The Spanish Personal Income Tax: Facts and Parametric Estimates

Transcription

1 The Spanish Personal Tax: Facts and Parametric Estimates Esteban García-Miralles Banco de España Nezih Guner CEMFI Roberto Ramos Banco de España October 2017 [PRELIMINARY AND INCOMPLETE. PLEASE DO NOT CIRCULATE] Abstract The purpose of this paper is twofold. First, we characterize the distribution of income and taxes pertaining to the personal income tax in Spain. Specifically, we use administrative data on tax returns to describe the distribution of pre- and after-tax income, as well as income sources, tax rates, tax liabilities and tax benefits, for both individuals and households. And second, we estimate a variety of tax functions that capture the underlying heterogeneity of the data in a parsimonious way. We do so for the years 2002 to 2013, as well as for microsimulated data for Moreover, we distinguish between individuals and households by sociodemographic groups, as well as between income types. These parametric functions can be readily used in applied work in macroeconomics and public finance. JEL codes: E62, H24, H31 Keywords: Personal Tax, Tax Functions, Distribution We thank seminar participants at Banco de España for comments and discussions. The views expressed in this paper are those of the authors and do not necessarily coincide with the views of the Banco de España or the Eurosystem. Corresponding author: 1

2 1 Introduction This paper presents two contributions. First, we characterize the distribution of income and taxes of the personal income tax in Spain, covering both pre- and after-tax income, as well as income sources, tax rates, tax liabilities, and tax benefits, for both individuals and households corresponding to And second, we provide estimates of effective tax functions for the years 2002 to 2013, as well as for microsimulated data of We estimate these functions for both individuals and households, as well as for different income types and tax credits. These functions capture the underlying heterogeneity of the data in a parsimonious way and constitute a readily available input to be used for applied work, such as in the flourishing dynamic macro models with heterogenous agents, see Heathcote et al. (2009). Our paper follows the work by Guner et al. (2014), who estimate tax functions of the US personal income tax. Our data come from an administrative dataset containing a stratified random sample of tax returns, which includes almost the complete set of fiscal and sociodemographic information taxpayers fill therein. The dataset is representative of the population of Spanish taxpayers and is not censored, making it ideal for our purposes. The data cover 15 regions and the 2 autonomous cities, for 2013 containing 2.16 million observations, which amounts to 11% of the population. For the household estimates we use the panel version of the dataset that allows us to link individual tax-fillers from the same household. This sample contains more than 500,000 observations, around 3.3% of the population. We find a large degree of heterogeneity in income and in taxes effectively paid. For 2013, the top quintile earns 46% of gross income and accounts for 70.9% of tax liabilities. Effective average tax rates are close to zero in the two lowest quintiles, while they are 20.2% on average in the top quintile, and 33.1% in the top 1%. Most deductions are regressive, with the tax credit for employed mothers, regional tax credits and the labor income tax credit accruing relatively more to taxpayers in the middle of the income distribution. Labor income is by far the largest source of income, but capital and self-employment income are relatively important for the lowest quintile. Also, capital income is relatively more concentrated in the well-off and the top 1%. Our estimates of the household income distribution are similar when compared to the Survey of Household Finances (EFF). For the estimation of the functional forms, we follow two different approaches. First, we estimate one single expression for the final tax liabilities as a function of gross income. We estimate four different specifications: a log function, used by Guner et al (2012a, 2012b), the HSV specification, used by Benabou (2002) and Heathcote et al. (2011), the Power specification, used by Guvenen et al. (2014) and the GS specification, used by Gouveia and Strauss (1994). In all these four specifications we account for the fact that low incomes are subject to zero effective tax rates. Therefore, we estimate an income threshold under which tax liabilities are zero. This threshold is chosen so as to minimize the mean squared error. Our results show that the HSV and 2

3 the GS functions provide the best fit of the data, both in terms of the amount and the distribution of tax liabilities. In the second approach we estimate three different functions, one for the general tax rates that apply to the general taxable income, one for the savings tax rates, applied to the savings taxable income and one for the tax credits. For the general tax rates we use the HSV and the GS functions; for the savings tax rate we use a piece-wise function that increases linearly up to a certain threshold and takes a constant value after it. For the tax credits we estimate a Ricker model. We also estimate parametric functions for households, distinguishing between single and married. Our functions cover the period 2002 to Moreover, in order to account for the 2015 tax reform, we construct microsimulated data for this year and estimate the set of tax functions. The rest of the paper is organized as follows. Section 2 describes the Spanish Personal Tax. Section 3 describes the dataset and lays out the definitions and sample restrictions. Section 4 presents the basic facts of the income and tax distributions. Section 5 presents the parametric estimates of the tax functions. Section 6 concludes. 2 The Spanish Personal Tax 2.1 Structure of the Tax The Spanish Personal Tax (PIT) or Impuesto sobre la Renta de las Personas Físicas (IRPF), taxes the Spanish residents income. In 2015, it amounted to 7.4% of GDP, being the largest source of revenue after social security contributions. 1 The tax is withheld at source. Every year, between April and June, taxpayers must fill in the tax return corresponding to the previous calendar year s income. As of 2016, this obligation applied, in general, to taxpayers whose labor income was above e22,000, or capital income was above e1,600, or real-estate income was above e1,000, or they were self-employed. Note however that most taxpayers below the labor income threshold decide to fill in the tax return (around 81%), as they are likely to obtain a refund from the taxes withheld at source. The structure of the tax is fairly complicated, see a simplified diagram in Figure I. subject to the tax can be of several types: labor income, capital income (both financial and realestate) and self-employment income. From these gross income source, a set of deductible expenses can be subtracted, including among other social security contributions paid by the employee, an amount for labor income earners, and expenses associated to the activity of the self-employed. The result of this subtraction gives rise to a concept that we name adjusted gross income. Adjusted gross income is then grouped into two categories, which are subsequently taxed at different rates. On the one hand, general income comprises labor income, self-employment 1 According to the OECD, taxes on income, profits and capital gains of individuals represent around 21% of total tax revenues in Spain (including social contributions). Compared to the Euro Area, the size of this source of revenue is around 2.6 percentage points below the average. 3

4 income and some capital income (each item being the result of gross income minus deductible expenses). On the other hand, saving income includes the largest portion of capital income. From each type of income a wide set of tax deductions can be applied. For example, general income can be reduced by a tax deduction for couples filling in a single tax return (joint filers) as well as by the amounts contributed to private pension plans (up to a limit), among many other. If the taxpayer is entitled to apply deductions exceeding the general income, he/she can apply the remnant to the saving income. The subtraction of these deductions from general and saving income results in concepts named general taxable income and saving taxable income. Each category of income is taxed according to different tax scales. Moreover, the tax scales are split into a state and a region portion. This is so because around half of the tax is transferred to the regions, which are entitled to design their tax scales and introduce their own tax benefits. In 2013 the state general tax scale consisted in 7 tax bands and a top marginal rate of 30.5%. The region general tax scale, which is applied on top of the state one, can vary substantially across regions. For example, that of the Community of Madrid comprised four tax bands and a top marginal rate of 21.4%, whereas that of Catalonia encompassed 6 tax bands and a top marginal tax rate of 25.5%. Therefore, Madrid taxpayers faced in 2013 a top marginal rate of 51.9% (30.5% %), while taxpayers in Catalonia were subject to a top marginal tax rate of 56% (30.5% %). The saving tax scale is much less progressive. In 2013 the state portion consisted in 3 bands and a top marginal rate of 16.5%, whereas the region portion comprised 2 bands and a top rate of 10.5%. In this case, the region layer does not vary across regions. In 2015, a significant reform of the tax was implemented, which led to a substantial decrease in marginal rates. See Figure II for a description of the tax scales in the two selected regions in 2013 and The resulting amount from applying the (state and region) tax scales to general taxable income is reduced by the family allowance, which again has a state and a region portion (although in practice they are very similar). The family allowance is calculated by applying the (state and region) general tax scale to a (state and region) amount that depends on the characteristics of the taxpayer and his/her family, such as age, number of dependent children, number of dependent parents, and disability of the taxpayer and the dependent members of his/her family. If the general taxable income is smaller than the family allowance, the remnant can be subtracted from the gross saving income, as long as it does not turn the saving taxable income into negative amount. 2 After subtracting the family allowance, the state general income and state saving income tax liabilities are pooled together. Similarly, the region tax liabilities (from general and saving income) are added up. To these two types of tax liabilities a set of non-refundable tax credits can be applied, some state or region specific, some pertaining to both. Among this set one can find that associated to house purchases (if they were carried out before 2013), the extended set of region tax credits and a tax credit for low labor income earners. After the application of the non-refundable tax 2 Starting in 2015, the family allowance applied to saving income was computed as in the general income. 4

5 credits, the state and region tax liabilities are pooled together. Finally, these pooled tax liabilities can be reduced by a set of refundable tax credits. In 2013, it was one, granted to employed mothers with children below 3 years old, while in 2015 the set was expanded to include mainly taxpayers with disable dependents and large families. The subtraction of this tax credit gives rise to the tax liabilities of the taxpayer. Figure I Structure of the Spanish Personal Tax (2013) 2.2 Recent Reforms ( ) The Spanish PIT has been subject to several reforms during the lapse of time encompassing our data. In this section we provide with a brief description of them, which will be helpful to interpret the over time evolution of the tax functions presented in Section In broad terms, the tax has zigzagged during the last decade and a half. Between 2003 and 2007 it underwent several reductions in a context of rapid economic growth and favorable macroeconomic conditions. Furthermore, in 2008 the government implemented additional cuts in order to counteract a slowdown in activity. Once the economic woes were confirmed, the sharp fall in GDP and the subsequent deterioration of the budget balance led to sizable tax increases in Finally, following the economic recovery, a major overhaul took place in 2015, resulting in a significant tax cut. The first major reform of the personal income tax during this century came into effect in It involved a reduction in the number of tax bands and tax rates, an increase in the family 5

6 Figure II Statutory Marginal Tax Rates (2013 & 2015) Panel A: General (2013): Panel B: Saving (2013) Statutory Marginal Tax Rates Statutory Marginal Tax Rates , , , , , , , ,000 Euros 0 10,000 20,000 30,000 40,000 50,000 Euros Madrid Catalonia All Regions Panel C: General (2015) Panel D: Saving (2015) Statutory Marginal Tax Rates Statutory Marginal Tax Rates , , , , ,000 Euros 0 10,000 20,000 30,000 40,000 50,000 60,000 70,000 Euros Madrid Catalonia All Regions Notes: This figure shows the statutory marginal tax rates of the personal income tax in 2013 and 2015 for residents in Madrid and Catalonia. Panels A and C display the rates applied to general income, consisting mainly in labor income, self-employment income and imputed rents. Panels B and D show the tax rates of saving income, comprised roughly by capital income and gains and losses. allowance and the introduction of a tax credit on employed mothers with at least one child below 3 years old. During the next two years there were no important actions, save a tax band update to keep up with inflation and the introduction of some one-off benefits to compensate for an increase in oil prices. In 2007 the government implemented a big reform, consisting in a reduction of tax bands and tax rates, an increase in the family allowance, which was redefined as a general income tax credit, a raise in saving tax rates and the introduction of a tax credit on births and adoptions. Moreover, a further tax cut was passed in 2008, involving e400 tax credit in order to spur private expenditures. Also in 2008 some regions lowered their tax rates, a new deduction for house tenants was introduced and tax bands were updated in order to accommodate the recent inflation developments. During the period the successive governments approved a set of tax increases in the 6

7 context of the economic crisis and the attendant deterioration of the budget balance. In 2010 the cabinet overrode the e400 tax benefit and raised the saving tax rates, while in 2011 it removed the tax credit on births and adoptions and it raised the marginal tax rates of the well-off. During these years there were also some changes associated to the new system of region financing, agreed on in Specifically, in 2010 tax rates shifted from the state to the region layer and the following year each region used their normative power to set their own tax scale. In 2012 the government approved a significant increase of marginal rates, affecting the whole income distribution. This tax raise, set initially to last two years, was extended to Furthermore, it suppressed in 2013 a large deduction on house purchases and the exemption of lottery prizes. After the crisis, the cabinet adopted a big reform of the tax code in 2015, leading to a tax cut of around 0.9% of GDP. It consisted mainly in a reduction of tax bands and tax rates, which overturned partly the 2012 tax raise. Also, the family allowance was increased and new refundable tax credits depending on family characteristics were introduced. The reform involved also a deep overhaul of tax deductions, some of them being decreased or repealed, partially offsetting the previous measures. 3 Data 3.1 Micro data on Tax Returns We use an administrative dataset containing a (stratified) random sample of tax returns, which includes almost the complete set of fiscal and sociodemographic information taxpayers fill therein. Hence, the dataset provides a very detailed account of income by income sources, tax benefits and family characteristics (number of dependent relatives, disability, location, etc.). Noteworthy, the dataset is not censored either at the bottom or at the top of the distribution. 3 The unit of observation in the dataset is the tax return, which can be of two types: single and joint. Single tax returns are filed at the individual level, whereas joint tax returns represent two spouses filing together, or single-parent families with at least one child. In joint tax returns incomes are pooled together and taxpayers are entitled to an additional tax deduction on top of those accruing to single filings. Save by these and other small differences, the computation of tax liabilities under both types is similar. The decision on which way to file is made by the taxpayer. In general, joint tax returns benefit couples in which one partner earns little or no income, as well as single-parent families with no child income. The different waves of the data comprise both a cross-section and a panel dimension. Repeated cross-sections are available for the years 2002 to 2013, and they have a large sample size. For 3 A detailed description of the dataset (in Spanish) and some statistics are provided every year in the working paper series of the Instituto de Estudios Fiscales. For example, see Pérez et al. (2016) for a description of the 2013 cross-section wave. 7

8 example, the 2013 dataset features 2.2 million observations, amounting to around 11% of the universe of tax returns; while for the periods and the sample size equals around 10% and 5% of the population, respectively. These cross-sections do not identify spouses filing single tax returns, hence they do not allow for a representation of taxes at the household level. The panel dataset covers the period and it renders a smaller sample size (around 3.3% of the universe of taxpayers in 2012). For this paper the main advantage of this dataset is that it provides an identifier of spouses filing single tax returns. Therefore, it allows to account for taxes paid by households. Note that we use the cross-section and the panel dimensions of the data to describe and estimate the functional forms of the personal income tax from the perspective of taxpayers and households, respectively. Moreover, the panel allows us to compare household income aggregates from tax data with those obtained in survey data, specifically the Survey of Household Finances (EFF). Table I provides a comparison between the cross-section sample aggregates and the population counterparts in As expected, the data render a very accurate representation of income and tax liabilities of the 19.2 million tax return filers, the differences being lower than 1% on the selected items. Table I Accuracy of the 2013 Cross-Section Data (emillion) Sample Aggregate Population Aggregate Difference (1) (2) (3) Number of Taxpayers (million) % Gross Labor % Gross Capital % Gross Self-Employment % Taxable % Tax Liabilities % Notes: The source of the population aggregates is the Spanish Tax Agency (Estadísticas de los declarantes del Impuesto sobre la Renta de las Personas Físicas (IRPF)) The definitions of the variables are described in Section 3.2. Gross capital income excludes some small items for which no population aggregates are reported. 3.2 Definitions and Sample Restrictions In this section we explain in detail the definition of the main variables used in the paper. Specifically, we describe the different income types we account for, the characterization of tax liabilities and tax benefits, and the computation of effective average tax rates. Regarding the income categories, our definition is threefold. First, gross income is the sum of labor income, capital income, and self-employment income. Labor income comprises benefits in cash 8

9 and in kind granted to individuals as employees. Capital income includes both financial income (interests, dividends, capital gains, etc.) and real-estate income (excluding the main residence). Self-employment income corresponds to the earnings of the self-employed associated to their business. It is worth noting that the data report self-employment income and part of capital income already net of some deductible expenses and tax deductions. This feature is specially relevant for the self-employed, as deductible expenses and other deductions can be relatively high, which may lead to an underestimation of their income. For this reason, we account for a second definition of income, named adjusted gross income, where all income categories are net of deductible expenses, such as employee s social security contributions, a deduction for labor-income earners, and business expenses. Nevertheless, we favor the first definition of income when it comes to computing average tax rates and total tax deductions, since doing so prevents an underestimation of these variables. The third income category is taxable income, which is income subject to the application of the (general and saving) tax scales. Note that we define also a general and saving taxable incomes, to which the corresponding general and saving tax scales are applied. Tax liabilities correspond to the amount that the taxpayer effectively has to pay, taking into account all the tax credits, including those refundable. For this reason, they can be a negative amount. Tax benefits can be of two types: tax deductions and tax credits. Tax deductions are amounts subtracted from the tax base, that is, before the application of the tax scales. Therefore, total tax deductions are defined as gross income minus taxable income. Tax credits are amounts subtracted from the tax liabilities resulting from the application of the tax scale. Hence, total tax credits can be computed as the difference between this amount and final tax liabilities. Lastly, average effective tax rates are computed as tax liabilities over gross income if the former are greater than or equal to zero, and zero otherwise. Note also that taxpayers with negative gross income are excluded from the sample, as we explain below. We also define the average effective general tax rate as tax liabilities resulting from the application of the general tax scale net of the family allowance (the box Intermediate Tax Liabilities 1 in Figure I) over general taxable income, provided that the latter is positive. We subtract the family allowance because for many (low income) taxpayers, this is equal to the general taxable income, hence by subtracting it from the numerator we avoid an artificial overestimation of the general tax rate (for these taxpayers the resulting average general tax rate is zero). Average saving tax rates are computed similarly, except for the fact the family allowance is a tax deduction regarding saving income, and then it is not subtracted from the numerator. 4 Hence, according to Figure I, the average effective saving tax rate is defined as Intermediate Tax Liabilities 2 over saving income. 5 4 In 2015, the family allowance regarding saving income was recoded as tax credit, as in general income. 5 According to the 2013 tax code, the cells (in Modelo 100 ) corresponding to each definition are the following. Gross income: 9 (labor) (capital) (self-employment). Adjusted gross income: 20 (labor)

10 Regarding the sample, we restrict it to taxpayers with positive total gross income, non-negative gross income sources (labor, capital and self-employment) and average tax rates below the maximum statutory marginal tax rate. Note also that we exclude taxpayers spared from the obligation to fill in the tax return and who decide not to do so, despite earning low income and hence, being likely entitled to obtain a tax refund from the taxes withheld at source. Our final sample leaves us with around 96% of tax filers and 85% of taxpayers. 3.3 Survey of Household Finances As mentioned above, we compare the estimated household income distribution from the tax return data with that obtained from the Survey of Household Finances (Encuesta Financiera de las Familias or EFF). This is a survey conducted by the Banco de España that collects information on income, assets, debts and spending of around 6,000 households each wave, several of the variables being reported for each household member, such as labor income and employment status. Moreover, the survey oversamples high-wealth households, in order to allow for a sufficient number of observations to study the financial behavior at the top of the wealth distribution and to accurately measure aggregate wealth. The EFF is undertaken every three years, the last wave corresponding to Each wave accounts for annual income pertaining to the previous year. A detailed description of the survey can be found in Bover et al. (2017b). Note that households in the tax data are defined as the taxpayer and his/her spouse. Therefore, we apply this definition to the EFF, by adding the gross income of the household s reference person and his/her spouse. Note that the EFF provides information on most of the labor and self-employment income items for each household member. On the contrary, the capital income items are mainly reported for the whole household. When this happens, we make the assumption that the item pertains to the household s reference person. 6 Note also that we classify the income sources provided by the EFF so as to mimic the labor, capital and self-employment groups defined in the tax data. 7 As with the tax data, we restrict the sample to households earning positive gross income and non-negative gross income sources (labor, capital and self-employment). This amounts to dropping (capital) ( ) (self-employment). Taxable income: 415 (general) + 2 (saving). General income: Saving income: Tax liabilities: Average effective general tax rate: general income. Average effective saving tax rate: saving income. 6 Note that since we focus on aggregate household income, it is irrelevant for two-person households to assign capital income to the reference person, his/her partner, or to split it between the two. 7 According to the 2014 wave of the EFF, we define gross income as: (p6 64 i + p6 66 i + p6 68 i p p6 74b i + p6 74 i) + p6 75d1 + p6 75d3 + p6 75d4 (labor) + ip6 72 i (self-employment) + p7 2 + p p p7 12a + p p7 4a + p7 4b + p7 6a + p7 6b + p7 8a + p7 8b + p6 76b + p6 75f (capital), where i indexes each household member (the reference person and his/her spouse in two-person households and the former in one-person households). 10 i

11 around 2% of the observations. 4 Basic Facts of the and Tax Distributions In this section we report basic facts of income, tax liabilities and tax benefits for both our sample of 2013 individuals and households. Moreover, we compare the results of the latter with survey data from the Survey of Household Finances. 4.1 Distribution Individuals Table II summarizes the distribution of taxpayers income and income sources in The data show a fair degree of income concentration at the top. For example, the top 20% of taxpayers earns about 46% of income, whereas the richest 10%, 5% and 1% earns about 30.2%, 20.0% and 8.1%, respectively. Nevertheless, this income concentration is substantially lower than that documented in the US by Guner et al. (2014), where the top 20% earns 61.4% of total income and the top 1% accumulates as much as 20.9%. Regarding the different concepts of income, the share of income accruing to the well-off increases once deductible expenses and tax deductions are subtracted from the gross measure, and taxpayers with zero income are excluded from the sample. For example, the share of income accounted for the top 20% increases from 46% of gross income, to 49.6% of adjusted gross income and taxable income. Likewise, the income share of the top 1% shows an increasing pattern across the different concepts of income. The Gini coefficients reveal that, conditional on positive income, taxable income inequality is larger than gross income inequality, while the levels are still substantially lower than in the US, compare the last two rows of Table II with Table 2 of Guner et al. (2014). With respect to the distribution of the different income sources, columns (4) to (6) of Table II show that capital income renders a higher degree of concentration at the bottom and top quintiles, when compared to gross income. For example, the bottom 20% accounts for just 4.8% of gross income, while it accumulates 6.6% of capital income; the top 1% accumulating 8.1% and 24.3%, respectively. Self-employment income is also concentrated at the very top, but the lower end of the income distribution accumulates a substantial amount. For instance, the first three quintiles earn more than 40% of this income source. In Table III we decompose the sources of income across the income distribution. Labor income is by far the largest source of income, its importance increasing monotonically from quintiles 1 to 4, where it represents between 80 and 90% of total income. In the top decile income from labor is less important, although even for the top 1% the share of labor income is very high, being close to 70%. In the lowest end of the distribution, specially in the bottom 1%, capital income appears very 11

12 Table II Distribution of Individuals and Sources Quantiles Share of Gross Sources Bottom Gross Adjusted Gross Taxable Labor Capital Selfemployment (1) (2) (3) (4) (5) (6) 1% 0.0% 0.0% 0.0% 0.0% 0.1% 0.0% 1-5% 0.3% 0.1% 0.2% 0.2% 1.1% 1.0% 5-10% 1.1% 0.5% 0.6% 0.9% 1.7% 2.4% Quintiles 1st (bottom 20%) 4.8% 2.9% 3.0% 4.2% 6.6% 10.1% 2nd (20-40%) 10.5% 9.0% 8.9% 10.2% 9.5% 15.9% 3rd (40-60%) 15.8% 15.3% 15.2% 16.2% 11.4% 16.4% 4th (60-80%) 23.0% 23.3% 23.3% 24.5% 15.1% 14.0% 5th (80-100%) 46.0% 49.6% 49.6% 44.9% 57.4% 43.6% Top 90-95% 10.2% 10.8% 10.8% 10.4% 9.3% 7.8% 95-99% 11.9% 13.0% 12.9% 11.6% 13.6% 12.0% 1% 8.1% 9.3% 9.4% 5.8% 24.3% 15.7% Other Statistics Gini Coefficient Var - Log Notes: This table displays the distribution of gross income, adjusted gross income and taxable income, as well as the distribution of gross income sources (labor, capital, and self-employment) for the sample of 2013 individuals. Note that observations with non-positive income values in columns (1) to (3) are excluded. The Gini coefficient in columns (4) to (6) are computed including the observations with zero income. significant. However, this is explained by the very low income levels of this group, see Table A.1 in the appendix. Excluding the lowest quintile, capital income accounts for around 6 to 9%, reaching 20.8% for the richest taxpayers. Self-employment income explains 8.6% of income in the second quintile, while it drops to around 4 to 6% for richer individuals. At the top of the distribution it accounts for slightly more than 10% of total income Households In Table IV we compare the household distribution of income estimated from the tax data with that obtained from the Survey of Household Finances (EFF). We do so for the two latest EFF waves, i.e and We can see that both sources provide a similar estimate of the income distribution. For example, in 2010 income accounted by the top 20% amounted to around 50%, the top 1% accumulating more than 9% of total income; while the bottom 20% received less than 5% of earnings. With respect to the most recent wave, estimates of the income distribution are also similar, although the differences at the right-end of the distribution are larger. For example, the EFF seems to under predict the share of income accruing to the top 1% by 1.7 percentage 12

13 Table III Individuals Gross Sources Quantiles Labor Capital Selfemployment (1) (2) (3) Bottom 1% 12.4% 79.3% 8.4% 1-5% 46.5% 36.5% 17.0% 5-10% 72.1% 15.3% 12.7% Quintiles 1st (bottom 20%) 66.8% 20.6% 12.6% 2nd (20-40%) 82.8% 8.7% 8.6% 3rd (40-60%) 87.3% 6.9% 5.9% 4th (60-80%) 90.4% 6.2% 3.4% 5th (80-100%) 87.5% 8.4% 4.1% Top 90-95% 87.2% 8.7% 4.2% 95-99% 83.8% 10.8% 5.4% 1% 68.8% 20.8% 10.4% Notes: This table shows the decomposition of gross income over income sources across the gross income distribution for the sample of 2013 taxpayers. Columns (1) to (3) add up to 100. points. 4.2 Tax Rates and Tax Liabilities In Table V we summarize the distribution of tax liabilities and tax rates of the 2013 taxpayers. In columns (1) and (2) we also depict the corresponding distributions of gross income and taxable income (already shown in Table II), in order to illustrate the progressivity of the tax code. Indeed, this feature of the tax becomes apparent. If the top quintile accumulates 46% of gross income, it shoulders more than 70% of tax liabilities, while the top 1% accounts for close to 20% of revenues. As a matter of fact, more than 90% of tax payments are concentrated in the top 40%, while the bottom two deciles account for only 1.3% of the tax. The high concentration of tax liabilities is reflected in small average tax rates at the lower end of the income distribution and larger rates for richer individuals, which average 20.2% in the top quintile and 33.1% in the top 1%. Average statutory marginal tax rates are also highest for the well-off, reaching 45% for the top 1%, while they are significantly lower as we move leftwards across the income distribution. Note that for each individual we computed the statutory marginal tax rate as the weighted average of the corresponding marginal rates of general income and saving income. These averages hide a substantial degree of heterogeneity across individuals. Panel A of Figure 13

14 Table IV Household Distribution: Tax Data Compared to EFF Quantiles Bottom Tax Data EFF Difference Tax Data EFF Difference (1) (2) (3) (4) (5) (6) 1% 0.0% 0.1% % 0.0% % 0.3% 0.7% % 0.5% % 1.0% 1.2% % 1.1% -0.1 Quintiles 1st (bottom 20%) 4.4% 4.9% % 4.7% nd (20-40%) 9.6% 9.4% % 9.1% 0.3 3rd (40-60%) 14.5% 14.5% % 14.3% 0.0 4th (60-80%) 21.7% 22.7% % 22.2% th (80-100%) 49.9% 48.5% % 49.6% 0.8 Top 90-95% 11.0% 11.1% % 10.8% % 13.2% 13.0% % 13.4% 0.2 1% 9.7% 8.0% % 9.1% 0.5 Other Statistics Gini coefficient Var - log income Notes: This table provides a summary of the household distribution of income according to the tax return data aggregated at the household level and the Survey of Household Financies (EFF). It compares the results from two waves: 2013 and III depicts the average effective tax rates across different multiples of mean gross income, alongside 2 standard error bands. 8 As can be seen, there is wide variation of tax rates even for individuals with the same gross income, being this the result of different family characteristics and tax benefit entitlements. The shape of this curve is what the parametric estimates of Section 5 are meant to approximate. In panel B we represent the corresponding curves of statutory and effective marginal tax rates. Statutory rates are again computed as the weighed average of the general and saving marginal rates. The effective rates are computed as in Guner et al. (2014), see Section 6 thereof, where we set the minimum (maximum) income level to 0.2 (10.4) times mean income and the increment to 0.4 times mean income. The findings shows that marginal rates increase rapidly with income, becoming flat at around 3 times mean income. The set of tax benefits renders the effective curve below the statutory one, being the difference roughly about 5 percentage points on average. It is worth noting one feature of the tax code which will be important in the parametric estimates later on, namely the large percentage of individuals facing zero tax rates, around 32% 8 Note that mean individual gross income in 2013 was e23.354, while household mean income in 2012 amounted to e

15 Table V Distribution of Individuals Tax Liabilities and Tax Rates Quantiles Gross Taxable Tax Liabilities Average Tax Rate Statutory Marginal Tax Rate (1) (2) (3) (4) (5) Bottom 1% 0.0% 0.0% 0.0% 0.0% 5.5% 1-5% 0.3% 0.2% 0.0% 0.0% 9.2% 5-10% 1.1% 0.6% 0.0% 0.0% 18.6% Quintiles 1st (bottom 20%) 4.8% 3.0% 0.0% 0.1% 18.4% 2nd (20-40%) 10.5% 8.9% 1.3% 2.0% 24.5% 3rd (40-60%) 15.8% 15.2% 7.9% 7.5% 24.8% 4th (60-80%) 23.0% 23.3% 19.9% 12.8% 28.9% 5th (80-100%) 46.0% 49.6% 70.9% 20.2% 36.4% Top 90-95% 10.2% 10.8% 13.8% 20.2% 38.0% 95-99% 11.9% 12.9% 20.4% 25.3% 42.7% 1% 8.1% 9.4% 18.7% 33.1% 45.0% Notes: This table shows the distribution of individuals tax liabilities (column 3), average effective tax rates (column 4) and statutory marginal tax rates (column 5) across the gross income distribution. In columns (1) and (2) the distribution of gross income and taxable income are summarized in order to illustrate the progressivity of the tax code. Figure III Individuals Effective Average and Marginal Tax Rates Panel A: Effective Average Tax Rates Panel B: Marginal Tax Rates Average Tax Rates Marginal Tax Rates Multiples of Mean Multiples of Mean Statutory Effective Notes: Panel A depicts the effective average tax rate (± 2 standard deviations) across different multiples of mean income. Note that each data point corresponds to the mean average tax rate of taxpayers with mean income within the corresponding point in the x-axis and the very next (excluding the right end point). For instance, the data point of mean income 1.4 is the mean average tax rate of taxpayers earning income within the interval [1.4,1.8). Panel B shows the statutory and effective marginal tax rates. Statutory rates are computed as the weighted average of general and saving marginal rates, while effective rates are described in Section 6 of Guner et al. (2014), where we set the minimum (maximum) income level to 0.2 (10.4) times mean income and the increment to 0.4 times mean income. 15

16 in Figure IV shows that until about 35% of mean income the percentage of taxpayers facing positive rates is below 10%. This share increases steeply, so that by 70% of mean income around 90% of taxpayers contributes to revenue, this number gradually converging towards 100% as income increases. Figure IV Share of Positive Effective Tax Rates (Individuals) Share of Positive Average Tax Rates Multiples of Mean Notes: Each data point corresponds to the share of taxpayers facing strictly positive average tax rates and earning income within the corresponding point in the x-axis and the very next (excluding the right end point). 4.3 Tax Benefits In Table VI we describe the distribution of the most relevant tax deductions, which, as mentioned in Section 3.2, correspond to tax benefits reducing the tax base. Also, in Table A.2 in the appendix we account for their relative weight across income groups. The biggest tax deduction is that granted to labor income earners, followed by Social Security contributions paid by the employees, the tax benefit associated to joint tax returns and the contributions to private pension plans. On the whole, the top quintile benefits from more than 25% of the total tax deductions, while the bottom quintile receives less than 15%, see the first column of Table VI. This feature of the tax system suggests that these benefits are overall regressive, being specially so in regard to benefits associated to private pension plan and Social Security contributors, where the top quintile represents 72% and 41% of the total amount, respectively. However, note that part of the regressive nature of tax deductions stems from the fact that the tax base of many low income earners goes to zero after making use of some tax benefits, exhausting the possibility of further deductions later on in the tax code. 16

17 Table VI Distribution of Individuals Tax Deductions Bottom Labor Joint Filing Quantiles Total Social Security Contributions Contributions to Private Pensions Other (1) (2) (3) (4) (5) (6) 1% 0.0% 0.0% 0.0% 0.0% 0.0% 0.2% 1-5% 1.2% 0.3% 1.4% 1.0% 0.1% 3.5% 5-10% 3.8% 0.7% 5.5% 1.7% 0.5% 5.0% Quintiles 1st (bottom 20%) 14.1% 3.7% 19.6% 8.2% 2.0% 20.0% 2nd (20-40%) 19.2% 10.5% 23.2% 19.5% 4.6% 21.2% 3rd (40-60%) 18.9% 18.6% 18.9% 25.7% 8.2% 14.0% 4th (60-80%) 20.4% 26.1% 19.2% 24.2% 13.2% 12.7% 5th (80-100%) 27.4% 41.1% 19.1% 22.4% 72.0% 32.0% Top 90-95% 6.8% 11.6% 4.8% 5.3% 16.7% 5.4% 95-99% 6.5% 10.1% 3.8% 4.2% 26.8% 7.1% 1% 2.8% 2.6% 0.9% 1.1% 15.3% 11.9% Notes: This table depicts the distribution of tax deductions over the gross income distribution. Tax deductions are amounts subtracted from the tax base. Total tax deductions are computed as gross income minus taxable income, where the latter refers to income to which the tax schedule applies. Tax credits, as mentioned in Section 3.2, correspond to tax benefits applied to the tax liabilities that result from the application of the tax schedule. Table VII depicts their distribution across income groups and Table A.3 in the appendix shows the relative weight of the most relevant ones. By far the family allowance is the largest tax credit, representing more than 95% of these benefits for the bottom 20% and more than 80% for the top 20%. Following in magnitude is the tax credit associated to house purchases, that granted to employed mothers and a battery of region-specific tax credits. 9 Finally, it is worth highlighting a tax credit granted to low labor income earners, which is a relevant tax benefit for the second quintile. As of the distribution of these benefits, the family allowance is roughly evenly distributed since it depends solely on family characteristics. Note that the smaller share accruing to the lower end of the income distribution is explained by the exhaustion of tax liabilities as a result of the application of (part of) this allowance. On the contrary, the tax credit associated to house purchases benefits more the well-off, where benefits granted to employed mothers, low labor income earners and the set of region-specific benefits goes mainly to the middle or the lower end of the income distribution. 9 The tax credit associated to house purchases was stopped in 2013, so that it only benefits transactions carried out before that year. 17

18 Quantiles Total Family Allowance Bottom Table VII Distribution of Individuals Tax Credits House Purchases Working Mothers Regional Labor (1) (2) (3) (4) (5) (6) (7) 1% 0.0% 0.0% 0.0% 0.1% 0.0% 0.0% 0.0% 1-5% 0.3% 0.3% 0.0% 0.8% 0.0% 0.0% 0.0% 5-10% 0.9% 1.1% 0.1% 1.8% 0.0% 0.3% -0.3% Quintiles 1st (bottom 20%) 6.2% 7.1% 0.7% 9.4% 0.6% 9.0% -1.0% 2nd (20-40%) 19.1% 19.9% 10.5% 21.7% 16.0% 87.1% -0.5% 3rd (40-60%) 23.8% 24.0% 23.0% 24.5% 35.0% 3.8% 24.0% 4th (60-80%) 24.7% 24.6% 28.7% 26.3% 31.1% 0.0% 18.1% 5th (80-100%) 26.2% 24.5% 37.1% 18.1% 17.3% 0.0% 59.3% Top 90-95% 6.4% 6.0% 9.3% 5.0% 3.3% 0.0% 10.7% 95-99% 5.4% 4.9% 8.9% 3.6% 4.6% 0.0% 13.1% 1% 1.9% 1.3% 2.2% 0.5% 1.3% 0.0% 24.3% Other Notes: This table characterizes the distribution of tax credits across income groups. Tax credits are amounts that directly reduce tax liabilities. Total tax credits are computed as the difference between tax liabilities following the application of the tax schedule and final tax liabilities. 5 Parametric Estimates In this section we provide estimates of effective average tax functions for applied use. Specifically, we estimate different parametric functions for our samples of individuals and households in 2013, differentiating by marital status for the latter. We also estimate separate parametric functions for the different components of the Spanish tax returns, namely the general tax rates, the savings tax rates and the tax credits. In addition, we provide estimates for the different cross-sections of individuals between 2002 and Finally, we estimate the effective average tax function under the 2015 legislation using the BdE-PIT microsimulation model. 5.1 Taxes in Individuals A significant number of Spanish taxpayers face zero tax rates, therefore we opt for estimating the following function: 0 if ỹ < t(ỹ) = Ī, (1) f(ỹ) if ỹ Ī, where t is the average tax rate, ỹ stands for multiples of mean gross income, Ī is chosen so as to minimize the mean squared error and f(ỹ) is a parsimonious non-linear function that takes one of 18

19 the following four specifications: The log specification which was used by Guner et al (2012a, 2012b) f(ỹ) = α + βlog(ỹ) (2) The HSV specification, used by Benabou (2002) and Heathcote et al. (2011) The Power specification, used by Guvenen et al. (2014) f(ỹ) = 1 λ(ỹ) τ (3) f(ỹ) = δ + γỹ ɛ (4) And the GS specification: Gouveia and Strauss (1994), in which ỹ is replaced by y, i.e. the income level. f(y) = b [ ] 1 (sy p + 1) 1 p Table VIII column (1) shows the parameter estimates for individuals in 2013 for the four specifications. Other columns of the table show the parameter estimates for other samples defined in the next subsections. We estimate the log and HSV functions by OLS and the Power and GS by NLS. Figure V illustrates the average tax rates predicted by the estimated parametric functions alongside the the data averages for each bin. Figure V Effective Average Tax Rates 2013: Functions (Individuals) (5) Average Tax Rates Multiples of Mean Data Log HSV Power GS Figure VI shows the mean squared error (MSE) under the four different specification, with the minimum of each function determining the corresponding threshold Ī. 19

20 Table VIII Parameter Estimates Functions Individuals 2013 Households 2013 Individuals 2015 (1) (2) (3) Log α (0.0000) (0.0001) (0.0000) β (0.0001) (0.0001) (0.0001) Ī 44% 34% 47% MSE HSV λ (0.0000) (0.0001) (0.0000) τ (0.0001) (0.0001) (0.0001) Ī 47% 36% 49% MSE Power δ (0.0000) (0.0001) (0.0000) γ (.) (.) (.) ɛ (0.0000) (0.0000) (0.0000) Ī 44% 33% 47% MSE GS b (0.0004) (0.0007) (0.0004) s (0.0000) (0.0000) (0.0000) p (0.0059) (0.0085) (0.0068) Ī 53% 42% 57% MSE Notes: Households For households we follow the same estimation approach as for individuals. Table VIII shows the parameter estimates and Figure VII plots our preferred tax functions (HSV and GS) for all households (Panel A) and distinguishing between single and married households (Panel B) Three-function Approach An alternative to the one-function approach is to estimate the three different tax functions that lead to final tax liabilities (c.f. Figure I): the general tax rates, applied to general taxable income (base liquidable general); the savings tax rates, applied to savings taxable income (base liquidable 20

21 Figure VI Evaluation of Tax Functions: Mean Squared Error Mean Squared Error Multiples of Mean Log Power HSV GS Figure VII Households Average Tax Rates 2013: Functions Panel A: All households Panel B: Single and married households Average Tax Rates Multiples of Mean Average Tax Rates Multiples of Mean Data HSV GS Singles Married Notes: del ahorro); and the tax credits (deducciones sobre la cuota). This approach allows to simulate more detailed reforms such as changes in capital taxes. Functional forms For the general taxable income we estimate the same functional form (1) as for the effective average tax function under our preferred specifications GS and HSV. For the savings income, we estimate the following parametric function 21

22 c + ζỹ s t s (ỹ s ) = κ if ỹ s < S, if ỹ s S, (6) where κ is the sample mean of the average savings tax rate if ỹ s S and S is chosen so as to minimize the MSE. Note that S = And for the tax credits we estimate the Ricker Model: t tc (ỹ) = β 0 + e β 1 e β 2ỹỹ β 3 (7) Table IX shows the estimated parameters for all these functions for individuals in Figure VIII plots the general income tax functions estimated, Figure IX shows the estimated function for savings income tax rates, and Figure X shows the Tax Credit function. Figure VIII Parametric Functions for the General Tax Rates General Average Tax Rates Multiples of Mean General Data HSV GS Notes: Evaluation of Tax Functions To evaluate the fit of different specifications across the tax liabilities distribution we calculate the percentage difference of the predicted total tax liabilities with respect to the data under the onefunction and three-function approaches for the HSV and GS specifications. The result is shown in Table X. Table XI shows a similar comparison but in terms of the distribution of tax liabilities. 22