Assessing the impacts of a major tax reform: A CGE-microsimulation analysis for Uruguay

In mid-2007, the Uruguayan government implemented a major tax reform with the explicit goals of promoting greater equity and efficiency of the tax scheme and stimulating investment and employment. Overall, the reform substantially changed the tax structure by increasing direct income taxation and lowering indirect taxation, corporate taxation and contributions to social security made by employers.

Four main changes were introduced: a) a dual direct income tax with higher progressive marginal rates on labor income and a range of flat rates on different sources of capital income; b) changes in the tax base and a reduction of the basic and minimum rates of the value added tax, along with the elimination of a distortionary tax on intermediate inputs and a tax on health services; c) modification of labor taxes by reducing and harmonizing employers’ contribution rate to social security across sectors; and d) a reduction of the corporate income tax. The shift towards direct income taxation in replacement of indirect instruments such as commodity or production taxes, in order to achieve equity and efficiency goals, follows some important results from optimal taxation theory (see Diamond & Mirrlees (1971); Atkinson & Stiglitz, (1976); Saez (2004)). Also, some studies point out the desirability of a more progressive schedule on direct income taxation and taxation on capital income (see Diamond & Saez (2011)).

Before the reform, the tax structure in Uruguay heavily relied on indirect tax instruments. Nearly 70% of fiscal revenues (excluding social security taxes) originated from the value added tax (VAT)and other indirect taxes. A corporate tax accounted for a further 13% of these revenues. The direct personal income tax amounted to only 5% of total fiscal revenues, and it only affected wages and pensions, while capital income was not taxed. As such, the system did not respect the criterion of horizontal equity (see Perazzo, Robino & Vigna, 2002; Barreix & Roca, 2003). The value added tax rate was one of the highest in Latin America (with a basic rate of 23%) and a large number of sales taxes only contributed to a small share of fiscal revenues. Finally, some studies pointed out the regressive nature of the pre-reform tax structure, showing that there was scope to use the tax reform as an instrument for redistribution (Grau & Lagomarsino, 2002; Barreix & Roca, 2003; 2006).

Existing evaluations of the Uruguayan tax reform are scarce and they mainly focus on the distributive impact of the modifications of the personal income tax and the value added tax through arithmetical microsimulations without taking into account behavioral responses. Using this methodology, Amarante, Arim and Salas (2007) assess the effects of these two modifications on poverty and inequality, finding small positive effects. The impact on poverty is largely explained by price reductions resulting from the change in the value added tax, assumed to be entirely translated into consumer prices.

However, a policy shock such as the 2007 tax reform can indeed lead to changes in agent behavior, induce the reallocation of resources and generate feedback effects on household income, consumption and savings. Household behavioral responses to variations in their budget constraint following the shock on the personal income tax include changes in consumption demand and labor supply. Modified VAT rates result in new market-clearing prices and quantities via consumer and producer optimization, and can lead to variations in the sectoral structure of the economy. The employer contribution to social security is part of the labor costs in each sector, so its variations affect labor demand. Finally, the reduction of the corporate tax should affect returns to capital. This would affect relative factor prices, as well as the level and allocation of new investment.

With the goal of assessing the effects of these four main components of the reform on aggregate and sectoral output, employment, fiscal balance, poverty and inequality, we use a static computable general equilibrium model (CGE) and a microsimulation model to capture the macro-micro links (top down approach). CGE models allow simulating behavioral responses and adjustments on several markets, while enabling some flexibility in setting macroeconomic rules to assess the impact of different government revenue allocation policies. Combining the CGE model with a microsimulation model makes it possible to include the general equilibrium effects of the tax reform and account for full household heterogeneity (see Colombo (2010) for an analysis of alternative CGE-microsimulation approaches). CGE models have been widely used to analyze tax policy since the seminal papers of Shoven and Whalley (1972; 1973; 1984). Recent examples of tax policy analysis for developing countries using CGE or CGE combined with microsimulation models include Chiripanhura, and Chifamba (2015) for an income tax reform in Namibia; Amir, Asafu-Adjaye and Ducpham (2013) for an income tax reform in Indonesia; or Ahmed, Ahmed and Abbas (2010) for alternative schemes for the sales tax in Pakistan.

The rest of the paper is organized as follows: Section 2 describes the main modifications introduced by the tax reform; Section 3 presents the methodology and data; Section 4 describes the simulated scenarios; Section 5 discusses the macro and sectoral results from the CGE model, including a sensitivity analysis, and Section 6 presents the main findings regarding poverty and inequality from the linked microsimulation model. Finally, we draw some concluding remarks.

The reform introduced the following main modifications. First, the tax on wages and pensions was eliminated and a dual personal income tax was introduced. The previous tax on wages and pensions established a minimum non-taxable income and two brackets of marginal rates that ranked from 2% to 6%. The reform increased the threshold for minimum non-taxable labor income and established increasing marginal rates ranging from 10 to 25%, according to tax bases defined in terms of the BPC (Basis for Contributive Benefits)1. Medical costs for children under the age of 18 and social security payments could be deducted. Finally, capital revenues were taxed at 3 to 12%, depending on the source of income. Pre and post reform schedules of the direct tax on labor, pensions and capital income are summarized in Tables 1 and 2.

Second, the reform reduced the corporate tax rate from 30% to 25%, standardizing it across all sectors. Dividends distributed to households were charged an additional 7% in the form of a personal income tax on capital.

Third, regarding the VAT, before the reform there was a basic rate of 23%, although some goods were taxed at a minimum rate of 14% and a third set of goods were exempted. The changes consisted in reducing the basic rate from 23% to 22% and the minimum rate from 14% to 10%. Also, some exempted goods before the reform were taxed at the minimum rate or at the basic rate in the post-reform situation. A 5% specific sales tax on health services was eliminated, and replaced by the minimum VAT rate. A specific tax of 3% on intermediate consumption of goods was also eliminated.

Finally, the employer contribution to social security (ECSS) stood at 17.5% for most commerce and service sectors, and consisted of a pension tax of 12.5% plus a health insurance tax equal to 5% of payroll. Manufacturing and passenger transport were only taxed by the health insurance tax of 5%, while other sectors were taxed at a considerably higher rate than the general regime (public sector activities and public enterprises were taxed at 19.5% and 24.5%, respectively). The reform established an employer pension contribution rate of 7.5% for industrial, commercial and services sectors, as well as for public enterprises. The health insurance rate remained unchanged at 5%, so that the overall ECSS rate stood at 12.5% in the post-reform situation. The 19.5% tax rate for public sector activities remained the same, as well as the exemption for passenger transportation. Pre-reform and post-reform ECSS rates by sector and value added tax rates by commodity (defined as any category of goods and services subject to the VAT) are summarized in Table 3.

It is worth mentioning that, due to the static nature of the model used here, the effects that the corporate tax has on firms’ capital accumulation are not assessed even though the main objective of modifying this tax has to do with capital accumulation. A full account of the effects of the corporate tax would require a dynamic model, a possible extension of this work.

We apply a static computable general equilibrium model (CGE) based on the IFPRI standard model by Löfgren, Harris and Robinson (2002), and a microsimulation procedure based on the methodology presented in Ganuza, Paes de Barros and Vos (2002) to assess the effects on poverty and inequality by accounting for the full distribution of income. This macro micro methodology has been applied in numerous works for various developing countries including Uruguay (Laens & Perera (2004), Terra et al. (2006), Laens & Llambí (2008)).

Several simulations were carried out with the CGE model. In each simulation, a shock for some (or all) of the specific tax rates involved in the reform was introduced. To start with, we simulated the four main components of the reform simultaneously. Each of these components was then simulated separately in order to assess the relative importance of their effects. The list of simulations carried out is then:

1. REFORM – Simulation of the full reform

2. VAT – Simulation of the modified VAT plus the elimination of the intermediate consumption tax and health tax

3. INCTAX – Simulation of the personal income tax that replaced the previous wage and pension tax

4. ECSS – Simulation of the changes to employers’ contribution to social security

5. FDIRTAX – Simulation of the modifications to the corporate tax

A savings-driven closure was adopted by maintaining a constant marginal propensity to save among all domestic non-government institutions. The trade balance is exogenous and the real exchange rate is the equilibrating variable.

Regarding fiscal balance, although when analyzing tax reforms it is generally assumed that government revenue does not change, three different government closures were tested. In the first two, we intend to capture the effects of the reform in the absence of compensating mechanisms, so we allow government income to vary endogenously. We considered two extreme government behaviors: a) the variation in government income due to the reform is completely absorbed by government expenditures, with constant government savings and b) government consumption is fixed and changes in income alter government savings.

We also considered a third government closure with a budget-neutral assumption, taking the VAT as the compensating tax. Note that the reduction of indirect taxes to compensate for the increase of the personal income tax was one of the main features of the reform. In the full reform scenario, the actual change in the VAT rate was simulated (together with the rest of the tax changes) and an additional proportional adjustment in VAT rates was introduced to compensate for variations in government revenues. So, this simulation allows us to estimate the number of additional points by which VAT rates could be reduced (or increased) if the full reform were set up to be revenueneutral.

Table 9 shows the results of the microsimulations according to the outcomes obtained from the CGE model for each of the three closures mentioned above. The results are obtained from the sequence of steps described in Section 3.3 and report the change in each indicator in each phase compared to the previous one. Table 10 shows the microsimulation results of the full reform effects on per capita household income by decile that result from the “next day” effects (arithmetical microsimulation) and due to changes in labor market indicators and factor prices provided by the CGE model (“general equilibrium” effects).

We first re-estimate disposable income under the direct income tax structure introduced by the reform. This leads to a 1.2% reduction in mean per capita household income and a 1.5% decrease in average labor income. Extreme poverty is reduced by 0.01 percentage points (hereafter pp) and moderate poverty by 0.33 pp. The poverty gap ratio and the severity of poverty also decline, respectively by 0.10 pp and 0.04 pp, as reported in Table 9.

We also find a decrease in income inequality. The Gini coefficient of per capita household income falls by approximately 0.010 points and the Gini coefficient of per capita labor income falls by 0.013 points. Average per capita household income increases moderately in the first eight deciles (Table 10) and decreases in the two richest deciles (-0.7% and −4.2% respectively). These results are clearly driven by the greater progressiveness of the direct income tax. Per capita household income grows by an average of 0.2% in the 1st decile and 0.6% in the 2nd decile. The income tax structure introduced by the reform is thus shown to have smaller effects on the poorest decile, a result that can be explained by the fact that there was already a minimum taxable income on labor.

We then introduced the counterfactual changes in labor market indicators as derived from the CGE model. The general equilibrium effects show an increase in mean per capita household income and mean labor income. These effects compensate for the “next day” reduction in mean income found above. The most important counterfactual changes driving these results are reduced unemployment and growth in average wages and capital income indicated by every CGE simulation. Moreover, general equilibrium effects strengthen the observed “next day” reduction of the incidence of poverty, the poverty gap and the severity of poverty. The size of these effects is much greater than the “next day” effect. The average 2% increase in the wage rate and the approximately 2 pp reduction in the unemployment rate, with particularly notable effects for the unskilled workers that are predominant in lower income households, are the most important labor market changes underlying this poverty reduction. As for inequality indicators, general equilibrium effects lead to a minor additional reduction of the Gini coefficient for total household income (in the same direction as the arithmetical simulation) and do not affect the Gini coefficient for labor income.

The simulated change in household income across the income distribution shows an interesting trend. The “next day” effects of the tax reform are progressive, with a significant reduction in the richest decile’s after-tax income. The “general equilibrium” effects with respect to per capita household income move in the opposite direction: every decile sees income increase regardless of the CGE closure used (see figure 2).

The counterfactual indicators of these key labor market parameters are very similar in each of the three government closures used in the CGE model, so the full reform’s general equilibrium effects on poverty and inequality are very similar in each simulation. In sum, substantial general equilibrium effects are robust to the type of government closure assumed in the CGE model and reinforce the observed poverty and inequality reduction obtained from the “next day” simulation.

Figure 1

Download asset Open asset

First, it is important to remark that we find significant general equilibrium effects from the full implementation of the 2007 tax reform. Taken together, they tend to reinforce the progressive nature of the “next day” effects of implementing its main policy, i.e. the introduction of a direct personal income tax with higher progressive marginal rates to replace the previous wage and pension tax. Although this is an expected result, it reinforces the importance of evaluating these types of macro reforms using methodologies that account for potential reallocation of resources due to changes in prices of goods and factors due to the policy.

Second, the main results relating to the effects of the full implementation of the reforms on aggregate economic activity, employment, poverty and inequality are robust to the alternative assumptions about government closure, although there are clear differences regarding the absorption structure and public-private shares of investment, in addition to possible dynamic effects that are not captured in this analysis. The results are robust to variations in key parameters.

An important result is that the full reform increases GDP and employment, even though it actually increases the tax burden. In other words, when government revenues increase due to the tax reform (and allow an increase in either government provision of public services or government savings), the reform also results in employment and wage growth and reduced unemployment, generating positive general equilibrium effects on average household income and poverty. This obviously does not mean that the government’s use of additional revenue is irrelevant. The simulations that alter either government consumption or savings accounts tend to crowd out private consumption or savings, with probable future negative effects on private capital accumulation. When the government budget is held fixed and additional reductions of the VAT rate are allowed, the reform generates the most positive effects in relation to economic activity, poverty and inequality.

This result is linked to the fact that the reform generally reduces or eliminates taxes on some goods and factors and harmonizes tax rates across categories for others. This is done through the elimination of the intermediate consumption tax and the health tax, and reductions in the VAT and ECSS, with any revenue losses being compensated for by an increase in direct taxation. Reduction, elimination or harmonization of these taxes tends to reduce price distortions of goods and factors. In a context of perfect factor mobility across economic activities, this leads to a better reallocation of resources, stimulating an expansion of economic activity.

Although direct taxation on household income could also be distortionary because of the efficiency loss associated with substitution between labor and leisure, the simulated models suggest that the proposed shift towards direct taxation in Uruguay is desirable from an efficiency perspective. Despite an elastic labor supply in the model, the final simulated changes in participation rates were insignificant and robust to sensitivity analysis. However, progressive marginal tax rates on household income could also have negative impacts on private savings, with negative implications for dynamic capital accumulation. In the long run, non-linear marginal rates on income may also induce negative effects on human capital accumulation by lowering the wage premium for the more educated. These aspects are not considered, and they would require further analysis using a dynamic framework.

We also find that the general equilibrium effects of the full reform include an increase in aggregate disposable household income, which compensates for the reduction obtained by the application of the personal income tax ignoring other components of the reform and economic agents’ responses to the reform. The magnitude of the general equilibrium effects on poverty indicators is significantly greater than the “next day” effects. The 1-point reduction of the Gini inequality coefficient is entirely due to the progressive nature of the personal income tax: households in the richest decile are the clear “losers” from the reform. However, the general equilibrium effects of the full reform do not play a significant role in this regard.

Finally, the results also indicate that VAT rates could be lowered further to make the reform budget neutral. When the full reform is compensated by changes in VAT rates, the VAT rates could be lowered by an additional percentage point, to 21% for the basic rate and 9% for the minimum rate. As budget-neutrality with additional reductions in the VAT rate generates the larger positive effects on aggregate economic activity, poverty and inequality, this suggests that further reductions in VAT rates are desirable both with respect to efficiency and equity.

1. 1
   ‘Taxation Reforms: A CGE-Microsimulation Analysis for Pakistan’. MPIA Working Paper 2010-12

   1. S Ahmed
   2. V Ahmed
   3. A Abbas

   (2010)

   ‘Taxation Reforms: A CGE-Microsimulation Analysis for Pakistan’. MPIA Working Paper 2010-12.

   • Google Scholar
2. 2
   ‘The impact of the Indonesian income tax reform: A CGE analysis’

   1. H Amir
   2. J Asafu-Adjaye
   3. T Ducpham

   (2013)

   Economic Modelling 31:492–501.

   • Google Scholar
3. 3
   ‘Wage inequality and the distribution of education: a study of the evolution of regional differences in Metropolitan Brazil’

   1. J Almeida dos Reis
   2. R Paes de Barros

   (1991)

   Journal of Development Economics 36:117–143.

   • Google Scholar
4. 4
   ‘The design of tax structure: direct versus indirect taxation’

   1. A Atkinson
   2. JE Stiglitz

   (1976)

   Journal of Public Economics 6:55–75.

   • Google Scholar
5. 5
   ‘Impacto distributivo de la reforma impositiva’

   1. V Amarante
   2. R Arim
   3. G Salas

   (2007)

   Draft prepared for Poverty and Social Impact Analysis (PSIA).

   • Google Scholar
6. 6
   Modelling labor markets in CGE models. Endogenous labor supply, unions and efficiency wages

   1. N Annabi

   (2003)

   Montreal, PEP – CIRPEE – Université Laval.

   • Google Scholar
7. 7
   ‘El sistema tributario en Uruguay: situación actual y propuestas’

   1. A Barreix
   2. J Roca

   (2003)
8. 8
   ‘Arquitectura de una propuesta de reforma tributaria y Siete pilares para mantener la reforma tributaria en Uruguay 2005’

   1. A Barreix
   2. J Roca

   (2006)
9. 9
   ‘An introduction to the wage curve’

   1. D Blanchflower
   2. A Oswald

   (1995)

   The Journal of Economic Perspectives 9:153–167.

   • Google Scholar
10. 10
    ‘MIDD: the microeconomics of income distribution dynamics. A comparative analysis of selected developing countries’

    1. F Bourguignon
    2. F Ferreira
    3. N Lustig

    (2001)

    Paper presented at the 2001 Latin American Meeting of the Econometric Society.

    • Google Scholar
11. 11
    ‘Fast development with stable income distribution: Taiwan, 1979-1994’

    1. F Bourguignon
    2. M Fournier
    3. M Gurgand

    (2001)

    Review of Income and Wealth 43:139–164.

    • Google Scholar
12. 12
    An estimation of wage curve for Uruguay

    1. M Bucheli
    2. C González

    (2007)

    Mimeo: Departamento de Economía de la Facultad de Ciencias Sociales.

    • Google Scholar
13. 13
    The Impact of Namibia’s Income Tax Reform: A CGE Analysis. AGRODEP Working Paper 0020

    1. B Chiripanhura
    2. R Chifamba

    (2015)

    Washington, DC: International Food Policy Research Institute.

    • Google Scholar
14. 14
    ‘Linking CGE and microsimulation models: a comparison of different approaches’

    1. G Colombo

    (2010)

    International Journal of Microsimulation 3:72–91.

    • Google Scholar
15. 15
    ‘Endogenous labor supply with several occupational categories in a bi-regional CGE Model’

    1. B Decaluwé
    2. A Lemelin
    3. D Bahan

    (2006)

    Regional Studies, Taylor and Francis Journals 44:1401–1414.

    • Google Scholar
16. 16
    ‘The Case for a Progressive Tax: From Basic Research to Policy Recommendations’

    1. P Diamond
    2. E Saez

    (2011)

    Journal of Economic Perspectives 25:165–90.

    • Google Scholar
17. 17
    ‘Optimal taxation and public production’

    1. P Diamond
    2. JA Mirrlees

    (1971)

    American Economic Review 61 pp. 8–27.

    • Google Scholar
18. 18
    ‘Sistemas tributarios alternativos y su impacto en la distribución del ingreso y en la oferta laboral. Una aproximación comportamental para el caso uruguayo’

    1. M De Rosa
    2. F Esponda
    3. S Soto

    (2011)

    Revista Quantum, VI, 1.

    • Google Scholar
19. 19
    Armington Elasticities: Estimates for Uruguayan Manufacturing Sectors

    1. M Flores
    2. A Cassoni

    (2010)

    Departamento de Economía, Universidad de la República Serie Documentos de Trabajo.

    • Google Scholar
20. 20
    Liberalization, inequality and poverty. Latin America and the Caribbean during the nineties

    1. E Ganuza
    2. R Paes de Barros
    3. R Vos

    (2002)

    Liberalization, inequality and poverty. Latin America and the Caribbean during the nineties, ed., UNDP.

    • Google Scholar
21. 21
    ‘An analysis of South Afica’s value added tax’. World Bank Policy Research Working Paper 3671

    1. D Go
    2. M Kearney
    3. S Robinson
    4. K Thierfelder

    (2005)

    ‘An analysis of South Afica’s value added tax’. World Bank Policy Research Working Paper 3671, August.

    • Google Scholar
22. 22
    Distribución del ingreso y crecimiento: La expansión del mercado interno vía la redistribución de los ingresos: un ejercicio de simulación para la economía uruguaya’, Trabajo de investigación monográfico para la obtención del título de Licenciado en Economía

    1. V González

    (2003)

    Montevideo: UdelaR.

    • Google Scholar
23. 23
    The Uruguayan tax structure and its incidence on income distribution

    1. C Grau
    2. G Lagomarsino

    (2002)

    Montevideo: Fundación de Cultura Universitaria.

    • Google Scholar
24. 24
    Políticas públicas para el desarrollo humano. ¿Cómo lograr los Objetivos de desarrollo del Milenio en América Latina y el Caribe?

    1. S Laens
    2. C Llambí

    (2008)

    ‘Uruguay’, et al, Políticas públicas para el desarrollo humano. ¿Cómo lograr los Objetivos de desarrollo del Milenio en América Latina y el Caribe?, Santiago, PNUD y Uqbar editores.

    • Google Scholar
25. 25
    ¿Quién se beneficia del libre comercio? Promoción de exportaciones y pobreza en América Latina y el Caribe en los 90

    1. S Laens
    2. M Perera

    (2004)

    ‘Uruguay: export growth, poverty and income distribution’, et al, ¿Quién se beneficia del libre comercio? Promoción de exportaciones y pobreza en América Latina y el Caribe en los 90, New York, UNDP.

    • Google Scholar
26. 26
    Microcomputers in Policy Research

    1. H Lofgren
    2. RL Harris
    3. S Robinson

    (2002)

    ‘A standard computable general equilibrium (CGE) model in GAMS’, Microcomputers in Policy Research, 5, Washington, DC, IFPRI.

    • Google Scholar
27. 27
    The tax system and income distribution in Uruguay. Monographic Research. Facultad de Ciencias Económicas y Administración- Universidad de la República, Montevideo

    1. I Perazzo
    2. C Robino
    3. A Vigna

    (2002)

    The tax system and income distribution in Uruguay. Monographic Research. Facultad de Ciencias Económicas y Administración- Universidad de la República, Montevideo.

    • Google Scholar
28. 28
    Crisis and Income Distribution, A Micro-Macro Model for Indonesia

    1. AS Robilliard
    2. F Bourguignon
    3. S Robinson

    (2001)

    Crisis and Income Distribution, A Micro-Macro Model for Indonesia, The World Bank (mimeo).

    • Google Scholar
29. 29
    ‘Direct or indirect tax instruments for redistribution: short-run versus long-run’

    1. E Saez

    (2004)

    Journal of Public Economics 88:503–518.

    • Google Scholar
30. 30
    ‘A general equilibrium calculation of the effects of differential taxation of income from capital in the U.S.’

    1. JB Shoven
    2. J Whalley

    (1972)

    Journal of Public Economics1 (3–4) pp. 281–321.

    • Google Scholar
31. 31
    ‘General equilibrium with taxes: A computational procedure and an existence proof’

    1. JB Shoven
    2. J Whalley

    (1973)

    The Review of Economic Studies pp. 475–489.

    • Google Scholar
32. 32
    ‘Applied general-equilibrium models of taxation and international trade: an introduction and survey’

    1. JB Shoven
    2. J Whalley

    (1984)

    Journal of Economic Literature pp. 1007–1051.

    • Google Scholar
33. 33
    The effects of increasing openness and integration to the Mercosur on the Uruguayan labor market: a CGE modelling analysis. MPIA Working Paper 2006-06

    1. I Terra
    2. S Laens
    3. M Bucheli
    4. C Estrades

    (2006)

    Canada: Laval University.

    • Google Scholar
34. 34
    ‘Tax System’

    1. Uruguayan Congress, Law Nº 18.083 (Uruguay)

    http://www.parlamento.gub.uy, Ley 18.083.

    • Google Scholar

Author details

1. Cecilia Llambi

   Centro de Investigaciones Económicas (CINVE), Montevideo-Uruguay

   For correspondence

   cllambi@cinve.org.uy

2. Silvia Laens

   Centro de Investigaciones Económicas (CINVE), Montevideo-Uruguay

   For correspondence

   slaens@cinve.org.uy

3. Marcelo Perera

   Centro de Investigaciones Económicas (CINVE), Montevideo-Uruguay

   For correspondence

   perera@cinve.org.uy