5.2. Assessment of the fiscal effects of the reforms
additional incentives at the intensive margin and therefore stimulates the
economy. At the cost of a small increase in income inequality, the ETC Table 2 shows the simulated effects of the 2008–2010 and 2010–2013
allows achieving significantly higher GDP in the long-run, while total policy packages. Looking at the fiscal effect of both packages first, the top
employment remains unaffected. panel of Table 2 shows that while both periods saw a net cut in PIT and
employer contributions, and an increase in the effective consumption tax
5. Tax and benefit reforms in Hungary between 2008 and 2013 rate, this was accompanied by different measures in other parts of the tax
and transfer system. Measures between 2008 and 2010 had a negative
In this section, we present a detailed analysis of the fiscal, macro- overall static fiscal effect of about 2.15% of GDP (HUF 576 billion),
economic and distributional impacts of the actual changes of the tax and although savings not accounted for in our simulations counterbalanced
benefit system which took place between 2008 and 2013 in Hungary. these measures to a considerable extent (including measures related to
Before discussing the quantitative results, we give account of the main old-age pensions). In contrast, measures in the period 2010–2013 had an
context and elements of fiscal policy developments in Hungary during the approximately neutral static fiscal effect, with cuts in the unemployment
period. benefit, increases of employee contributions and taxes on sales and
consumption making up for foregone PIT revenue. As both the reforms in
the period of 2008–2010 and in 2010–2013 gave a boost to the economy
5.1. Policy context and main changes
(see the next sub-section), the fiscal situation improves considerably in
the long-run (dynamic impact).
This paper focuses on tax and benefit reforms adopted between 2008
and 2013.17 This period was characterised by intensive reform activity as
5.3. Assessment of the macro effects of the reforms
a response to long-standing structural challenges and the financial crisis.
The 2008 financial crisis found Hungary in a vulnerable position,
The middle panel of Table 2 shows that the policy packages of both
including large public and private external debt levels and foreign ex-
periods differ in their macroeconomic effect. Mainly due to the cut in
change exposures (see, e.g., MNB, 2010). The Government had no fiscal
employer contributions, the changes between 2008 and 2010 increase
room after a period of high budget deficits (averaging about 8% of GDP
long-run employment by 1.24% and GDP by 0.97%. Since top marginal
between 2002 and 2006 and reaching 9% of GDP in 2006) and increasing
rates remained unchanged during this period, there is less adjustment at
public debt (surpassing 70% of GDP in 2008). Experiencing financial
the intensive margin. In contrast, the combination of cuts in corporate
market stress in October 2008, Hungary received financial assistance
taxes, PIT and unemployment benefits in the period 2010–2013 are
from the European Union in the framework of a Balance of Payments
estimated to increase long-run employment by 2.12% and the long-run
assistance programme, in conjunction with loans from the IMF, sup-
level of GDP by 5.18%. Here, the employment effect is almost entirely
ported by a stand-by arrangement, and the World Bank.
due to the cuts in unemployment benefits. Since the PIT cuts are
We analyse reforms adopted between 2008 and 2013, divided into
concentrated at high incomes, they increase effective labour supply, and
two sub-periods based on a change in Government. Reforms between late
thus GDP, but employment to a much smaller degree. The employment
2008 and early 2010 were adopted by Governments supported by a
effects of the elimination of the ETC and the introduction of the targeted
Parliamentary majority of Socialists and Liberals. Reforms in the period
employer contribution relief roughly cancel out.
2010–2013 were adopted by a Conservative government that came into
power in April 2010.
The main reforms between 2008 and 2010 included a tax shift from 5.4. Effects on inequality
employers' contributions to the VAT, and a tax relief focused on middle-
income earners. The main reforms between 2010 and 2013 included the The bottom panel of Table 2 shows that both sets of tax policy changes
introduction of the flat personal income tax, other tax shifts and cuts to increase income inequality: the long-run Gini index is estimated to in-
unemployment benefits. Appendix 7 provides a more detailed descrip- crease by somewhat less than a point (from 27.18 to 28.07, a 3.27%
tion of these changes, while the exact parameters used in the simulations increase) due to the changes up to 2010, and by a bit less than four
of actual changes between 2008 and 2013 are documented in Table A3 in additional points (to 32.04, an additional 14.14% increase) due to the
Appendix 6.18 changes up to 2013.
From the point of view of labour supply incentives, the most signifi- This is a truly significant increase in inequality. The same cumulative
cant tax-benefit reform over the period was the introduction of the flat percentage increase in Hungary's official (Eurostat) 25.2 Gini coefficient
personal income tax. Fig. 2 shows the effect of tax reforms between 2008 would have taken the country from the sixth lowest position (after
and 2013 on the tax wedge, defined as total taxes and contributions Slovenia, Slovakia, Sweden, the Czech Republic and Denmark) to the
payed as a percentage of total wage cost at various levels of income. A median (29.71). Actual developments have broadly borne out this pre-
number of conclusions can be drawn. First, Hungary was a high-tax diction: 2013 and 2014 Gini values were 28.3 and 28.6, both at the
country throughout the period. The tax wedge is above the 2008 EU eleventh lowest position in the EU. Of course, actual inequality has been
average for all income levels between 2008 and 2013. Second, the shaped by other elements, beyond those we have modelled. However,
changes in the period 2008–2010 reduced the tax wedge at all income many analyses have confirmed that policy changes have significantly
levels. Finally, the flat-tax reform introduced between 2010 and 2013 has contributed to increased inequality in Hungary (De Agostini et al., 2016;
improved incentives at high earnings levels (reducing the tax wedge European Commission, 2017, pp. 59–62).
because of the reduction in the tax rate), but dampened incentives at low The other three inequality measures in Table 2 also show growing
earnings levels (increasing the tax wedge, mostly because of the elimi- income inequality but they also help to identify which part of the income
nation of the employee tax credit). distribution the change comes from. The measure P50/P10 shows that
the gap between the median earner and individuals around the 10th
percentile of the income distribution grows only a little as a consequence
17
We take into account measures entering into force between 2010 and 2013 of the reforms.19 There is more significant growth in the gap between the
that were announced before August 2012.
18
A detailed description of Hungarian labour market policies and benefits, as
19
well as their changes, is available in the various editions of the annual series Our percentile ratios are similar to the ones published by Eurostat for 2008:
"The Hungarian Labour Market”, published by the Hungarian Academy of according to Eurostat data, the P90/P50 ratio was 1.70, while the P50/P10 was
Sciences. 1.76.
448
P. Benczúr et al. Economic Modelling 75 (2018) 441–457
Fig. 2. Tax wedge in Hungary and the EU at various levels of earnings, single worker without children, 2008, 2010 and 2013.
Source: European Commission and OECD tax and benefits database.
449
P. Benczúr et al. Economic Modelling 75 (2018) 441–457
second exercise focuses on the robustness of the results to the main adjustment are “switched off”. Here there is no change in labour supply at
calibrated macro parameters. all (since it is perfectly inelastic); but changes of taxes on corporate in-
The first row of Table 3 repeats the baseline simulations of the come, sales and consumption affect the macroeconomic equilibrium.
2010–2013 policy package presented in Table 2. In contrast, the second GDP increases by 0.42%.
row shows the results of the simulation if both margins of individual In the third row, individual adjustment at the extensive margin is
Table 3
Robustness to labour supply elasticities and to macro parameters: the changes of 2010–2013.
Effective Empl. Capital GDP Average gross Disp. Dynamic* fiscal impact (% of Gini index
labour stock wage income GDP) (2010 ¼ 28.07)
No behavioural change (only 0.00 0.00 1.19 0.42 0.12 1.13 0.54 31.45
macro)
Only extensive response 2.00 2.07 2.85 2.30 0.27 0.02 1.08 31.14
- Baseline elasticities 2.89 0.00 3.64 3.15 0.31 0.69 1.30 32.37
- With income effect 0.57 0.00 1.69 0.96 0.15 0.85 0.71 31.40
- Higher elasticity with 3.73 0.00 4.35 3.95 0.37 1.17 1.54 32.50
income effect
- rK/Y ¼ 0.3 5.04 2.11 5.28 5.11 0.55 1.87 1.90 32.03
- rK/Y ¼ 0.4 5.07 2.13 5.57 5.27 0.39 2.00 1.89 32.03
- sigma ¼ 0.7 4.76 2.14 5.50 5.02 0.27 1.88 1.87 31.95
- sigma ¼ 0.9 5.03 2.11 5.24 5.10 0.56 1.85 1.86 32.03
Notes: Results from dynamic simulations. See the text for parameter values of the alternative intensive-margin elasticities. *Dynamic effects include labour supply
reaction of individuals as well as long-run, general-equilibrium macroeconomic effects.
450
P. Benczúr et al. Economic Modelling 75 (2018) 441–457
“switched on”. The policy package is estimated to increase long-term output, wages and effective labour) react more sensitively the more
employment by 2.07%, thereby raising GDP by 2.3% and improving elastic the supply of capital is. It is apparent that our baseline calibration
the dynamic fiscal balance. of η ¼ 15 is closer to the perfectly elastic case than to the perfectly in-
In the following three rows, focus is on the adjustment at the intensive elastic case. Simulation results also show that, within a plausible range,
margin: on the one hand, this is a controversial issue in the literature (see, the capital-income share parameter does not affect results very sensi-
for example, the discussion by Meghir and Phillips, 2010); on the other tively. Finally, the last block of Table 3 shows how the results are affected
hand, the impact of the 2010–2013 tax reform crucially depends on the by moving the factor-substitutability parameter. The results show that
elasticity of taxable income. In this exercise, adjustment at the extensive moving away from the extreme Cobb-Douglas case makes capital adjust
margin is “switched off” but adjustment at the intensive margin does take more sensitively to changes in its marginal product.
place. The first of these rows uses the “baseline elasticity” as taken from
the study of Kiss and Mosberger (2015). The study finds that 7. Conclusion
higher-income earners are somewhat responsive to the marginal
net-of-tax rate (the estimated elasticity is 0.2). In the simulation model In this paper, we present a simple, tractable, yet versatile general-
this elasticity is applied to the top 20% of earners. Individuals with lower equilibrium behavioural microsimulation model, designed to assess
incomes are assumed not to adjust along the intensive margin. The point long-run output, employment, fiscal and distributional consequences of
estimate of the income effect in the study of Kiss and Mosberger (2015) is reforms to the tax and the transfer system. We estimate and calibrate the
sizable but very imprecisely estimated, therefore, the baseline income model to Hungary, a country characterised by significant structural
effect was chosen to be zero. challenges and intensive reform activity in the wake of the 2008 financial
The results from this “baseline elasticity” simulation support the view crisis. We simulate two sets of scenarios. First, we analyse four different
that most of the stimulative effect of the 2010–2013 policy package ways of tax relief in the personal income tax system. Second, we analyse
comes from labour supply adjustment at the intensive margin. Due actual changes of the tax and transfer system between 2008 and 2013.
mainly to the cuts in the marginal tax rate of high earners, the policy The results of the first exercise show that the desirability of different
package is estimated to add, under these assumptions, 2.73–2.89 per- tax policy scenarios depends greatly on the criterion used. If the main
centage points to the level of GDP and to the effective labour supply in objective is employment growth, policymakers should keep (or decrease)
the long run (compare row 4 – “baseline elasticities” – to the second row). the average tax burden of low incomes as low as possible, since the lower-
The next row of Table 3 (“With income effect”) repeats this analysis wage, mostly low-skilled, groups — where the large part of the inactivity
including a non-zero income effect. The parameter chosen in the simu- is concentrated in most countries — are the most sensitive to changes in
lation is 0.5 (applied to the top 20% of the income distribution); this is their financial gains to work. The most efficient way of keeping the
the point estimate by Kiss and Mosberger (2015) of the parameter (the average tax rate low for low-income earners is the Employee Tax Credit
elasticity of reported income with respect to the average net-of-tax rate). (ETC): similarly to a system with a zero lower tax rate, the ETC keeps the
The simulation shows that an income effect of this magnitude dampens average tax rate low at low incomes, but it imposes a smaller burden on
the stimulative effects of the policy package to about one-fifth: GDP in- the budget. This budget surplus can be used for various other policy
crease only by 0.96% instead of 5.18%. measures such as, for example, lowering marginal tax rates for higher
The next row of Table 3 (“Higher elasticity with income effect”) earners.
shows results from a simulation where the elasticities describing the Our simulation exercises underline that the tax and the transfer sys-
adjustment at the intensive margin are taken from the other available tem should be considered simultaneously. For example, in countries
Hungarian study (Bakos et al. (2008)). For higher income earners, they where some transfers are taxed as income, a tax rate cut (or increase) will
estimated a higher substitution elasticity than Kiss and Mosberger (0.34 also affect positively (negatively) the net transfers one can get at zero
instead of 0.2), with an income effect of 0.27. In later follow-ups, as hours worked, which will, ceteris paribus, decrease (increase) individuals'
indicated by Benczúr et al. (2013), the marginal tax rate elasticity was incentives to work.
found to be closer to the Kiss and Mosberger (2015) results. Nevertheless, If the main objective is to boost labour productivity and GDP,
it is still a reasonable alternative to postulate an even larger substitution different policy measures should be applied. As individuals at the top of
effect (also coinciding with the “consensus” intensive margin elasticity of the income distribution are relatively responsive at the intensive margin,
Chetty et al., 2012) and a moderately negative income effect which leads decreasing the marginal tax rates improves incentives at the top of the
to an almost-zero uncompensated elasticity. We find that the estimated income distribution and boosts effective labour and GDP in the long-run.
effects of the 2010–2013 policy package with these alternative elastici- In the second exercise, we evaluate the impact of actual tax and
ties are similar to, but somewhat higher than, the results estimated with transfer changes in the two years before the general elections in 2010 and
the baseline elasticities. The two parametric differences between the three years after. We find that the packages of the two governments
estimations of Bakos et al. (2008) and Kiss and Mosberger (2015) have during this period have different macroeconomic effects. The changes
the opposite effect, partially cancelling out each other. between 2008 and 2010 increase long-run employment and GDP; but
In the second set of robustness checks, we analyse how the simulation there is no significant adjustment at the intensive margin. In contrast, the
results depend on the most relevant calibrated macro parameters. The changes between 2010 and 2013 produce a large gain at the intensive
bottom three blocks of Table 3 present results from three types of alter- margin of labour supply, but employment is expected to increase only
native simulations. First, the elasticity of capital supply (parameter η) is due to cuts in the unemployment benefit. Both policy packages are found
modified: the corner cases of perfectly elastic capital supply ðη ¼ ∞Þ and to increase income inequality in the long run (the changes after 2010
perfectly inelastic capital supply ðη ¼ 0Þ are analysed. Second, the capital about four times more than the changes before 2010). The cumulative
income share parameter is modified from the baseline value of 0.35 to 0.3 change has the potential to move Hungary to the median of EU member
and 0.4. Finally, the production-function parameter governing the sub- states from its original ranking as the country with the 6th most equal
stitutability between labour and capital (parameter β) is modified. This income distribution among the EU-27.
parameter is given in our model as β ¼ σ1 σ : While we chose σ ¼ 0:8 as It is important to keep in mind that our model is solely based on
our baseline, as a robustness check we ran simulations with σ ¼ 0:7 and financial incentives and does not take into account some potentially
σ ¼ 0:9. important features of the economy such as labour market frictions and
All three types of changes in the parameters can ultimately be skill mismatches. For example, the shortening of the unemployment
described by how they affect the reaction of the capital stock to changes benefit period will make the individuals want to find a job earlier and
in the marginal return of capital. Table 3 confirms that capital (and thus therefore they increase their labour supply, but a shorter job-search
period may also impair the quality of employee-employer matches.
451
P. Benczúr et al. Economic Modelling 75 (2018) 441–457
This effect is not incorporated in the model and, as a result, we might model of a small open economy.
overestimate the employment effect of such a measure. Furthermore, This linked micro-macro modelling approach allows us to trace
such a policy may have adverse effects on income inequality and poverty. general-equilibrium feedback effects through labour demand, capital
At a general, methodological level, our microsimulation model has demand and capital supply. This approach also enables us to assess the
two key features. First, it is behavioural, which means that it takes into macroeconomic and labour market effects of changes to the corporate
account the labour supply response of individuals at the intensive and side of taxation. Still, the macro model is parsimonious and transparent:
extensive margin. This is done by merging the elasticity of taxable- it consists of an aggregate production function and a capital supply curve,
income literature and the discrete choice labour supply literature. Sec- ensuring that input prices equal their marginal products and that capital
ond, the labour supply shocks that result from individual behavioural supply is elastic. The overall simplicity and tractability of the model
responses to tax and transfer reforms are fed into a simple neoclassical makes it easily extendable to fiscal analysis in other countries as well.
APPENDIX
Personal income tax. Hungary has an individual, rather than family-based, personal income tax (PIT) system. In 2008, the PIT was progressive, with
three tax brackets. The lowest rate was 18% and applied to income up to approximately the average yearly income; a rate of 36% applied to income
above that up to the pension contribution ceiling; and a rate of 40% applied to income above that. There was no basic allowance (income exempt from
taxation), but in 2008 workers could benefit from a relatively generous Employee Tax Credit (ETC), which reduced the PIT on the minimum wage to
(nearly) zero. The tax credit was gradually withdrawn (at a rate of 9%) as the taxable income increased, increasing the marginal effective tax rate of
medium income earners. Besides this tax credit, and some other minor ones (e.g. for charitable giving), the main tax credit in the PIT system was the
family tax credit.
Social security contributions. Social contributions, used to cover a pay-as-you-go pension system and the costs of a national healthcare insurance, are
paid by both employers and employees. The employers' rates are higher than the employees' rates. Individuals paid social security contributions at a rate
of 17% up to the pension contribution ceiling and 7.5% above that. Given the comparatively high contribution rates, the tax wedge, especially of low-
income earners, has been high.
Unemployment benefits. In 2008, the insurance-based job-seekers’ benefit had a maximum duration of nine months. After this benefit, individuals
with a relatively long work history were eligible for an additional three months of job-seekers' assistance.
Family benefits. Family benefits included a universal benefit, the family supplement. Its level was set per children. Higher rates applied to single
parents, and there was a special rate for children with serious health impairment or disability. Hungary also had a system of generous parental leave
benefits, both insurance-based and universal.
Other welfare benefits. The Hungarian welfare system had a means-tested welfare benefit of last resort. The means test was family-based. The benefit
was conditional on cooperation with the public employment services. The Hungarian welfare system also included a number of smaller benefits,
including those for individuals with a reduced capacity to work, with care responsibilities, as well as some local welfare benefits (including a housing
and heating benefit).
452
P. Benczúr et al. Economic Modelling 75 (2018) 441–457
Appendix 3. Estimation results on the labour supply adjustment along the extensive margin
As explained in Galuscak and Katay (2017), the estimated translog random utility model is a simple logit model that can be estimated using standard
maximum likelihood techniques:
Pi;j¼1 ¼ P Vi;j¼1 Vi;j¼0 > εi;j¼0 εi;j¼1 ;
where Pi;j¼1 represents the probability that individual i is economically active; εij are i.i.d. type I extreme value distributed random terms; Vij is a positive
deterministic term that represents the mean utility across observationally identical agents when alternative j (1 if active, 0 if inactive) is chosen.
Vij is expressed as Vij ¼ αi logcij þ βi ðlogcij Þ2 þ γ ' Zi þ ηi , and includes an alternative-specific log of total disposable income (if the individual works full
time and in the absence of remunerated work), their squared terms (logcij and ðlogcij Þ2 ), and their interactions with the following individual controls
(taste shifters in αi and βi ): gender, age, family status, a dummy for the employment status of the husband or wife, and the presence of dependent
children in the household. Zi includes education dummies (five categories) interacted with age and age squared, gender, attending full-time education,
household head, family status (single, married living together, widow(er), divorced), household membership status (husband, wife, child, single, other),
a dummy indicating whether the husband or wife is employed, the presence of dependent children in the household, household size, county (NUTS3)
and year dummies. Finally, ηi is an i.i.d. random term.
The estimated conditional marginal elasticities are presented in Table A1. The second column (“reduced-form”) present the results from an alter-
native specification (for details, see Galuscak and Katay (2017).
Table A1
Conditional marginal effects at the extensive margin.
Structural Reduced-forma
453
P. Benczúr et al. Economic Modelling 75 (2018) 441–457
The demand for labour and capital is determined by firm behaviour. The production function of the representative firm exhibits constant elasticity of
substitution (CES).20 The profit-maximisation problem of the firm can be formulated as:21
1 r
max αK β þ ð1 αÞLβ β ð1 τs Þ wð1 þ τW ÞL K:
1 τK
Here, τs is the effective tax rate on sales (representing, in the baseline, the effects of the local business tax), w is the gross wage, τW is the rate of
employer-side social security contributions (equivalent to a payroll tax), τK is the effective tax rate on capital and 1rτK is the net user cost of capital.
The presence of the effective sales and corporate tax rate enables us to assess the implications of changes to non-labour taxation. The changes in these
effective measures enter into our simulations as expert estimates, based on proposed legislation (see Table A3 for specific values).
The model is closed by the equation that determines the aggregate supply of capital. Capital is provided by an international capital market. Its supply
is modelled in a reduced form: K b ¼ ηbr ; where η is the elasticity of capital supply K with respect to the after-tax rate of return r (and x
b denotes the
percentage change of variable x).
Let us log-linearize the model around the equilibrium. Deriving the first order condition for labour (L), introducing the capital-labour ratio k ¼ K=L
and rearranging yields
∂ 11
: ð1 αÞLβ1 αK β þ ð1 αÞLβ β ð1 τS Þ ¼ wð1 þ τw Þ
∂L
11 wð1 þ τw Þ
ð1 αÞ αk β þ ð1 αÞ β ¼ :
1 τS
x ¼ dx
Rearranging further and then obtaining the log-differential (using that b x
, where x is the starting equilibrium value of variable x), we obtain
β
1 wð1 þ τw Þ 1β
αk β þ ð1 αÞ ¼
1 α 1 τS
1β
β β !
β 1 β wð1 þ τw Þ 1β wð1d
þ τw Þ
αβk bk ¼
1α 1β 1 τS 1 τS
1β
β β
1 1 1 wð1 þ τw Þ 1β
b
k ¼ þ τw 1 d
w þ 1d τS :
αk
β
1α 1β 1 τS
∂ 11 r
: αK β1 αK β þ ð1 αÞLβ β ð1 τS Þ ¼
∂K 1 τK
β11 r
αk β1 αk β þ ð1 αÞ ¼ :
ð1 τK Þð1 τS Þ
Rearranging further and then obtaining the log-differential:
1β
β
1 r
αk β þ ð1 αÞ ¼ k 1β
α ð1 τK Þð1 τS Þ
1β
β 1β
β !
β 1 β r d r β
αβk bk ¼ þ βk kb
α 1 β ð1 τK Þð1 τS Þ ð1 τK Þð1 τS Þ
β 1β
β 1β
1
1 1 1β 1 r 1
b
k ¼ br 1 d
τ K 1 d
τ S þ b
k
αk α 1 β ð1 τK Þð1 τS Þ α
β
0 11β
β
1β
1 @w 1 þ τW A
β
1 1
b
k¼ b þ ð1 d
w þ τW Þ ð1d
τs Þ
αk
β
1α 1β 1 τs
0 11β
β
1β
β
11β
1
11 1 r
b
k¼ β @ A br ð1 d
τK Þ ð1d
τs Þ þ b
k
αk α 1β 1 τK 1 τs α
20
Previous estimations of factor demand and the substitutability between labour and capital rejected that the Cobb-Douglas production function can be used. See
Katay and Wolf (2004) for an estimation of the demand for capital.
21
We write the firm's problem in net terms, so it does not contain the value-added tax (VAT). The VAT is, however, included in the net wage entering the labour
supply decision.
454
P. Benczúr et al. Economic Modelling 75 (2018) 441–457
b b b
k¼K L
b ¼ ηbr
K
The first equation ensures that wages are equal to the marginal product of labour, while the second equation ensures that the return on capital is
equal to its marginal product. The third equation defines the capital-labour ratio, where the labour supply shock is the result of microsimulation
(reflecting both the exogenous labour supply response to the policy shock and the endogenous response to the change in wages).22 A balanced budget
restriction is not imposed (see a discussion of this point in Section 3.7).
Appendix 5. Calibration of the macro model
The parameters of the model are calibrated based on previous estimations, and simple statistics taken from national accounts.
1) Taxes on capital (corporate income tax and other, less significant, corporate taxes) and consumption (VAT and other, less significant, consumer
taxes) are calculated as effective tax burdens on aggregates taken from national accounts. The initial (2008) effective tax rate on capital is τK ¼
0:073, while the initial effective tax rate on sales (calculated as total tax revenue divided by GDP) is τs ¼ 0:0174. The effective tax rate on con-
sumption is τVAT ¼ 0:182:
2) The elasticity of substitution between capital and labour in the production function is chosen based on estimations of Katay and Wolf (2004): β ¼
ðσ 1Þ=σ ¼ ð0:8 1Þ=0:8 ¼ 0:25.
3) The net user cost of capital is computed as described in Katay and Wolf (2004). Its average value for the period 2005–2008 is 0.155.
4) The capital income share is calibrated based on averages from national accounts:
r
K Y ¼ 0:35:
ð1 τK Þð1 τs Þ
Using the production function and the net user cost, the capital income share pins down also the value of k, the capital labour ratio.
5) Parameter α is obtained by rearranging the first order condition of profit maximisation with respect to capital:
1β1 r
α αK β þ ð1 αÞLβ K β1 ð1 τs Þ ¼
ð1 τK Þ
β
r
ð1τK Þð1τs Þ
K 1β r
⇒α¼ ¼ 0:429
Y ð1 τK Þð1 τs Þ
6) One can choose the units such that L ¼ 1, thus K ¼ k, and the first order condition of profit maximisation with respect to labour determines the wage
w.
7) It is left to calibrate η, the elasticity of capital supply with respect to the after-tax return on capital. The two extreme cases are perfect capital mobility
ðη ¼ ∞Þ and perfectly inelastic supply of capital ðη ¼ 0Þ. The former is a reasonable assumption in the long run, supported either by a small open
economy assumption (the rental rate is set by the world rate) or a closed economy Ramsey model (the rental rate is determined by the rate of time
preference); while the latter is probably a good description of the very short run. While the model can be run with any of these values, the baseline
results in the paper are based on a quite elastic capital supply ðη ¼ 15Þ.
Table A2
Parameters of hypothetical tax scenarios from Table 1
Baseline (2008) Across-the-board tax cut Flat tax Two rates, no ETC One rate, with ETC
22
VAT and labour taxes directly affect labour supply.
455
P. Benczúr et al. Economic Modelling 75 (2018) 441–457
Table A3
Parameters of actual tax scenarios from Table 2
Appendix 7. Main changes to the tax and benefit system between 2008 and 2013
Personal income tax. In the period 2008–2010, the main PIT reform reduced the tax burden of medium-income earners by reducing the number of tax
brackets from three to two. In the period 2010–2012, a flat PIT of 16% was gradually introduced, resulting in a significant tax relief for high earners.
Meanwhile, the Employee Tax Credit was eliminated and the Family Tax Credit was significantly increased, especially for families with at least three
children.
Social security contributions. In the period 2008–2010, a significant tax shift took place, reducing employers' contributions from 32% to 27% (financed
by increasing the VAT from 20% to 25%).23 In the period 2010–2013, employees' contributions were somewhat increased (while the VAT increased to
27%). Finally, as a partial compensation for the elimination of the Employee Tax Credit, targeted rebates to employer social security contributions were
adopted under the Job Protection Act in 2012.
Unemployment benefits. The main unemployment benefit reform was adopted in the period 2010–2013, reducing the maximum duration to 3 months,
reducing the benefit cap and the rate at which employment history is translated to benefit entitlement.
Family benefits. The level of the universal family benefits remained nominally frozen during the period of study. A reform in the 2008–2010 period
shortened the maximum duration of one type of maternity leave benefits, and tightened the eligibility conditions of another, but the reform was revoked
after the 2010 elections, before it could have taken effect.
Other welfare benefits. The main last-resort welfare benefit went through a number of reforms in the period of the study, including changes to its
name. In the period 2008–2010, it changed from a top-up benefit (topping up a family's income to a certain level taking into account family compo-
sition) to a lump sum benefit. In the period 2010–2013, its level was reduced, and its conditionality tightened.
References Arntz, Melanie, Boeters, Stefan, Gürtzgen, Nicole, Schubert, Stefanie, 2008. Analysing
welfare reform in a microsimulation-AGE model: the value of disaggregation. Econ.
Modell. 25, 422–439.
Aaberge, Rolf, Dagsvik, John K., Strom, Steinar, 1995. Labor supply responses and
Ayala, Luis, Paniagua, Milagros, 2016. Behavioural Microsimulation of the Impact of In-
welfare effects of tax reforms. Scand. J. Econ. 97 (4), 635–659.
work Benefits on Female Labour Supply and Income Distribution: Evidence from
Aaberge, Rolf, Colombino, Ugo, 2014. Labour supply models. In: O'Donoghue, Cathal
Spain. EQUALITAS Working Paper No. 39.
(Ed.), Handbook of Microsimulation Modelling. Emerald Group, pp. 167–221.
Bakos, Peter, Benczúr, Peter, Benedek, Dora, 2008. The Elasticity of Taxable Income:
Aaberge, Rolf, Colombino, Ugo, Holmøy, Erling, Strøm, Birger, Wennemo, Tom, 2004.
Estimates and Flat Tax Predictions Using the Hungarian Tax Changes in 2005. MNB
Population ageing and fiscal sustainability: integrating detailed labour supply models
Working Paper 2008/7. Central Bank of Hungary.
with CGE models. In: Harding, Ann, Gupta, Anil (Eds.), Modelling Our Future: Social
Bargain, Oliver, Orsini, Kristian, Peichl, Andreas, 2014. Comparing labour supply
Security and Taxation, vol. I. Elsevier, pp. 259–290.
elasticities in Europe and the United States: new results. J. Hum. Resour. 49 (3),
Aaberge, Rolf, Colombino, Ugo, Strom, Steinar, 2000. Labor supply responses and welfare
723–838.
effects from replacing current tax rules by a flat tax: empirical evidence from Italy,
Benczúr, Peter, Katay, Gabor, 2017. The Gains to Work Model of Labour Supply.
Norway and Sweden. J. Popul. Econ. 13 (4), 595–621.
Manuscript.
23
Because of the linked micro-macro structure of the model, it is possible to include a simplified assessment of changes to non-labour taxation. This is done through
including effective tax rates on consumption, capital, and sales, which exert their effects through the macro module. This allows the complete, if simplified, assessment
of tax shifts like the one from 2009 mentioned above, shifting the tax burden from employers' contributions to consumption taxes. Also in this case, the exact pa-
rameters used in the related simulations are summarised in Table A3 in Appendix 6.
456
P. Benczúr et al. Economic Modelling 75 (2018) 441–457
457